Guide to Climatological Practices

GUIDE TO CLIMATOLOGICAL PRACTICES

2011 edition

P-CLW_101264

WMO-No. 100

www.wmo.int WMO-No. 100

 Guide to
Climatological Practices

WMO-No. 100

2011
 © 2011, World Meteorological Organization, Geneva

ISBN 978-92-63-10100-6

Note

The designations employed and the presentation of material in this publication do not imply the expression of
any opinion whatsoever on the part of the Secretariat of the World Meteorological Organization concerning the
legal status of any country, territory, city or area, or of its authorities, or concerning the delimitation of its
frontiers or boundaries.
 CONTENTS

Page

PREFACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

CHAPTER 1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–1

1.1 Purpose and content of the Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–1
1.2 Climatology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–1
1.2.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–1
1.2.2 The climate system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–2
1.2.3 Uses of climatological information and research . . . . . . . . . . . . . . . . . . . . . . . . . . 1–5
1.3 International climate programmes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–6
1.4 Global and regional climate activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–6
1.5 National climate activities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–7
1.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–9
1.6.1 WMO publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–9
1.6.2 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–9
CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS. . . . . . . . . . . . . . . . . . . . . . 2–1
2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–1
2.2 Climatic elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–1
2.2.1 Surface and subsurface elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–2
2.2.2 Upper-air elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–3
2.2.3 Elements measured by remote-sensing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–5
2.3 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–5
2.3.1 Basic surface equipment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–6
2.3.2 Upper-air instruments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–7
2.3.3 Surface-based remote-sensing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–7
2.3.4 Aircraft-based and space-based remote-sensing . . . . . . . . . . . . . . . . . . . . . . . . . . 2–8
2.3.5 Calibration of instruments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–10
2.4 The siting of climatological stations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–10
2.5 The design of climatological networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–12
2.6 Station and network operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–13
2.6.1 Times of observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–13
2.6.2 Logging and reporting of observations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–13
2.6.3 On-site quality control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–14
2.6.4 Overall responsibilities of observers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–14
2.6.5 Observer training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–14
2.6.6 Station inspections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–15
2.6.7 Preserving data homogeneity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–16
2.6.8 Report monitoring at collection centres. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–16
2.6.9 Station documentation and metadata. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–16
2.7 References and additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–17
2.7.1 WMO publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–17
2.7.2 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–18
CHAPTER 3. CLIMATE DATA MANAGEMENT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–1
3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–1
3.2 The importance and purpose of managing data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–2
3.3 Climate data management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–2
3.3.1 CDMS design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–3
3.3.2 CDMS data acquisition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–3
iv GUIDE TO CLIMATOLOGICAL PRACTICES

Page

3.3.3 CDMS data documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–4

3.3.4 CDMS data storage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–5
3.3.5 CDMS data access and retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–5
3.3.6 CDMS archives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–6
3.3.7 CDMS security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–6
3.3.8 CDMS management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–6
3.3.9 International CDMS standards and guidelines. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–6
3.4 Quality control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–7
3.4.1 Quality control procedures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–8
3.4.2 Quality control documentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–9
3.4.3 Types of error. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–9
3.4.4 Format tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–9
3.4.5 Completeness tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–9
3.4.6 Consistency tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–10
3.4.7 Tolerance tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–11
3.5 Exchange of climatic data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–11
3.6 Data rescue. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–12
3.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–13
3.7.1 WMO publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–13
3.7.2 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–14
CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–1
4.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–1
4.2 Dataset evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–1
4.3 Qualitative visual displays of data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–1
4.4 Quantitative summary descriptors of data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–2
4.4.1 Data modelling of frequency distributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–3
4.4.2 Measures of central tendency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–8
4.4.3 Measures of variability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–9
4.4.4 Measure of symmetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–11
4.4.5 Measure of peakedness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–11
4.4.6 Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–11
4.5 Correlation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–12
4.5.1 Contingency tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–12
4.5.2 Measures of correlation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–12
4.6 Time series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–13
4.7 Interpretation of summary characteristics of climate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–14
4.8 Normals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–15
4.8.1 Period of calculation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–16
4.8.2 Stations for which normals and averages are calculated . . . . . . . . . . . . . . . . . . . . 4–16
4.8.3 Homogeneity of data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–16
4.8.4 Missing data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–17
4.8.5 Average daily temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–17
4.8.6 Precipitation quintiles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–18
4.8.7 Dissemination of normals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–18
4.9 References and additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–18
4.9.1 WMO publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–18
4.9.2 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–18
CHAPTER 5. STATISTICAL METHODS FOR ANALYSING DATASETS. . . . . . . . . . . . . . . . . . . . . . . . 5–1
5.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–1
5.2 Homogenization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–1
5.2.1 Evaluation of homogenized data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–4
5.3 Model-fitting to assess data distributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–5
 CONTENTS v

Page

5.4 Data transformation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–6

5.5 Time series analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–6
5.6 Multivariate analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–7
5.7 Comparative analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–8
5.8 Smoothing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–9
5.9 Estimating data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–10
5.9.1 Mathematical estimation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–11
5.9.2 Estimation based on physical relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–11
5.9.3 Spatial estimation methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–11
5.9.4 Time series estimation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–12
5.9.5 Validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–12
5.10 Extreme value analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–13
5.10.1 Return period approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–13
5.10.2 Probable maximum precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–14
5.11 Robust statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–14
5.12 Statistical packages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–14
5.13 Data mining. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–15
5.14 References and additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–15
5.14.1 WMO publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–15
5.14.2 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–16
CHAPTER 6. SERVICES AND PRODUCTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–1
6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–1
6.2 Users and uses of climatological information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–1
6.3 Interaction with users. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–2
6.4 Information dissemination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–3
6.5 Marketing of services and products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–4
6.6 Products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–5
6.6.1 General guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–5
6.6.2 Climatological data periodicals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–6
6.6.3 Occasional publications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–7
6.6.4 Standard products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–7
6.6.5 Specialized products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–7
6.6.6 Climate monitoring products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–8
6.6.7 Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–8
6.7 Climate models and climate outlooks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–9
6.7.1 Climate outlook products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–9
6.7.2 Climate predictions and projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–10
6.7.3 Climate scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–10
6.7.4 Global climate models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–10
6.7.5 Downscaling: Regional climate models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–11
6.7.6 Local climate models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–11
6.8 Reanalysis products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–11
6.9 Examples of products and data displays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–12
6.10 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–20
6.10.1 WMO publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–20
6.10.2 Additional reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–20
ANNEX 1. ACRONYMS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ann–1

ANNEX 2. INTERNATIONAL CLIMATE ACTIVITIES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ann–2

 PREFACE

Since 1983, when the second edition of the Guide to a substantially revised edition in the light of the
Climatological Practices (WMO-No. 100) was released, progress made in climatology during the preceding
climate-related activities have expanded in virtually decade and in the use of climatological information
every area of human life, and this is particularly true and knowledge in various areas of meteorology and
in science and public policy. The Guide to other disciplines. The seventh session of the
Climatological Practices is a key resource that is Commission re-established the working group,
designed to help Members provide a seamless stream which continued to finalize the work on the second
of crucial information for daily practices and opera- edition on a chapter-by-chapter basis, producing
tions in National Meteorological Services (NMSs). the version that was ultimately published in 1983.

One of the purposes of the World Meteorological The work on the third edition of the Guide started
Organization, as laid down in the WMO in 1990 when the content and authorship were
Convention, is to promote the standardization of approved by the advisory working group of CCl at a
meteorological and related observations, including meeting in Norrköping, Sweden. An Editorial Board
those that are applied to climatological studies and on the CCl Guide was subsequently established to
practices. With this aim, the World Meteorological supervise individual lead authors and chapter
Congress adopts from time to time Technical editors. Nevertheless, it was not until 1999 that the
Regulations that lay down the meteorological prac- lead authors received a draft summary to further
tices and procedures to be followed by the develop the text for the Guide. In the following
Organization’s Member countries. The Technical year, the Editorial Board met in Reading, United
Regulations are supplemented by a number of Kingdom of Great Britain and Northern Ireland,
Guides, which describe in more detail the practices, and defined further details and content for each
procedures and specifications that Members are chapter. In 2001, the thirteenth session of the
expected to follow or implement in establishing Commission for Climatology (CCl-XIII) decided to
and conducting their arrangements in compliance establish an Expert Team on the Guide with clear
with the Technical Regulations and in otherwise terms of reference to expedite the process. While
developing their meteorological and climatological Part I of the publication had been substantially
services. One of the publications in this series is the completed and made available on the Web, a major
Guide to Climatological Practices, the aim of which is effort was needed to finalize Part II and the presen-
to provide, in a convenient form for all concerned tation of information on specialized requirements
with the practice of climatology, information about for the provision of climate services. The fourteenth
those practices and procedures that are of the great- session of the Commission (CCl-XIV) re-established
est importance for the successful implementation the Expert Team on the Guide and agreed that some
of their work. Complete descriptions of the theo- overarching activities would be the responsibility of
retical bases and the range of applications of the Management Group. Those included the further
climatological methods and techniques are beyond development of Part II of the Guide and further
the scope of this guide, although references to such work towards the review and designation of
documentation are provided wherever applicable. Regional Climate Centres (RCCs). The Expert Team
met in Toulouse, France, in 2005 and decided to
The first edition of the Guide to Climatological compile a full, integrated draft text, including
Practices was published in 1960 on the basis of annexes, of the third edition of the Guide.
material developed by the Commission for
Climatology (CCl); it was edited by a special work- With the collective effort and expertise rendered by
ing group, with the assistance of the Secretariat. a large number of authors, editors and internal and
The second edition of the Guide originated at the external reviewers, the text of the third edition of
sixth session of the Commission for Special the Guide was finally approved by the President of
Applications of Meteorology and Climatology. The CCl just before the fifteenth session of the
Commission instructed the working group respon- Commission for Climatology (CCl-XV), held in
sible for the Guide to arrange for the preparation of Antalya, Turkey, in February 2010.
viii GUIDE TO CLIMATOLOGICAL PRACTICES

This edition of the Guide will be published in the Director of the National Climatic Data Center,
six official languages of WMO to maximize dissemi- Asheville (United States of America), and Mr Ned
nation of knowledge. As with the previous versions, Guttman (United States), the lead of the Expert
WMO Members may translate this Guide into their Team on the Guide, who served as a consultant
national languages. with patience and close attention to bring the work
on this publication to a successful conclusion.
It is a pleasure to express my gratitude to the WMO
Commission for Climatology for taking the
initiative to oversee this long process. On behalf of
the World Meteorological Organization, I also wish
to express thanks to all those who have contributed
to the preparation of this publication. Special
recognition is due to Mr Pierre Bessemoulin, former
president of the Commission for Climatology, who
guided and supervised the preparation of the text
during the fourteenth intersessional period of the
Commission. I also wish to recognize the significant M. Jarraud
contributions of Mr Kenneth Davidson, Deputy Secretary-General
 CHAPTER 1

INTRODUCTION

1.1 PURPOSE AND CONTENT OF THE 1.2 CLIMATOLOGY

GUIDE
Climatology is the study of climate, its variations
This publication is designed to provide guidance and extremes, and its influences on a variety of
and assistance to World Meteorological activities including (but far from limited to) human
Organization (WMO) Members in developing health, safety and welfare. Climate, in a narrow
national activities linked to climate information sense, can be defined as the average weather condi-
and services. There have been two previous tions for a particular location and period of time.
editions of the Guide: the original publication, Climate can be described in terms of statistical
which appeared in 1960, and the second edition, descriptions of the central tendencies and variabil-
which was published in 1983. While many basic ity of relevant elements such as temperature,
fundamentals of climate science and climatologi- precipitation, atmospheric pressure, humidity and
cal practices have remained consistent over time, winds, or through combinations of elements, such
scientific advances in climatological knowledge as weather types and phenomena, that are typical
and data analysis techniques, as well as changes of a location or region, or of the world as a whole,
in technology, computer capabilities and instru- for any time period.
mentation, have made the second edition
obsolete.
1.2.1 History

The third edition describes basic principles and Early references to the weather can be found in the
modern practices important in the development poems of ancient Greece and in the Old Testament
and implementation of all climate services, and of the Judaeo-Christian Bible. Even older references
outlines methods of best practice in climatology. It appear in the Vedas, the most ancient Hindu scrip-
is intended to describe concepts and considera- tures, which were written about 1800 B.C. Specific
tions, and provides references to other technical writings on the theme of meteorology and clima-
guidance and information sources, rather than tology are found in Hippocrates’ Air, Waters and
attempting to be all-inclusive in the guidance Places, dated around 400 B.C., followed by Aristotle’s
presented. Meteorologica, written around 350 B.C. To the early
Greek philosophers, climate meant “slope” and
This first chapter includes information on clima- referred to the curvature of the Earth’s surface,
tology and its scope, the organization and which gives rise to the variation of climate with
functions of a national climate service, and inter- latitude due to the changing incidence of the Sun’s
national climate programmes. The remainder of rays. Logical and reliable inferences on climate are
the Guide is broken down into five chapters to be found in the work of the Alexandrian philoso-
(Climate Observations, Stations and Networks; phers Eratosthenes and Aristarchus.
Climate Data Management; Characterizing
Climate from Datasets; Statistical Methods for With the onset of extensive geographical explora-
Analysing Datasets; Services and Products) and tion in the fifteenth century, descriptions of the
two annexes (Acronyms and International Climate Earth’s climates and the conditions giving rise to
Activities). the climates started to emerge. The invention of
meteorological instruments such as the thermome-
Procedures in the Guide have been taken, where ter in 1593 by Galileo Galilei and the barometer in
possible, from decisions on standards and recom- 1643 by Evangelista Torricelli gave a greater impulse
mended practices and procedures. The main to the establishment of mathematical and physical
decisions concerning climate practices are relationships between the different characteristics
contained in the WMO Technical Regulations, of the atmosphere. This in turn led to the establish-
Manuals, and reports of the World Meteorological ment of relationships that could describe the state
Congress and the Executive Council, and origi- of the climate at different times and in different
nate mainly from recommendations of the places.
Commission for Climatology. For additional
assistance and information, lists of appropriate The observed pattern of circulation linking the
WMO and other publications of particular inter- tropics and subtropics, including the trade winds,
est to those working in climatology are provided tropical convection and subtropical deserts, was
in the references. first interpreted by George Hadley in 1735, and
1–2 GUIDE TO CLIMATOLOGICAL PRACTICES

subsequently became known as the Hadley cell. being directed at other aspects of climatology as
Julius von Hann, who published the first of three well. These efforts include improved measurement
volumes of the Handbook of Climatology in 1883, and monitoring of climate, increased understand-
wrote the classic work on general and regional ing of the causes and patterns of natural variability,
climatology, which included data and eyewitness more reliable methods of predicting climate over
descriptions of weather and climate. In 1918 seasons and years ahead, and better understanding
Wladimir Köppen produced the first detailed clas- of the linkages between climate and a range of
sification of world climates based on the vegetative social and economic activities and ecological
cover of land. This endeavour was followed by more changes.
detailed developments in descriptive climatology.
The geographer E.E. Federov, for example, attempted
1.2.2 The climate system
to describe local climates in terms of daily weather
observations. The climate system (Figure 1.1) is a complex, inter-
active system consisting of the atmosphere, land
In the first thirty years of the twentieth century, the surface, snow and ice, oceans and other bodies of
diligent and combined use of global observations water, and living organisms. The atmosphere is the
and mathematical theory to describe the atmos- gaseous envelope surrounding the Earth. The dry
phere led to the identification of large-scale atmosphere consists almost entirely of nitrogen
atmospheric patterns. Notable in this field was Sir and oxygen, but also contains small quantities of
Gilbert Walker, who conducted detailed studies of argon, helium, carbon dioxide, ozone, methane
the Indian monsoon, the Southern Oscillation, the and many other trace gases. The atmosphere also
North Atlantic Oscillation and the North Pacific contains water vapour, condensed water droplets in
Oscillation. the form of clouds, and aerosols. The hydrosphere
is that part of the Earth’s climate system comprising
Other major works on climatology included those liquid water distributed on and beneath the Earth’s
by Tor Bergeron (on dynamic climatology in 1928) surface in oceans, seas, rivers, freshwater lakes,
and Wladimir Köppen and Rudolf Geiger (who underground reservoirs and other water bodies. The
produced a climatology handbook in 1936). Geiger cryosphere collectively describes elements of the
first described the concept of microclimatology in Earth system containing water in its frozen state
some detail in 1927, but the development of this and includes all snow and ice (sea ice, lake and river
field did not occur until the Second World War. ice, snow cover, solid precipitation, glaciers, ice
During the war, for planning purposes, a probabil- caps, ice sheets, permafrost and seasonally frozen
ity risk concept of weather data for months or even ground). The surface lithosphere is the upper layer
years ahead was found to be necessary and was of the solid Earth, including both the continental
tried. C.W. Thornthwaite established a climate clas- crust and the ocean floor. The biosphere comprises
sification in 1948 based on a water budget and all ecosystems and living organisms in the atmos-
evapotranspiration. In the following decades devel- phere, on land (terrestrial biosphere) and in the
opment of theories on climatology saw major oceans (marine biosphere), including derived dead
progress. organic matter, such as litter, soil organic matter
and oceanic detritus.
The creation of WMO in 1950 (as the successor to
the International Meteorological Organization, Under the effects of solar radiation and the radia-
which was founded in 1873) established the organi- tive properties of the surface, the climate of the
zation of a system of data collection and led to the Earth is determined by interactions among the
systematic analysis of climate and to conclusions components of the climate system. The interaction
about the nature of climate. During the latter of the atmosphere with the other components plays
decades of the twentieth century, the issue of a dominant role in forming the climate. The atmos-
climate change began to focus attention on the phere obtains energy directly from incident solar
need to understand climate as a major part of a radiation or indirectly via processes involving the
global system of interacting processes involving all Earth’s surface. This energy is continuously redis-
of the Earth’s major domains (see section 1.2.2). tributed vertically and horizontally through
Climate change is defined as a statistically signifi- thermodynamic processes or large-scale motions
cant variation in either the average state of the with the unattainable aim of achieving a stable and
climate or in its variability, persisting for an balanced state of the system. Water vapour plays a
extended period, typically decades or longer. It may significant role in the vertical redistribution of heat
be caused by natural internal processes, external through condensation and latent heat transport.
forcing or persistent anthropogenic (resulting from The ocean, with its vast heat capacity, limits the
or produced by human activity) changes in the rate of temperature change in the atmosphere and
composition of the atmosphere or in land use. supplies water vapour and sensible heat to the
Considerable national and international efforts are atmosphere. The distribution of the continents
 CHAPTER 1. INTRODUCTION 1–3

Figure 1.1. The climate system

affects oceanic currents, and mountains redirect linked to the nature and scope of human health
atmospheric motions. The polar, mountain and sea and welfare. It also defines and determines the
ice reflects solar radiation back to space. In high impact of major features of the global circulation
latitudes the sea ice acts as an insulator and protects such as the El Niño–Southern Oscillation (ENSO),
the ocean from rapid energy loss to the much colder the monsoons and the North Atlantic Oscillation.
atmosphere. The biosphere, including its human
activities, affects atmospheric components such as A temporal scale is an interval of time. It can range
carbon dioxide, as well as features of the Earth’s from minutes and hours to decades, centuries, and
surface such as soil moisture and albedo. longer. The characteristics of an element over an
hour are important, for example, in agricultural
Interactions among the components occur on all operations such as pesticide control and in moni-
scales (Figures 1.2 and 1.3). Spatially, the micro­ toring energy usage for heating and cooling. The
scale encompasses features of climate characteristics characteristics of an element over a day might
over small areas such as individual buildings and determine the human activities that be safely
plants or fields. A change in microclimate can be of pursued. The climate over months or years will
major importance when the physical characteristics determine, for example, the crops that can be
of an area change. New buildings may produce grown or the availability of drinking water and
extra windiness, reduced ventilation, excessive food. Longer timescales of decades and centuries
runoff of rainwater, and increased pollution and are important for studies of climate variation caused
heat. Natural variations in microclimate, such as by natural phenomena such as atmospheric and
those related to shelter and exposure, sunshine and oceanic circulation changes and by the activities of
shade, are also important: they can determine, for humans.
example, which plants will prosper in a particular
location or the need to make provisions for safe Climate change has become a major issue for the
operational work and leisure activities. The meso­ human community. Human activities, especially
scale encompasses the climate of a region of limited the burning of fossil fuels, have led to changes in
extent, such as a river catchment area, valley, the composition of the global atmosphere. Marked
conurbation or forest. Mesoscale variations are increases in tropospheric carbon dioxide and meth-
important in applications including land use, irri- ane during the industrial era, along with increased
gation and damming, the location of natural energy aerosols and particulate emissions, are significantly
facilities, and resort location. The macroscale affecting global climate. Chlorofluorocarbons
encompasses the climate of large geographical extensively used in the past as aerosol propellants,
areas, continents and the globe. It determines cleaning fluids and refrigerants are a main cause of
national resources and constraints in agricultural stratospheric ozone depletion. Over one fifth of the
production and water management, and is thus world’s tropical forests were cleared between 1960
1–4 GUIDE TO CLIMATOLOGICAL PRACTICES

100 000 Years

10 000 Years
Millenniums
1 000 km

Soil micro-organism

Figure 1.2. Temporal and spatial scales (courtesy Todd Albert, United States)

Figure 1.3. Lifetime of atmospheric phenomena (after J.W. Zillman, WMO Bulletin, Vol. 48(2), 1999)
 CHAPTER 1. INTRODUCTION 1–5

and 2000, likely altering the complex mesoscale applications of climate knowledge for societal
and global hydrological cycles. Artificial canyons benefits. Previously, the study of climate provided
formed in cities by buildings, together with asphalt basic data, information and techniques to delineate
road surfaces, increase the amount of radiation local, mesoscale and global climates. While these
absorbed from the Sun and form urban heat islands. are primary deliverables, they are also the raw mate-
Accelerated runoff of rainwater and the removal of rial for deeper analysis and services when coupled
trees and other vegetation reduce the amount of and analysed with other social, economic and
transpired water vapour that would otherwise help physical data. The crucial roles of climate data and
to moderate temperature. Pollution from vehicles climate predictions in planning for disaster mitiga-
and buildings accumulates, especially in calm tion and sustainable development, and in
conditions, and causes many human health prob- addressing all the consequences of climate change,
lems and damage to structures. are now firmly established within various conven-
tions, such as the United Nations Framework
Conscious of the growing worldwide concern about Convention on Climate Change.
the danger of irreversible changes occurring in the
natural environment, WMO has taken the lead in Applied climatology makes maximum use of mete-
promoting studies of changes in the climate system orological and climatological knowledge and
and their effects on mankind, on world energy and information for solving practical social, economic
food production, and on water reserves. Climate and environmental problems. Climatological
change and its potential consequences, as well as services are designed for a variety of public,
those effects that have already occurred, have commercial and industrial users. Assessments of the
become key topics occupying decision-makers in effects of climate variability and climate change on
recent years, and in some countries these concerns human activities, as well as the effects of human
are second only to economic and defence matters. activities on climate, are major factors in local,
Even in these latter two areas, climate is involved in national and global economic development, social
strategic planning and tactical decision-making. programmes, and resource management.
Many international conferences have been held to
devise ways of reducing the human impact on Current interest in the impact of economic devel-
climate and to design strategies to exploit climate for opment and other human activities on climate and
social and economic benefit. These meetings include how climate variability and change influence
the World Climate Conferences in Geneva in 1979, human societies highlights the need for further
1990 and 2009; the United Nations Conference on research into the physical and dynamical processes
Environment and Development in Rio de Janeiro in involved in the climate system, as well as the need
1992; and the World Summit on Sustainable for their statistical description. Understanding of
Development in Johannesburg in 2002. The estab- natural climate variability, appreciation of climate
lishment of the Intergovernmental Panel on Climate sensitivity to human activities, and insight on the
Change (IPCC) and the Intergovernmental predictability of weather and climate for periods
Negotiating Committee for the United Nations ranging from days to decades are fundamental to
Framework Convention on Climate Change also improving our capability to respond effectively to
marked important milestones in addressing changes economic and societal problems. Physical climatol-
in climate that are related to human activities. ogy embraces a wide range of studies that include
interactive processes of the climate system.
Dynamic climatology is closely related to physical
1.2.3 Uses of climatological information
climatology, but it is mainly concerned with
and research
patterns in the general circulation of the atmos-
Climatology has become a dynamic branch of phere. Both involve the description and study of
science with a broad range of functions and appli- the properties and behaviour of the atmosphere.
cations. New techniques are being developed and
investigations are being undertaken to study the Improving the prediction of climate is now a substan-
application of climate in many fields, including tial global activity. Initially, predictions were based
agriculture, forestry, ecosystems, energy, industry, on empirical and statistical techniques, but more
production and distribution of goods, engineering and more now they derive from expanded numeri-
design and construction, human well-being, trans- cal weather prediction techniques. Increasingly
portation, tourism, insurance, water resources and complex models that represent and couple together
disaster management, fisheries, and coastal devel- the atmosphere, oceans, land interface, sea ice, and
opment. To improve the ability of climatologists to atmospheric aerosols and gases are being developed.
inform and advise users and to answer a myriad of The models can be used to simulate climate change
questions about climate, there is a continuing need over several decades and also to predict seasonal or
for viable and useful research programmes on the interannual variations in climate. Such seasonal
climate system and its broad influence, and on the outlooks generally take the form of the probability
1–6 GUIDE TO CLIMATOLOGICAL PRACTICES

that the value of an element, such as the mean This understanding requires climate observations,
temperature or aggregated rainfall over a period, will management and transmission of data, various data
be above, near or below normal. Seasonal outlooks services, climate system monitoring, practical appli-
presently show skill for regions where there is a cations and services for different user groups,
strong relationship between sea surface temperature forecasts on subseasonal and interannual time­
and weather, such as in many tropical areas. Because scales, climate projections, policy-relevant
of their probabilistic nature, however, much care is assessments of climate variability and change, and
needed in their communication and application. research priorities that increase the potential bene-
Decision-making that incorporates climate informa- fits of all these activities. Many countries, especially
tion is a growing area of investigation. the developing and least developed countries, may
not have sufficient individual capacity to perform
All climate products and services, from information all of these services. The World Climate
derived from past climate and weather data to esti- Conference-3 in Geneva in 2009 proposed the crea-
mations of future climate, for use in research, tion of a Global Framework for Climate Services to
operations, commerce and government, are under- strengthen the production, availability, delivery
pinned by data obtained by extensively and and application of science-based climate prediction
systematically observing and recording a number of and services. The Framework is intended to provide
key variables that enable the characterization of a mechanism for developers and providers of
climate on a wide range of timescales. The adequacy climate information, as well as climate-sensitive
of a climate service is highly dependent on the sectors around the world, to work together to help
spatial density and accuracy of the observations the global community better adapt to the chal-
and on the data management processes. Without lenges of climate variability and change.
systematic observations of the climate system, there
can be no climate services. The World Meteorological Organization has devel-
oped a network of Global Producing Centres for
The need for more accurate and timely information Long-Range Forecasts (GPCs) and Regional Climate
continues to increase rapidly as the diversity of users’ Centres (RCCs) to assist Member countries cope
requirements continues to expand. It is in the inter- effectively with their climate information needs. The
est of every country to apply consistent practices in definitions and mandatory functions of GPCs and
performing climate observations, in handling RCCs are contained in the Manual on the Global Data-
climate records and in maintaining the necessary processing and Forecasting System (Volume I – Global
quality and utility of the services provided. Aspects, WMO-No. 485) and are part of the WMO
Technical Regulations. The Manual also provides the
criteria for the designation of GPCs, RCCs and other
operational centres by WMO.
1.3 INTERNATIONAL CLIMATE
PROGRAMMES The designated GPCs produce global long-range fore-
casts according to the criteria defined in the Manual
The WMO Commission for Climatology (CCl) is on the Global Data-processing and Forecasting System
concerned with the overall requirements of WMO and are recognized by WMO based on the recommen-
Members for advice, support and coordination in dation of the Commission for Basic Systems. In
many climate activities. The Commission has been addition, WMO has established the Lead Centre for
known by slightly different names and has seen its Long-Range Forecast Multi-Model Ensembles and the
Terms of Reference change in accordance with Lead Centre for the Standardized Long-Range Forecast
changing demands and priorities, but it has effec- Verification System, which provide added value to the
tively been in operation since it was established in operational services of GPCs.
1929 under the International Meteorological
Organization. It provides overall guidance for the The Regional Climate Centres are designed to assist
implementation of the World Climate Programme WMO Members in a given region to deliver better
within WMO. Additional details about interna- and more consistent climate services and products,
tional climate programmes are contained in such as regional long-range forecasts, and to
Annex 2. strengthen the capability of Members to meet
national climate information needs. The primary
clients of an RCC are National Meteorological and
Hydrological Services (NMHSs) and other RCCs in
1.4 GLOBAL AND REGIONAL CLIMATE the region and in neighbouring areas. The services
ACTIVITIES and products from the RCCs are provided to the
NMHSs for further definition and dissemination and
All countries should understand and provide for are not distributed to users without the permission
the climate-related information needs of the public. of the NMHSs within the region. The responsibilities
 CHAPTER 1. INTRODUCTION 1–7

of an RCC do not duplicate or replace those of Recognizing that climate information can be of
NMHSs. It is important to note that NMHSs retain substantial benefit in adapting to and mitigating
the mandate and authority to provide the liaison the impacts of climate variability and change,
with national user groups and to issue advisories and WMO has helped establish Regional Climate
warnings, and that all RCCs are required to adhere to Outlook Forums. Using a predominantly consensus-
the principles of WMO Resolution 40 concerning based approach, the forums have an overarching
the exchange of data and products. responsibility to produce and disseminate a
regional assessment of the state of the regional
The complete suite of products and services of RCCs climate for the upcoming season. The forums
can vary from one region to another, based on the bring together national, regional and interna-
priorities established by the relevant Regional tional climate experts on an operational basis to
Association. There will, however, be certain essen- produce regional climate outlooks based on input
tial functions all WMO-designated RCCs must from NMHSs, regional institutions, RCCs and
perform to established criteria, thus ensuring a GPCs. They facilitate enhanced feedback from
certain uniformity of service around the globe in the users to climate scientists, and catalyse the
RCC mandatory functions. These functions include: development of user-specific products. They also
(a) Operational activities for long-range forecast- review impediments to the use of climate infor-
ing, including interpretation and assessment mation, share successful lessons regarding
of relevant output products from GPCs, gener- applications of past products and enhance sector-
ation of regional and subregional tailored specific applications. The forums often lead to
products, and preparation of consensus state- national forums for developing detailed national-
ments concerning regional or subregional scale climate outlooks and risk information,
forecasts; including warnings, for decision-makers and the
(b) Climate monitoring, including regional and public.
subregional climate diagnostics, analysis of
climate variability and extremes, and imple- The Regional Climate Outlook Forum process,
mentation of regional Climate Watches for which can vary in format from region to region,
extreme climate events; typically includes at least the first of the following
(c) Data services to support long-range forecast- activities and, in some instances, all four:
ing, including the development of regional (a) Meetings of regional and international climate
climate datasets; experts to develop a consensus for regional
(d) Training in the use of operational RCC prod- climate outlooks, usually in a probabilistic
ucts and services. form;
(b) A broader forum involving both climate
In addition to these mandatory RCC functions, a scientists and representatives from user
number of activities are highly recommended. sectors, for the presentation of consensus
Some of these activities are downscaling of climate climate outlooks, discussion and identifica-
change scenarios, non-operational data services tion of expected sectoral impacts and impli-
such as data rescue and data homogenization, coor- cations, and the formulation of response
dination functions, training and capacity-building, strategies;
and research and development. (c) Training workshops on seasonal climate
prediction to strengthen the capacity of
Regional associations may also establish centres national and regional climate scientists;
that conduct various climate functions as speci- (d) Special outreach sessions involving media
fied in the Manual on the Global Data-processing experts to develop effective communications
and Forecasting System (Volume II – Regional strategies.
Aspects). This volume does not fall under the
WMO Technical Regulations, so these centres are
not required to go through the formal designa-
tion procedure. Regional associations have full 1.5 NATIONAL CLIMATE ACTIVITIES
responsibility for developing and approving the
requirements for such centres. These centres In most countries, NMHSs have long held key
often play an important participatory role in responsibilities for national climate activities,
regional climate networks. It must be noted, including the making, quality control and storage
however, that the term “WMO RCC” is reserved of climate observations; the provision of climato-
exclusively for those entities formally designated logical information; research on climate; climate
under Volume I – Global Aspects of the Manual on prediction; and the applications of climate knowl-
the Global Data-processing and Forecasting System, edge. There has been, however, an increasing
and therefore should not be used to refer to any contribution to these activities from academia and
other centre. private enterprise.
1–8 GUIDE TO CLIMATOLOGICAL PRACTICES

Some countries have within their NMHSs a single promote and market the use of the information;
division responsible for all climatological activities. make available its expertise to interpret the data;
In other countries the NMHS may find it beneficial and advise on the use of the data (see Chapter 6).
to assign responsibilities for different climatological
activities (such as observing, data management and An NMHS should maintain a continuing research
research) to different units within the Service. The and development programme or establish working
division of responsibilities could be made on the relationships with an institution that has research
basis of commonality of skills, such as across synop- and development capabilities directly related to the
tic analysis and climate observation, or across climatological functions and operations of the
research in weather and climate prediction. Some NMHS. The research programme should consider
countries establish area or branch offices to handle new climate applications and products that increase
subnational activities, while in other cases the user understanding and application of climate infor-
necessary pooling and retention of skills for some mation. Studies should explore new and more
activities are achieved through a regional coopera- efficient methods of managing an ever-increasing
tive entity serving the needs of a group of volume of data, improving user access to the archived
countries. data, and migrating data to digital form. Quality
assurance programmes for observations and summa-
When there is a division of responsibility within an ries should be routinely evaluated with the goal of
NMHS, or in those cases when responsibilities are developing better and timelier techniques. The use
assigned to another institution altogether, it is of information dissemination platforms such as the
essential that a close liaison exist between those Internet should also be developed.
applying the climatological data in research or serv-
ices and those responsible for the acquisition and The meeting of national and international respon-
management of the observations. This liaison is of sibilities, and the building of NMHS capacity
paramount importance in determining the relevant to climate activities, can be achieved only
adequacy of the networks and of the content and if adequately trained personnel are available. Thus,
quality control of the observations. It is also essen- an NMHS should maintain and develop links with
tial that the personnel receive training appropriate training and research establishments dealing with
to their duties, so that the climatological aspects are climatology and its applications. In particular, it
handled as effectively as would be the case in an should ensure that personnel attend training
integrated climate centre or division. If data are programmes that supplement general meteorolog-
handled in several places, it is important to estab- ical training with education and skills specific to
lish a single coordinating authority to ensure that climatology. The WMO Education and Training
there is no divergence among datasets. Programme fosters and supports international
collaboration that includes the development of a
Climatologists within an NMHS should be directly range of mechanisms for continued training, such
accountable for, or provide consultation and advice as fellowships, conferences, familiarization visits,
regarding: computer-assisted learning, training courses and
(a) Planning of station networks; technology transfer to developing countries. In
(b) Location or relocation of climatological addition, other WMO Programmes, such as the
stations; World Climate Programme, the Hydrology and
(c) Care and security of the observing sites; Water Resources Programme and the Agricultural
(d) Regular inspection of stations; Meteorology Programme, undertake capacity-
(e) Selection and training of observers; building activities relevant to climate data,
(f) Instruments or observing systems to be monitoring, prediction, applications and services.
installed so as to ensure that representative
and homogeneous records are obtained (see To be successful, a national climate services
Chapter 2). programme must have a structure that works effec-
tively within a particular country. The structure
Once observational data are acquired, they must be must be one that allows the linkage of available
managed. Functions involved in the management applications, scientific research, technological
of information from observing sites include data capabilities and communications into a unified
and metadata acquisition, quality control, storage, system. The essential components of a national
archiving and access (see Chapter 3). Dissemination climate services programme are:
of the collected climatic information is another (a) Mechanisms to ensure that the climate infor-
requirement. An NMHS must be able to anticipate, mation and prediction needs of all users are
investigate and understand the needs for climato- recognized;
logical information among government (b) Collection of meteorological and related
departments, research institutions and academia, observations, management of databases and
commerce, industry and the general public; the provision of data;
 CHAPTER 1. INTRODUCTION 1–9

(c) Coordination of meteorological, oceano- World Climate Research Programme (WCRP), 2005:
graphic, hydrological and related scientific The World Climate Research Programme Strategic
research to improve climate services; Framework 2005–2015. Coordinated Observation
(d) Multidisciplinary studies to determine national and Prediction of the Earth System (WMO/TD-
risk and sectoral and community vulnerability No. 1291, WCRP-No. 123), Geneva.
related to climate variability and change, to World Meteorological Organization, 1983: Guide to
formulate appropriate response strategies, and Climatological Practices. Second edition (WMO-
to recommend national policies; No. 100), Geneva.
(e) Development and provision of climate infor- ———, 1986: Report of the International Conference on
mation and prediction services to meet user the Assessment of the Role of Carbon Dioxide and of
needs; Other Greenhouse Gases in Climate Variations and
(f) Linkages to other programmes with similar or Associated Impacts (Villach, Austria, 9–15 October
related objectives to avoid unnecessary dupli- 1985) (WMO-No. 661), Geneva.
cation of efforts. ———, 1990: Forty Years of Progress and Achievement:
A Historical Review of WMO (Sir Arthur Davies,
It is important to realize that a national climate serv- ed.) (WMO-No. 721), Geneva.
ices programme constitutes an ongoing process that ———, 1990: The WMO Achievement: 40 Years in the
may change in structure through time. An integral Service of International Meteorology and Hydrology
part of this process is the continual review and incor- (WMO-No. 729), Geneva.
poration of user requirements and feedback in order ———, 1991: Manual on the Global Data-processing
to develop useful products and services. Gathering System. Vol. I – Global Aspects. Supplement
requirements and specifications is vital in the proc- No. 10, October 2005 (WMO-No. 485), Geneva.
ess of programme development. Users may ———, 1992: International Meteorological Vocabulary
contribute through the evaluation of products, (WMO-No. 182), Geneva.
which invariably leads to refinements and the devel- ———, 1992: Manual on the Global Data-processing
opment of improved products. Measuring the System. Vol. II – Regional Aspects. Supplement
benefits of the application of products can be a diffi- No. 2, August 2003 (WMO-No. 485), Geneva.
cult task, but interaction with users through ———, 1997. Report of the GCOS/GOOS/GTOS Joint
workshops, training and other outreach activities Data and Information Management Panel, Third
will aid the process. The justification for a national session (Tokyo, Japan, 15–18 July 1997)
climate services programme, or requests for interna- (WMO/TD-No. 847, GCOS-No. 39, GOOS-
tional financial support for aspects of the programme, No. 11, GTOS-No. 11), Geneva.
can be greatly strengthened by well-documented ———, 2003: Climate: Into the 21 st Century.
user requirements and positive feedback. The docu- Cambridge, Cambridge University Press.
mented endorsement of the programme, by one or ———, 2008: Final Report of the CCl/CBS
more representative sections of the user community, Intercommission Technical Meeting on Designation
is essential to guide future operations and to assist in of Regional Climate Centres (Geneva, 21–22
the promotion of the service as a successful entity. January 2008), Geneva.
———, 2000: WMO – 50 Years of Service (WMO-
No. 912), Geneva.
———, 2009: WMO Statement on the Status of the
1.6 REFERENCES Global Climate in 2008 (WMO-No. 1039), Geneva.
———, 2003: Proceedings of the Meeting on Organization
and Implementation of Regional Climate Centres
1.6.1 WMO publications
(Geneva, 27–28 November 2003) (WMO/TD-
Global Climate Observing System (GCOS), 2004: No. 1198, WCASP-No. 62), Geneva.
Implementation Plan for the Global Observing ———, 2004. Implementation Plan for the Global
System for Climate in Support of the UNFCCC Observing System for Climate in support of the
(WMO/TD-No. 1219, GCOS-No. 92), Geneva. UNFCCC (WMO/TD-No. 1219, GCOS-No. 92),
Integrated Global Observing Strategy (IGOS), Geneva.
2007: Cryosphere Theme Report: For the
Monitoring of our Environment from Space and
1.6.2 Additional reading
from Earth (WMO/TD-No. 1405), Geneva.
Intergovernmental Panel on Climate Change Aristotle, circa 350 B.C.: Meteorologica.
(IPCC), 2004: 16 Years of Scientific Assessment in Bergeron, T., 1930: Richtlinien einer dyna-
Support of the Climate Convention, Geneva. mischen Klimatologie. Meteorologische Zeitung,
———, 2007: The Fourth Assessment Report: Climate 47:246–262.
Change 2007 (AR4), Vols. 1–4. Cambridge, Federov, E.E., 1927: Climate as totality of the weather.
Cambridge University Press. Monthly Weather Rev., 55:401–403.
1–10 GUIDE TO CLIMATOLOGICAL PRACTICES

Geiger, R., 1927: Das Klima der bodennahen Luftschicht. Hippocrates, circa 400 B.C.: On Airs, Waters, and Places.
Ein Lehrbuch der Mikroklimatologie. Second Köppen, W. and G. Geiger (eds.), 1930–1939.
edition, 1942; third edition, 1942; fourth Handbuch der Klimatologie, 5 vols. Berlin,
edition, 1961. Braunschweig, Vieweg. Gebruder Borntraeger.
Geiger, R., R.H. Aron and P. Todhunter, 2003. The Köppen, W., 1918: Klassification der Klimate
Climate Near the Ground. Sixth edition. Lanham, n a c h Te m p e r a t u r, N i e d e r s c h l a g a n d
Maryland, Rowman and Littlefield Publishers. Jahreslauf. Petermanns Geog. Mitt., 64:193–
Group on Earth Observations (GEO), 2007: GEO 2007- 203, 243–248.
2009 Work Plan. Toward Convergence, Geneva. Landsberg, H., 1962: Physical Climatology. Second
———, 2005: Global Earth Observation System of edition. Dubois, Pennsylvania, Gray Printing.
Systems (GEOSS): 10-year Implementation Plan. Mann, M.E., Bradley, R.S. and M.K. Hughes, 1999:
Reference Document GEO 1000R/ESA SP-1284. Northern Hemisphere temperatures during
Noordwijk, European Space Agency Publications the past millennium: Inferences, uncertain-
Division, ESTEC. ties, and limitations. Geophys. Res. Lett.,
Hadley, G., 1735: Concerning the cause of the general 26(6):759.
trade-winds. Royal Soc. London Philos. Trans., Thornthwaite, C.W., 1948. An Approach toward a
29:58–62. rational classification of climate. Geographical
Hann, J. von, 1883: Handbuch der Klimatologie. Rev., 38(1):55–94.
Second edition, 1897, 3 vols.; third edition, Walker, G.T., 1923-24: World weather, I and II.
1908–11, 3 vols. Stuttgart, Englehorn. Indian Meteorol. Dept. Mem., 24(4):9.

.
 CHAPTER 2

CLIMATE OBSERVATIONS, STATIONS AND NETWORKS

2.1 INTRODUCTION Guidance is also based on ten climate monitoring

principles set forth in the Report of the GCOS/GOOS/
All national climate activities, including research GTOS Joint Data and Information Management Panel
and applications, are primarily based on (Third Session, Tokyo, 15–18 July 1997, WMO/TD-
observations of the state of the atmosphere or No. 847):
weather. The Global Observing System provides 1. The impact of new systems or changes to
observations of the state of the atmosphere and existing systems should be assessed prior to
ocean surface. It is operated by National implementation.
Meteorological and Hydrological Services, national 2. A suitable period of overlap for new and old
or international satellite agencies, and several observing systems is required.
organizations and consortiums dealing with 3. The details and history of local conditions,
specific observing systems or geographic regions. instruments, operating procedures, data
The WMO Global Observing System is a coordinated processing algorithms, and other factors perti-
system of different observing subsystems that nent to interpreting data (metadata) should
provides in a cost-effective way high-quality, be documented and treated with the same
standardized meteorological and related care as the data themselves.
environmental and geophysical observations, from 4. The quality and homogeneity of data should
all parts of the globe and outer space. Examples of be regularly assessed as a part of routine
the observing subsystems relevant to climate are operations.
the Global Climate Observing System (GCOS) 5. Consideration of the needs for environmental
Surface Network (GSN), the GCOS Upper-Air and climate monitoring products and assess-
Network (GUAN), Regional Basic Climatological ments should be integrated into national,
Networks, Global Atmosphere Watch (GAW), regional, and global observing priorities.
marine observing systems, and the satellite-based 6. Operation of historically uninterrupted
Global Positioning System. The observations from stations and observing systems should be
these networks and stations are required for the maintained.
timely preparation of weather and climate analyses, 7. High priority for additional observations
forecasts, warnings, climate services, and research should be focused on data-poor areas, poorly
for all WMO Programmes and relevant observed parameters, areas sensitive to
environmental programmes of other international change, and key measurements with inad-
organizations. equate temporal resolution.
8. Long-term requirements should be specified
This chapter on observations follows the sequence to network designers, operators, and instru-
of specifying the elements needed to describe the ment engineers at the outset of system design
climate and the stations at which these elements and implementation.
are measured, instrumentation, siting of stations, 9. The conversion of research observing systems
network design and network operations. The guid- to long-term operations in a carefully planned
ance is based on the WMO Guide to Meteorological manner should be promoted.
Instruments and Methods of Observation (WMO- 10. Data management systems that facilitate
No. 8, fifth, sixth and seventh editions), the Guide access, use, and interpretation of data and
to the Global Observing System (WMO-No. 488) and products should be included as essential
the Guidelines on Climate Observation Networks and elements of climate monitoring systems.
Systems (WMO/TD-No. 1185). Each edition of the
Guide to Meteorological Instruments and Methods of These principles were established primarily for surface-
Observation has a slightly different emphasis. For based observations, but they also apply to data for all
example, the sixth edition contains valuable infor- data platforms. Additional principles specifically for
mation on sensor calibration, especially of the basic satellite observations are listed in section 2.3.4.
instrumentation used at climate stations, but Tables
2 and 3 of the fifth edition provide more informa-
tion about the accuracy of measurements that are
needed for general climatological purposes. Cross 2.2 CLIMATIC ELEMENTS
references are provided in the sections below to
other WMO publications containing more detailed A climatic element is any one of the properties of the
guidance. climate system described in section 1.2.2. Combined
2–2 GUIDE TO CLIMATOLOGICAL PRACTICES

with other elements, these properties describe the observe the same elements as principal climatologi-
weather or climate at a given place for a given period cal stations, with the addition of air pollution data
of time. Every meteorological element that is such as low-level ozone and other chemicals and
observed may also be termed a climatic element. The particulate matter.
most commonly used elements in climatology are
air temperature (including maximum and mini- Marine observations can be generally classified into
mum), precipitation (rainfall, snowfall and all kinds physical-dynamical and biochemical elements. The
of wet deposition, such as hail, dew, rime, hoar frost physical-dynamical elements (such as wind, tempera-
and precipitating fog), humidity, atmospheric ture, salinity, wind and swell waves, sea ice, ocean
motion (wind speed and direction), atmospheric currents and sea level) play an active role in changing
pressure, evaporation, sunshine, and present weather the marine system. The biochemical elements (such
(for example, fog, hail and thunder). Properties of as dissolved oxygen, nutrients and phytoplankton
the land surface and subsurface (including hydro- biomass) are generally not active in the physical-
logical elements, topography, geology and dynamical processes, except perhaps at long
vegetation), of the oceans, and of the cryosphere are timescales, and thus are called passive elements. From
also used to describe climate and its variability. the perspective of most NMHSs, high priority should
generally be given to the physical-dynamical
The subsections below describe commonly observed elements, although in some cases biochemical
elements for specific kinds of stations and networks of elements could be important when responding to the
stations. Details are in the Manual on the Global needs of stakeholders (for example, observations
Observing System, the WMO Technical Regulations related to the role of carbon dioxide in climate
(WMO-No. 49, in particular Volume III – Hydrology), change).
and the Guide to Agricultural Meteorological Practices
(WMO-No. 134). These documents should be kept In some NMHSs with responsibilities for monitor-
readily available and consulted as needed. ing hydrological events, hydrological planning, or
hydrological forecasting and warning, it is neces-
sary to observe and measure elements specific to
2.2.1 Surface and subsurface elements
hydrology. These elements may include combina-
An ordinary climatological station provides the tions of river, lake and reservoir level; streamflow;
basic land area requirements for observing daily sediment transport and deposition; rates of abstrac-
maximum and minimum temperature and amount tion and recharge; water and snow temperatures;
of precipitation. A principal climatological station ice cover; chemical properties of water; evapora-
usually provides a broader range of observations of tion; soil moisture; groundwater level; and flood
weather, wind, cloud characteristics, humidity, extent. These elements define an integral part of
temperature, atmospheric pressure, precipitation, the hydrologic cycle and play an important role in
snow cover, sunshine and solar radiation. In order the variability of climate.
to define the climatology of precipitation, wind, or
any other specific element, it is sometimes neces- In addition to surface elements, subsurface elements
sary to operate a station to observe one or a subset such as soil temperature and moisture are particu-
of these elements, especially where the topography larly important for application to agriculture,
is varied. Reference climatological stations (see forestry, land-use planning and land-use manage-
section 2.5) provide long-term, homogeneous data ment. Other elements that should be measured to
for the purpose of determining climatic trends. It is characterize the physical environment for agricul-
desirable to have a network of these stations in each tural applications include evaporation from soil
country, representing key climate zones and areas and water surfaces, sunshine, short- and long-wave
of vulnerability. radiation, plant transpiration, runoff and water
table, and weather observations (especially hail,
In urban areas weather can have a significant lightning, dew and fog). Ideally, measurements of
impact. Heavy rains can cause severe flooding; agriculturally important elements should be taken
snow and freezing rain can disrupt transportation at several levels between 200 cm below the surface
systems; and severe storms with accompanying and 10 m above the surface. Consideration should
lightning, hail and high winds can cause power fail- also be given to the nature of crops and vegetation
ures. High winds can also slow or stop the progress when determining the levels.
of automobiles, recreational vehicles, railcars, tran-
sit vehicles and trucks. The urban zone is especially Proxy data are measurements of conditions that are
susceptible to land falling tropical storms because indirectly related to climate, such as phenology, ice
of the large concentrations of people at risk, the core samples, varves (annual sediment deposits),
high density of man-made structures, and the coral reefs and tree ring growth. Phenology is the
increased risk of flooding and contamination of study of the timing of recurring biological events in
potable water supplies. Urban stations usually the animal and plant world, the causes of their
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–3

timing with regard to biotic and abiotic forces, and climate activities that require upper-air observations
the interrelation among phases of the same or includes monitoring and detecting climate variabil-
different species. Leaf unfolding, flowering of plants ity and change, climate prediction on all timescales,
in spring, fruit ripening, colour changing and leaf climate modelling, studies of climate processes, data
fall in autumn, as well as the appearance and depar- reanalysis activities, and satellite studies concerning
ture of migrating birds, animals and insects are all calibration of satellite retrievals and radiative
examples of phenological events. Phenology is an transfer.
easy and cost-effective system for the early detec-
tion of changes in the biosphere and therefore The longest record of upper-air observations has
complements the instrumental measurements of been obtained from balloon-based instruments
national meteorological services very well. combined with ground tracking devices in a radio-
sonde network. These radiosonde measurements
An ice core sample contains snow and ice and provide a database of atmospheric variables dating
trapped air bubbles. The composition of a core, back to the 1930s, although coverage is generally
especially the presence of hydrogen and oxygen poor before 1957. The radiosonde data record is
isotopes, relates to the climate of the time the ice characterized by many discontinuities and biases
and snow were deposited. Ice cores also contain resulting from instrument and operational proce-
inclusions such as windblown dust, ash, bubbles of dural changes and incomplete metadata. Satellite
atmospheric gas, and radioactive substances in the observations have been available since the 1970s,
snow deposited each year. Various measurable and some have been assembled and reprocessed to
properties along the core profiles provide proxies create continuous records. Just as the radiosonde
for temperature, ocean volume, precipitation, record has deficiencies, however, the satellite data
chemistry and gas composition of the lower atmos- also suffer from, among other things, limited verti-
phere, volcanic eruptions, solar variability, sea cal resolution, orbit drift, satellite platform changes,
surface productivity, desert extent and forest fires. instrument drift, complications with calibration
The thickness and content of varves are similarly procedures, and the introduction of biases through
related to annual or seasonal precipitation, stream- modifications of processing algorithms. Other
flow, and temperature. upper-air measurements have come from moving
platforms such as aircraft. Observations from some
Tropical coral reefs are very sensitive to changes in high-mountain locations have also been consid-
climate. Growth rings relate to the temperature of ered part of the upper-air measurement system.
the water and to the season in which the rings grew.
Analysis of the growth rings can match the water The main observational requirements for monitor-
temperature to an exact year and season. Data from ing long-term upper-air changes are:
corals are used to estimate past El Niño–Southern (a) A long-term (multidecadal), stable, temporally
Oscillation (ENSO) variability, equatorial upwelling, homogeneous record so that changes can confi-
changes in subtropical gyres, trade wind regimes dently be identified as true atmospheric changes
and ocean salinity. rather than changes in the observing system or
as artefacts of homogenization methods;
Tree-ring growth shows great interannual variability (b) Good vertical resolution to describe the verti-
and also large spatial differences. Some of the varia- cal structure of temperature, water vapour,
tion can be related to weather and climate conditions and ozone changes, and of changes in the
in the microscale and macroscale; plants can be tropopause;
viewed as integrative measurement devices for the (c) Sufficient geographical coverage and resolu-
environment. Since trees can live for centuries, tion, so that reliable global and area trends
annual growth rings in some tree species can provide can be determined;
a long historical indication (predating instrumental (d) Observational precision finer than the
measurements) of climate variability. Because of the expected atmospheric variations to allow
close relationship between plant development and clear identification of both variability and
weather and climate, phenological observation long-term changes. This requirement is
networks are run by the NMHSs in many countries. particularly important for water vapour
observations in the upper troposphere and
Table 2.1 summarizes the most common surface stratosphere.
and subsurface climatic elements that are observed
for various networks or types of station. The essential climate elements from upper-air
observations are given in the Second Report on the
Adequacy of the Global Observing Systems for Climate
2.2.2 Upper-air elements
in Support of the UNFCCC (WMO/TD-No. 1143) and
Upper-air observations are an integral component of the Implementation Plan for the Global Observing
the Global Observing System. The spectrum of System for Climate in Support of the UNFCCC
2–4 GUIDE TO CLIMATOLOGICAL PRACTICES

Table 2.1. Examples of surface and subsurface elements for different station networks .
or types of stations

Element Ordinary Principal Marine Hydrometeorological Agrometeorological Urban Proxy

climate climate

Air temperature • • • • •
Soil temperature •
Water • •
temperature
Precipitation • • • • • •
Weather • • • •
Clouds • • • •
Pressure • • • •
Visibility • • • •
Humidity • • • •
Wind • • • •
Solar radiation • • •
Sunshine • • •
Salinity •
Currents •
Sea level •
Waves •
Air–sea •
momentum
Air–sea fluxes •
Ice • •
Dissolved oxygen •
Nutrients •
Bathymetry •
Biomass •
Streamflow •
River stages •
Sediment flow •
Recharge •
Evaporation • • •
Soil moisture • • •
Runoff • •
Groundwater • •
Plant development • •
Pollen •
Ice and sediment •
composition
Tree ring growth •
Coral ring growth •
Atmospheric •
chemicals
Particulate matter •

(WMO/TD-No. 1219) as temperature, water vapour, plant and human health and well-being (Chapter
pressure, wind speed and direction, cloud properties, B2 in the Technical Regulations, the Plan for the
radiance and radiation (net, incoming and Global Climate Observing System (GCOS), Version 1.0
outgoing). Because the chemical composition of (WMO/TD-No. 681), and the GCOS/GTOS Plan for
the atmosphere is of major importance in climate Terrestrial Climate-related Observations, Version 2.0
prediction, climate change monitoring, ozone and (WMO-No. 796)), it is important to understand the
other air-quality predictions, and in application vertical structure of the composition of the global
areas such as the study and forecasting of animal, atmosphere. Chemical composition elements
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–5

requiring measurement both in the free atmosphere of the Earth’s skin temperature are not the same as
and near the ground include concentrations of temperature measurements taken in a standard
ozone and other greenhouse gases, such as carbon screen, and the relationship between radar
dioxide and methane, atmospheric turbidity measurements of reflectivity and precipitation
(aerosol optical depth), aerosol total load, reactive amounts collected in raingauges may be quite
gas species, and radionuclides. Measurements of complex. It is possible with care, however, to
acid rain (or more generally precipitation and construct homogeneous series that combine
particulate chemistry) and ultraviolet radiation are remotely sensed and in situ measurements.
also needed. (See the Integrated Global Atmospheric
Chemistry Observations (IGACO) Report of IGOS-
WMO-ESA (WMO/TD-No. 1235) for details
concerning the chemical composition of the 2.3 INSTRUMENTATION
atmosphere).
Climatological stations that are part of a national
Upper-air measurements should capture the full network should be equipped with standard
range of climate regimes and surface types. Radiative approved instruments; the NMHS may supply the
transfer codes used to convert raw satellite radiances instruments. When equipment is supplied by other
to geophysical parameters depend upon assump- agencies or purchased by the observer, every effort
tions about the surface conditions. Therefore, should be made by a climate office to ensure compli-
different local environmental conditions should be ance with national standards.
represented, including both land and ocean areas.
This section gives guidance on some basic surface
instrumentation and on the selection of instru-
2.2.3 Elements measured by remote-
ments. There are several other WMO publications
sensing
that are necessary companions to this Guide, and
Satellites and other remote-sensing systems such as should be readily available and consulted as needed.
weather radar provide an abundance of additional A thorough survey of instruments suitable for meas-
information, especially from otherwise data-sparse uring climate and other elements at land and
areas, but are not as yet capable of providing, with marine stations is provided in the Guide to
the required accuracy and homogeneity, measure- Meteorological Instruments and Methods of Observation.
ments of many of the elements that are reported Details of instrumentation needed for the measure-
from land-based surface stations. The spatial cover- ment of chemical composition are given in the
age they offer makes them complementary to, but International Operations Handbook for Measurement of
not a substitute for, the surface networks. The Background Atmospheric Pollution (WMO-No. 491),
elements that can be measured or estimated for agrometeorological elements in the Guide to
remotely are precipitation (with limited accuracy Agricultural Meteorological Practices, and for hydro-
over small areas, ocean–atmosphere interfaces, logical purposes in the Guide to Hydrological Practices
highlands or steep orography); cloud amount; radi- (WMO-No. 168).
ation fluxes; radiation budget and albedo; upper
oceanic biomass, ocean surface topography and When selecting instrumentation, including any
wave height; sea-ice cover; sea surface temperature; associated data-processing and transmission
ocean surface wind vectors and wind speed; atmos- systems, the 10 climate monitoring principles
pheric temperature, humidity and wind profiles; (section 2.1) should be followed. Several concerns
chemical constituents of the atmosphere; snow should be considered when complying with these
cover; ice sheet and glacier extent; vegetation and principles:
land cover; and land surface topography. (a) Reliability;
(b) Suitability for the operational environment at
Greater spatial and temporal coverage can be the station of use;
achieved with remote-sensing than with in situ (c) Accuracy;
observations. Remotely sensed data also supplement (d) Simplicity of design;
observations from other platforms and are especially (e) Reasons for taking observations.
useful when observations from other platforms are
missing or corrupted. Although this is an advantage, Reliability requires that an instrument functions
there are problems in using remotely sensed data within its design specifications at all times.
directly for climate applications. Most importantly, Unreliable instrumentation leads to data gaps,
the short period of record means that remotely biases and other inhomogeneities. Reliable instru-
sensed data cannot be used to infer long-term ments need to be robust enough to cope with the
climate variability and change. Also, remotely full range of weather and physical extremes
sensed data may not be directly comparable to in expected at the site, and possibly the handling that
situ measurements. For example, satellite estimates is part of manual observations.
2–6 GUIDE TO CLIMATOLOGICAL PRACTICES

Instruments must be suited both to the climate in Whenever possible, trained local personnel, such
which they must function, and to other equip- as a caretaker, should examine an observing site
ment with which they must operate. For example, on a regular basis to keep surface conditions (such
an anemometer head at a cold location will need as grass growth) in check, perform basic mainte-
to withstand icing, while one in a desert area will nance of instruments (such as simple cleaning),
need to be protected against dust ingression. examine for damage, and detect breaches of secu-
Sensors for use in an automatic weather station rity. These tasks should be performed at least
need to provide output suitable for automatic weekly at accessible, manned land stations.
processing. The standard mercury-in-glass ther- Inspection of sites and instruments at remote
mometer used at a manual recording site, for locations should be made as often as possible.
example, will need to be substituted by a tempera- Personnel should also be available to provide
ture-sensitive probe, such as a thermocouple, rapid maintenance response when critical systems
whose response can be converted into an elec- fail.
tronic signal. Instruments must also be sited so
that they can be accessed and maintained. Autographic and data-logger equipment exists for
the recording of many climatic elements, such as
Ideally, all instruments should be chosen to provide temperature, humidity, wind and rates of rainfall.
the high level of accuracy and precision required Data need to be transferred from autographic
for climatological purposes. It is also important that records to tables or digital form. Observers should
the instrument can continue to provide the required ensure that the equipment is operating properly
level of accuracy for a long period of time, as instru- and that information recorded on charts, for
ment “drift” can lead to serious inhomogeneities in example, is clear and distinct. Observers should be
a climate record; accuracy is of limited use without responsible for regularly verifying and evaluating
reliability. the recorded data (by checking against direct­
reading equipment) and for making time marks at
The simpler the instrumentation, the easier it is to frequent, specified intervals. The recorded data
operate and to maintain, and the easier it is to can be effectively used to fill gaps and to complete
monitor its performance. It is sometimes necessary the record when direct observations are missed
to install redundant sensors (for example, triple because of illness and other causes of absence from
thermistors at automated data-logger stations) to the observing station. Sections 1.4.2 and 1.4.3 of
properly track performance and reliability over the Guide to Meteorological Instruments and Methods
time. Complex systems can easily lead to data inho- of Observation (fifth edition) give specific guidance
mogeneities, data loss, high maintenance cost and on the maintenance and operation of recording
changing accuracy. instruments, drums and clocks.

The purpose of the observations generally dictates Data from automatic weather stations (AWSs), at
requirements for measurements. Types of instru- which instruments record and transmit observa-
ments, installation of sensors, and characteristics of tions automatically, are usually restricted to those
the instruments should be considered to ensure readily obtained in digital form, although the range
that measurement requirements can be met. of sensors is wide and continues to evolve. Such
stations have been used to supplement manned
Details about these concerns, including instru- stations and to increase network densities, report-
ment and measurement standards and ing frequencies and the quantities of elements
recommended practices, are can be found in the observed, especially in remote and largely unpopu-
Guide to Meteorological Instruments and Methods of lated areas where human access is difficult. Some of
Observation. the sensitivity and accuracy requirements of these
automated stations are given in the Guide to
2.3.1 Basic surface equipment Meteorological Instruments and Methods of Observation;
others are being developed, especially for studies of
There may be a variety of options for obtaining climate variability.
climate observations from surface stations. These
options include equipping a station with, for exam- In many countries, AWSs have lowered operational
ple, basic instruments, autographic or automated costs. NMHSs choosing between manned and AWS
output available for unmanned periods, or totally observation programmes need to consider a number
automated sensors. When considering the options, of issues. Notwithstanding the considerable poten-
it is important to compare costs of personnel, main- tial for AWSs to provide high-frequency data, as
tenance and replacement. Price negotiation is often well as additional data from remote locations, there
possible with manufacturers on the basis of, for are several significant costs associated with operat-
example, quantities purchased, among other ing an AWS, including labour costs for maintenance
things. and ensuring AWS reliability, labour availability,
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–7

accessibility for installation and maintenance, the and sensor and exposure differences; metadata
availability of suitable power sources, security of concerning instrumentation, data-reduction and
the site, and communications infrastructure. These data-processing procedures are crucial to utilizing
aspects must be carefully weighed against the radiosonde data in climate applications. New refer-
significant benefits, such as a denser or more exten- ence radiosondes are being developed to mitigate
sive network. AWSs can be powerful alternatives to the deficiencies of the current standard radio-
manned observational programmes and sometimes sondes. A limited network of these will be used to
are the only option, but they require a strong calibrate and validate various satellite observations
organizational commitment to manage them. of both temperature and water vapour.

Marine instrumentation includes drifting and An upper-air observing system may change over
moored data buoys, ice floats and subsurface floats. time with technological advances. Hence, a key
Although the data are collected remotely, the requirement of the network is sufficient overlap of
instruments are generally performing in situ meas- systems to maintain continuity and allow full
urements. They are a cost-effective means for comparison of the accuracy and precision of the old
obtaining meteorological and oceanographic data and new systems. Measurement systems should be
from remote ocean areas. As such, they form an calibrated regularly at the site. It is imperative that
essential component of marine observing systems instrument replacement strategies take into account
and meteorological and oceanographic operational changes in other networks, such as the use of satel-
and research programmes. For example, the Tropical lites. The Climate Monitoring Principles (see section
Atmosphere Ocean array of moorings has enabled 2.1) should guide the development and operation
timely collection of high-quality oceanographic of an upper-air observing system.
and surface meteorological data across the equato-
rial Pacific Ocean for the monitoring, forecasting
2.3.3 Surface-based remote-sensing
and understanding of climate swings associated
with El Niño and La Niña. Remote-sensing can use either active or passive
sensors. Active sensor systems emit some form of
radiation, which is scattered by various targets; the
2.3.2 Upper-air instruments
sensors detect the backscatter. Passive sensors meas-
Historically, most climatological data for the upper ure radiation being emitted (or modified) by the
air have been derived from measurements made for environment.
synoptic forecasting by balloon-borne radiosondes.
A variety of techniques and instruments are used The most common surface-based active remote-
for the measurement of pressure, temperature, sensing technique is weather radar. A short pulse
humidity and wind, and for processing instrumen- of high-power microwave energy is focused by an
tal output into meteorological quantities. It is antenna system into a narrow beam. This beam is
important for each NMHS to issue suitable instruc- scattered back by the target precipitation, with the
tion manuals to each upper-air station for the backscattered radiation received, generally, by the
proper use of equipment and interpretation of data. same antenna system. The location of the precipi-
The Manual on the Global Observing System (WMO- tation can be determined from the azimuth and
No. 544, section 2.10.4.5) requires that prompt elevation of the antenna and the time between
reports be made to the WMO Secretariat of changes transmitting and receiving the reflected energy.
in radiosonde type or changes in wind systems in The power of the received radiation depends on
operational use at a station. the nature of the precipitation, and the signal can
be processed to estimate its intensity. Atmospheric
There are several issues concerning the quality of and environmental conditions can adversely affect
radiosonde measurements for climate monitoring radar data, and caution should be exercised when
and climate change detection purposes. Radiation interpreting the information. Some of these effects
errors cause uncertainties in temperature. Standard include returns from mountains, buildings, and
radiosondes are not capable of measuring water other non-meteorological targets; attenuation of
vapour at low temperatures with sufficient accu- the radar signal when viewing weather echoes
racy. Sensor types, especially humidity sensors, through intervening areas of intense precipita-
have changed over time. The spatial coverage of tion; temperature inversions in the lower layers of
radiosonde observations is not uniform; most the atmosphere, which bend the radar beam in
stations are located on the land-surface territories such a way that ground clutter is observed where
of the northern hemisphere, while the southern normally not expected; and the bright band,
hemisphere and ocean networks are much less which is a layer of enhanced reflectivity caused by
dense. The historical record of radiosonde observa- the melting of ice particles as they fall through the
tions has innumerable problems relating to a lack freezing level in the atmosphere, which can result
of intercomparisons among types of radiosondes in overestimation of rainfall. Use of radar data in
2–8 GUIDE TO CLIMATOLOGICAL PRACTICES

climate studies has been limited by access and analysis systems and, through programs of
processing capabilities, uncertainties in calibra- reanalysis, ultimately contribute substantially to
tion and calibration changes, and the complex the broader climate record.
relationship between reflectivity and
precipitation. Aircraft meteorological data-relay systems operate
on aircraft that are equipped with navigation and
Wind profilers use radar to construct vertical other sensing systems. There are sensors for meas-
profiles of horizontal wind speed and direction uring air speed, air temperature and air pressure.
from near the surface to the tropopause. Fluctuations Other data relating to aircraft position, acceleration
of atmospheric density are caused by turbulent and orientation are obtained from the aircraft navi-
mixing of air with different temperatures and mois- gation system. The aircraft also carry airborne
ture content. Fluctuations in the resulting index of computers for the flight management and naviga-
refraction are used as a tracer of the mean wind. tion systems by which navigation and
Although they work best in clear air, wind profilers meteorological data are computed continuously
are capable of operating in the presence of clouds and are made available to the aircrew. The data are
and moderate precipitation. When equipped with a automatically fed to the aircraft communication
radio-acoustic sounding system, profilers can also system for transmission to the ground, or alterna-
measure and construct vertical temperature profiles. tively, a dedicated processing package can be used
The speed of sound in the atmosphere is affected by on the aircraft to access raw data from the aircraft
temperature. Acoustic energy is tracked through systems and derive the meteorological variables
the atmosphere, and the temperature profile is esti- independently. Normally, messages transmitted to
mated from the speed of the sound wave ground stations contain horizontal wind speed and
propagation. direction, air temperature, altitude (related to a
reference pressure level), a measure of turbulence,
Lightning detection is the most common passive time of observation, phase of flight, and the aircraft
surface-based remote-sensing. Lightning sensors position. The data are used by aviation controllers
scan a range of electromagnetic frequencies to to ensure flight safety and by weather forecasters.
detect electrical discharges inside clouds, between
clouds, or between clouds and the ground. There are potentially a large number of error sources
Characteristics of the received radiation (such as contributing to aircraft measurement uncertainty.
the amplitude, time of arrival, source direction, An uncertainty of about 5 to 10 per cent in the
sign and other wave form characteristics) are meas- calculation process can be expected. A further
ured, and from them characteristics of the lightning complication arises over the choices of sampling
flash are inferred. One sensor cannot accurately interval and averaging time. Examination of typical
locate lightning events; data from several sensors time series of vertical acceleration data often indi-
are concentrated in a central location in a central cates a high variability of statistical properties over
lightning processor. The processor computes and short distances. Variation of air speed for a single
combines data from multiple sensors to calculate aircraft and between different aircraft types alters
the location and characteristics of the observed the sampling distances and varies the wavelengths
lightning flashes. The accuracy and efficiency of a filtered. While not as precise and accurate as most
lightning-detector network drops progressively on ground observing systems, aircraft data can provide
its outer boundaries. The detection wave will prop- useful supplemental information to meteorological
agate without too much attenuation with distance databases.
depending on the frequency band used, but if a
lightning flash is too far away from the network Satellite data add valuable information to climate
(this distance varies with the stroke amplitude and databases due to their wide geographical coverage,
the network configuration), the stroke may no especially over areas with sparse or completely
longer be detected. missing in situ data. Satellites are very useful for
monitoring phenomena such as polar sea-ice
2.3.4 Aircraft-based and space-based extent, snow cover, glacial activity, sea level
remote-sensing changes, vegetation cover and moisture content,
and tropical cyclone activity. They also help
Many long-distance aircraft are fitted with improve synoptic analyses, an important compo-
automatic recording systems that report temperature nent of synoptic climatology.
and wind, and in some cases humidity, regularly
while en route. Some aircraft record and report Sensing techniques make use of the emission, absorp-
frequent observations during takeoff and descent to tion and scattering properties of the atmosphere and
significantly augment the standard radiosonde the surface. The physical equations for radiative trans-
data, at least throughout the troposphere. Such fer provide information about the radiative properties
data are assimilated into operational meteorological of the atmosphere and the Earth’s surface and,
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–9

through inversion of the radiative transfer equation, climatological purposes. For example, the Global
geophysical properties such as temperature and mois- Positioning System uses a network of dozens of
ture profiles, surface skin temperature, and cloud satellites to assist in navigation. But by measuring
properties. the propagation delay in Global Positioning System
signals, it is possible to estimate atmospheric water
The figures and specifications of satellite plat- vapour content.
forms and sensors are in the GCOS Guide to
Satellite Instruments for Climate (WMO/TD- Two complementary orbits have been used for
No. 685). The elements, accuracy and spatial and operational environmental satellites: geostationary
temporal resolution of data measured by satellites and polar-orbiting. In geostationary orbit, about
are in the Preliminary Statement of Guidance 36 000 km above the Equator, a satellite will orbit
Regarding How Well Satellite Capabilities Meet the Earth once every 24 hours. The satellite there-
WMO User Requirements in Several Application fore remains stationary relative to the Earth and
Areas (WMO/TD-No. 913). The histories and can thus provide a constant monitoring capability
future plans of satellite platforms and sensors are and the ability to track atmospheric features and
in the GCOS Plan for Space-based Observations infer winds. Polar-orbiting satellites are typically
(Version 1.0, WMO/TD-No. 684) and Systematic about 800 km above the surface, moving almost
Observation Requirements for Satellite-based Products north–south relative to the Earth. Most of the globe
for Climate (WMO/TD-No. 1338). The technology is observed by the suite of instruments on the oper-
of remote-sensing is progressing rapidly and the ational polar-orbiting satellites twice per day about
operational plans of platforms and sensors may 12 hours apart. The inconvenience of only two
be changed occasionally. Therefore, the latest passes each day is balanced by the higher spatial
documents should be referred to in using remote- resolution and greater range of instruments carried,
sensing data. Reports published by the Committee and the ability to see high latitudes that are poorly
on Earth Observation Satellites, available on the captured from geostationary orbit.
Internet, are helpful in seeking the latest infor-
mation about satellites. As is the case with the treatment of in situ data, the
climate community must recognize the need to
As in the case of surface-based remote-sensing, provide scientific data stewardship of both the
satellite and other airborne sensors can be classified “raw” remotely sensed measurements and the data
into two groups: passive and active. Passive sensors processed for climate purposes. In addition to the
include imagers, radiometers and sounders. They ten principles listed in section 2.1, satellite systems
measure radiation emitted by the atmosphere or should also adhere to the following principles:
the Earth’s surface. Their measurements are 1. Constant sampling within the diurnal cycle
converted into geophysical information such as (minimizing the effects of orbital decay and
vertical profiles of water vapour, temperature and orbit drift) should be maintained.
ozone; cloud information; surface ocean and land 2. Overlapping observations should be ensured
temperatures; and ocean and land colour. The for a period sufficient to determine inter-
wavelength at which a sensor operates influences satellite biases.
the resulting information, with different wave- 3. Continuity of satellite measurements (elimi-
lengths having different advantages and nation of gaps in the long-term record)
disadvantages. through appropriate launch and orbital strat-
egies should be ensured.
Active sensors include radar, scatterometers and 4. Rigorous pre-launch instrument characteri-
lidar. They measure the backscattered signal from zation and calibration, including radiance
an observing target when it is illuminated by a confirmation against an international radi-
radiation source emitted from the platform. Their ance scale provided by a national metrology
advantage is that the accurate range of an observing institute, should be ensured.
target can be obtained by measuring a time lag 5. On-board calibration adequate for climate
between an emission and its return, while the use system observations should be ensured and
of a tightly focused and directional beam can associated instrument characteristics moni-
provide positional information. Backscattered tored.
signals can be converted into wind speed and 6. Operational production of priority climate
direction, ocean dynamic height and wave products should be sustained and peer-
spectrum, ocean wind stress curl and geostrophic reviewed new products should be introduced
flow, cloud properties, precipitation intensity, and as appropriate.
inventories of glacial extent. 7. Data systems needed to facilitate user access
to climate products, metadata and raw data,
Sometimes, information can be derived from satel- including key data for delayed-mode analysis,
lite data that was not originally intended for should be established and maintained.
2–10 GUIDE TO CLIMATOLOGICAL PRACTICES

8. Use of functioning baseline instruments that organizing instrument evaluations, and providing
meet the calibration and stability require- advice about instrumental performance.
ments stated above should be maintained for
as long as possible, even when these exist on The GCOS Plan for Space-Based Observations details
decommissioned satellites. calibration, inter-calibrational overlapping records
9. Complementary in situ baseline observations and metadata requirements for space-based remote
for satellite measurements should be main- sensors. The plan of the Global Space-Based Inter-
tained through appropriate activities and Calibration System is to compare the radiances
cooperation. simultaneously measured by satellite pairs at the
10. Random errors and time-dependent biases in crossing points of their ground track, in particular
satellite observations and derived products where a polar orbiter and a geostationary satellite
should be identified. cross paths. This inter-calibration will give a globally
consistent calibration on an operational basis.
2.3.5 Calibration of instruments
Weather radar calibration requires the measure-
It is of paramount importance, for determining the ment of system characteristics such as transmitted
spatial and temporal variations of climate, that the frequency and power, antenna gain, beam widths,
relative accuracy of measurement of individual receiver output and filter losses. Performance moni-
sensors in use in a network at one time be measured toring ensures that other system characteristics,
and periodically checked, and similarly, that the such as antenna orientation, side lobes, pulse dura-
performance of replacement sensors and systems tion and pulse shape, beam patterns and receiver
can be related to that of those replaced. The Manual noise levels, are within acceptable limits.
on the Global Observing System states that all stations
shall be equipped with properly calibrated instru- Drifts of lightning detection sensors or central light-
ments. Details on calibration techniques can be ning processor parameters are problems that should
found in the Guide to Meteorological Instruments and be detected with regular data analyses (for example,
Methods of Observation. For climatology it is not cross-checking of sensor behaviour and analysis of
generally sufficient to rely upon manufacturers’ stroke parameters). Comparison should also be made
calibrations and it is wrong to assume that a calibra- with other observations of lightning activity, such as
tion will not drift or otherwise change with time. manual observations of “thunder heard” or “light-
ning seen”, or observations of cumulonimbus clouds.
Comparisons of instrumental or system measure- As with weather radars, monitoring and calibration of
ments should be made with portable standard the characteristics of the system should be a routine
instruments when replacement instruments are process.
issued to a station and at each regular inspection of
the station (see section 2.6.6). Travelling standards
should be checked against national reference stand-
ards before and after each period of travel, and they 2.4 THE SITING OF CLIMATOLOGICAL
should be robust in transport and withstand calibra- STATIONS
tion changes. Records of instrument changes and
calibration drifts must be kept and made available as The precise exposure requirements for specific
metadata, as they are essential to the assessment of instruments used at climatological stations, aimed
true climate variations (see section 2.6.9). at optimizing the accuracy of the instrumental
measurements, are discussed in Part III of the
During inspections of remotely sited AWSs, observa- Manual on the Global Observing System, the Guide to
tions should be taken using the travelling standards Meteorological Instruments and Methods of Observation,
for later comparison with the recorded AWS output and Representativeness, Data Gaps and Uncertainties
as received at the data reception point. Some NMHSs in Climate Observations (WMO/TD-No. 977). These
have automated fault or instrumental drift detection publications are a necessary companion to this
procedures in place, which compare individual Guide.
measurements with those from a network and with
values analysed from numerically fitted fields. These The representativeness and homogeneity of clima-
automated procedures are useful for detecting not tological records are closely related to the location
only drift, but also anomalous step changes. of the observing site. A station sited on or near a
steep slope, ridge, cliff, hollow, building, wall or
Some NMHSs operate their own calibration facili- other obstruction is likely to provide data that are
ties, or use accredited calibration companies. more representative of the site alone and not of a
Regional calibration facilities within WMO are wider area. A station that is or will be affected by
responsible for keeping and calibrating standards, the growth of vegetation, including even limited
certifying an instrument’s conformity to standards, tree growth near the sensor, growth of tall crops or
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–11

woodland nearby, erection of buildings on adjacent Observing sites and instruments should be properly
land, or increases (or decreases) in road or air traffic maintained so that the quality of observations does
(including those due to changes in the use of not deteriorate significantly between station
runways or taxiways) will provide neither broadly inspections. Routine, preventive maintenance
representative nor homogeneous data. schedules include regular “housekeeping” at
observing sites (for example, grass cutting and
A climatological observing station should be sited at cleaning of exposed instrument surfaces, including
a location that permits the correct exposure of the thermometer screens) and manufacturers’
instrumentation and allows for the widest possible recommended checks on instruments. Routine
view of the sky and surrounding country if visual quality control checks carried out at the station or
data are required. Ordinary and principal climato- at a central point should be designed to detect
logical stations should be sited on a level piece of equipment faults at the earliest possible stage.
ground covered with short grass; the site should be Depending on the nature of the fault and the type
well away from trees, buildings, walls and steep of station, the equipment should be replaced or
slopes and should not be in a hollow. A plot size of repaired according to agreed priorities and time
about 9 metres by 6 metres is sufficient for outdoor intervals. It is especially important that a log be
temperature and humidity-sensing instruments, and kept of instrument faults and remedial action taken
an area of 2 metres by 2 metres of bare ground within where data are used for climatological purposes.
the plot is ideal for observations of the state of the This log will be the principal basis for the site’s
ground and soil temperature measurements. A metadata and hence becomes an integral part of the
slightly larger plot (10 metres by 7 metres) is prefer- climate record. Detailed information on site
able if the site is to enclose a raingauge in addition to maintenance can be found in the Guide to
the other sensors. Meteorological Instruments and Methods of Observation
(sixth edition).
A rule used by many NMHSs is that the distance of
any obstruction, including fencing, from the rain- Additional constraints on siting apply to GAW
gauge must be more than twice, and preferably stations established to provide data on atmospheric
four times, the height of the object above the chemical composition, as discussed in Chapter B2
gauge. In general terms, anemometers require of the Technical Regulations. These constraints
exposure at a distance from any obstruction of at include the need for no significant changes in land-
least 10, and preferably 20, times the height of the use practices within 50 kilometres of the site, and
obstruction. The different exposure requirements freedom from the effects of local and area pollution
of various instruments may give rise to a split site, from, for example, major population centres, indus-
where some elements are observed from one point trial and extensive farming activities, highways,
while others are observed nearby, with data from volcanic activity and forest fires. Both global and
all the elements combined under the one site regional GAW stations should be within 70 kilo­
identifier. metres of an upper-air synoptic station.

Prevention of unauthorized entry is a very impor- The nature of urban environments makes it impossi-
tant consideration, and may require enclosure by ble to conform to the standard guidance for site
a fence. It is important that such security meas- selection and exposure of instrumentation required
ures do not themselves compromise the site for establishing a homogeneous record that can be
exposure. Automatic stations will normally need used to describe the larger-scale climate. None­theless,
a high level of security to protect against animal urban sites do have value in their own right for moni-
and unauthorized human entry; they also require toring real changes in local climate that might be
the availability of suitable and robust power significant for a wide range of applications. Guidelines
supplies, and may possibly need additional for the selection of urban sites, installation of equip-
protection against floods, leaf debris and blowing ment and interpretation of observations are given in
sand. Initial Guidance to Obtain Representative Meteorological
Observations at Urban Sites (WMO/TD-No. 1250).
Ordinary and principal climatological stations Fundamental to the guidance is the need to clearly
should be located at such sites and should be subject understand the purpose of making the observations
to such administrative conditions that will allow and to obtain measurements that are representative
the continued operation of the station, with the of the urban environment. In many urban situations
exposure remaining unchanged, for a decade or it will be possible to conform to standard practices,
more. For stations used or established to determine but flexibility in siting urban stations and in instru-
long-term climate change, such as reference clima- mentation may be necessary. These characteristics
tological stations and other baseline stations in the further heighten the importance of maintaining
GCOS network, constancy of exposure and opera- metadata that accurately describe the setting of the
tion is required over many decades. station and instrumentation.
2–12 GUIDE TO CLIMATOLOGICAL PRACTICES

2.5 THE DESIGN OF CLIMATOLOGICAL Stations should be located to give representative

NETWORKS climatic characteristics that are consistent with all
types of terrain, such as plains, mountains, plateaus,
A network of stations is several stations of the coasts and islands, and surface cover such as forests,
same type (such as a set of precipitation stations, urban areas, farming areas and deserts within the
radiation measuring stations or climatological area concerned. Station density should be depend-
stations), which are administered as a group. ent upon the purposes for making the observations
Each network should be optimized to provide the and the uses of the data. For data used in sectoral
data and perform as required at an acceptable applications within an area, there may be a need for
cost. Most optimizing methods rely on data from a greater density of stations where activities or
a pre-existing network, available over a long health are sensitive to climate, and a lesser density
enough period to correctly document the proper- in locations with fewer people. When planning a
ties of the meteorological fields. They are based land network, compromises often have to be made
on both temporal and spatial statistical analyses between the ideal density of stations and the
of time series. It is difficult to assess a priori how resources available to install, operate and adminis-
long the data series must be because the number ter the stations.
of years necessary to capture variability and
change characteristics may vary with the climatic The distribution of stations in the Regional Basic
element. It has been common practice to assume Synoptic Network from which monthly surface
that at least ten years of daily observations are climatological data are collected should be such
necessary to produce the relevant base statistical that every 250 000 square kilometres are repre-
parameters for most elements, and at least thirty sented by at least one station and by up to 10
years for precipitation. Observed global and evenly spread stations if possible. The distribu-
regional climatic trends and variability in many tion of stations from which monthly upper-air
areas of the globe over the past century suggest, climatological data are collected should be such
however, that such short periods of record may that every 1 000 000 square kilometres are repre-
not be particularly representative of similar sented by at least one station. Networks of
periods to follow. principal climatological stations should have a
maximum average separation of 500 kilometres,
The identification of redundant stations allows and for climate purposes upper-air stations should
network managers to explore options for optimiz- have a maximum average separation of 1 000
ing the network, for example, by eliminating the kilometres.
redundant stations to reduce costs or by using the
resources to establish stations at locations where Each Member should establish and maintain at
observations are needed for a more effective reali- least one reference climatological station for
zation of the network objectives. Network determining climate trends. Such stations need to
managers should take advantage of the relatively provide more than 30 years of homogeneous
high spatial coherence that exists for some mete- records and should be situated where anthropogenic
orological fields, such as temperature. Techniques environmental changes have been and are expected
used to evaluate the level of redundancy of infor- to remain at a minimum. Information on
mation include the use of the spatial agrometeorological and hydrometeorological
variance–covariance matrix of the available networks and sites can be found in the Guide to
stations, multiple linear regression, canonical Agricultural Meteorological Practices and the Guide to
analysis and observation system simulation exper- Hydrological Practices, respectively, and additional
iments (see Chapter 5). guidance is provided in the Manual on the Global
Observing System.
The density and distribution of climatological
stations to be established in a land network within A nation’s environmental information activities are
a given area depend on the meteorological elements often conducted by many parties whose contribu-
to be observed, the topography and land use in the tions are complementary and at times overlapping.
area, and the requirements for information about A nation benefits from environmental information
the specific climatic elements concerned. The rate collected and disseminated by both governmental
of variation of climatic elements across an area will agencies and non-governmental entities (including
differ from element to element. A sparse network is private companies, utilities and universities).
sufficient for the study of surface pressure, a fairly Formal partnerships between the NMHS and these
dense network for the study of maximum and mini- other parties are highly desirable for optimizing
mum temperature, and very dense networks for resources. Because data and information obtained
examining the climatology of precipitation, wind, from non-NMHS sources are not usually under the
frost and fog, especially in regions of significant control of the NMHS, metadata are critical for the
topography. most effective use of the information. As for stations
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–13

maintained by the NMHS, metadata on instrumen- observers’ working day, usually one morning obser-
tation, siting, processing procedures, methodologies vation and one afternoon or evening observation. If
and anything else that would enhance the use of daylight saving time is used for a part of the year, the
the information should be obtained and docu- observations should continue to be made according
mented. The metadata should also be maintained to the fixed local time; the dates when daylight
and accessible. To promote the open and unre- saving time commences and ends must be recorded.
stricted exchange of environmental information, If at all possible, the times of observation should
including weather observations, it is highly desira- coincide with either the main or intermediate stand-
ble that the NMHS be granted full use of all the ard times for synoptic observations (0000, 0300,
climate data and information obtained from part- 0600 Coordinated Universal Time (UTC), and so
nerships, without restriction, as if they were its own on). If conditions dictate that only one observation
data. An appropriate contract or “memorandum of a day is possible, this observation should be taken
understanding” between the NMHS and other between 0700 and 0900 local standard time.
organizations may need to be drafted and signed at
the senior management level. In selecting the schedule for climatological observa-
tions, times at or near the normal occurrence of
In addition to data from standard and private daily minimum and maximum temperatures
networks of climatological stations, there are some- should be avoided. Precipitation amounts and
times observational data from networks of maximum temperatures noted at an early morning
temporary stations established in connection with observation should be credited to the previous
research and study programmes, as well as measure- calendar day, while maximum temperatures
ments made in mobile transects and profiles. The recorded at an afternoon or evening observation
NMHS should endeavour to obtain these data and should be credited to the day on which they are
associated metadata. Although the data may not be observed.
ideal for typical archiving, they will often prove to
be quite valuable as supplementary information, Times of observation are often different among
for example, for investigations of specific extreme networks. Summary observations such as temperature
events. When these observations are collected from extremes or total precipitation made for one 24-hour
data-poor areas, they are highly valuable. period (such as from 0800 on one day to 0800 on the
next day) are not equivalent to those made for a
different 24-hour period (such as from 0000 to 2400).

2.6 STATION AND NETWORK If changes are made to the times of observations
OPERATIONS across a network, simultaneous observations should
be carried out at a basic network of representative
Guidance material in this section concerns mainly stations for a period covering the major climatic
observations at ordinary climatological stations (at seasons in the area at the old and new times of
which observations are usually made twice a day, observation. These simultaneous observations
but in some cases only once a day, and include should be evaluated to determine if any biases
readings of extreme temperature and precipitation). result from the changed observation times. The
Guidance is also given regarding precipitation station identifiers for the old and new times of
stations (stations at which one or more observa- observations must be unique for reporting and
tions of precipitation only are made each day). archiving.
Regulatory and guidance material for principal
climatological stations (which usually also function
2.6.2 Logging and reporting of
as synoptic observing stations) and other types of
observations
climatological stations is found in the Manual on
the Global Observing System. Immediately after taking an observation at a manual
station, the observer must enter the data into a
logbook, journal or register that is kept at the station
2.6.1 Times of observations
for this purpose. Alternatively, the observation may
Observations at ordinary climatological and precipi- be entered or transcribed immediately into a compu-
tation stations should be made at least once (and ter or transmission terminal and a database.
preferably twice) each day at fixed hours that remain Legislation or legal entities (such as courts of law) in
unchanged throughout the year. At principal clima- some countries may require that a paper record or a
tological stations, observations must be made at least printout of the original entry be retained for use as
three times daily in addition to an hourly tabulation evidence in legal cases, or there may be difficulties
from autographic records, but non-autographic associated with the acceptance of database-generated
observations are usually taken hourly. From a practi- information. The observer must ensure that a
cal viewpoint, times of observation should fit the complete and accurate record has been made of the
2–14 GUIDE TO CLIMATOLOGICAL PRACTICES

observation. At a specified frequency (ranging from sequence of observations, for consistency in the
immediately to once a month), depending on the sequence of dates and times of observation, for
requirements of the NMHS, data must be transferred consistency with other elements and calculations,
from the station record (including a computer data- and of the accuracy of copies and of encoded
base) to a specific report form for transmittal, either reports. These checks can be done either manually
by mail or electronically, to a central office. or by using automated procedures. If there are
errors, remedial action such as correcting the origi-
Climatological station personnel must ensure that nal data and the report should be taken before
there is a correct copy of the pertinent informa- transmission. Errors detected after transmission
tion in the report form. In the case of paper should also be corrected, with the corrected report
records, the need for good, clear handwriting and then transmitted. Checks should also be made,
“clean” journals and report forms should be and any necessary amendments recorded and
emphasized. It is quite common for more informa- corrections transmitted, if a query about data qual-
tion, perhaps pertaining to unusual weather ity is received from an outside source. Records of
phenomena and occurrences, to be entered in the an original observation containing an error should
local record than is required by the central office. include a notation or flag indicating that the origi-
The on-station record must be retained and readily nal value is erroneous or suspect. On-site quality
accessible so that the station personnel can control must also include the maintenance of the
respond to any inquiries made by the central office standard exposure of the sensors, of the site, and
regarding possible errors or omissions in the report of the proper procedures for reading the instru-
form. Some services request observers to send mentation and checking autographic charts.
logbooks to the national climate centre for perma-
nent archiving. Any patterns of measurement error should be
analysed, for example, to see if they relate to instru-
Some national climate centres require that station ment drift or malfunction, and summaries of data
personnel calculate and insert monthly totals and or report deficiencies should be prepared monthly
means of precipitation and temperature so that or annually.
the data may be more easily checked at the section
or central office. In addition, either the climate 2.6.4 Overall responsibilities of
centre or observer should encode data for the observers
CLIMAT messages, as described in the Handbook on
CLIMAT and CLIMAT TEMP Reporting (WMO/TD- In general, the NMHS of each Member will specify
No. 1188), if appropriate. Software to encode the the responsibilities of observers. The responsibilities
data has been developed by WMO. The observer should include the competent execution of the
should note in the station logbook and on the following:
report forms the nature and times of occurrence of (a) Making climatological observations to the
any damage to or failure of instruments, mainte- required accuracy with the aid of appropriate
nance activities, and any change in equipment or instruments;
exposure of the station, since such events might (b) Maintaining instruments and observing sites
significantly affect the observed data and thus the in good order;
climatological record. Where appropriate, instruc- (c) Performing appropriate quality checks;
tions should be provided for transmitting (d) Coding and dispatching observations in the
observations electronically. If mail is the method absence of automatic coding and communica-
of transmission, instructions for mailing should be tion systems;
provided to the station, as well as pre-addressed, (e) Maintaining in situ recording devices and
stamped envelopes for sending the report forms to electronic data loggers, including the chang-
the central climate office. ing of charts when provided;
(f) Making or collating weekly or monthly records
of climatological data, especially when auto-
2.6.3 On-site quality control
matic systems are unavailable or inadequate;
General guidance on on-site quality control of (g) Providing supplementary or backup observa-
observations and reports is given in Part V of the tions when automatic equipment does not
Manual on the Global Observing System and detailed observe all required elements, or when the
guidance is given in Part VI of the Guide to the Global equipment is out of service.
Observing System. The procedures described below
should be followed when there is an observer or
2.6.5 Observer training
other competent personnel on site.
Observers should be trained or certified by an
Checks should be made for gross errors, against appropriate meteorological service to establish their
existing extremes, for internal consistency in a competence to make observations to the required
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–15

standards. They should have the ability to interpret visits provide the opportunity to address siting or
instructions for the use of instrumental and manual instrument problems and to further the training of
techniques that apply to their own particular the observer.
observing systems. Guidance on the instrumental
training requirements for observers is given in Some climate centres arrange special training
Chapter 4, Part III, of the Guide to Meteorological courses for groups of volunteer observers. Such
Instruments and Methods of Observation (sixth courses are especially useful in creating a uniform
edition). high standard of observations, as a result of the
training given and the availability of time to address
Often, observers are either volunteers or part-time a wider range of problems than may be raised by a
employees, or take observations as part of their single observer at an on-site visit.
other duties. They may have little or no training in
climatology or in taking scientific observations, 2.6.6 Station inspections
and thus will depend on a good set of instructions.
Instructional booklets for ordinary climatological Principal climatological stations should be
and precipitation station observers should be care- inspected once a year. Ordinary climatological
fully prepared and made available to observers at all stations and precipitation stations should be
stations. The instructions should be unambiguous inspected at least once every three years, or more
and simply outline the tasks involved, being limited frequently if necessary, to ensure the maintenance
to that information which the observer actually and correct functioning of the instruments and
needs to know in order to perform the tasks satis- thus a high standard of observations. Automated
factorily. Illustrations, graphs and examples could stations should be inspected at least every six
be used to stimulate the interest of the observer and months. Special arrangements for the inspection of
facilitate the understanding of the tasks to be ship-based instruments are described in the Guide to
undertaken every day. Sample copies of correctly Meteorological Instruments and Methods of Observation
completed pages of a logbook or journal and of a (fifth edition).
report form should be included in the instruction
material available to an observer. Ideally, a climate Before each inspection, the inspector should deter-
centre representative should visit the site, install mine to the fullest extent possible the quality of
the station and instruct the observer. information and data received from each station on
the itinerary. At each inspection, it should be
An observer must gain familiarity with the instru- confirmed that:
ments, and should be aware in particular of the (a) The observer’s training is up to date;
sources of possible error in reading them. The (b) The observer remains competent;
instructions should include a descriptive text with (c) The siting and exposure of each instrument are
simple illustrations showing the functioning of known, recorded and still the best obtainable;
each instrument. Detailed instructions regarding (d) The instruments are of an approved pattern,
methods to be used for day-to-day care, simple in good order and verified against relevant
instrument maintenance and calibration checks standards;
should be given. If correction or calibration tables (e) There is uniformity in the method of obser-
are necessary for particular observing and recording vation and procedures for calculating derived
tasks, the observer should be made thoroughly quantities from the observations;
familiar with their use. Instructions should also (f) The station logbook is well maintained;
cover the operation of computer terminals used for (g) The required report forms are sent punctually
data entry and transmissions. and regularly to the climate centre.

Instructions must cover visual as well as instrumen- Inspection reports should include sketches, photo-
tal observations. Visual observations are particularly graphs or diagrams of the immediate observing site,
prone to subjective error and their accuracy depends indicating physical objects that might influence the
on the skill and experience acquired by the observer. observed values of the climatic elements. The
Since it is very difficult to check the accuracy or reports must also list any changes in instruments
validity of an individual visual observation, as and any differences in readings between instru-
much guidance as possible should be given so that ments and travelling standards, changes in exposure
correct observations can be made. and site characteristics from the previous visit, and
dates of appropriate comparisons and changes.
To complement the instruction material, personnel Inspectors must also be prepared to advise observ-
responsible for station management in the climato- ers on any problems arising in the transmission of
logical service should contact observing stations data, including automated data-entry and transmis-
regarding any recurring observing errors or misin- sion systems. Inspection reports are an important
terpretation of instructions. Regular inspection source of metadata for use in determining the
2–16 GUIDE TO CLIMATOLOGICAL PRACTICES

homogeneity of a climate record and should be stations, a variety of statistical checks, checks
retained indefinitely, or the information therein against preset limits, temporal consistency and
should be transferred to a computerized database inter-element consistency. Chapters 4 and 5
(see section 3.1). describe some of the techniques for checking data.

Monitoring shortly after observations are taken,

2.6.7 Preserving data homogeneity
either on site or remotely, is of limited value unless
Unlike observations taken solely to support the action is initiated to quickly remedy problems.
preparations of forecasts and warnings, the availa- Information must be fed back to the observers,
bility of a continuous, uninterrupted climate record caretakers, inspectors, and instrument or system
is the basis for many important studies involving a maintainers or manufacturers, and then informa-
diverse array of climatological communities. tion on the actions taken must be fed back again to
Homogeneous climate datasets are of the utmost the monitoring centre. Copies of all reports must be
importance for meeting the needs of climate kept.
research, applications and user services.
2.6.9 Station documentation and
Changes to a site, or its relocation, are major causes
metadata
of inhomogeneities. The 10 principles of climate
monitoring (see section 2.1) should be followed The efficient use of climatological data requires the
when relocation of a climatological station is neces- climatological or other responsible section to main-
sary, when one station is to be replaced by another tain complete documentation of all stations in the
nearby, or when instrument systems change. Where country for all networks and observing platforms.
feasible and practical, both the old and new observ- These metadata are essential and should be kept
ing stations and instrumentation should be current and be easily obtainable in the form of
operated for an overlapping period of at least one station catalogues, data inventories and climate
year, and preferably two or more years, to deter- data files. The World Meteorological Organization
mine the effects of changed instruments or sites on is currently developing metadata standards based
the climatological data. The old and new sites on International Organization for Standardization
should have unique station identifiers for both (ISO) metadata standards, especially the ISO 19100
reporting and archiving. Specific guidance is given series. The guidance given below should be followed
in the Guidelines for Managing Changes in Climate unless it is superseded by published climate meta-
Observation Programmes (WMO/TD-No. 1378). data standards.

Basic station metadata should include station name

2.6.8 Report monitoring at collection
and station index number (or numbers); geographi-
centres
cal coordinates; elevation above mean sea level;
Data collection or archiving centres need to check administrator or owner; types of soil, physical
the availability and quality of information at the constants and profile of soil; types of vegetation
time when it is due from observers, and they should and condition; local topography description;
have additional responsibilities concerning data description of surrounding land use; photographs
from automated measuring or transmission and diagrams of the instrumentation, site and
systems. Since such centres normally process large surrounding area; type of AWS, manufacturer,
volumes of information, computerized checking model and serial number; observing programme of
systems save much effort. the station (elements measured, reference time,
times at which observations and measurements are
The first task is to check that the expected observa- made and reported, and the datum level to which
tions have arrived and that they have been atmospheric pressure data of the station refer); and
submitted at the correct time. If the expected obser- contact information, such as name and mailing
vations are not available, the observer should be address, electronic mail address, and telephone
contacted to determine the reason. In the case of numbers.
automated systems, “caretakers” must provide
information on visible signs of failure as soon as Documentation should contain a complete history
possible to the authority responsible for mainte- of the station, giving the dates and details of all
nance of the observing and transmission systems. changes. It should cover the establishment of the
station, commencement of observations, any inter-
Checks on the quality of data received from manned ruptions to operation, and eventually the station’s
or automated sites should include those described closure. Comments from inspection visits (see 2.6.6)
in section 2.6.3. Other checks are useful and can be are also important, especially comments about the
readily made in computerized monitoring. They site, exposure, quality of observations and station
include checks against data from neighbouring operations.
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–17

Instrument metadata should include sensor type, Practices; Vol. II – Meteorological Service for
manufacturer, model and serial number; principle of International Air Navigation; Vol. III – Hydrology
operation; method of measurement and observa- (WMO-No. 49), Geneva.
tion; type of detection system; performance ———, 1990: Manual on Marine Meteorological
characteristics; unit of measurement and measuring Services, Vol. I – Global Aspects; Vol. II –Regional
range; resolution, accuracy (uncertainty), time Aspects (WMO-No. 558), Geneva.
constant, time resolution and output averaging ———, 1991, 1992: Manual on the Global Data-
time; siting and exposure (location, shielding and processing System, Vols. I and II (WMO-No. 485),
height above or below ground); date of installation; Geneva.
data acquisition (sampling interval and averaging ———, 1994: Guide to the Applications of Marine
interval and type); correction procedures; calibration Climatology (WMO-No. 781), Geneva.
data and time of calibration; preventive and correc- ———, 1994: Guide to Hydrological Practices. Fifth
tive maintenance (recommended and scheduled edition (WMO-No. 168), Geneva.
maintenance and calibration procedures, including ———, 1994: Observing the World’s Environment:
frequency, and a description of procedures); and Weather, Climate and Water (J.P. Bruce) (WMO-
results of comparison with travelling standards. No. 796), Geneva.
———, 1995: GCOS Guide to Satellite Instruments for
For each individual meteorological element, meta- Climate (WMO/TD-No. 685, GCOS-No. 16),
data related to procedures for processing Geneva.
observations should include the measuring and ———, 1995: GCOS Plan for Space-Based Observations.
observing programme (time of observations, report- Version 1.0 (WMO/TD-No. 684, GCOS-No. 15),
ing frequency and data output); the data processing Geneva.
method, procedure and algorithm; formulae for ———, 1995: Manual on the Global Observing
calculations; the mode of observation and measure- System, Vol. II – Regional Aspects (WMO-
ment; the processing interval; the reported No. 544), Geneva.
resolution; the input source (instrument and ———, 1995: Plan for the Global Climate Observing
element); and constants and parameter values. System (GCOS). Version 1.0 (WMO/TD-No. 681,
GCOS-No. 14), Geneva.
Data-handling metadata should include quality ———, 1997: Report of the GCOS/GOOS/GTOS Joint Data
control procedures and algorithms, definitions of and Information Management Panel, third session
quality control flags, constants and parameter (Tokyo, Japan, 15–18 July 1997) (WMO/TD-
values, and processing and storage procedures. The No. 847, GCOS-No. 39), Geneva.
transmission-related metadata of interest are ———, 1998: Preliminary Statement of Guidance
method of transmission, data format, transmission Regarding How Well Satellite Capabilities Meet
time and transmission frequency. WMO User Requirements in Several Application
Areas (WMO/TD-No. 913, SAT-21), Geneva.
Upper-air stations have metadata requirements that ———, 1999: Representativeness, Data Gaps and
are similar to those of surface stations. In addition, Uncertainties in Climate Observations (invited
they must maintain metadata on each of the expend- lecture given by Chris Folland to the Thirteenth
able instruments used (such as radiosondes). World Meteorological Congress, 21 May 1999)
(WMO/TD-No. 977, WCDMP-No. 44), Geneva.
———, 2000: SPARC Assessment of Upper Tropospheric
and Stratospheric Water Vapour (WMO/TD-
2.7 REFERENCES AND ADDITIONAL No. 1043, WCCRP-No. 113), Geneva.
READING ———, 2000: WMO Technical Conference on
Meteorological and Environmental Instruments and
Methods of Observation (TECO 2000), Lightning
2.7.1 WMO publications
Detection Systems, (WMO/TD-No. 1028),
World Meteorological Organization, 1978: Geneva.
International Operations Handbook for Measure­ ———, 2001: Guide to Marine Meteorological Services.
ment of Background Atmospheric Pollution Third edition (WMO-No. 471), Geneva.
(WMO-No. 491), Geneva. ———, 2002: Guide to the GCOS Surface and Upper-
———, 1981: Guide to Agricultural Meteorological Air Networks. GSN and GUAN, Version 1.1
Practices. Second edition (WMO-No. 134), Geneva. (WMO/TD-No. 1106, GCOS-No. 73), Geneva.
———, 1983, 1996, 2008: Guide to Meteorological ———, 2003: Aerosol Measurement Procedures
Instruments and Methods of Observation. Fifth, Guidelines and Recommendations (WMO/TD-
sixth and seventh editions (WMO-No. 8), No. 1178, GAW-No. 153), Geneva.
Geneva. ———, 2003: Guidelines on Climate Observation,
———, 1988: Technical Regulations, Vol. I – General Network and Systems (WMO/TD-No. 1185,
Meteorological Standards and Recommended WCDMP-No. 52), Geneva.
2–18 GUIDE TO CLIMATOLOGICAL PRACTICES

———, 2003: Manual on the Global Observing System, J. Lanzante, S. Klein, J. Christy, S. Schroeder,
Vol. I – Global Aspects (WMO-No. 544), Geneva. B. Soden and L.M. McMillin, 2002: Creating
———, 2003: The Second Report on the Adequacy of the climate reference datasets: CARDS workshop on
Global Observing System for Climate in Support of the adjusting radiosonde temperature data for
UNFCCC (WMO/TD-No. 1143, GCOS-No. 82), climate monitoring. Bull. Amer. Meteorol. Soc.,
Geneva. 83:891–899.
———, 2004: The Changing Atmosphere: An Integrated Free, M. and D.J. Seidel, 2005: Causes of differing
Global Atmospheric Chemistry Observation Theme temperature trends in radiosonde upper-air
for the IGOS Partnership (WMO/TD-No. 1235, datasets. J. Geophys. Res., 110:D07101.
ESA SP-1282), Geneva. Haas, T.C.: 1992. Redesigning continental-scale
———, 2004: Guidelines on Climate Data Rescue, monitoring networks. Atmospheric Environment,
(WMO/TD-No. 1210, WCDMP-No. 55), Geneva. 26A:3323–3333.
———, 2004: Handbook on CLIMAT and CLIMAT Global Ocean Observing System (GOOS), 2000:
TEMP Reporting (WMO/TD-No. 1188), Geneva. Report of the First GOOS Users’ Forum (GOOS-
———, 2004: Implementation Plan for the Global No. 92). UNESCO, Paris.
Climate Observing System for Climate in Support of ———, 2001: Principles of the Global Ocean Observing
the UNFCCC (WMO/TD-No. 1219, GCOS-No. 92), System (GOOS) Capacity Building (GOOS-
Geneva. No. 69). UNESCO, Paris.
———, 2006: Initial Guidance to Obtain Representative ———, 2001: GOOS Capacity Building Implementation
Meteorological Observations at Urban Sites Strategy (GOOS-No. 106). UNESCO, Paris.
(WMO/TD-No. 1250), Geneva. Karl, T.R., C.N. Williams Jr, P.J. Young and
———, 2006: Systematic Observation Requirements for W.M. Wendland, 1986: A model to estimate the
Satellite-based Products for Climate (WMO/TD- time of observation bias associated with monthly
No. 1338, GCOS-No. 107), Geneva. mean maximum, minimum, and mean temper-
———, 2007: Guide to the Global Observing System ature for the United States. J. Climate Appl.
(WMO-No. 488), Geneva. Meteorol., 25:145–160.
———, 2007: Guidelines for Managing Changes in Leroy, M., 1998: Meteorological measurements
Climate Observation Programmes (WMO/TD- representativity: nearby obstacles influence. In
No. 1378, WCDMP-No. 62), Geneva. Tenth Symposium on Meteorological Observations
———, 2007: RA IV Training Seminar on Capacity and Instrumentation, Phoenix, Arizona, 11–
Building in Climate-related Matters (WMO/TD- 16 January 1998. Boston, American
No. 1386, WCDMP-No. 63), Geneva. Meteorological Society.
———, 2008: Plant Phenological Guidelines (WMO/TD- ———, 1998: Climatological site classification. In:
No. 1484, WCDMP-No. 70), Geneva. Tenth Symposium on Meteorological Observations
and Instrumentation, Phoenix, Arizona, 11–
16 January 1998. Boston, American
2.7.2 Additional reading
Meteorological Society.
Bénichou, P., 1992: Optimisation de l’implantation Lieth, H. (ed.), 1974: Phenology and Seasonality
d’un réseau complémentaire en zone acciden- Modeling. New York, Springer.
tée: application au réseau automatisé National Climatic Data Center (NCDC), 2002: U.S.
d’Auvergne. La Météorologie, 43-44: 3–17. Climate Reference Network Site Information
Christian, H., 2003: Global lightning activity. In: Handbook. Asheville, NCDC.
Proceedings of the 12th International Conference on Panel on Reconciling Temperature Observations,
Atmospheric Electricity, Versailles, France, 9–11 Climate Research Committee, Board on
June 2003 (S. Chauzy and P. Laroche, eds). Atmospheric Sciences and Climate, National
Paris, ONERA. Research Council, 2000: Reconciling Observations
Cronin, T.M., 1999: Principles of Paleoclimatology. of Global Temperature Change. Washington, DC,
New York, Columbia University Press. National Academy Press.
Der Mégréditchian, G., 1985: Méthodes statistiques ———, 2000: Improving Atmospheric Temperature
d’analyse et d’interpolation spatiales des Monitoring Capabilities: Letter Repor t.
champs météorologiques. La Météorologie, Washington, DC, National Academy Press.
17:51–66. Panel on Climate Change Feedbacks, Climate
Dover, J. and L.J. Winans, 2002: Evaluation of Research Committee, National Research
windshields for use in the automated surface Council, 2003: Understanding Climate Change
observing system (ASOS). In: Proceedings of the Feedbacks. Washington, DC, National Academy
Sixth Symposium on Integrated Observing Systems, Press.
Orlando, Florida, 14–17 January 2002. Boston, Ohring, G., B. Wielicki, R. Spencer, B. Emery and
American Meteorological Society. R. Datla, 2005: Satellite instrument calibration for
Free, M., I. Durre, E. Aguilar, D. Seidel, T.C. Peterson, measuring global climate change: report of a work-
R.E. Eskridge, J.K. Luers, D. Parker, M. Gordon, shop. Bull. Amer. Meteorol. Soc., 86:1303–1313.
 CHAPTER 2. CLIMATE OBSERVATIONS, STATIONS AND NETWORKS 2–19

Oke, T.R., 1987: Boundary Layer Climates. Second Thompson, D.W.J. and J.M. Wallace, 2000: Annular
edition. Cambridge, Cambridge University Press. modes in the extratropical circulation. Part I:
———, 2006: Towards better scientific communica- Month-to-month variability. J. Climate,
tion in urban climate. Theor. Appl. Climatol., 13:1000–1016.
84:179-190. Thorne, P.W., D.E. Parker, J.R. Christy and
Quadrelli, R. and J.M. Wallace, 2004: A simplified C.A. Mears, 2005: Uncertainties in climate
linear framework for interpreting patterns of trends: Lessons from upper-air temperature
Northern Hemisphere wintertime climate vari- records. Bull. Amer. Meteorol. Soc.,
ability. J. Climate, 17:3728–3744. 86:1437–1444.
Santer, B.D., M.F. Wehner, T.M.L. Wigley, Thorne, P.W., D.E. Parker, S.F.B. Tett, P.D. Jones,
R. Sausen, G.A. Meehl, K.E. Taylor, C. Ammann, M. McCarthy, H. Coleman and P. Brohan, 2005:
J. Arblaster, W.M. Washington, J.S. Boyle and Revisiting radiosonde upper-air temperatures
W. Bruggemann, 2003: Contributions of anthro- from 1958 to 2002. J. Geophys. Res., 110
pogenic and natural forcing to recent tropopause (D18):D18105.
height changes. Science, 301(5632):479–483. Unninayar, S. and R.A. Schiffer, 1997: In-Situ
Schwartz, M.D. (ed.), 2003: Phenology: An Integrative Observations for the Global Observing Systems: A
Environmental Science. Tasks for Vegetation Science, Compendium of Requirements and Systems.
Vol. 39. Dordrecht, Kluwer Academic Publishers. Washington, DC, National Aeronautics and
Seidel, D.J. and M. Free, 2006: Measurement require- Space Administration Office of Mission to
ments for climate monitoring of upper-air Planet Earth, NP-1997(01)-002-GSFC.
temperature derived from reanalysis data. Wang, J., D.J. Carlson, D.B. Parsons, T.F.Hock,
J. Climate, 19:854–871. D. Lauritsen, H.L. Cole, K. Beierle and
Singleton, F. and E.A. Spackman, 1984: Climatological E. Chamberlain, 2003: Performance of opera-
network design. Meteorol. Mag., 113:77–89. tional radiosonde humidity sensors in direct
Sneyers, R., 1973: Sur la densité optimale des comparison with a chilled mirror dew-point
r é s e a u x météorologiques. Archiv für hygrometer and its climate implication.
Meteorologie, Geophysik und Bioklimatologie Serie Geophys. Res. Lett., 30:10.1029/2003GL016985.
B, 21:17–24. Zhang, H.-M. and L.D. Talley, 1998: Heat and
Soden, B.J. and I.M. Held, 2006: An assessment of buoyancy budgets and mixing rates in the
climate feedbacks in coupled ocean–atmosphere upper thermocline. J. Phys. Oceanogr.,
models. J. Climate, 19:3354–3360. 28:1961–1978.

.
 CHAPTER 3

CLIMATE DATA MANAGEMENT

3.1 INTRODUCTION temperature measurements and raingauges

facilitated the measurement of precipitation. The
development of clock-driven mechanisms
For thousands of years historians have recorded permitted the establishment of intensity and
information about the weather. In the past, duration values and the recording of these data.
however, this information was often based on Other types of recording instruments provided
accounts from others and was not drawn from the autographic or analog records. With each new
historians’ personal observations. Such accounts improvement or addition to the tools of observation,
may have been vague, truncated or affected by the number of items or variables entered in journals
memory lapses. This type of weather information and logbooks increased and specially prepared
was embedded within an immense array of other formats were developed. Even as formats have
kinds of information, and much of it is contained changed, regularity and consistency, or continuity
in national libraries and archives. Specialized of the record-keeping, have always been highly
national meteorological archives are a relatively desirable. A good chronological record should be
recent phenomenon, with the earliest typically kept current and in sequential order. Methodical
being established during the first half of the twenti- and careful observation and recording permit easier
eth century. collection, archiving and subsequent use of the
records.
Early records in manuscript form were kept in daily,
weekly or monthly journals. Notes were made of In most countries, manuscript forms were sent peri-
extreme or catastrophic events such as high or low odically to a central location. Until the 1970s, these
temperatures, abnormal wind speeds, excessive original forms constituted the bulk of all the hold-
rainfall or prolonged drought, dates of frost or ings of climatological information at most
freezing, hurricanes, and tornadoes. Storms, calms, collection centres. These centres may have been a
winds, currents, types of cloud and cloudiness were section of the local or national government or the
noted in marine logbooks. Freezing and thawing central office of an industry such as mining, agri-
dates of rivers, lakes and seas, as well as the first and culture or aviation. Gradually, the climatological
last dates of snowfall, were often recorded as an data-gathering activities affecting national life were
important part of any journal. Pride of work and assembled within a concerted programme of obser-
accomplishment has always been an important vation and collection to serve national and
feature in weather observing and recording. The international interests.
person responsible for an observation who signs or
seals the logbook still lends authority to records Since the late twentieth century, most weather
and serves as the personal source of the recorded information has been transmitted digitally to
history. centralized national collection centres. As the
messages have been intended primarily for opera-
Specific journals for the collection and retention of tional weather forecasting, it has been common
climatological information have been established practice to rely on the original observing docu-
within the last two or three centuries. Even up to ments for the creation of the climate record in
the 1940s, the forms developed, printed and used climate centres around the world. The collection,
in various countries were often different, and the transmission, processing and storage of operational
observations were almost always recorded by hand. meteorological data, however, are being dramati-
Since the 1940s and especially following the estab- cally improved by rapid advances in computer
lishment of WMO, standardized forms and technology, and increasingly meteorological
procedures have gradually become prevalent, and archives are being populated with data that have
national meteorological archives have been desig- never been recorded on paper. The power and ease
nated as the storage site for these records. of use of computers, the ability to record and trans-
fer information electronically, and the development
The quantification of climatological data developed of international exchange mechanisms such as the
as improved instrumentation facilitated the Internet have given climatologists new tools to
observation of continuous, as well as discrete, rapidly improve the understanding of climate.
variables and the recording of appropriate values in
journals or logbooks. For example, thermometers Every effort should be made to obtain, in electronic
enabled the systematic recording of quantitative digital form, a complete collection of all primary
3–2 GUIDE TO CLIMATOLOGICAL PRACTICES

observed data. Collection of data electronically at 3.3 CLIMATE DATA MANAGEMENT

the source allows rapid and automatic control
measures to be applied, including error checking,
prior to the data’s transmission from the observation Climatological data are most useful if they are
site. In many cases, the collection of climate data edited, quality-controlled and stored in a national
by mail may still be cheaper and more reliable, archive or climate centre and made readily accessi-
especially in less technologically advanced regions, ble in easy-to-use forms. Although technological
but unless the data have been recorded on some innovations are occurring at a rapid pace, many
form of electronic media prior to postage, they will climatological records held by NMHSs are still in
need to be scanned or digitized centrally. The non-digital form. These records must be managed
management of the vast variety of data collected along with the increasing quantity of digital
for meteorological or climatological purposes records. A climate data management system
requires a systematic approach that encompasses (CDMS) is a set of tools and procedures that allows
paper records, microform records and digital all data relevant to climate studies to be properly
records. stored and managed.

The primary goals of database management are to

maintain the integrity of the database at all times, and
3.2 THE IMPORTANCE AND PURPOSE OF to ensure that the database contains all the data and
MANAGING DATA metadata needed to meet the requirements for which
it was established, both now and into the future.
Database management systems have revolutionized
The basic goal of climate data management is to climate data management by allowing efficient stor-
preserve, capture and provide access to climate age, access, conversion and updating for many types
data and products for use by planners, decision- of data, and by enhancing data security.
makers and researchers. Permanent archiving is an
important objective. The data management system A major step forward in climate database manage-
of a climate archive must provide the information ment occurred with the World Climate Data and
to describe the climate of the domain of interest Monitoring Programme (WCDMP) Climate
for which the archive has been established, be it Computing (CLICOM) project in 1985. This project
national, regional or global. Data produced from led to the installation of climate database software
meteorological and climatological networks and on personal computers, thus providing NMHSs in
various research projects represent a valuable and even the smallest of countries with the capability of
often unique resource, acquired with substantial efficiently managing their climate records. The
expenditure of time, money and effort. Many of project also provided the foundation for demonstra-
the ultimate uses for climate data cannot be fore- ble improvements in climate services, applications
seen when the data acquisition programmes are and research. In the late 1990s, WCDMP initiated a
being planned, and frequently new applications CDMS project to take advantage of the latest tech-
emerge, long after the information is acquired. nologies to meet the varied and growing data
The initial utilization of meteorological and management needs of WMO Members. Aside from
related data is often only the first of many applica- advances in database technologies such as relational
tions. Subsequent analysis of the data for many databases, query languages and links with
and diverse purposes leads to a significant and Geographical Information Systems (GIS), more effi-
ongoing enhancement of the return on the origi- cient data capture was made possible with the
nal investment in the data acquisition programmes. increase in AWSs, electronic field books, the Internet
The global climate change issue, for example, is and other advances in technology.
stretching the requirements for climate data and
data management systems far beyond those origi- It is essential that both the development of climate
nally conceived when the original networks were databases and the implementation of data manage-
established. To meet these expanding needs, it is ment practices take into account the needs and
critically important that climate information, capabilities of existing and future data users. While
both current and historical, be managed in a this requirement may seem intuitive, information
systematic and comprehensive manner. important for a useful application is sometimes omit-
Conventional meteorological data are now ted, or data centres commit insufficient resources to
augmented by data from a wide array of new checking the quality of data for which users explicitly
instruments and systems, including satellites, or implicitly demand high quality. For instance, a
radar systems and other remote-sensing devices, database without both current and past weather codes
thus making effective and comprehensive climate could lead to underestimates of the prevalence of
data management systems essential for modern observed phenomena. In all new developments, data
climate centres. managers should attempt to have at least one key
 CHAPTER 3. CLIMATE DATA MANAGEMENT 3–3

data user as part of the project team or to undertake constantly reviewed to ensure that it is meeting the
some regular consultative process with user stake- requirements of users and archivists. Figure 3.1
holders to keep abreast of both changes in needs and depicts the relationships and flow among the func-
any issues that user communities may have. Examples tional components of a generalized data
of stakeholder communities are climate prediction, management system.
climate change, agriculture, public health, disaster
and emergency management, energy, natural resource Metadata do not appear as a distinct box in the
management, urban planning, finance and diagram because they are gathered and consoli-
insurance. dated from all components of the data management
system. For example, at each observing site, instru-
ment documentation and site environmental
3.3.1 CDMS design
conditions must be gathered, and during the qual-
All CDMSs are based on some underlying model of ity control phase, the algorithms and methodology
the data. This model design is very important for must be thoroughly documented. The entirety of
the quality of the resulting system. An inappropriate information about the data values and the data
model will tend to make the system harder to processing represents the metadata for the system.
develop and maintain. In general, a database
designed for current meteorological data will allow
3.3.2 CDMS data acquisition
rapid retrieval of recent data from a large number of
stations. By contrast, many climate data applications Data that are already in digital form can be readily
involve the retrieval of data for one or a few stations ingested directly by the system. Non-digital records
over a long period. It is essential to document the are generally digitized during an entry process. A
overall design and underlying data model of the fundamental goal of a data-entry process is to
CDMS to facilitate subsequent extension or duplicate, with a minimum of error, the raw data as
modification by computer programmers. Similar they were recorded in the capture process. A
considerations apply to a metadata model. Details key-based entry system should be efficient and easy
about data models can be found in the Guidelines on for a data-entry operator to use. The system could
Climate Data Management (WMO/TD-No. 1376). also be designed to validate the data as they are
entered and to detect likely errors. It is also possible
Once a data management system has been estab- to set default values for some elements, thus saving
lished and becomes operational, it should be unnecessary keystrokes.

Figure 3.1. Generalized data management system

3–4 GUIDE TO CLIMATOLOGICAL PRACTICES

Where AWSs are in use, climate data, including any synoptic conditions precluded the occurrence of
error control messages, should be transferred elec- precipitation. In other cases, lost data can be esti-
tronically to the CDMS. Manually observed data mated with reasonable certainty using the
should be transferred to the CDMS as soon as possi- techniques discussed in section 5.9. In all cases,
ble by whatever means are most practical. It is dataset documentation should flag the recon-
advantageous to collect data at least daily because structed or estimated data appropriately (see 3.4.1).
data quality is likely to be improved, the manual
effort for quality control will likely decrease, tech-
3.3.3 CDMS data documentation
nical errors will be detected faster, and there will be
greater opportunities for improved access to more An adequate set of metadata must be available to
data. Nevertheless, the submission of data for a inform future users about the nature of the data in
month is an acceptable alternative when daily data the system, how the various datasets were collected,
transmission is not practicable. For example, many and any inherent problems. It is recommended that
of the 6 000 or so voluntary observers in Australia database management include all information that
continue to forward monthly rainfall reports that can affect the homogeneity of a dataset or series,
contain the daily observations for the month. including those factors outlined in section 2.6.9.

Many weather observations are recorded by institu- The structure of metadata in an ideal system is
tions or organizations other than NMHSs, and generally more complex than the data themselves.
acquisition of the data in their original form may be For example, a rainfall observation essentially gives
difficult. In these cases efforts should be made to the quantity of precipitation over a certain period
collect copies of the original report forms. If it is at a certain station. The associated metadata that
impossible to secure either the original or a copy of can be applied to this observation, and which
the record, a note to this effect should be made in the might be needed to interpret the data fully, could
inventory of the centre’s holdings, stating informa- include information such as the reference date used
tion pertaining to the existence and location of the by the database (for example, Greenwich Mean
data, volume available, period of record covered, Time (GMT) or time zone); quality indicators or
stations in the network as applicable, and elements flags that have been ascribed to the observation;
observed. history of changes made to the values and any asso-
ciated flags; instrument used to record the
Though not a formal requirement, it is recom- observation, together with details about mainte-
mended that the CDMS also contain information nance programmes, tolerances, internal parameters
about media reports, pictures and other similar and similar information; name and contact infor-
information beyond the traditional data and meta- mation of the observer; full details of the location
data. Such information could be captured by and siting of a station and its history; programme
imaging the report from the print media with a of observations in effect at the time and its history;
digital camera or scanner; defining the date, area and topographical and ground-cover details of the
and type of event (such as flood, drought or heavy site. A detailed treatment of station-specific meta-
precipitation); identifying the media; and data can be found in the Guidelines on Climate
writing additional comments about the event. Metadata and Homogenization (WMO/TD-No. 1186).
As another example, the metadata associated with a
It is important to retain both the data value that gridded dataset of satellite observations of high-
was originally received, as well as the latest quality- resolution solar exposure should include the
controlled value. The original value will likely pass geographical extent of the observations, the period
through an initial automated quality control proc- of record of the dataset, a history of revisions to and
ess at ingestion and as necessary, a more extensive maintenance of the dataset, the satellites from
quality control process, and even if rejected by which the data were obtained, transfer functions
either of these processes, it must still be retained. and averaging procedures to obtain grid values,
Some CDMSs retain not only the original and latest satellite positional accuracy, information about the
values, but also all modifications. accuracy of the data, and contact information.

Another aspect of data acquisition is the recording Metadata are also needed for the CDMS itself. Each
of occurrences when data were expected but not process within the system (for example, key entry or
received. Loss of data can occur as a result of situa- quality control) should be completely described. A
tions such as inoperable instrumentation, data history of any changes made to any part of the
transmission errors and acquisition processing system (for example, software, hardware or manual
errors. Lost data can be reconstructed with varying procedures) should be documented and maintained.
levels of certainty. For example, a missing precipita- Since observing practices, quality control techniques
tion measurement can be considered to be zero and data-handling procedures change over time,
when it is known from other data that local and these metadata are critical in the climatological
 CHAPTER 3. CLIMATE DATA MANAGEMENT 3–5

analysis of historical data. The analyst uses the would be storage in acid-free boxes in air-conditioned,
metadata to identify and understand how a data secure storerooms. A maintenance programme should
value was observed and processed in order to separate be established to rescue deteriorating documents and
meteorological from possible non-meteorological particularly the data they contain.
influences in the data record.
With an ever-increasing amount of information
Another category of metadata is the record of data being generated and retained, the problem arises as
holdings in the CDMS. Inventories of the data to whether or not to continue to store all the records
contained in the CDMS should be prepared in their original manuscript form. All too often,
routinely. Stratification could be, for example, by climatological records are stored in basements,
data element, station location, or time. Lists of sheds and other undesirable facilities. They are
contents should be established and maintained to frequently not catalogued, and they may be inac-
describe and define the data content of the indi- cessible and subject to deterioration. As a means of
vidual files and to provide information on the codes reducing paper costs, making better use of space,
and observational practices used. Knowing what is and providing security for original documents, it is
contained in the CDMS is important for efficient recommended that the manuscript data be scanned
retrieval of information from the system. The WMO into a digital file and carefully preserved. The elec-
core profile of the ISO 19100 series for data and tronic images of the paper records can then be
metadata should be used unless it is superseded by stored and retrieved using computer technologies
published climate metadata standards. such as optical character recognition software. The
specifications of computer hardware required for
storing and retrieving documents depend on data
3.3.4 CDMS data storage
needs and the limits of financial resources, and also
An important function of the data manager is to esti- on technological advances, so there are no global
mate data storage requirements, including the standards or a preferred single storage medium. It is
estimation of future growth. Account must be taken important to remember that no storage medium is
of the additional information to be included in data permanent, and therefore regular review of archival
records (for example, data quality flags, original arrangements should be undertaken. Computer-
messages, and date and time of record updates), based archives must be securely and regularly
metadata needs, and any redundancy necessary to backed up with at least one copy stored at a site
ensure that databases can be restored. Some data separate from the main archive.
types (such as those from remote-sensing, oceanog-
raphy and AWSs with high temporal resolution) Microform refers to document images photographi-
require large amounts of storage. Unconventional cally reduced to a very small fraction of their
data (such as soil moisture, phenological observa- original size. A variety of microform formats exist.
tions and vegetation indices) may have storage needs Examples are microfilm reels, microfiche sheets,
that are different from the more traditional observa- jackets, film folios, aperture cards, cartridges and
tions. AWSs will often generate data that are relevant cassettes. Significant advances in digital storage
to the quality of observations, but are not strictly capabilities, however, now make it highly desirable
climate data (for example, information on the that paper documents be directly scanned or digit-
battery-level voltage for an AWS). Generally, this ally photographed into a computer system along
information should be utilized prior to data archiv- with superseded microform images. This process
ing; if it is not included in the CDMS, it should be facilitates access and ensures preservation for future
permanently retained elsewhere and made accessi- generations.
ble to data managers. The quality control process
often generates values and information that may be
3.3.5 CDMS data access and retrieval
different from the original data, so there is a storage
requirement for keeping both the original data and An important aspect of any CDMS is the power of
any different data generated by quality control proc- the facilities to perform data retrieval and analysis.
esses. Estimating future growth can be very difficult, Graphical user interface retrieval facilities should
as it is hard to determine what data types may be provided for most users, and command line
become available as time and technology progress. facilities should be available for the small number
The data manager must consider all of these factors of knowledgeable users who have a need for non-
when determining storage requirements. standard retrievals. Users should be able to specify
their own retrieval criteria, and the system docu-
Non-digital records should be stored in a way that mentation should be clear and provide as much
minimizes their deterioration. They should be stored information as necessary to support the users.
in a controlled environment to avoid temperature
and humidity extremes, insects, pests, fire, flood, acci- Output options should be extensive and include
dents or deliberate destruction. An ideal example facilities for customizing stations, times and details
3–6 GUIDE TO CLIMATOLOGICAL PRACTICES

of output presentations. Access should be given to (j) Every so often, typically monthly, a
listings of data, tabular summaries, statistical analy- complete backup of the data tables should
ses and graphical presentation. be put in a safe, secure, fireproof location,
remote from the physical location of the
climate database. It is common to have
3.3.6 CDMS archives
three copies of the same archive in differ-
An archive is the permanent retention of the data and ent secure places and if possible in different
metadata in the CDMS. The archive structure, whether towns or cities;
simple or complex, physical or electronic, should be (k) Backups of the CDMS must be performed
guided by financial resources, the degree of training of prior to any changes to the system software,
archive personnel, the volume of data to be archived, to the system design or to the applications
the media of the data (such as paper documents or contained in the CDMS.
digital format), the ease of putting information into
and retrieving information from the archive, user-
3.3.8 CDMS management
friendliness in accessing information, the ease of
maintenance of the archive, and the ease of expan- A CDMS should be monitored routinely to deter-
sion as data holdings increase. All aspects of the mine how well the processes using and supporting
CDMS should be archived, including not just the data the database are performing. Examples of the proc-
values but also catalogues, inventories, histories, esses that support the data are metadata maintenance,
dictionaries, and other similar information. database ingest, quality control actions that modify
the database, and information retrieval. Each process
should be monitored, evaluated, and if necessary,
3.3.7 CDMS security
improved. It is strongly recommended that data
The main goal of a security policy and associated managers think in terms of end-to-end data manage-
activities is to prevent loss of or damage to the ment, with information on systemic data quality
CDMS. To satisfy this goal, the requisites are: issues, loss of data, or other practices that harm the
(a) All personnel must be aware of their profes- climate record being referred back to observation
sional responsibilities; managers for rectification.
(b) The archives and database environment must be
secured and protected against physical hazards Typical monitoring reports would include the
to the records, such as fire and excess humidity; number and type of stations in the database, the
(c) For digital data, user-level security should be quantity of data in the database grouped by stations
enforced with respect to the database and and by observation element types, and information
its components. Only a small and registered about missing data. This information can be
group of people should have the right to compared to observation schedules to identify
perform data manipulations such as inser- when and where data are being lost. Other reports
tions, updates or deletions; could include quality control actions to ensure that
(d) Personnel with write access to a database must the quality control process is performing properly
agree not to perform any transactions besides with new data or to identify any groupings of data
the operations and practices approved by the with excessive quality problems. It is useful to track
data manager; the quantity and range of data being retrieved for
(e) All changes to data tables should have an user inquiries, since this information is helpful in
audit trail, and controls on access to this trail indicating both the most important datasets and
should be in place; the areas to be developed in the future.
(f) Password security principles should be
applied, including not sharing passwords, not The frequency and reporting period of monitoring
writing passwords on paper, changing pass- reports depend on the needs of the NMHS. Reports
words regularly, and using “strong” passwords on data ingestion may be made automatically,
consisting of seemingly unrelated letters, perhaps every day. Monthly reports on the quantity
numbers and characters; and quality of data usually match the monthly
(g) All unnecessary services should be disabled on cycle of many climate products.
the database computer;
(h) The database must be protected against attacks 3.3.9 International CDMS standards and
from viruses and hackers; guidelines
(i) Regular backups must be made, noting that
work done after the most recent backup will There is no agreement on the optimal structure of a
likely be lost and need to be repeated should climatological database, as the design depends on
a computer failure occur. Typically, an incre- the specific needs of the NMHSs and stakeholders.
mental backup should be made daily and a One need may be to access all data for a specified
full backup weekly; element over a region for a given time, but another
 CHAPTER 3. CLIMATE DATA MANAGEMENT 3–7

need may be to access a data time series for the (i) Data and system administration: The system
same element for a single location. The particular should be capable of being backed up
needs will have a strong impact on the required routinely and on demand without being shut
storage space or the response time to load or to down; of being restored as needed in a timely
access data. General principles that should be manner; of logging individual transactions; of
followed in any design, however, include: maintaining security; of being monitored for
(a) User documentation: Manuals should include system performance (for example, memory
an overview and guides for installation, users, usage, available storage space, number of
system administrators and programmers; transactions and status of system logs); and of
(b) Key entry: On-screen layouts of data input being copied at regular intervals to a separate
forms should be similar to the layout of the physical location;
paper forms from which the data are being (j) Level of assistance: Users should be able to
copied, customizing of layouts should be solve problems using available documenta-
possible, procedures for making entries into tion, interact with other users to exchange
the database should meet the needs of the questions and comments, and obtain advice
NMHS, validation of entries (for example, from developers of the system as needed and
permissible values or station identifiers) in a timely manner;
should be automatically performed in the key (k) Flexibility: The system should be capable of
entry process, and default values should be being expanded and modified as hardware
entered automatically; and software technologies evolve and data
(c) Ingest of digital data: The system should be sources change, and as the need for output
able to automatically ingest standard format- products changes.
ted data that are transmitted via the Global
Telecommunication System (GTS), files
containing data for multiple stations, multiple 3.4 QUALITY CONTROL
files containing data for a single station, data
from AWSs, data with user-defined formats, The objective of quality control is to verify whether
CLICOM data, and metadata; a reported data value is representative of what was
(d) Validation and quality control: The system intended to be measured and has not been contam-
should provide flags indicating the data inated by unrelated factors. It is important therefore
source, level of quality assurance performed to be clear from the outset what the readings of a
(such as key entry process or end-of-month particular data series are meant to represent. Data
processing), quality assurance results, and should be considered as satisfactory for permanent
the reason for the decision to accept, reject archiving only after they have been subjected to
or estimate a value. It should retain original, adequate quality control.
estimated and changed data values; it should
also evaluate data temporally and spatially for The observer or automated observing system should
permissible values, meteorological consist- apply quality control to ensure that the time and
ency and physical reasonableness; station identification are correct, that the recorded
(e) Technical documentation: There must be values reliably reflect current conditions, and that
listings defining each table in the database there is consistency among the observed elements.
and the relationships among tables; naming These steps must all be taken prior to the recording
conventions should be consistent among all or transmission of an observation.
tables, indexes, entities and views;
(f) Data access: The interface between the user The archiving centre should also apply quality
and the database should be easy to use, and control to the observations received. If manuscript
procedures for extracting information should records constitute the source document, trained
be documented with clear instructions and personnel should, upon receipt at the archiving
examples; centre, scrutinize them before any digitization
(g) Metadata: The system must be able to manage takes place. The forms should be reviewed to
the full range of metadata (as described in ensure proper identification (for example, station
section 2.6.9); name, identifying number and location),
(h) Output products: CDMSs should be able to legibility, and proper recording of data (for
produce standard output products that meet example, to the correct precision and in the
the needs of the NMHS, such as data listings proper columns). If any problems are discovered,
and tabulations of hourly, daily, monthly and the observation sites should be contacted to
longer-period data; statistical summaries; and clarify the issues or correct the problems. If
graphical depictions such as contour analyses, resources do not permit the quality control of all
windroses, time series, upper-air soundings data, priority should be given to the most
and station model plots; important climate elements.
3–8 GUIDE TO CLIMATOLOGICAL PRACTICES

3.4.1 Quality control procedures entered into the database. The original data,
however, must also be retained in the database.
When observed data are available in digital form, After the data have been quality-controlled,
the archiving centre should subject them to full, corrected and edited, the final dataset once again
elaborate quality control procedures on a regular, should be cycled through the quality control
systematic basis. Computer programs can examine checks. This last step will help ensure that errors
all the available data and list those that fail pre- have not been introduced during the quality
defined tests, but are not so adept at identifying the control procedures. Further manual review should
underlying problem. A skilled human analyst can help identify patterns of errors that may have
often make judgments about the cause of errors and resulted from, for example, software errors, and
determine any corrections that should be applied, inadequate or improper adherence to instruc-
but is generally overwhelmed by the vast quantity tions or procedures. The patterns should be
of observations. The best technique is a combina- relayed to managers of the NMHS observation
tion of the two, with computer-generated lists of program.
potential errors presented to the human analyst for
further action. In a database a given value is generally available at
different stages of quality control. The original data
Statistical techniques (described in Chapters 4 and 5) as received in the database must be kept, but valida-
are invaluable for detecting errors, and in some cases tion processes often lead to modifications of the
for suggesting what the “correct” value should be. data. These different stages of the value are reflected
Objective, automated screening of data is essential in quality flags. A multitude of flags could be
when validating large quantities of data. A manual constructed, but the number of flags should be kept
review of the automated output is needed, however, to to the minimum needed to describe the quality
ensure that the automated procedures are indeed assessment and reliability of the raw data or esti-
performing as expected. Graphical and map displays of mated values.
data and data summaries are excellent tools for visual
examinations. These techniques integrate and assimi- A quality flag code using two digits, one for the
late large quantities of data and enable a trained analyst type of data and one for the stage of validation,
to recognize patterns for assessing physical reasonable- meets most requirements. When data are
ness, identifying outliers, noticing suspect data and acquired from multiple sources, a third flag for
evaluating the performance of automated procedures. the source of the data is often useful. Examples
of “type of data”, “validation stage” and “acqui-
All observations should be appropriately flagged. sition method” codes are given in Tables 3.1, 3.2
Corrections or estimated correct data should be and 3.3.

Table 3.1. Example of data type codes

Type of data code Meaning

0 Original data
1 Corrected data
2 Reconstructed (such as by interpolation, estimation or disaggregation)
3 Calculated value

Table 3.2. Example of a stage of validation flag code.

.
Validation stage code Meaning
1 Missing data (data not received or observation not made)
2 Data eliminated once controls completed
3 Not controlled (newly inserted data or historical data not subject to any control)
4 Declared doubtful as identified as an outlier by preliminary checks, awaiting controls
(data possibly wrong)
5 Declared doubtful after automatic controls or human supervision (data probably wrong)
6 Declared validated after automatic controls or human supervision (but further
modification allowed, for example, if a subsequent study reveals that the data can still
be improved)
7 Declared validated after automatic controls and human supervision and no further
modification allowed
 CHAPTER 3. CLIMATE DATA MANAGEMENT 3–9

Table 3.3. Example of a data acquisition method flag code

Acquisition method code Meaning

1 Global Telecommunication System
2 Key entry
3 Automated weather station telecommunication network
4 Automated weather station digital file
5 Manuscript record

3.4.2 Quality control documentation 3.4.4 Format tests

Quality control procedures and algorithms should Checks should be made for repeated observations
be documented in detail for each stage of data or impossible format codes such as alpha characters
processing from observation to archiving. The in a numeric field, embedded or blank fields within
checks that are performed by the observer, initial an observation, impossible identification codes,
validation by the collection centre, final validation, and impossible dates. The actual causes of a format
quality control of changed formats for archiving or error could include miskeying, the garbling of a
publication, and checks on summarized data all message in transmission, or a mistake by an
require documentation. operator.

Detailed records and documentation should be Procedures should be introduced to eliminate, or at

accessible to users of the data. Knowledge of the least reduce, format errors. Two methods commonly
data-processing and quality control procedures employed that reduce key entry errors are double
allows a user of the data to assess the validity of entry (where the same data are entered independ-
the observation. With proper documentation and ently by two operators) and error detection
retention of the original data, future users are able algorithms. Which method is better depends on the
to assess the impact of changes in procedures on skills of the data-entry personnel, the complexity of
the validity, continuity or homogeneity of the the observation, and the resources available. Digital
data record; apply new knowledge in atmospheric error detection and correction techniques should
science to the older data; and perhaps revalidate be used to eliminate, or at least detect, transmission
the data based on new techniques and errors. Careful design of data-entry systems can
discoveries. minimize operator errors, but proper training and
performance checks are needed even with the most
user-friendly system.
3.4.3 Types of error
Metadata errors often manifest themselves as data
3.4.5 Completeness tests
errors. For example, an incorrect station identifier
may mean that data from one location apparently For some elements, missing data are much more crit-
came from another; an incorrect date stamp may ical than for others. For monthly extremes or event
mean the data appear to have been observed at a data such as the number of days with precipitation
different time. Data that are missing for the correct greater than a certain threshold, missing daily data
place and time should be detected by completeness may render the recorded value highly questionable.
tests; data that have been ascribed to an incorrect Total monthly rainfall amounts may also be strongly
place or time should be detected by consistency compromised by a few days of missing data, particu-
and tolerance tests. larly when a rain event occurred during the missing
period. On the other hand, monthly averaged
Data errors arise primarily as a result of instru- temperature may be less susceptible to missing data
mental, observer, data transmission, key entry than the two previous examples. For some applica-
and data validation process errors, as well as tions data completeness is a necessity.
changing data formats and data summarization
problems. When establishing a set of quality Data should be sorted by type of observation into a
control procedures, all potential types, sources prescribed chronological order by station. An
and causes of error should be considered and inventory should be compared to a master station
efforts should be made to reduce them. It is identifier file. Comparison should be made between
recommended that, in developing automated the observations actually received and the observa-
and semi-automated error-flagging procedures, tions that are expected to be received. The absence
system designers work closely with operational of any expected observation should be flagged for
quality control personnel. future review.
3–10 GUIDE TO CLIMATOLOGICAL PRACTICES

3.4.6 Consistency tests elements should be visualized at the same time in

order to facilitate diagnostics. For example, it will
The four primary types of consistency checks are be easier to validate a temperature drop if the infor-
internal, temporal, spatial and summarization. mation showing the veering of winds associated
Since data values are interrelated in time and space, with the passage of a cold front, or heavy rain from
an integrated procedure should be developed to a thunderstorm, is also available.
examine consistency. All consistency tests should
be completely documented with procedures, formu- Spatial consistency compares each observation
lae and decision criteria. with observations taken at the same time at other
stations in the area. Each observation can be
Internal consistency relies on the physical relation- compared to what would be expected at that site
ships among climatological elements. All elements based on the observations from neighbouring
should be thoroughly verified against any associ- stations. Those data for which there is a significant
ated elements within each observation. For difference between the expected and actual obser-
example, psychrometric data should be checked to vations should be flagged for review, correction or
ensure that the reported dry bulb temperature deletion as necessary. It is important to recognize
equals or exceeds the reported wet bulb tempera- that only like quantities should be directly
ture. Similarly, the relationship between visibility compared, such as wind speeds measured at the
and present weather should be checked for adher- same height; values measured at similar eleva-
ence to standard observation practices. tions, such as flat, open topography; or values
measured within a climatologically similar area.
Data should be checked for consistency with defi- Section 5.9 details the data estimation techniques
nitions. For example, a maximum value must be that are required for this type of quality control
equal to or higher than a minimum value. Physical process.
bounds provide rules for further internal consist-
ency checks. For example, sunshine duration is Summarization tests are among the easiest to
limited by the duration of the day, global radia- perform. By comparing different summaries of
tion cannot be greater than the irradiance at the data, errors in individual values or in each
top of the atmosphere, wind direction must be summary can be detected. For example, the sums
between 0° and 360°, and precipitation cannot be and means of daily values can be calculated for
negative. various periods such as weeks, months or years.
Checking that the total of the twelve monthly
Temporal consistency tests the variation of an reported sums equals the sum of the individual
element in time. Many climatological datasets daily values for a year provides a quick and simple
show significant serial correlation. A check should cross-check for an accumulation element like rain-
be made by comparing the prior and subsequent fall. Systematic errors in upper-air station data can
observations with the one in question. Using expe- sometimes be revealed by comparing monthly
rience or analytical or statistical methodologies, averages with the averages derived for the same
data reviewers can establish the amount of change location and height from a numerical analysis
that might be expected in a particular element in system. The cause of any inconsistencies should be
any time interval. This change usually depends on reviewed and corrected.
the element, season, location and time lag between
two successive observations. For example, a temper- Marine observations, in general, can be subjected
ature drop of 10°C within one hour may be suspect, to procedures similar to those used for surface
but could be quite realistic if associated with the land stations, with slight modification for addi-
passage of a cold front or onset of a sea breeze. The tional elements, assuming that there is a ship
suspicious value will have to be compared to present identifier present in each observation to allow
weather at that time, and perhaps to other types of data to be sorted chronologically and by ship
observations (such as wind direction or satellite, order. Upper-air observations must be verified in
radar or lightning detection) before a decision is a somewhat different manner. Some cross-checks
made to validate or modify it. For some elements, a should be made on surface-level conditions with
lack of change could indicate an error. For example, those at a nearby or co-located surface station. A
a series of identical wind speeds may indicate a quality control programme to check upper-air
problem with the anemometer. data should compute successive level data from
the preceding level starting with the surface data.
Temporal consistency checks can be automated Limits on the difference allowed between the
easily. Section 5.5 describes some of the techniques computed and reported values should be estab-
of time series analysis, which can be adapted for lished. Any level with reported elements that fail
quality control purposes. Graphical displays of data the test should be flagged as suspect, reviewed or
are also excellent tools for verification. Several corrected.
 CHAPTER 3. CLIMATE DATA MANAGEMENT 3–11

3.4.7 Tolerance tests resolutions address the concepts of “essential” and

“additional” data, with a specification of a minimum
Tolerance tests set upper or lower limits on the set of data that should be made available in a non-
possible values of a climatological element (such as discriminatory manner and at a charge of no more
wind direction, cloud cover, and past and present than the cost of reproduction and delivery, without
weather), or in other cases where the theoretical requiring payment for the data and products them-
range of values is infinite, the limits outside of selves. Members may decide to declare as “essential”
which it is unlikely for a measurement to lie. In the more than the minimum set. The use of agreed-upon
latter case, the limits are usually time- and location- international standard formats for data exchange is
dependent and should be established by recourse to critical.
the historical values or by spatial interpolation
methods. It is also important to identify and then Beyond CLIMAT and related messages (see section
quickly address systematic biases in the outputs 4.8.7), Members are also asked to provide additional
from instrumentation. Documentation must be data and products that are needed to sustain WMO
maintained regarding which tolerance tests have programmes at the global, regional and national
been applied, the climate limits established for each levels and to assist other Members in providing
inspected element, and the rationale for determin- meteorological and climatological services in their
ing these limits. countries. Members supplying such additional data
and products may place conditions on their
In general, tolerance tests compare a value in ques- re-export. Research and educational communities
tion against some standard using a statistical should be provided with free and unrestricted
threshold. Some simple tolerance tests include access to all data and products exchanged under
comparing an observed value to the extreme or the auspices of WMO for their non-commercial
record value or to some multiple of standard devia- activities.
tions around the average value for that date. In the
latter case, one must take into consideration the Members of WMO voluntarily nominate subsets of
possibility that the element may not necessarily their stations to be parts of various networks,
have a symmetrical or Gaussian distribution, and including the Global Climate Observing System
that some extreme values identified from the stand- (GCOS) Upper-Air Network (GUAN), the GCOS
ard deviation multiplier may be incorrect. Surface Network (GSN), the Regional Basic Synoptic
Network and the Regional Basic Climatological
When using long-term historical data for quality Network. Nomination of stations to participate in
control, it is preferable to use a standardized refer- these networks implies an obligation to share the
ence (for example, standard deviations or a data internationally.
non-parametric rank order statistic) rather than an
absolute reference. Section 4.4 discusses the various Data are also shared through International Council
summary descriptors of data, including the restric- for Science (ICSU) World Data Centres (WDCs). The
tions on their appropriateness. WDC system works to guarantee access to solar,
geophysical and related environmental data. It
It may be possible to perform some tolerance tests serves the whole scientific community by assem-
using completely different data streams, such as bling, scrutinizing, organizing and disseminating
satellite or radar data. For example, a very simple data and information. The centres collect, docu-
test for the occurrence or non-occurrence of precip- ment and archive measurements and the associated
itation using satellite data would be to check for the metadata from stations worldwide and make these
presence of clouds in a satellite image. data freely available to the scientific community. In
some cases, WDCs also provide additional prod-
ucts, including data analyses, maps of data
distributions, and data summaries. There are
3.5 EXCHANGE OF CLIMATIC DATA climate-related ICSU World Data Centres covering
meteorology, paleoclimatology, oceanography,
Exchange of data is essential for climatology. For atmospheric trace gases, glaciology, soils, marine
WMO Members, the obligation to share data and geology and geophysics, sunspots, solar activity,
metadata with other Members, and the conditions solar–terrestrial physics, airglow, aurora, and cosmic
under which these may be passed to third parties, are rays, as well as other disciplines.
covered under Resolution 40 of the Twelfth World
Meteorological Congress (with regard to meteoro- The World Meteorological Organization is actively
logical data), Resolution 25 of the Thirteenth World involved in the provision of data to a number of these
Meteorological Congress (for hydrological data), and WDCs, and there are a number of associated centres
Intergovernmental Oceanographic Commission operated directly through WMO. The WMO centres
Resolution XXII-6 (for oceanographic data). These deal with ozone and ultraviolet radiation, greenhouse
3–12 GUIDE TO CLIMATOLOGICAL PRACTICES

gases, aerosols, aerosol optical depth, radiation and International data exchange agreements allow for
precipitation chemistry. There are differences in data the global compilation of publications such as
access policy for ICSU and WMO centres. International Climatic Normals, World Weather Records, and
Council for Science data centres exchange data among Monthly Climatic Data for the World. Bilateral or
themselves without charge and provide data to scien- multilateral agreements are also important in creat-
tists in any country free of charge. Data centres ing and exchanging long-term datasets, such as the
operated through WMO must abide by Resolutions Global Historical Climate Network, Comprehensive
40 and 25 referred to above, which allow for some Aerological Reference, and Comprehensive Ocean–
data or products to be placed in the WDCs with Atmosphere Data Sets compiled by the United
conditions attached to their use. States, and the Hadley Centre global observations
datasets compiled by the United Kingdom. These
In addition to the International Council for Science datasets are generally provided to research centres.
WDCs, there are many other centres that operate
under cooperative agreements with WMO or with The current WMO information systems have been
individual NMHSs. These centres include the Global developed to meet a diverse set of needs for many
Precipitation Climatology Centre and Global Runoff different programmes and commissions. The multi-
Data Centre (Germany); Australia’s National Climate plicity of systems has resulted in incompatibilities,
Centre; the World Ozone and Ultraviolet Radiation inefficiencies, duplication of effort and higher over-
Data Centre (Canada); the Met Office Hadley Centre all costs for Members. An alternative approach
(United Kingdom of Great Britain and Northern planned to improve efficiency of the transfer of data
Ireland); and in the United States of America, the and information among countries is the WMO
Lamont-Doherty Earth Observatory of Columbia Information System (WIS). It is envisioned that this
University, the National Climatic Data Center, the system will be used for the collection and sharing of
National Oceanographic Data Center, the National information for all WMO and related international
Geophysical Data Center, the National Aeronautics programmes. Non-meteorological and non-climatic
and Space Administration (NASA) Goddard environmental and geophysical data such as ecologi-
Distributed Active Archive Center, the Tropical cal, earthquake and tsunami data could be included.
Pacific Ocean Observing Array, and the University The WIS vision provides guidance for the orderly
Corporation for Atmospheric Research. evolution of existing systems into an integrated
system that efficiently meets the international envi-
Exchange of digital data is simple for many ronmental information requirements of Members.
Members because of the range of computer commu-
nications systems available. The Global The WMO Information System will provide an inte-
Telecommunication System is a meteorological grated approach to routine collection and automated
communication system with connections to virtu- dissemination of observed data and products, timely
ally all countries of the world. As an operational delivery of data and products, and requests for data
system with a critical role in global weather fore- and products. It should be reliable, cost-effective and
casting, it provides reliable communication services, affordable for developing as well as developed
albeit sometimes with low bandwidth. Like the Members. It should also be technologically sustaina-
Internet, the Global Telecommunication System is ble and appropriate to local expertise, modular,
based on a confederation of interconnected scalable, flexible and extensible. It should be able to
networks. As a closed system, however, it is free adjust to changing requirements, allow dissemination
from the security breaches that often plague the of products from diverse data sources, and allow
Internet. Open communication linkages such as participants to collaborate at levels appropriate to
the Internet should be protected by the best avail- their responsibilities and budgetary resources. The
able security software systems to minimize the WMO Information System should also support differ-
danger of unwanted access and file manipulation or ent user groups and access policies such as those
corruption. outlined in Resolutions 40 and 25 referred to above,
data as well as network security, and integration of
It is highly unlikely that archived formats used for diverse datasets.
climatological data by one country would be the
same as those used by another. The format docu-
mentation describing the data organization,
element types, units and any other pertinent infor- 3.6 DATA RESCUE
mation should accompany the data. In addition, if
the digital data are compacted or in a special non- Data rescue is recognized now as a two-part process,
text format, it is extremely useful for the as defined in the Report of the CLICOM-DARE
contributing archive centre to provide “read” Workshop (WMO/TD-No. 1128). First there is the
routines to accompany digital data requested from ongoing process of preserving all data at risk of
an archive. being lost due to deterioration of the medium, and
 CHAPTER 3. CLIMATE DATA MANAGEMENT 3–13

then there is the digitization of data into computer- ———, 1993. Advisory Committee on Climate
compatible form for easy access. Each NMHS should Applications and Data (ACCAD) (Report of the
establish and maintain a data rescue programme. second session, Geneva, 16–17 November 1992)
(WMO/TD-No. 529, WCASP-No. 22, WCDMP-
In the mid-1990s technological advancements No. 22), Geneva.
made it possible to optically scan climate data as a ———, 1993: Final Report of the CCl Working Group
new method of creating digital climate archives. on Climate Data and its Rapporteurs to the Eleventh
This technology permits the data not only to be Session of the Commission for Climatology
preserved, but also to be in a form suitable for (Havana, 15–26 February 1993) (WMO/TD-
exchange via computer media. Optically scanning No. 523, WCDMP-No. 21), Geneva.
images preserves the data and is a major improve- ———, 1996: Report of the Fifth Session of the
ment over hard-copy media, but these data should Advisory Committee on Climate Applications
be moved into digital databases for use in analyses and Data (ACCAD) (Geneva, 26 September
and product development (see section 3.3.4). To 1995) (WMO/TD-No. 712, WCASP-No. 35,
ensure that digital files will always be accessible: WCDMP-No. 25), Geneva.
(a) Data should be stored as image files on media ———, 1996: Report on the Status of the Archival
such as cartridges, CDs and DVDs that can Climate History Survey (ARCHISS) Project
be regularly renewed to prevent loss from (WMO/TD-No. 776, WCDMP-No. 26), Geneva.
deterioration of the medium; ———, 1997: Expert Meeting to Review and Assess the
(b) Data should be digitized and entered into a Oracle-Based Prototype for Future Climate Database
CDMS; Management Systems (CDMS) (Toulouse, 12–16
(c) Data already in computer-compatible media May 1997) (WMO/TD-No. 902, WCDMP-No. 34),
should be migrated as soon as possible to Geneva.
storage facilities that conform to new and ———, 1997: Meeting of the CCl Working Group on
accepted technologies; Climate Data: Summary Report, (Geneva,
(d) Data formats should also be migrated to 30 January–3 February 1995) (WMO/TD-
formats that conform to software changes. No. 841, WCDMP-No. 33), Geneva.
———, 1997: Reports for CCl-XII from Rapporteurs that
Relate to Climate Data Management (WMO/TD-
3.7 REFERENCES No. 833, WCDMP-No. 31), Geneva.
———, 1997: Summary Notes and Recommendations
Assembled for CCl-XII from Recent Activities
3.7.1 WMO publications
Concerning Climate Data Management (WMO/TD-
World Meteorological Organization, 1988: WMO No. 832, WCDMP-No. 30), Geneva.
Region III/IV Training Seminar on Climate Data ———, 1999: Report of the Meeting of the WMO
Management and User Service (Barbados, 22–26 Commission for Climatology Task Group on a Future
September 1986; Panama, 29 September–3 October WMO Climate Database Management System
1986) (WMO/TD-No. 227, WCDP-No. 1), Geneva. (Ostrava, Czech Republic, 10–13 November 1998)
———, 1989: CLICOM Project (Climate Data and Follow-up Workshop to the WMO CCl Task
Management System) (WMO/TD-No. 299, WCDP- Group Meeting on a Future WMO CDMS (Toulouse,
No. 6), Geneva. 30 March–1 April 1999) (WMO/TD-No. 932,
———, 1989: Report of the Meeting of CLICOM Experts WCDMP-No. 38), Geneva.
(Paris, 11–15 September 1989) (WMO/TD- ———, 1999: Proceedings of the Second Seminar for
No. 342, WCDP-No. 9), Geneva. Homogenization of Surface Climatological Data
———, 1990: Report of the Expert Group on Global (Budapest, Hungary, 9–13 November 1998)
Baseline Datasets (Asheville, North Carolina, (WMO/TD-No. 962, WCDMP-No. 41), Geneva.
22–26 January 1990) (WMO/TD-No. 359, ———, 1999: Meeting of the CCl Working Group on
WCDP-No. 11), Geneva. Climate Data (Geneva, 30 November–4 December
———, 1990: Report of the Meeting on Archival Survey 1998) (WMO/TD-No. 970, WCDMP-No. 39),
for Climate History (Paris, 21–22 February 1990) Geneva.
(WMO/TD-No. 372, WCDP-No. 12), Geneva. ———, 1999: Report of the Training Seminar on
———, 1992: CCl Working Group on Climate Data Climate Data Management Focusing on CLICOM/
(Geneva, 11–15 November 1991) (WMO/TD- CLIPS Development and Evaluation (Niamey,
No. 488, WCDMP-No. 18), Geneva. Niger, 3 May–10 July 1999) (WMO/TD-No. 973,
———, 1992: Report of the First Session of the Advisory WCDMP-No. 43), Geneva.
Committee on Climate Applications and Data ———, 2000: CLICOM 3.1. Release 2. Geneva.
(ACCAD) (Geneva, 19–20 November 1991) ———, 2000: GCOS/GOOS/GTOS Joint Data and
(WMO/TD-No. 475, WCASP-No. 18, WCDMP- Information Management Plan, 2000 (WMO/TD-
No. 17), Geneva. No. 1004, GCOS-No. 60), Geneva.
3–14 GUIDE TO CLIMATOLOGICAL PRACTICES

———, 2000: Representativeness, Data Gaps and ———, 2007: Guidelines on Climate Data Management
Uncertainties in Climate Observations, Invited (WMO/TD-No. 1376, WCDMP-No. 56), Geneva.
Scientific Lecture given by Chris Folland to the
Thirteenth World Meteorological Congress
3.7.2 Additional reading
(Geneva, 21 May 1999) (WMO/TD-No. 977,
WCDMP-No. 44), Geneva. Cornford, D., 1998: An overview of interpolation.
———, 2000: Task Group on Future WMO Climate In: Seminar on Data Spatial Distribution in
Database Management Systems (Geneva, 3–5 May Meteorology and Climatology (M. Bindi and
2000) (WMO/TD-No. 1025, WCDMP-No. 46), B. Gozzini, eds). Volterra, European Union
Geneva. COST Action 79.
———, 2002: Guide to the GCOS Surface and Upper- De Gaetano, A.T., 1997: A quality control proce-
Air Networks. GSN and GUAN, Version 1.1 dure for hourly wind data. J. Atmos. Oceanic
(WMO/TD-No. 1106, GCOS-No. 73), Geneva. Technol., 14:137–151.
———, 2002: Report of the CLICOM-DARE Workshop Graybeal, D.Y., A.T. De Gaetano and K.L. Eggleston,
(San José, Costa Rica, 17–28 July 2002); Report of 2004: Complex quality assurance of historical
the International Data Rescue Meeting (Geneva, hourly surface airways meteorological data.
11–13 September 2001) (WMO/TD-No. 1128, J. Atmos. Oceanic Technol., 21:1156–1169.
WCDMP-No. 49), Geneva. Hungarian Meteorological Service, 1997: Proceedings
———, 2002: Report of the Climate Database of the First Seminar for Homogenization of Surface
Management Systems Evaluation Workshop Climatological Data (Budapest, 6–12 October
(Geneva, 27 May–1 June 2002) (WMO/TD- 1996), Budapest.
No. 1130, WCDMP-No. 50), Geneva. Kresse, W. and K. Fadaie, 2004: ISO Standards for
———, 2003: Guidelines on Climate Metadata and Geographic Information. Berlin, Springer.
Homogenization (WMO/TD-No. 1186, WCDMP- Merlier, C., 2001: Interpolation des Données Spatiales
No. 53), Geneva. en Climatologie, et Conception Optimale des
———, 2004: Guidelines on Climate Data Rescue Réseaux Climatologiques. Annexe du rapport de
(WMO/TD-No. 1210, WCDMP-No. 55), Geneva. Météo-France concernant ses activités en rapport
———, 2004: Fourth Seminar for Homogenization and avec la Commission for Climatology (CCl) de
Quality Control in Climatological Databases l’OMM. WMO report, CCl Lead Rapporteur on
(Budapest, 6–10 October 2003) (WMO/TD- Statistical Methods with Emphasis on Spatial
No. 1236, WCDMP-No. 56), Geneva. Interpolation.

.
 CHAPTER 4

����CHARACTERIZING CLIMATE FROM DATASETS

4.1 INTRODUCTION precipitation is falling, and others do not have

quantitative values but only a descriptive category,
Each year more and more data are added to the such as a cloud type or present weather description.
climatological archives of NMHSs. Climatologists The population is all possible values for an element.
must be able to express all the relevant information If an element is continuous, then the population is
contained in the data by means of comparatively infinite. If the element is not continuous, then the
few values derived through the application of a population is all the specific values that the element
range of statistical methods. When chosen and can have within boundaries defined by the analyst.
applied with care, statistical processes can isolate
and bring to the fore relevant information A sample is a set of observations from the popula-
contained in the data. tion, which is taken to represent the entire
population. Datasets are samples. The larger the
This chapter concentrates on descriptive statistics, sample size, the more accurate the estimate of the
the tool used to reduce to a comprehensible form descriptive features of the population will be. Much
the properties of an otherwise large amount of data. of climatology is concerned with the study of
Many of the methods described in this chapter are samples, but the analyst must recognize that a data-
best used with computers to process and display the set may be representative of only a part of the
data. It is necessary, though, to draw attention to population. The influence of, for example, inhomo-
the dangers of an overly mechanical application of geneities, dependence in time, and variations in
automated methods of analysis, because of the ease space complicate the interpretation of what the
with which automated procedures can be misused dataset represents.
and the results misinterpreted. While there are
unquestionable advantages to using a computer, Prior to the description or use of a dataset, the data
assumptions implicit in most analysis software run should be checked for accuracy and validity.
the risk of being ignored or not clearly articulated, Accuracy refers to the correctness of the data, while
hence leading potentially to erroneous results. validity refers to the applicability of the data to the
purpose for which the values will be used. The user
Chapter 5 concentrates on statistical methods and of a dataset should never assume without confirma-
should be used in conjunction with this chapter. tion that a dataset is accurate and valid, especially
Both chapters are intended to describe basic without relevant information from the quality
concepts rather than to provide detailed specifics of control processes applied during the assembling of
complex subjects. The references at the end of the the dataset. It is also important to know how the
chapter and textbooks on statistical theory and data have been collected, processed and compiled,
methods provide more detailed information. and sometimes even to know why the data were
initially collected. Chapter 3 covers climate data
management, with a discussion of the importance
of metadata and quality control issues in
4.2 DATASET EVALUATION sections 3.3 and 3.4, respectively.

A dataset consists of a collection of observations of

elements. An observation is a single estimate of
some quantity. Simple observations include reading 4.3 QUALITATIVE VISUAL DISPLAYS OF
a thermometer and reading the level of water in a DATA
raingauge. Other observations are more complex.
For example, obtaining barometric pressure from a Some of the basic features of a dataset that are often
mercury barometer involves taking observations of desired are the middle or typical value, the spread
both the length of the column of mercury and the or range of the observations, the existence of unex-
temperature of the barometer. The pressure is pected observations, how the observations trail off
considered a single observation, however. from either side of the middle value, and the clus-
tering of observations. Without systematic
Some elements are continuous, that is, there are no organization, large quantities of data cannot be
discontinuities in the state of the phenomenon easily interpreted to find these and other similar
observed, such as air temperature. Some elements features. The first step when organizing the data is
are not continuous, for example, whether or not to gain a general understanding of the data through
4–2 GUIDE TO CLIMATOLOGICAL PRACTICES

visual displays of the distribution of the observed

values.

There are many ways to portray data to obtain a

qualitative appreciation for what the data are tell-
ing the climatologist. One way to organize a
dataset is to sort the observations by increasing
or decreasing magnitude. The ordered observa-
tions can then be displayed graphically or as a
table, from which some characteristics, such as Figure 4.2. Cumulative frequency distribution
extreme values and the range, will become
apparent.
and so on. Similarly, a proportion based on tenths
A second way to organize a dataset is to group the is called a decile; one tenth of the observations are
data into intervals. Counts are made of the number below the first decile, two tenths are below the
of observations in each interval. A graphical second decile, and so on. One based on quarters is
display of the number of cases or percentage of the called a quartile. One based on fifths is called a
total number of observations in each interval quintile and has particular use in CLIMAT messages
immediately gives an indication of the shape of (see section 4.8.6).
the distribution of the population values, and is
called a frequency distribution or histogram Other visualizations include box plots (Figure 4.3),
(Figure 4.1). The number of intervals is arbitrary, stem and leaf diagrams (Figure 4.4), and data arrays
and the visual appearance of the distribution on a (Figure 4.5). If the time order of the data is
graph is affected by the number of intervals. Some important, then a graph of observed values against
information contained in the original dataset is time can be plotted to produce a time series (see
lost when the observations are grouped, and in 4.6). For data with two elements, such as wind
general, the fewer the number of intervals, the speed and direction, scatter diagrams can be
greater the loss. The number of intervals should be constructed by plotting the value of the first
a balance among accuracy, ease of communica- element against the value of the second (see 4.5.2).
tion, the use to which the information will be put, Windroses also provide excellent depictions of
and the statistical tests to which the data will be wind information. Double-mass curves, which are
subjected. frequently used by hydrologists and for data
homogeneity detection, are constructed by plotting
the cumulative value of one element against the
cumulative value of the second element (see 5.2).
Visualization techniques are limited only by the
imagination of the analyst, but all techniques
involve the sorting and classifying of the data. No
matter which technique is used, the resulting
graphic should be informative and should not
inadvertently draw users to unsupported
conclusions.

Figure 4.1. Frequency distribution (histogram) 4.4 QUANTITATIVE SUMMARY

DESCRIPTORS OF DATA

A third approach to organization is to form a Rather than presenting the entire dataset to illus-
cumulative frequency distribution, also called an trate a particular feature, it is often useful to extract
ogive. A graph is constructed by plotting the several quantitative summary measures. The
cumulative number or percentage of observations summary measures help describe patterns of varia-
against the ordered values of the element (Figure tion of observations. Understanding these patterns
4.2). The ogive representation of the data is useful furthers the knowledge of the physical processes
for determining what proportion of the data is that underlie the observations, and improves infer-
above or below a certain value. This proportion of ences that can be made about past, current and
values below a certain value, expressed as a future climate conditions.
percentage, is called a percentile; 1 per cent of the
observations are smaller than the first percentile, 2 Care must be taken to ensure that the contents of
per cent are smaller than the second percentile, a dataset that are summarized by quantitative
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–3

Figure 4.3. Box plot

stem leaf
0 5 7 8
1 2 3 3 7 9
2 1 3 4 4 5 5 7 8
3 2 2 4 5 6 7 8 9
4 4 5 5 6 6 6 6 7
5 2 3 3 3 4 8 8
6 1 1 1 3 4 5
7 3 2 2 5 7
8 4 6 7 7

Key: 2|5 means 25

Figure 4.4. Example of a stem and leaf data display. The leading digit of the value of an observation is
the stem, and the trailing digit is the leaf. In the table there are, for example, two observations of 25.

measures are really comparable. For example, a induced or if they reflect real effects of the climate
series of temperature observations may be compa- system.
rable if they are all are taken with the same
instrumentation, at the same time each day, at the
4.4.1 Data modelling of frequency
same location, and with the same procedures. If
distributions
the procedures change, then artificial variations
can be introduced in the dataset (see sections 3.3, Data visualizations (section 4.3) provide a quali-
3.4 and 5.2). Sometimes the summary descriptors tative view of the structure of a series of
of a dataset identify variations that are unex- observations. Shapes and patterns become appar-
pected; any unexpected patterns should be ent. Frequency distributions can be classified by
examined to determine if they are artificially their shape:
4–4 GUIDE TO CLIMATOLOGICAL PRACTICES

Year Month
1 2 3 4 5 6 7 8 9 10 11 12
1961 23 33 44 50 59 64 79 76 61 50 44 32
1962 26 31 40 54 60 67 78 73 60 49 40 30
1963 27 35 43 55 58 68 77 72 58 52 43 32
1964 24 37 47 58 57 64 79 74 59 54 46 34
1965 27 32 43 56 57 65 76 74 58 53 47 44
1966 30 38 44 53 58 67 80 75 58 55 46 32
1967 19 35 47 55 61 66 74 73 60 56 43 30
1968 22 33 46 56 60 69 78 70 56 52 45 30
1969 28 37 43 51 56 70 76 72 54 52 44 34
1970 25 34 46 56 58 63 73 71 54 50 43 31

Figure 4.5. Example of a data array

(a) Unimodal symmetrical curves: these curves statistical relationships among weather patterns in
are common for element averages, such as for different parts of the Earth).
annual and longer-term average temperature.
Generally, the longer the averaging period, One approach for summarizing the distribution of
the more symmetrical the distribution; a set of observations is to fit a probability distribu-
(b) Unimodal moderately asymmetrical curves: tion to the observations. These distributions are
many curves of averaged data are mostly (but functions with known mathematical properties
not quite) symmetrical; that are characterized by a small number of param-
(c) Unimodal strongly asymmetrical curves: these eters (typically no more than three). The functions
shapes depart far from symmetry and exhibit are always constructed so as to have non-negative
a high degree of skew; they are common for values everywhere, so that the relative magnitudes
precipitation amounts and wind speeds; of different values reflect differences in the relative
(d) U-shaped  curves: these curves are common likelihoods of observing those values. Several
for elements that have two-sided boundaries, common probability distributions, such as the
such as the fraction of cloud cover (there are normal (or Gaussian) distribution and the general-
greater tendencies for skies to be mostly clear ized extreme value distribution, describe situations
or mostly overcast). that often occur in nature. If an observed frequency
or conditional frequency distribution can be
Multimodal or complex curves: these curves are described by these known probability density func-
common for elements observed daily in areas with tions, then the properties and relationships can be
strong seasonal contrasts. In such a case the exploited to analyse the data and make probabilis-
frequency distribution curve built using the whole tic and statistical inferences. Some examples of
dataset may show a very characteristic bimodal probability density functions that may approxi-
shape. Datasets with very complex frequency distri- mate observed frequency distributions of
butions are likely to be better understood by continuous data (where any value in a continuum
stratifying the data a priori to reflect the different may be observed) are shown in Figures 4.6 to 4.12.
underlying processes.
Statistical frequency distributions are also defined
More than one series of observations can be simpli- for describing data that can attain only specific,
fied by examining the distribution of observations discrete values. An example is the number of days
of one variable when specified values of the other precipitation occurs in a month; only integer values
variables are observed. The results are known as are possible up to a maximum value of 31. Some of
conditional frequencies. The conditions are often the statistical frequency distributions that describe
based on prior knowledge of what can be expected discrete data are shown in Figures 4.13 and 4.14.
or on information about the likelihood of certain
events occurring. Conditional frequency analysis is The visual, qualitative display of an observed
especially useful in developing climate scenarios frequency distribution often suggests which of the
and in determining local impacts of events such as statistical frequency distributions may be appropri-
the El Niño–Southern Oscillation (ENSO) phenom- ate to describe the data. If a statistical frequency
enon and other teleconnection patterns (strong distribution with known probabilistic relationships
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–5

Figure 4.6. Normal or Gaussian distribution: Values over the range for the property being observed tend
to cluster uniformly around a single value, such as annual average temperatures.

Figure 4.7. Exponential distribution: Describes the times between events in a process in which events
occur continuously and independently at a constant average rate and has been used in daily .
precipitation amount analyses.

Figure 4.8. Weibull distribution: Describes the times between events in a process in which events occur
continuously and independently at a varying rate and has been used in wind speed analyses.
4–6 GUIDE TO CLIMATOLOGICAL PRACTICES

Figure 4.9. Generalized extreme value distribution: .

Used to model extreme values in a distribution.

Lognormal Distribution
4.5
(0, 0.1)
4 (0, 1)
(1, 1)

3.5

2.5
P (x)

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3

Figure 4.10. Lognormal distribution: Used when the logarithm of a distribution follows the normal
distribution, such as the distribution of air pollution particle concentrations.

can be used to describe the data, then inferences of the observations. If the fit is acceptable, then with
about the data can be made. There are specific just a few parameters, the summary function of the
approaches to fitting distributions to observations, observations should provide a realistic description of
such as the methods of moments, probability- the data that is compatible with the underlying phys-
weighted moments (L-moments) and maximum ics in a smoothed form that ignores data errors. A
likelihood. An understanding of the statistical secondary intent is to describe the data within a theo-
theory underlying a distribution function is neces- retical framework that is sufficiently simple so that
sary for making proper inferences about the data to statistical inferences can be made. Overfitting with a
which it is being applied. mathematical model may lead to an unrealistic
description of the data with too much weight placed
Any series of observations can be fitted by mathemati- on data errors or random factors that are extraneous
cal functions that reproduce the observations. One to the processes being studied. The degree of smooth-
needs to take care in applying such techniques for a ing is usually determined by how the data are to be
number of reasons. For example, the application of a used and by what questions the climatologist is trying
mathematical function generally assumes that the to answer.
observational dataset, usually a sample, fairly repre-
sents the population from which it is drawn, and that How well a summary function describes the
the data contain no errors (see section 3.4). The intent observations can be determined by examining the
of fitting a function is to approximate the distribution differences between the observations and the values
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–7

Gamma Distribution
1
(1, 1)
0.9 (3, 1)
(3, 3)
0.8

0.7

0.6
P (x)

0.5

0.4

0.3

0.2

0.1

0
0 2 4 6 8 10 12 14 16 18 20

Figure 4.11. Gamma distribution: Describes distributions that are bounded .

at one end and skewed, such as precipitation data.

Beta Distribution
8
(5, 15)
(15, 5)
7 (0.5, 0.5)
(2, 0.5)
6 (2, 2)

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Figure 4.12. Beta distribution: Describes distributions that are bounded .

at both ends, such as cloud amounts.

Figure 4.13. Binomial distribution: Describes two discrete outcomes, such as .

the occurrence or non-occurrence of an event.
4–8 GUIDE TO CLIMATOLOGICAL PRACTICES

Figure 4.14. Poisson distribution: Describes rare events, such as the frequency.
of occurrence of tropical storms.

produced by the function. Objective goodness-of- in climatology. It is calculated by simply dividing

fit tests should be applied. Datasets can usually be the sum of the values by the number of values. For
modelled by more than one function, and the test observations that tend to cluster around a central
measures can be compared to find the best or most value, the mean represents a number towards which
useful fit. The chi-square and Kolmogorov-Smirnov the average in a very long time series of observa-
tests are commonly used. The chi-square goodness- tions or other large dataset would converge as the
of-fit test assumes that data values are discrete and number of data values increases. The mean is not
independent (no observation is affected by any representative of the central tendency of strongly
other observation). If the sum of the squares of the asymmetrical distributions.
differences between observed and fitted frequencies
exceeds a threshold that depends on the sample A weighted mean is calculated by assigning different
size, then the fit should be regarded as inappropriate. levels of importance to the individual observations
This test is sensitive to the number of intervals. The so that, for example, more trustworthy or more
Kolmogorov-Smirnov test hypothesizes that if the representative observations can be given more influ-
maximum absolute difference between two ence in the calculation of a mean. Weights can be
continuous cumulative frequencies of independent determined by many methods. A common example
observations is larger than a critical value, then the is distance weighting, where weights are inversely
distributions are likely to be different. This test is related to a measure of distance. For example,
effective if the dataset has a large number of distance weighting is often used when estimating a
observations. mean value representative of a specific location from
observations taken in a region surrounding that
location. The weights are generally mathematical
4.4.2 Measures of central tendency
relations that may have no inherent relation to the
Observations often tend to cluster around a particu- physical processes that are measured, but whenever
lar value. Measures of central tendency are designed possible the choice of weighting methods should try
to indicate a central value around which data tend to take into account physical considerations.
to cluster. Measures of central tendency do not Generally speaking, weighting methods give good
substitute for all the detailed information contained results when both physical and statistical properties
in the complete set of observations. Calculation of vary continuously and quite slowly over the studied
a single measure is often inadequate to describe the space and time.
manner in which the data tend to concentrate
because it does not consider variation of the obser- The advantages of a mean are that it is a very
vations. Any measure of central tendency should be convenient standard of reference for fluctuations in
accompanied by a measure of the degree of varia- the observations (since the sum of the departures
tion in the values of the observations from which from a mean is zero), that it is easily calculated, that
the central tendency is derived. means for different non-overlapping subsets of the
whole observational record can be combined, and
The arithmetic mean or, as commonly termed, that the error of an estimate of a mean from a
average, is one of the most frequently used statistics sample is smaller than other measures of central
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–9

tendency (On the Statistical Analysis of Series of For elements having a circular nature, such as wind
Observations, WMO-No. 415). direction, the concept of mean can be ambiguous.
The modal value, such as prevailing wind direction,
Means, however, have limitations. Any single is often a more useful measure of central tendency
value may prove misleading when used to describe for elements that are measured by direction.
a series of observations. Very similar means may
be computed from datasets or distributions that A quantity that has a magnitude only is called a
are totally different in their internal structure. For scalar. A quantity that associates a direction with its
example, the mean of a bimodal distribution of magnitude is called a vector. For example, wind
cloud cover may be the same as the mean of a velocity is a vector as it has both speed and direc-
unimodal distribution, but the interpretation of tion. Mathematically, a vector can be transformed
the two means would be very different. The mean into independent components, and these compo-
is greatly affected by exceptional and unusual nents can then be averaged and combined into a
values; a few extreme observations may destroy resultant mean vector. For example, the wind can
the representative character of the mean. be expressed as a combination of two different
Observations that do not cluster towards a central scalars, eastward speed and northward speed, with
value are not well represented by a mean (for westward and southward speeds, respectively,
example, cloud cover, which often tends to cluster having negative values. The central tendency of the
at either 0 or 8 oktas). A mean, to be useful, must wind velocity is the resultant vector formed from
convey a valid meaning with respect to the actual the central tendencies of the eastward and north-
conditions described by the dataset and not be ward speeds. A mathematical resultant vector
merely the result of a mathematical calculation. calculated from data with opposing directions and
equal speeds will have a magnitude of zero; this
The median is the middle of a cumulative calculation may not be meaningful in the context
frequency distribution; half the data are above of describing a climate. An alternative approach
the median and the other half are below. It is that may be more meaningful for climate descrip-
calculated by ordering the data and selecting the tions is to calculate an average scalar direction,
middle value. If there are an odd number of ignoring the speed but accounting for circularity
values, the median is the middle value. For an (for example, wind directions of 355 and 5 degrees
even number of values the median is located are separated by 10 degrees and not 350 degrees),
between the two middle values, generally as the and an average scalar magnitude ignoring direc-
mean (or a weighted mean) of the two. If the two tion. An alternative is to combine the resultant
middle values are identical, then this value is vector direction with the scalar average
chosen as the median. magnitude.

Extreme variations have less of an influence on the In a perfectly symmetrical frequency distribution
median than on the mean because the median is a with one mode, such as the Gaussian distribution,
measure of position. Since the median is based on the values of the mean, median and mode will be
the number of observations, the magnitude of exactly the same. If the frequency distribution is
extreme observations does not influence the skewed towards high values, then the mean will
median. The median is especially useful when have the highest value, followed by the median and
observations tend to cluster around the centre but then the mode. This sequence is reversed if the
some of the observations are also very high or very frequency distribution is skewed towards low
low. As with the mean, data that do not cluster values. These relationships (Figure 4.15) and the
towards a central value are not well represented by features of the measures (Table 4.1) should be
the median. considered whenever a measure of central tendency
is selected to represent a dataset.
The mode is the value in the dataset that occurs
most often. Like the median, it is a positional
4.4.3 Measures of variability
measure. It is affected neither by the value (as is
the mean) nor by the position of other observa- Once a suitable estimate of the central tendency is
tions (as is the median). Modes from small samples chosen, it is possible to measure the variability of
or from samples that have more than one cluster individual observations around that value. The
of observations are unreliable estimates of central measurement of variation and its explanation is of
tendency. If multiple concentrations of the obser- fundamental importance. A record of only a few
vations are really typical (a multimodal observations generally gives a poor basis for judg-
distribution), then the dataset may possibly be ing the variability, however.
comprised of dissimilar factors, each of which has
a different central value around which the obser- Variability can be measured absolutely or relatively.
vations tend to cluster. The deviation of each individual observation from
4–10 GUIDE TO CLIMATOLOGICAL PRACTICES

Figure 4.15. Relationships among the mean, median and mode

the central tendency can be reduced to a value that be misleading. The range imparts no information
represents and describes the entire dataset. This about the nature of the frequency distribution
single number is the absolute variability. within the extreme limits. The range also ignores the
degree of concentration of the values almost entirely,
The simplest measure of absolute variability is the and fails to characterize in a useful manner the data-
range of the observations. The range is the difference set as a whole. Also, the range offers no basis for
between the highest and lowest values. Although judging the reliability of the central tendency.
easy to calculate, the range has many limitations. If
the extreme values are very rare or they fall well The interquartile range is another common meas-
beyond the bulk of observations, then the range will ure of absolute variability. It is the difference

Table 4.1 Comparison of some features of measures of central tendencies

Feature Mean Median Mode

Affected by outliers Yes No No

Representative of central tendency when Yes Yes Yes

frequency distributions are narrowly spread

Representative of central tendency when No Maybe No

frequency distributions are widely spread

Representative of central tendency when No Maybe No

observations are clustered into more than one
group

Representative of central tendency when No Yes Yes

frequency distributions with one cluster are
skewed

Ease of calculation Easiest Easy from Easy from histogram

ordered data

Departures sum to zero Yes Not always Not always

Possibility for more than 1 No No Yes

Indicator of variability No No Only if more than one

mode
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–11

between the third and first quartiles. The first 4.4.4 Measure of symmetry
quartile is the value of the ordered observations
such that 25 per cent are below this value and 75 Skewness is a measure of the departure from symme-
per cent are above, and the third quartile is the try. It is a relative and dimensionless measure,
value of the ordered data such that 75 per cent are therefore allowing for comparisons among datasets.
below this value and 25 per cent are above. The One simple measure of skewness is the difference
interquartile range is thus the range of the central between the mean and mode, divided by the stand-
50 per cent of the ordered observations. When ard deviation. Skewness is positive when the mean
coupled with the median, it describes some of the is greater than the mode and negative when the
characteristics of the frequency distribution. Other mode is greater than the mean. Other measures
central ranges can be calculated in a similar way. have also been defined, such as one based on the
The interdecile range, for example, is the differ- interquartile range and median.
ence between the 90th and 10th percentiles, and is
the range of the middle 80 per cent of Positive skewness is characteristic of some precipi-
observations. tation datasets that have a lower limit of zero but
an unbounded upper limit. Daily maximum
The average deviation is the mean of the absolute temperature datasets also often tend towards posi-
value of all the deviations of individual observa- tive skewness, but daily minimum temperatures
tions from the chosen measure of central tendency. often tend towards negative skewness.
While deviations may be calculated from the
mean, median or mode, they should theoretically
4.4.5 Measure of peakedness
be calculated from the median because the sum of
the absolute deviations from the median is less Symmetrical frequency distributions may have
than or equal to the sum from either the mean or different degrees of flatness in the central part of
mode. the distribution. Kurtosis is a dimensionless ratio
that provides a relative measure for comparative
The standard deviation is the square root of the purposes of the flatness or peakedness. Positive
mean of the square of all the individual deviations kurtosis indicates a narrow maximum in the centre
from the mean. Deviations are taken from the mean of the frequency distribution, with frequencies fall-
instead of the median or mode because the sum of ing sharply to low values away from the mean.
squares from the mean is a minimum. Squaring Negative values indicate a large, flat central region,
deviations gives greater weight to extreme varia- and are typical of many meteorological distribu-
tions. The standard deviation is used in the tions, such as upper-air humidity.
derivation of many statistical measures. It is also
used extensively as a normative quantity to stand-
4.4.6 Indices
ardize different distributions for comparative
purposes. The purpose of an index is to reduce complex
conditions to a single number that retains some
For comparative purposes, absolute measures of physical meaning and can be used to monitor a
variability may have serious limitations. particular process. It expresses the relationship
Comparisons should be made only if averages from between observed and baseline conditions as a
which the deviations have been measured are single value. The baseline is usually, but not always,
approximately equal in value, and when the units the average climatic state. An example is the Palmer
of measurement are the same. For example, compar- Drought Severity Index, which is a summary
ing standard deviations calculated from a comparison of a complex water balance system of
temperature dataset and from a heating degree-day precipitation, evaporation, runoff, recharge and
dataset is meaningless. soil properties to climatically average conditions.
Development of an index has four components: the
Often, comparisons are required when the means selection of the elements that are to be included in
are not approximately equal or when the units of the index; the selection and calculation of the base-
measurement are not the same. Some measure is line; the method of construction of the index; and
therefore required that takes into account the mean the weights or importance of each of the included
from which deviations are measured, and that elements.
reduces different units of measurement to a
common basis for the purposes of comparison. The Examination and selection of the data to be
relation of the absolute variability to the magnitude included in the index often constitute a more
of the central tendency is the relative variability. complicated task than the actual calculation of the
One such measure is the coefficient of variation, index. One of the concerns when choosing a base-
which is the ratio of the standard deviation to the line is that the characteristics of the observations
mean of a dataset. used to define the baseline may change over time; it
4–12 GUIDE TO CLIMATOLOGICAL PRACTICES

is essential that the observations used are homoge- 4.5.1 Contingency tables
neous (see section 5.2). Another concern is that the
baseline should represent normal, standard or Contingency tables are a simple yet effective way of
expected conditions, since most users of an index discovering important relationships among factors,
assume that the baseline represents such condi- especially for large datasets. Contingency tables are
tions. When selecting a baseline, care should be most often formed from qualitative descriptors
exercised to explicitly define what is to be compared (such as mild, moderate or severe), or from dichoto-
and for what purpose. mous variables (an event did or did not occur).
They can also be formed from the joint frequency
Selection of weights is a critical consideration. Care of two elements, such as wind speed and direction
should be taken to weight the importance of each or the diurnal distribution of visibility. Table 4.2 is
element contributing to the index relative to the an example of a contingency table.
purpose of calculating the index. If the index is to
be calculated in the future, care must also be taken Independence between the elements of a contin-
to periodically review the contribution of each gency table is often assessed by using a chi-square
element for changes in, for example, importance, test. When this test is used, the serial dependence
data accuracy, measurement and processing. often found in climatological time series, according
to which an observation is more likely to be similar
to its preceding observation than dissimilar (see
4.6), violates the assumptions of the test, so that
4.5 CORRELATION conclusions drawn from the test may be suspect.

One often needs to detect or describe the relation-

4.5.2 Measures of correlation
ship between or among elements. A relationship
may be evident from visual displays of data, but A scatter diagram is another simple yet useful tool
quantitative measures are often calculated. for visualizing relationships. It can show a relation-
Correlation is a measure that quantifies a relation- ship between two elements or the trend of one
ship. No matter which measure is calculated, it is element over time, or whether any useful relation-
important to note that correlation does not imply ship exists at all. Figures 4.16 and 4.17 are examples
a cause and effect relationship, but only that of scatter diagrams. The association among elements
elements behave in a similar way. Often, factors and temporal patterns can sometimes be summa-
other than those being investigated could be rized by a correlation measure. The correlation
responsible for the observed association; many coefficient is the most commonly used measure of
apparent relationships in meteorology and clima- association. Another measure, the Spearman rank
tology are generally too complex to be explained correlation, is also sometimes used.
by a single cause. Just as a positive or negative
correlation does not imply causation, a zero corre- The correlation coefficient is a number between –1
lation does not necessarily imply the absence of a and +1. It measures the linear relationship between
causative relationship. two elements. A coefficient of zero implies no

September 2009

Figure 4.16. Scatter diagram with weak correlation

 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–13

September 2009

Figure 4.17. Scatter diagram with strong positive correlation

similarity of behaviour between elements. Figure coefficient will be high and positive. The Spearman
4.16 is an example of a pattern that is expected measure is less sensitive to extremes than the corre-
when two elements are very weakly correlated. A lation coefficient; it measures linear association, and
coefficient of +1 indicates that as the value of one sometimes indicates non-linear association.
element increases, the value of the other element
also increases in direct proportion. Figure 4.17 is an
example of a pattern that is expected when two
elements have a strong positive correlation. A 4.6 TIME SERIES
coefficient of –1 indicates that as the value of the
first element increases, the value of the other An ordering of observations by their time of occur-
element decreases in inverse proportion. One of the rence is called a time series (Figure 4.18). A graph of
problems with using a simple correlation coefficient data values plotted against time is an important
is that the implied relationship is linear. Often, qualitative tool for identifying time-related varia-
meteorological elements are related in a non-linear tions. In climatology, a trend is an interesting
manner, and the dataset may need to be transformed characteristic because it summarizes the historical
(see section 5.4) prior to the calculation of a behaviour of observations of an element. Linear
correlation coefficient. trends are examined most often for a given time
series, but a trend may sometimes be better described
The Spearman rank correlation measures agreement in non-linear terms, such as a curve, or even as an
between the ordered ranks of two datasets. The abrupt upward or downward shift. Trends, whether
measure is again a number between –1 and +1. If the linear or non-linear, are generally sustained in
observations in the two sets do not maintain the climate series for a finite period, which may be quite
same relative order, then the measure will be low or long. Over time the climate system has often shown
negative; if they have a similar relative order, the trends in one direction, which are eventually

Table 4.2. Contingency table of highway accidents and visibility observations

Visibility below Visibility above Total

200 metres 200 meters

Accident occurred 16 4 20

No accident occurred 13 332 345

Total 29 336 365

4–14 GUIDE TO CLIMATOLOGICAL PRACTICES

Figure 4.18. Time series of monthly average temperature

followed by a reversal. What might appear as a of the mathematical variables defining the equation
sustained trend in the most recent period of a climate for a curve fit, such as the coefficients of a
record could be part of a slow oscillation related to polynomial function. Similarly, periodic features
multidecadal variations that cannot be clearly seen are also represented by the coefficients of the
because the time interval of the apparent sustained mathematical variables defining the oscillations,
trend is only a part of the whole oscillation, or such as the frequency, phase and amplitude of
because the nature of the series projected into the trigonometric functions.
future is unknown. Anthropogenic climate change
poses a particularly difficult challenge in this regard
since human decisions will likely play a part in deter-
mining, for example, how long the global warming 4.7 INTERPRETATION OF SUMMARY
trend observed over the past century will be CHARACTERISTICS OF CLIMATE
sustained. Complete study of a time series requires
the identification not only of the trends, but also the Although it is possible to calculate numerous
periodic or quasi-periodic oscillations, and irregular summary measures, it may not be appropriate to
or apparently random variations exhibited in the use them to describe the dataset. All measures that
data. The goal of time series analysis is to understand reduce observations with the purpose of detecting
how the variability in a time series is distributed as a and describing a climate signal or relationship are
function of the timescale. based on assumptions, and if these assumptions are
not valid, then the summary measures may be
Commonly in meteorology and climatology, succes- misleading. There are four issues that must be
sive observations tend to be more similar to each considered in detail before using summary meas-
other than dissimilar. The measure used to summa- ures: dataset errors, inhomogeneity, independence
rize the relationship between each observation and of observations and neglect of important factors.
the prior one is the autocorrelation coefficient. This
measure is calculated in the same way as the correla- Often, data are erroneous because of recording
tion coefficient (see section 4.5.2), with the exception errors (such as the transposition of numbers),
that the second series is the same as the first, but garbled communications, misunderstanding of
shifted by one or more time steps. coding practices by an observer, processing errors
(for example improper conversion from degrees
Measures that summarize trends depend on the Fahrenheit to degrees Celsius), computer program
kind of trend being isolated. Linear trends are logic and coding errors, and incorrect identifier
represented by the slope of a straight line. (location or time) information of a value (see
Non-linear trends are represented by the coefficients section 3.4). These types of errors do not relate to
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–15

the physical conditions being observed, and they inhomogeneous, then the assumption that all data
can contaminate data so that improper conclusions are comparable is violated. This violation, and its
are drawn from a data analysis. effect on the calculation of the mean, should be
disclosed.
Faulty inferences are often made when quantitative
measures are used to compare data that are not
really comparable, such as when comparing inho-
mogeneous observations. If possible, any dataset 4.8 NORMALS
being analysed should be made homogeneous (see
section 5.2). Climate normals are used for two principal
purposes. They serve as a benchmark against which
Many meteorological datasets violate the assump- recent or current observations can be compared,
tion of independence. Prior to summarizing a including providing a basis for many anomaly-
dataset, care should be taken to remove, if possible, based climate datasets (for example, global mean
the dependence among observations. For example, temperatures). They are also widely used, implicitly
the effect of known annual cycles can be largely or explicitly, as a prediction of the conditions most
removed by summarizing departures from the likely to be experienced in a given location.
known cycle. For another example, if persistence
(autocorrelation) is known to affect a series of obser- Historical practices regarding climate normals (as
vations, as sometimes occurs with daily temperatures described in previous editions of this Guide (WMO-
observed during a synoptic surface high-pressure No. 100), the Technical Regulations (WMO-No. 49)
event, it should be taken into account by the analyti- and the Handbook on CLIMAT and CLIMAT TEMP
cal model. If dependencies are not accounted for by Reporting (WMO/TD-No. 1188)) date from the first
the models, subsampling by selecting only one half of the twentieth century. The general recom-
observation of the several available during the mendation was to use 30-year periods of reference.
persistent event would remove the persistence affect- The 30-year period of reference was set as a stand-
ing all the observations taken during the event. Care ard mainly because only 30 years of data were
must also be exercised in this process, however, so as available for summarization when the recommen-
not to introduce aliasing (an effect that causes differ- dation was first made. The early intent of normals
ent signals to be indistinguishable when sampled), was to allow comparison among observations from
which can lead to an incorrect analysis of any under- around the world. The use of normals as predictors
lying oscillations. slowly gained momentum over the course of the
twentieth century.
An incomplete or erroneous explanation can result
from presenting quantitative evidence concerning Traditionally, climatological normals have focused
only one factor while ignoring other important on the mean value of a climate element over a
factors. An example is comparing temperatures period of time. As discussed in section 4.4.2, the
over a cold season at a coastal location and a conti- mean is an incomplete description of the climate,
nental location. The averages may be similar and many applications require information about
enough to suggest that the climates are the same, other aspects of that element’s frequency distribu-
but such a conclusion would not be drawn if the tion and statistical behaviour, such as the frequency
greater variability at the continental location were of extended periods when a value is above a thresh-
not ignored. old. Extreme values of an element over a specified
period, and other statistical descriptors of the
Specific statistical assumptions concerning, for frequency distribution of an element (such as the
example, the consistency and homogeneity of the standard deviation of daily or monthly values), are
data or the nature of the dependence between useful descriptors of the climate at a location and
observations, are implicit in all statistical analysis should be included with datasets of normals.
techniques. These assumptions should be clearly
identified and assessed by the analyst, and the Many NMHSs calculate daily normals along with
interpretation of a summary measure should be monthly and annual normals. Although not
tempered by the extent to which the assumptions required by WMO, daily normals illustrate the non-
are satisfied. If any of the assumptions are violated, random pattern of daily variations of an element
then the interpretation of a summary measure that cannot be captured with monthly normals.
should be changed to account for the violations. They are calculated by averaging the values of an
The usual interpretation of the measure may still element for a specified calendar date over a period
suffice, but the real or suspected violations should of time. The observed values are usually smoothed
be disclosed with the measure. For example, if by 3- to 7-day moving averages or binomial smooth-
annual average temperatures are calculated from a ing to reduce the effects of random high-frequency
dataset that is known from the metadata to be temporal variability of weather systems. Another
4–16 GUIDE TO CLIMATOLOGICAL PRACTICES

smoothing approach is to fit the series of daily aver- Secular trends reduce the representativeness of
ages calculated from the observations with spline, historical data as a descriptor of the current, or
trigonometric or polynomial smoothing functions, likely future, climate at a given location.
and these smoothed series become the daily Furthermore, the existence of multidecadal varia-
normals (see section 5.8). bility in the climate system causes differences in
climate normals from one reference period to the
next such that the representativeness of any given
4.8.1 Period of calculation
normal for the present climate is reduced. For
Under the WMO Technical Regulations, climatological predictive uses, NMHSs are encouraged to prepare
standard normals are averages of climatological data averages and period averages. The optimal length
computed for the following consecutive of record for predictive use of normals varies with
periods of 30 years: 1 January 1901 to 31 December element, geography and secular trend. In general,
1930, 1 January 1931 to 31 December 1960, and so the most recent 5- to 10-year period of record has as
forth. Countries should calculate climatological much predictive value as a 30-year record. Shorter
standard normals as soon as possible after the end of reference periods allow normals to be calculated for
a standard normal period. It is also recommended a much wider range of stations than is usually
that, where possible, the calculation of anomalies be possible for a standard normals reference period.
based on climatological standard normal periods, in For elements that show a substantial underlying
order to allow a uniform basis for comparison. trend (such as mean temperature), predictive accu-
Averages (also known as provisional normals) may be racy is improved by updating the averages and
calculated at any time for stations not having 30 years period averages frequently.
of available data (see 4.8.4). Period averages are aver-
ages computed for any period of at least ten years In any publication of normals and averages, and
starting on 1 January of a year ending with the digit 1 also in any publication that uses them for analysis
(for example, 1 January 1991 to 31 December 2004). and display of climate variability, it is important to
Although not required by WMO, some countries document the period used for the calculation and
calculate period averages every decade. the calculation methodologies. The CLIMAT and
CLIMAT SHIP codes contain a provision for a start
Where climatological standard normals are used as year and an end year to be incorporated in normals
a reference, there are no clear advantages to updat- distributed through such messages.
ing the normals frequently unless an update
provides normals for a significantly greater number
4.8.2 Stations for which normals and
of stations. Frequent updating carries the disadvan-
averages are calculated
tage that it requires recalculation not only of the
normals themselves, but also numerous datasets Climate normals and averages should be calculated
that use the normals as a reference. For example, for as wide a range of stations as possible, subject to
global temperature datasets have been calculated as the requirement that a station meet standards for
anomalies from a reference period such as 1961– the amount and completeness of available data. As
1990 (Figure 4.19). Using a more recent averaging a minimum, they should be calculated, if possible,
period, such as 1971–2000, results in a slight for all stations whose data are distributed on the
improvement in ”predictive accuracy” for elements Global Telecommunication System (section
that show a secular trend (that is, where the time B.1.3.1.2 of the Technical Regulations).
series shows a consistent rise or fall in its values
when measured over a long term). Also, 1971–2000
4.8.3 Homogeneity of data
normals would be viewed by many users as more
“current” than 1961–1990. The disadvantages of As far as possible, the data used in the calculation of
frequent updating, however, could be considered to climate normals and averages should be homoge-
offset this advantage when the normals are being neous. The issue of homogeneity is addressed more
used for reference purposes. fully in section 5.2. In the context of climate
normals and averages, homogeneity issues that
A number of studies have found that 30 years is require particular attention are changes of site loca-
not generally the optimal averaging period for tion; changes of observation procedure, including
normals used for prediction. The optimal period changes of observation time; changes of instrument
for temperatures is often substantially shorter than type; changes of instrument exposure over time;
30 years, but the optimal period for precipitation and changes in the processing of data.
is often substantially greater than 30 years. The
Role of Climatological Normals in a Changing Climate In practice, at many locations it will not be possible
(WMO/TD-No. 1377) and other references at the to construct a suitably homogenous dataset. It may
end of this chapter provide much detail on the instead be necessary to produce normals from a
predictive use of normals of several elements. composite of two or more parts of an inhomogene-
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–17

Figure 4.19. Global average temperature anomalies .

(courtesy Met Office Hadley Centre, United Kingdom)

ous record. An option is to make adjustments to the It is recommended that a monthly value should not
earlier part of a record to make it as homogeneous be calculated if more than ten daily values are miss-
as possible with the most recent data. ing or five or more consecutive daily values are
missing. In the case of elements for which the
monthly value is a sum of daily values rather than a
4.8.4 Missing data
mean (such as for rainfall or sunshine), a monthly
Normals calculated from incomplete datasets can value should be calculated only if either all daily
be biased. For example, if one year in a period was observations are available, or if any missing days are
particularly cold, a normal calculated without data incorporated in an observation accumulated over
from that year would be higher than a normal that the period of missing data on the day when observa-
did include that year. As there is often considerable tions resume. The Calculation of Monthly and Annual
autocorrelation in climatological data, consecutive 30-Year Standard Normals (WMO/TD-No. 341) recom-
missing observations can have more impact on mends stricter criteria for calculating averages, with
normals than the same number of missing observa- the limits being more than five missing days in total,
tions scattered randomly through the period in or more than three consecutive missing days.
question.
4.8.5 Average daily temperature
As a guide, normals or period averages should be
calculated only when values are available for at least There are many methods for calculating an average
80 per cent of the years of record, with no more than daily temperature. These include methods that use
three consecutive missing years. An alternative a daily maximum and daily minimum, 24 hourly
option, when there is an extended period of missing observations, synoptic observations and
data but reasonably complete data after that time, is observations at certain specified hours during a day.
to calculate a period average using only data from The best statistical approximation of an average is
the years following the break in the record. based on the integration of continuous observations
over a period of time; the higher the frequency of
Annual normals or averages should be calculated as observations, the more accurate the average.
the mean or sum (as appropriate) of the 12 monthly Practical considerations generally preclude the
normals or averages, without consideration of the calculation of a daily average from a large number
varying lengths of the months (section B.1.4.2.4 of of observations evenly distributed over a 24-hour
the Technical Regulations). No missing monthly period because many observing sites do not measure
normals are permitted in the calculation of annual an element continuously. For comparative purposes,
normals. a standard processing methodology is desirable for
4–18 GUIDE TO CLIMATOLOGICAL PRACTICES

all stations worldwide, with the number of stations surface stations) and CLIMAT SHIP (for ship-based
maximized. surface observations) coded messages sent on the
Global Telecommunication System. Coding and
All ordinary climatological stations observe a reporting procedures are described in the Handbook on
daily maximum and minimum temperature (see CLIMAT and CLIMAT TEMP Reporting (WMO/TD-
section 2.2.1). Hence, the recommended method- No. 1188).
ology for calculating average daily temperature is
to take the mean of the daily maximum and
minimum temperatures. Even though this
method is not the best statistical approximation, 4.9 REFERENCES AND ADDITIONAL
its consistent use satisfies the comparative READING
purpose of normals. An NMHS should also calcu-
late daily averages using other methods if these
4.9.1 WMO publications
calculations improve the understanding of the
climate of the country. World Meteorological Organization, 1983: Guide to
Climatological Practices. Second edition (WMO-
No. 100), Geneva.
4.8.6 Precipitation quintiles
———, 1988: Technical Regulations, Vol. I – General
Quintiles of precipitation are used to relate an Meteorological Standards and Recommended
observed monthly precipitation total to the Practices; Vol. II – Meteorological Service for
frequency distribution of values observed over the International Air Navigation; Vol. III – Hydrology
period for which normals have been calculated. No (WMO-No. 49), Geneva.
universally accepted method exists for the calcula- ———, 1989: Calculation of Monthly and Annual
tion of quintile boundaries, and the choice of 30-Year Standard Nor mals (WMO/ T D -
method can make a substantial difference to the No. 341, WCDP-No. 10), Geneva.
calculated values. The recommended procedure for ———, 1990: On the Statistical Analysis of Series of
calculating the boundaries, however, is as follows: Observations (WMO/TN-No. 143, WMO-
No. 415), Geneva.
For any month the 30 monthly values of precipita- ———, 1995: Manual on Codes (WMO-No. 306),
tion recorded during the 30-year normal period are Geneva.
listed in ascending order. The list is then divided ———, 1996: Climatological Normals (CLINO) for the
into five groups of quintiles of six values each. The Period 1961–90 (WMO-No. 847), Geneva.
first quintile contains the six lowest values for the ———, 2004: Handbook on CLIMAT and CLIMAT
month in question that have been observed during TEMP Reporting (WMO/TD-No. 1188), Geneva.
the 30-year period, the second quintile the next six ———, 2007: The Role of Climatological Normals in a
lowest values, and so on to the fifth quintile, which Changing Climate (WMO/TD-No. 1377,
contains the six highest values. WCDMP-No. 61), Geneva.

The boundary between two adjacent quintiles is set

4.9.2 Additional reading
halfway between the top value of the one quintile
and the first value of the next. The quintile index is Angel, J.R., W.R. Easterling and S.W. Kirtsch, 1993:
the number of the lowest quintile containing the Towards defining appropriate averaging periods
monthly precipitation in the month for which the for climate normals. Clim. Bull., 27:29–44.
report is being prepared, with the following special Asnani, G.C., 1993: Tropical Meteorology (2 vols.),
rules: Pune, Sind Society.
(a) If the precipitation is 0: use index 0 if this has Barnston, A.G. and R.E. Livezey, 1987: Classification,
not occurred during the reference period; use 1 seasonality and persistence of low-frequency
if it has occurred but fewer than 6 times; use 2 atmospheric circulation patterns. Monthly
if it has occurred between 7 and 12 times; use 3 Weather Rev., 115:1083–1126.
if it has occurred 13 to 18 times, and so on; Bhalme, H.N. and D.A. Mooley, 1980: Large-scale
(b) If the precipitation is less than any value in droughts/floods and monsoon circulation.
the reference period: use index 0 (regardless of Monthly Weather Rev., 108:1197–1211.
whether the precipitation is 0); Bhalme, H.N., D.A. Mooley and S.K. Jadhav, 1984:
(c) If the precipitation is greater than any value Large-scale April pressure index connected with
in the reference period: use index 6. the southern oscillation and its potential for
prediction of large-scale droughts over India.
4.8.7 Dissemination of normals Mausam, 35(3):355–360.
Brohan, P., J.J. Kennedy, I. Harris, S.F.B. Tett and
A main avenue for international dissemination of P.D. Jones, 2006: Uncertainty estimates in
climate normals is through CLIMAT (for land-based regional and global observed temperature
 CHAPTER 4. CHARACTERIZING CLIMATE FROM DATASETS 4–19

changes: a new dataset from 1850. J. Geophysical Rao, G.A., S.V. Datar and H.N. Srivastava, 1992:
Research 111, D12106. Study of drought indices in relation to rice crop
Dixon, K.W. and M.D. Shulman, 1984: A statistical production over some States of India. Mausam,
evaluation of the predictive abilities of climatic 43(2):169–174.
averages. J. Clim. Appl. Meteorol., 23:1542–1552. Srivastava, A.K., P. Guhathakurta and S.R. Kshirsagar,
Guttman, N.B., 1989: Statistical descriptors of 2003: Estimation of annual and seasonal
climate. Bull. Amer. Meteor. Soc., 70:602–607. temperatures over Indian stations using opti-
Huang, J., H.M. van den Dool and A.G. Barnston, mal normals. Mausam, 54:615–622.
1996: Long-lead seasonal temperature Tukey, J.W., 1977: Exploratory Data Analysis.
prediction using optimal climate normals. Reading, Massachusetts, Addison Wesley.
J. Climate, 9:809–817. Von Storch, H. and F.W. Zwiers, 1999: Statistical
Lamb, P.J. and S.A. Changnon, 1981: On the “best” Analysis in Climate Research. Cambridge,
temperature and precipitation normals: the Illinois Cambridge University Press.
situation. J. Appl. Meteorol., 20:1383–1390. Wilks, D.S., 1995: Statistical Methods in the
Matheron, G., 1973: The intrinsic random functions Atmospheric Sciences. San Diego, Academic
and their applications. Adv. Appl. Prob., 5:207–221. Press.

.
 CHAPTER 5

����STATISTICAL METHODS FOR ANALYSING DATASETS

5.1 INTRODUCTION associated with making a wrong decision. Observed

data represent only a single realization of the
This chapter introduces some of the statistical physical system of climate and weather, and,
concepts and methods available to climatologists, further, are generally observed with some level of
but does not provide detailed specifics of complex error. Conclusions can be correct or incorrect.
subjects. Some statistical methods are given only Quantitative factors that describe the confidence of
cursory treatment, while others are ignored. The the decisions are therefore necessary to properly
references at the end of the chapter and textbooks use the information contained in a dataset.
on statistical theory and methods provide more
detailed information. Two references that should
be on every climatologist’s bookshelf are Some
Methods in Climatological Analysis (WMO-No. 199) 5.2 HOMOGENIZATION
and On the Statistical Analysis of Series of Observations
(WMO-No. 415). Since new and improved statisti- Analysis of climate data to detect changes and
cal and analytical methodologies are rapidly trends is more reliable when homogenized datasets
emerging, climatologists should maintain aware- are used. A homogeneous climate dataset is one in
ness of current techniques that have practical which all the fluctuations contained in its time
applications in climatology. series reflect the actual variability and change of the
represented climate element. Most statistical meth-
The main interest in the use of observed meteoro- ods assume the data under examination are as free
logical or climatological data is not to describe the from instrumentation, coding, processing and
data (see Chapter 4), but to make inferences from a other non-meteorological or non-climatological
limited representation (the observed sample of errors as possible. Meteorological or climatological
data) of complex physical events that are helpful to data, however, are generally not homogeneous nor
users of climatological information. The interpreta- are they free from error. Errors range from system-
tion of climatological data usually involves both atic (they affect a whole set of observations the
spatial and temporal comparisons among charac- same way, such as constant instrument calibration
teristics of frequency distributions both. These errors or improper conversion of units), to random
comparisons answer common questions such as: (any one observation is subject to an error that is as
Are average temperatures taken over a specific time likely to be positive as negative, such as parallax
interval at different locations the same? differences among observers reading a mercury
(a) Is the variability of precipitation the same at barometer).
different locations?
(b) Is the diurnal temperature range at a location The best way to keep the record homogeneous is to
changing over time, and if so, how? avoid changes in the collection, handling, trans-
(c) What is the likelihood of occurrence of tropi- mission and processing of the data. It is highly
cal storms in an area? advisable to maintain observing practices and
instruments as unchanged as possible (Guide to the
Inferences are based directly on probability theory, GCOS Surface and Upper-Air Networks: GSN AND
and the use of statistical methods to make infer- GUAN, WMO/TD-No. 1106). Unfortunately, most
ences is therefore based on formal mathematical long-term climatological datasets have been
reasoning. Statistics can be defined as the pure affected by a number of factors not related to the
and applied science of creating, developing and broader-scale climate. These include, among other
applying techniques such that the uncertainty of things, changes in geographical location; local land
inductive inferences may be evaluated. Statistics is use and land cover; instrument types, exposure,
the tool used to bridge the gap between the raw mounting and sheltering; observing practices;
data and useful information, and it is used for calculations, codes and units; and historical and
analysing data and climate models and for climate political events. Some changes may cause sharp
prediction. Statistical methods allow a statement of discontinuities such as steps (for example, a change
the confidence of any decision based on applica- in instrument or site), while others may cause grad-
tion of the procedures. ual biases (for example, increasing urbanization in
the vicinity of a site). In both cases, the related time
The confidence that can be placed in a decision is series become inhomogeneous, and these inhomo-
important because of the risks that might be geneities may affect the proper assessment of
5–2 GUIDE TO CLIMATOLOGICAL PRACTICES

climatic trends. Note that site changes do not series. A reference time series ideally has to have
always affect observations of all elements, nor do experienced all of the broad climatic influences of
changes affect observations of all elements equally. the candidate, but none of its possible and artificial
The desirability of a homogeneous record stems biases. If the candidate is homogeneous, when the
primarily from the need to distil and identify candidate and reference series are compared by
changes in the broader-scale climate. There are differencing (in the case of elements measured on
some studies, however, that may require certain an interval scale, like temperature) or by calculating
“inhomogeneities” to be reflected in the data, such ratios or log ratios (for elements measured on a
as an investigation of the effects of urbanization on proportional scale, like precipitation), the resulting
local climate or of the effects of vegetation growth time series will show neither sudden changes nor
on the microclimate of an ecosystem. trends, but will oscillate around a constant value. If
there are one or more inhomogeneities, however,
Statistical tests should be used in conjunction with they will be evident in the difference or ratio time
metadata in the investigation of homogeneity. In series. An example of an observed candidate series
cases in which station history is documented well and a reference series is shown in Figure 5.1, and an
and sufficient parallel measurements have been example of a difference series revealing an
conducted for relocations and changes of instru- inhomogeneity in a candidate series is shown in
mentation, a homogenization based on this Figure 5.2.
qualitative and quantitative information should be
undertaken. Therefore, the archiving of all histori- Reference time series work well when the dataset
cal metadata is of critical importance for an effective has a large enough number of values to ensure a
homogenization of climatological time series and good climatological relation between each candi-
should be of special concern to all meteorological date and the neighbouring locations used in
services (see Chapters 2 and 3). building the reference series, and when there are no
inhomogeneities that affect all or most of the
After the metadata analysis, statistical tests may stations or values available. In general, a denser
find additional inhomogeneities. The tests usually network is needed for climatic elements or climatic
depend on the timescale of the data; the tests used types with a high degree of spatial variability (for
for daily data are different from those used for examples, more data points are needed for precipi-
monthly data or other timescales. The results of tation than for temperature, and more data points
such statistical homogenization procedures then are needed to homogenize precipitation in a highly
have to be checked again with the existing meta- variable temperature climate than in a less variable
data. In principle, any statistical test that compares temperature climate). When a change in instru-
a statistical parameter of two data samples may be ments occurs at about the same time in an entire
used. But usually, special homogeneity tests that network, the reference series would not be effective
check the whole length of a time series in one run because all the data points would be similarly
are used. Both non-parametric tests (in which no affected. When a suitable reference series cannot be
assumptions about statistical distributions are constructed, possible breakpoints and correction
made) and parametric tests (in which frequency factors need to be evaluated without using any data
distribution is known or correctly assumed) can be from neighbouring stations.
used effectively.
The double-mass graph is often used in the field of
When choosing a homogeneity test, it is very hydrometeorology for the verification of measures
important to keep in mind the shape of the of precipitation and runoff, but can be used for
frequency distribution of the data. Some datasets most elements. The accumulated total from the
have a bell-shaped (normal or Gaussian) distribu- candidate series is plotted against the accumulated
tion; for these a parametric approach works well. total from the reference series for each available
Others (such as precipitation data from a site with period. If the ratio between the candidate and refer-
marked interannual variability) are not bell- shaped, ence series remains constant over time, the resultant
and rank-based non-parametric tests may be better. double-mass curve should have constant slope. Any
Effects of serial autocorrelation, the number of important variation in the slope or the shape of the
potential change points in a series (documented curve indicates a change in the relationship between
with metadata and undocumented), trends and the two series. Since variations may occur naturally,
oscillations, and short periods of record that may it is recommended that the apparent changes of the
be anomalous should also be considered when slope occur for a well-defined continuous period
assessing the confidence that can be placed in the lasting at least five years and that they be consistent
results from any test. with events referenced in the metadata records of
the station before concluding inhomogeneity.
Many approaches rely on comparing the data to be Figure 5.3 shows a double-mass graph for the same
homogenized (the candidate series) with a reference data used in Figures 5.1 and 5.2. Because it is often
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–3

Figure 5.1. Example of a candidate time series (dashed line) and a reference (solid line) time series

Figure 5.2. Example of a difference time series

difficult to determine where on a double-mass over time). One is a runs tests, which hypothesizes
graph the slope changes, a residual graph of the that trends and other forms of persistence in a
cumulative differences between the candidate and sequence of observations occur only by chance. It is
reference station data is usually plotted against based on the total number of runs of directional
time (Figure 5.4). The residual graph more clearly changes in consecutive values. Too small a number
shows the slope change. The double-mass graph of runs indicates persistence or trends, and too large
can be used to detect more than one change in a number indicates oscillations. Stationarity of
proportionality over time. When the double-mass central tendencies and variability between parts of a
graph reveals a change in the slope, it is possible to series are important. Techniques for examining these
derive correction factors by computing the ratio of characteristics include both parametric and non-
the slopes before and after a change point. parametric methods.

There are several tests of stationarity (the hypothesis Caution is needed when data are in sub-monthly
that the characteristics of a time series do not change resolution (such as daily or hourly observations)
5–4 GUIDE TO CLIMATOLOGICAL PRACTICES

Figure 5.3. Example of a double-mass graph with the dashed line representing a slope of 1

because one of the uses of homogeneous daily 5.2.1 Evaluation of homogenized data
data is assessing changes in extremes. Extremes,
no matter how they are defined, are rare events Evaluation of the results of homogeneity detection
that often have a unique set of weather conditions and adjustment is time-consuming but unavoida-
creating them. With few extreme data points avail- ble, no matter which approach has been used. It is
able for the assessment, determining the proper very important to understand which adjustment
homogeneity adjustment for these unique condi- factors have been applied to improve the reliability
tions can be difficult. Extremes should be of the time series and to make measurements
considered as part of the whole dataset, and they comparable throughout their entire extent.
should therefore be homogenized not separately Sometimes, one might need to apply a technique
but along with all the data. Homogenization tech- that has been designed for another set of circum-
niques for monthly, seasonal or yearly temperature stances (such as another climate, meteorological or
data are generally satisfactory, but homogeniza- climatological element, or network density), and it
tion of daily data and extremes remains a is important to analyse how well the homogeniza-
challenge. tion has performed. For example, most techniques
used to homogenize monthly or annual precipita-
Although many objective techniques exist for tion data have been designed and tested in rainy
detecting and adjusting the data for inhomogenei- climates with precipitation throughout the year,
ties, the actual application of these techniques and may have serious shortcomings when applied
remains subjective. At the very least, the decision to data from climates with very dry seasons.
about whether to apply a given technique is
subjective. This means that independent attempts To assess corrections, one might compare the
at homogenization may easily result in quite adjusted and unadjusted data to independent infor-
different data. It is important to keep detailed and mation, such as data from neighbouring countries,
complete documentation of each of the steps and gridded datasets, or proxy records such those from
decisions made during the process. The adjusted phenology, observation journals, or ice freeze and
data should not be considered absolutely “correct”, thaw dates. When using such strategies, one also has
nor should the original data always be considered to be aware of their limitations. For example, grid-
“wrong”. The original data should always be ded datasets might be affected by changes in the
preserved. number of stations across time, or at a particular grid
point they might not be well correlated with the
Homogeneity assessment and data adjustment original data from a co-located or nearby station.
techniques are an area of active development, and
both the theory and practical tools are continuing Another approach is to examine countrywide, area-
to evolve. Efforts should be made to keep abreast of averaged time series for adjusted and unadjusted
the latest techniques. data and to see if the homogenization procedure
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–5

has modified the trends expected from knowledge that are used to estimate the parameters of a distri-
of the station network. For example, when there bution are called degrees of freedom. Generally, the
has been a widespread change from afternoon higher the number of degrees of freedom, the better
observations to morning observations, the unad- the estimate will be. When the smooth theoreti-
justed temperature data have a cooling bias in the cally derived curve is plotted with the data, the
time series, as the morning observations are typi- degree of agreement between the curve fit and the
cally lower than those in the afternoon. The data can be visually assessed.
adjusted series accounting for the time of observa-
tion bias, as one might predict, shows more Examination of residuals is a powerful tool for
warming over time than the unadjusted dataset. understanding the data and suggests what changes
to a model or data need to be taken. A residual is the
More complete descriptions of several widely used difference between an observed value and the corre-
tests are available in the Guidelines on Climate sponding model value. A residual is not synonymous
Metadata and Homogenization (WMO/TD-No. 1186) with an anomalous value. An anomalous value is a
and in several of the references listed at the end of strange, unusual or unique value in the original
this chapter. If the homogenization results are valid, data series. A graphical presentation of residuals is
the newly adjusted time series as a whole will useful for identifying patterns. If residual patterns
describe the temporal variations of the analysed such as oscillations, clusters and trends are noticed,
element better than the original data. Some single then the model used is usually not a good fit to the
values may remain incorrect or made even worse by data. Outliers (a few residual values that are very
the homogenization, however. different from the majority of the values) are indi-
cators of potentially suspicious or erroneous data
values. They are usually identified as extremes in
later analyses. If no patterns exist and if the values
5.3 MODEL-FITTING TO ASSESS DATA of the residuals appear to be randomly scattered,
DISTRIBUTIONS then the model may be accepted as a good fit to the
data.
After a dataset is adjusted for known errors and
inhomogeneities, the observed frequency distribu- If an observed frequency distribution is to be fitted
tions should be modelled by the statistical by a statistical model, the assumptions of the model
distributions described in section 4.4.1 so that and fitting process must be valid. Most models
statistical methods can be exploited. A theoretical assume that the data are independent (one observa-
frequency distribution can be fit to the data by tion is unaffected by any other observation). Most
inserting estimates of the parameters of the distri- comparative tests used in goodness-of-fit tests
bution, where the estimates are calculated from the assume that errors are randomly and independently
sample of observed data. The estimates can be based distributed. If the assumptions are not valid, then
on different amounts of information or data. The any conclusions drawn from such an analysis may
number of unrelated bits of information or data be incorrect.

Figure 5.4. Example of a residual double-mass graph

5–6 GUIDE TO CLIMATOLOGICAL PRACTICES

Once the data have been fitted by an acceptable offset must be added before taking the square root
statistical frequency distribution, meeting any so that all values are greater than or equal to 0. The
necessary independence, randomness or other cube root has a similar effect to the square root, but
sampling criteria, and the fit has been validated (see does not require the use of an offset to handle nega-
section 4.4), the model can be used as a representa- tive values. Logarithmic transformations compresses
tion of the data. Inferences can be made that are the range of values, by making small values rela-
supported by mathematical theory. The model tively larger and large values relatively smaller. A
provides estimates of central tendency, variability constant offset must first be added if values equal to
and higher-order properties of the distribution 0 or lower are present. An inverse makes very small
(such as skewness or kurtosis). The confidence that numbers very large and very large numbers very
these sample estimates represent real physical small; values of 0 must be avoided.
conditions can also be determined. Other charac-
teristics, such as the probability of an observation’s These transformations have been described in the
exceeding a given value, can also be estimated by relative order of power, from weakest to strongest.
applying both probability and statistical theory to A good guideline is to use the minimum amount of
the modelled frequency distribution. All of these transformation necessary to improve normality. If a
tasks are much harder, if not impossible, when meteorological or climatological element has an
using the original data rather than the fitted inherent highly non-normal frequency distribu-
frequency distribution. tion, such as the U-shape distribution of cloudiness
and sunshine, there are no simple transformations
allowing the normalization of the data.

5.4 DATA TRANSFORMATION The transformations all compress the right side of a
distribution more than the left side; they reduce
The normal or Gaussian frequency distribution is higher values more than lower values. Thus, they
widely used, as it has been studied extensively in are effective on positively skewed distributions
statistics. If the data do not fit the normal distribu- such as precipitation and wind speed. If a distribu-
tion well, applying a transform to the data may tion is negatively skewed, it must be reflected
result in a frequency distribution that is nearly (values are multiplied by –1, and then a constant is
normal, allowing the theory underlying the normal added to make all values greater than 0) to reverse
distribution to form the basis for many inferential the distribution prior to applying a transformation,
uses. Transforming data must be done with care so and then reflected again to restore the original
that the transformed data still represent the same order of the element.
physical processes as the original data and that
sound conclusions can be made. Data transformations offer many benefits, but they
should be used appropriately in an informed
There are several ways to tell whether a distribution of manner. All of the transformations described above
an element is substantially non-normal. A visual attempt to improve normality by reducing the rela-
inspection of histograms, scatter plots, or probability– tive spacing of data on the right side of the
probability (P–P) or quantile–quantile (Q–Q) plots is distribution more than the spacing on the left side.
relatively easy to perform. A more objective assess- The very act of altering the relative distances
ment can range from simple examination of skewness between data points, which is how these transfor-
and kurtosis (see section 4.4) to inferential tests of mations aim to improve normality, raises issues in
normality. the interpretation of the data, however. All data
points remain in the same relative order as they
Prior to applying any transformation, an analyst were prior to transformation, which allows inter-
must make certain that the non-normality is caused pretation of results in terms of the increasing value
by a valid reason. Invalid reasons for non-normal- of the element. The transformed distributions will
ity include mistakes in data entry and missing data likely become more complex to interpret physi-
values not declared missing. Another invalid reason cally, however, due to the curvilinear nature of the
for non-normality may be the presence of outliers, transformations. The analyst must therefore be
as they may well be a realistic part of a normal careful when interpreting results based on trans-
distribution. formed data.

The most common data transformations utilized

for improving normality are the square root, cube
root, logarithmic and inverse transformations. The 5.5 TIME SERIES ANALYSIS
square root makes values less than 1 relatively
greater, and values greater than 1 relatively smaller. The principles guiding model-fitting (see section
If the values can be positive or negative, a constant 5.3) also guide time series analysis. A model is fitted
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–7

to the data series; the model might be linear, curvi- techniques are emerging fields, and although the
linear, exponential, periodic or some other mathematics has been defined, future refinements
mathematical formulation. The best fit (the fit that in techniques and application methodology may
minimizes the differences between the data series mitigate the limitations.
and the model) is generally accomplished by using
least-squares techniques (minimizing the sum of Other common techniques for analysing time series
squared departures of the data from the curve fit). are autoregression and moving average analyses.
Residuals from the best fit are examined for patterns, Autoregression is a linear regression of a value in a
and if patterns are found, then the model is adjusted time series against one or more prior values in the
to incorporate the patterns. series (autocorrelation). A moving average process
expresses an observed series as a function of a
Time series in climatology have been analysed random series. A combination of these two meth-
mainly with harmonic and spectral analysis tech- ods is called a mixed autoregressive and moving
niques that decompose a series into time domain or average (ARMA) model. An ARMA model that
frequency domain components. A critical assump- allows for non-stationarity is called a mixed autore-
tion of these models is that of stationarity gressive integrated moving average (ARIMA) model.
(characteristics of the series such as mean and vari- These regression-based models can be made more
ance do not change over the length of the series). complex than necessary, resulting in overfitting.
This condition is generally not met by climatologi- Overfitting can lead to the modelling of a series of
cal data even if the data are homogeneous (see values with minimal differences between the model
section 5.2). and the data values, but since the data values are
only a sample representation of a physical process,
Gabor and wavelet analysis are extensions of the a slight lack of fit may be desirable in order to repre-
classical techniques of spectral analysis. By allow- sent the true process. Other problems include
ing subintervals of a time series to be modelled with non-stationarity of the parameters used to define a
different scales or resolutions, the condition of model, non-random residuals (indicating an
stationarity is relaxed. These analyses are particu- improper model), and periodicity inherent in the
larly good at representing time series with data but not modelled. Split validation is effective
subintervals that have differing characteristics. in detecting model overfitting. Split validation
Wavelet analysis gives good results when the time refers to developing a model based on a portion of
series has spikes or sharp discontinuities. Compared the available data and then validating the model on
to the classical techniques, they are particularly the remaining data that were not used in the model
efficient for signals in which both the amplitude development.
and frequency vary with time. One of the main
advantages of these “local” analyses is the ability to Once the time series data have been modelled by an
present time series of climate processes in the coor- acceptable curve, and the fit validated, the mathe-
dinates of frequency and time, studying and matical properties of the model curve can be used
visualizing the evolution of various modes of vari- to make assessments that would not be possible
ability over a long period. They are used not only as using the original data. These include measuring
a tool for identifying non-stationary scales of varia- trends, cyclical behaviour, or autocorrelation and
tions, but also as a data analysis tool to gain an persistence, together with estimates of the confi-
initial understanding of a dataset. There have been dence of these measures.
many applications of these methods in climatol-
ogy, such as in studies of the El Nino–Southern
Oscillation (ENSO) phenomenon, the North
Atlantic Oscillation, atmospheric turbulence, 5.6 MULTIVARIATE ANALYSIS
space–time precipitation relationships and ocean
wave characteristics. Multivariate datasets are a compilation of
observations of more than one element or a
These methods do have some limitations. The most compilation of observations of one element at
important limitation for wavelet analysis is that an different points in space. These datasets are often
infinite number of wavelet functions are available studied for many different purposes. The most
as a basis for an analysis, and results often differ important purposes are to see if there are simpler
depending on which wavelet is used. This makes ways of representing a complex dataset, if
interpretation of results somewhat difficult because observations fall into groups and can be classified,
different conclusions can be drawn from the same if the elements fall into groups, and if
dataset if different mathematical functions are interdependence exists among elements. Such
used. It is therefore important to relate the wavelet datasets are also used to test hypotheses about the
function to the physical world prior to selecting a data. The time order of the observations is generally
specific wavelet. Gabor and wavelet analysis not a consideration; time series of more than one
5–8 GUIDE TO CLIMATOLOGICAL PRACTICES

element are usually considered as a separate analysis that for three points forming a triangle, the length
topic with techniques such as cross-spectral of one side should be less than or equal to the sum
analysis. of the lengths of the other two sides (triangle
inequality). A fourth criterion should be that if the
Principal components analysis, sometimes referred distance from A to B is zero, then A and B are the
to as empirical orthogonal functions analysis, is a same (definiteness). Most techniques iteratively
technique for reducing the dimensions of multi- separate the data into more and more clusters,
variate data. The process simplifies a complex thereby presenting the problem for the analyst of
dataset and has been used extensively in the analy- determining when the number of clusters is suffi-
sis of climatological data. Principal components cient. Unfortunately, there are no objective rules
analysis methods decompose a number of corre- for making this decision. The analyst should there-
lated observations into a new set of uncorrelated fore use prior knowledge and experience in deciding
(orthogonal) functions that contain the original when a meteorologically or climatologically appro-
variance of the data. These empirical orthogonal priate number of clusters has been obtained. Cluster
functions, also called principal components, are analysis has been used for diverse purposes, such as
ordered so that the first component is the one constructing homogeneous regions of precipita-
explaining most of the variance, the second compo- tion, analysing synoptic climatologies, and
nent explains the second-largest share of the predicting air quality in an urban environment.
variance, and so on. Since most of the variance is
usually explained by just a few components, the Canonical correlation analysis seeks to determine
methods are effective in reducing “noise” from an the interdependence between two groups of
observed field. Individual components can often be elements. The method finds the linear combination
related to a single meteorological or climatological of the distribution of the first element that produces
element. The method has been used to analyse a the correlation with the second distribution. This
diversity of fields that include sea surface tempera- linear combination is extracted from the dataset
tures, regional land temperature and precipitation and the process is repeated with the residual data,
patterns, tree-ring chronologies, sea level pressure, with the constraint that the second linear combina-
air pollutants, radiative properties of the atmos- tion is not correlated with the first combination.
phere, and climate scenarios. Principal components The process is again repeated until a linear combi-
have also been used as a climate reconstruction nation is no longer significant. This analysis is used,
tool, such as in estimating a spatial grid of a climatic for example, in making predictions from telecon-
element from proxy data when actual observations nections, in statistical downscaling (see section
of the element are not available. 6.7.3), in determining homogeneous regions for
flood forecasting in an ungauged basin, and in
Factor analysis reduces a dataset from a larger set of reconstructing spatial wind patterns from pressure
observations to a smaller set of factors. In the mete- fields.
orological and climatological literature, factor
analysis is often called rotated principal compo- These methods all have assumptions and limita-
nents analysis. It is similar to principal components tions. The interpretation of the results is very much
analysis except that the factors are not uncorre- dependent on the assumptions being met and on
lated. Since a factor may represent observations the experience of the analyst. Other methods, such
from more than one element, meteorological or as multiple regression and covariance analysis, are
climatological interpretation of a factor is often even more restrictive for most meteorological or
difficult. The method has been used mainly in climatological data. Multivariate analysis is
synoptic climatology studies. complex, with numerous possible outcomes, and
requires care in its application.
Cluster analysis attempts to separate observations
into groups with similar characteristics. There are
many methods for clustering, and different meth-
ods are used to detect different patterns of points. 5.7 COMPARATIVE ANALYSIS
Most of the methods, however, rely on the extent
to which the distance between means of two groups By fitting a model function to the data, be it a
is greater than the mean distance within a group. frequency distribution or a time series, it is possible
The measure of distance does not need to be the to use the characteristics of that model for further
usual Euclidean distance, but it should obey certain analysis. The properties of the model characteristics
criteria. One such criterion should be that the meas- are generally well studied, allowing a range of
ure of distance from point A to point B is equal to conclusions to be drawn. If the characteristics are
the distance from point B to point A (symmetry). A not well studied, bootstrapping may be useful.
second criterion is that the distance should be a Bootstrapping is the estimation of model character-
positive value (non-negativity). A third criterion is istics from multiple random samples drawn from
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–9

the original observational series. It is an alternative When the null hypothesis is rejected but it is
to making inferences from parameter-based actually true, a Type I error has been made. When
assumptions when the assumptions are in doubt, the null hypothesis is accepted and it is actually
when parametric inference is impossible, or when false, a Type II error has been made. Unfortunately,
parametric inference requires very complicated reducing the risk of a Type I error increases the risk
formulas. Bootstrapping is simple to apply, but it of making a Type II error, so that a balance between
may conceal its own set of assumptions that would the two types is necessary. This balance should be
be more formally stated in other approaches. based on the seriousness of making either type of
error. In any case, the confidence of the conclusion
In particular, there are many tests available for can be calculated in terms of probability and should
comparing the characteristics of two models to be reported with the conclusion.
determine how much confidence can be placed in
claims that the two sets of modelled data share
underlying characteristics. When comparing two
models, the first step is to decide which characteris- 5.8 SMOOTHING
tics are to be compared. These could include the
mean, median, variance or probability of an event Smoothing methods provide a bridge between
from a distribution, or the phase or frequency from making no assumptions based on a formal structure
a time series. In principle, any computable charac- of observed data (the non-parametric approach)
teristic of the fitted models can be compared, and making very strong assumptions (the paramet-
although there should be some meaningful reason ric approach). Making a weak assumption that the
(based on physical arguments) to do so. true distribution of the data can be represented by a
smooth curve allows underlying patterns in the
The next step is to formulate the null hypothesis. data to be revealed to the analyst. Smoothing
This is the hypothesis considered to be true before increases signals of climatic patterns while reducing
any testing is done, and in this case it is usually that noise induced by random fluctuations. The applica-
the modelled characteristics are the same. The alter- tions of smoothing include exploratory data
native hypothesis is the obverse, that the modelled analysis, model-building, goodness-of-fit of a repre-
characteristics are not the same. sentative (smooth) curve to the data, parametric
estimation, and modification of standard
A suitable test to compare the characteristics from the methodology.
two models is then selected. Some of these tests are
parametric, depending on assumptions about the Kernel density estimation is one method of smooth-
distribution, such as normality. Parametric tests ing; examples include moving averages, Gaussian
include the Student’s t-test (for comparing means) smoothing and binomial smoothing. Kernel
and the Fisher’s F-test (for comparing variability). smoothers estimate the value at a point by combin-
Other tests are non-parametric, so they do not make ing the observed values in a neighbourhood of that
assumptions about the distribution. They include point. The method of combination is often a
sign tests (for comparing medians) and the weighted mean, with weights dependent on the
Kolmogorov-Smirnov test for comparing distribu- distance from the point in question. The size of the
tions. Parametric tests are generally better (in terms of neighbourhood used is called the bandwidth; the
confidence in the conclusions), but only if the larger the bandwidth, the greater the smoothing.
required assumptions about the distribution are valid. Kernel estimators are simple, but they have draw-
backs. Kernel estimation can be biased when the
The seriousness of rejecting a true hypothesis (or region of definition of the data is bounded, such as
accepting a false one) is expressed as a level of confi- near the beginning or end of a time series. As one
dence or probability. The selected test will show bandwidth is used for the entire curve, a constant
whether the null hypothesis can be accepted at the level of smoothing is applied. Also, the estimation
level of confidence required. Some of the tests will tends to flatten peaks and valleys in the distribu-
reveal at what level of confidence the null hypoth- tion of the data. Improvements to kernel estimation
esis can be accepted. If the null hypothesis is include correcting the boundary biases by using
rejected, the alternative hypothesis must be special kernels only near the boundaries, and by
accepted. Using this process, the analyst might be varying the bandwidths in different sections of the
able to make the claim, for example, that the means data distribution. Data transformations (see section
of two sets of observations are equal with a 99 per 5.4) may also improve the estimation.
cent level of confidence; accordingly, there is only a
1 per cent chance that the means are not the same. Spline estimators fit a frequency distribution piece-
wise over subintervals of the distribution with
Regardless of which hypothesis is accepted, the null polynomials of varying degree. Again, the number
or the alternative, the conclusion may be erroneous. and placement of the subintervals affects the degree
5–10 GUIDE TO CLIMATOLOGICAL PRACTICES

of smoothing. Estimation near the boundaries of identify a value as an outlier because the intent is to
the data is problematic as well. Outliers can severely smooth all the observations. Outliers could be a
affect a spline fit, especially in regions with few valid meteorological or climatological response, or
observations. they could be aberrant; additional investigation of
the outlier is necessary to ensure the validity of the
A range of more sophisticated, often non-paramet- value. Regression estimates are also affected by
ric smoothers, are also available. These include correlation. Estimates are based on the assumption
local maximum likelihood estimation, which is that all errors are statistically independent of each
particularly useful when prior knowledge of the other; correlation can affect the asymptotic
behaviour of the dataset can lead to a good “first properties of the estimators and the behaviour of the
guess” of the type of curve that should be fitted. bandwidths determined from the data.
These estimators are sometimes difficult to inter-
pret theoretically.

With multivariate data, smoothing is more complex 5.9 ESTIMATING DATA

because of the number of possibilities of smoothing
and the number of smoothing parameters that need One of the main applications of statistics to clima-
to be set. As the number of data elements increases, tology is the estimation of values of elements when
smoothing becomes progressively more difficult. few or no observed data are available or when
Most graphs are limited to only two dimensions, so expected data are missing. In many cases, the plan-
visual inspection of the smoother is limited. Kernel ning and execution of user projects cannot be
density can be used to smooth multivariate data, delayed until there are enough meteorological or
but the problems of boundary estimation and fixed climatological observations; estimation is used to
bandwidths can be even more challenging than extend a dataset. Estimation also has a role in qual-
with univariate data. ity control by allowing an observed value to be
compared to its neighbours in both time and space.
Large empty regions in a multivariate space usually Techniques for estimating data are essentially appli-
exist unless the number of data values is very large. cations of statistics, but should also rely on the
Collapsing the data to a smaller number of dimen- physical properties of the system being considered.
sions with, for example, principal components In all cases, it is essential that values statistically
analysis, is a smoothing technique. The dimension estimated be realistic and consistent with physical
reduction should have the goal of preserving any considerations.
interesting structure or signal in the data in the
lower-dimension data while removing uninterest- Interpolation uses data that are available both
ing attributes or noise. before and after a missing value (time interpola-
tion), or surrounding the missing value (space
One of the most widely used smoothing tools is interpolation), to estimate the missing value. In
regression. Regression models, both linear and non- some cases, the estimation of a missing value can
linear, are powerful for modelling a target element be performed by a simple process, such as by
as a function of a set of predictors, allowing for a computing the average of the values observed on
description of relationships and the construction of both sides of the gap. Complex estimation methods
tests of the strength of the relationships. These are also used, taking into account correlations with
models are susceptible, however, to the same prob- other elements. These methods include weighted
lems as any other parametric model in that the averages, spline functions, linear regressions and
assumptions made affect the validity of inferences kriging. They may rely solely on the observations of
and predictions. an element, or take into account other information
such as topography or numerical model output.
Regression models also suffer from boundary Spline functions can be used when the spatial vari-
problems and unrealistic smoothing in subintervals ations are regular. Linear regression allows the
of the data range. These problems can be solved by inclusion of many kinds of information. Kriging is
weighting subintervals of the data domain with a geostatistical method that requires an estimation
varying bandwidths and by applying polynomial of the covariances of the studied field. Cokriging
estimation near the boundaries. Regression estimates, introduces into kriging equations the information
which are based on least-squares estimation, can be given by another independent element.
affected by observations with unusual response
values (outliers). If a data value is far from the Extrapolation extends the range of available data
majority of the values, the smooth curve will be tend values. There are more possibilities for error of
to be drawn closer to the aberrant value than may be extrapolated values because relations are used
justified. When using adjusted non-parametric outside the domain of the values from which the
smoothing, it is often difficult to unambiguously relationships were derived. Even if empirical
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–11

relations found for a given place or period of time estimation. When simultaneous values at two
seem reasonable, care must be taken when applying stations close to each other are compared, some-
them to another place or time because the times either the difference or the quotient of the
underlying physics at one place and time may not values is approximately constant. This is more
be the same as at another place and time. The same often true for summarized data (for months or
methods used for interpolation can be used for years) than for those over shorter time intervals
extrapolation. (such as daily data). The constant difference or
ratio can be used to estimate data. When using
these methods, the series being compared should
5.9.1 Mathematical estimation methods
be sufficiently correlated for the comparison to be
Mathematical methods involve the use of only meaningful. Then, the choice of the method
geometric or polynomial characteristics of a set of should depend on the time structure of the two
point values to create a continuous surface. Inverse series. The difference method can be used when
distance weighting and curve fitting methods, such the variations of the meteorological or climato-
as spline functions, are examples. The methods are logical element are relatively similar from one
exact interpolators; observed values are retained at station to the other. The ratio method can be
sites where they are measured. applied when the time variations of the two series
are not similar, but nevertheless proportional (this
Inverse distance weighting is based on the distance is usually the case when a series has a lower bound
between the location for which a value is to be of zero, as with precipitation or wind speed, for
interpolated and the locations of observations. example). In the event that those assumptions are
Unlike the simple nearest neighbours method not fulfilled, particularly when the variances of
(where the observation from the nearest location is the series at the two stations are not equal for the
chosen), inverse distance weighting combines method using the differences, these techniques
observations from a number of neighbouring loca- should not be used. More complex physical
tions. Weights are given to the observations consistency tools include regression, discriminant
depending on their distance from the target loca- analysis (for the occurrence of phenomena) and
tion; close stations have a larger weight than those principal components analysis.
farther away. A “cut-off” criterion is often used,
either to limit the distance to observation locations Deterministic methods are based upon a known
or the number of observations considered. Often, relation between an in situ data value (predictand)
inverse squared distance weighting is used to and values of other elements (predictors). This rela-
provide even more weight to the closest locations. tion is often based on empirical knowledge about
With this method no physical reasoning is used; it the predictand and the predictor. The empirical
is assumed that the closer an observation location relation can be found by either physical or statisti-
is to the location where the data are being esti- cal analysis, and is frequently a combination in
mated, the better the estimation. This assumption which a statistical relation is derived from values
should be carefully validated since there may be no based on the knowledge of a physical process.
inherent meteorological or climatological reason to Statistical methods such as regression are often
justify the assumption. used to establish such relations. The deterministic
approach is stationary in time and space and must
Spline fits suffer from the same limitation as inverse therefore be regarded as a global method reflecting
distance weighting. The field resulting from a spline the properties of the entire sample. The predictors
fit assumes that the physical processes can be repre- may be other observed elements or other geographic
sented by the mathematical spline; there is rarely parameters, such as elevation, slope or distance
any inherent justification to this assumption. Both from the sea.
methods work best on smooth surfaces, so they
may not result in adequate representations on
5.9.3 Spatial estimation methods
surfaces that have marked fluctuations.
Spatial interpolation is a procedure for estimating
the value of properties at unsampled sites within
5.9.2 Estimation based on physical
an area covered by existing observations. The
relationships
rationale behind interpolation is that observation
The physical consistency that exists among differ- sites that are close together in space are more likely
ent elements can be used for estimation. For to have similar values than sites that are far apart
instance, if some global radiation measurements (spatial coherency). All spatial interpolation
are missing and need to be estimated, elements methods are based on theoretical considerations,
such as sunshine duration and cloudiness could be assumptions and conditions that must be fulfilled
used to estimate a missing value. Proxy data may in order for a method to be used properly.
also be used as supporting information for Therefore, when selecting a spatial interpolation
5–12 GUIDE TO CLIMATOLOGICAL PRACTICES

algorithm, the purpose of the interpolation, the combining principal components analysis, linear
characteristics of the phenomenon to be multiple regression and kriging. Depending on the
interpolated, and the constraints of the method method used, topography is described by the
have to be considered. elevation, slope and slope direction, generally
averaged over an area. The topographic character-
Stochastic methods for spatial interpolation are istics are generally at a finer spatial resolution than
often referred to as geostatistical methods. A the climate data.
feature shared by these methods is that they use a
spatial relationship function to describe the corre- Among the most advanced physically based methods
lation among values at different sites as a function are those that incorporate a description of the dynam-
of distance. The interpolation itself is closely ics of the climate system. Similar models are routinely
related to regression. These methods demand that used in weather forecasting and climate modelling
certain statistical assumptions be fulfilled, for (see section 6.7). As the computer power and storage
example: the process follows a normal distribu- capacity they require becomes more readily available,
tion, it is stationary in space, or it is constant in all these models are being used more widely in climate
directions. monitoring, and especially to estimate the value of
climate elements in areas remote from actual observa-
Even though it is not significantly better than tions (see section 5.13 on reanalysis).
other techniques, kriging is a spatial interpolation
approach that has been used often for interpolat-
5.9.4 Time series estimation
ing elements such as air and soil temperature,
precipitation, air pollutants, solar radiation, and Time series often have missing data that need to be
winds. The basis of the technique is the rate at estimated or values that must be estimated at time­
which the variance between points changes over scales that are finer than those provided by the
space and is expressed in a variogram. A variogram observations. One or just a few observations can be
shows how the average difference between values estimated better than a long period of continuous
at points changes with distance and direction missing observations. As a general rule, the longer
between points. When developing a variogram, it the period to be estimated, the less confidence one
is necessary to make some assumptions about the can place in the estimates.
nature of the observed variation on the surface.
Some of these assumptions concern the constancy For single-station analysis, one or two consecutive
of means over the entire surface, the existence of missing values are generally estimated by simple
underlying trends, and the randomness and inde- linear, polynomial or spline approximations that
pendence of variations. The goal is to relate all are fitted from the observations just before and
variations to distance. Relationships between a after the period to be estimated. The assumption is
variogram and physical processes may be accom- that conditions within the period to be estimated
modated by choosing an appropriate variogram are similar to those just before and after the period
model (for example, spherical, exponential, to be estimated; care must be taken that this
Gaussian or linear). assumption is valid. An example of a violation of
this assumption in the estimation of hourly temper-
Some of the problems with kriging are the compu- atures is the passage of a strong cold front during
tational intensity for large datasets, the complexity the period to be estimated. Estimation of values for
of estimating a variogram, and the critical assump- longer periods is usually accomplished with time
tions that must be made about the statistical nature series analysis techniques (see section 5.5)
of the variation. This last problem is most impor- performed on parts of the series without data gaps.
tant. Although many variants of kriging allow The model for the values that do exist is then
flexibility, the method was developed initially for applied to the gaps. As with spatial interpolation,
applications in which distances between observa- temporal interpolation should be validated to
tion sites are small. In the case of climatological ensure that the estimated values are reasonable.
data, the distances between sites are usually large, Metadata or other corollary information about the
and the assumption of smoothly varying fields time series is useful for determining the
between sites is often not realistic. reasonableness.

Since meteorological or climatological fields such

5.9.5 Validation
as precipitation are strongly influenced by topog-
raphy, some methods, such as Analysis Using Any estimation is based on some underlying
Relief for HYdrometeorology (AURELHY) and structure or physical reasoning. It is therefore
Parameter-elevation Regressions on Independent very important to verify that the assumptions
Slopes Model (PRISM), incorporate the topogra- made in applying the estimation model are
phy into an interpolation of climatic data by fulfilled. If they are not fulfilled, the estimated
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–13

values may be in error. Furthermore, the error 5.10 EXTREME VALUE ANALYSIS
could be serious and lead to incorrect conclu-
sions. In climatological analysis, model Many practical problems in climatology require
assumptions often are not met. For example, in knowledge of the behaviour of extreme values of
spatial analysis, interpolating between widely some climatological elements. This is particularly
spaced stations implies that the climatological true for the engineering design of structures that
patterns between stations are known and can be are sensitive to high or low values of meteorologi-
modelled. In reality, many factors (such as topog- cal or climatological phenomena. For example,
raphy, local peculiarities or the existence of water high precipitation amounts and resulting stream-
bodies) influence the climate of a region. Unless flows affect sewerage systems, dams, reservoirs and
these factors are adequately incorporated into a bridges. High wind speed increases the load on
spatial model, the interpolated values will likely buildings, bridges, cranes, trees and electrical
be in error. In temporal analysis, interpolating power lines. Large snowfalls require that roofs be
over a large data gap implies that the values repre- built to withstand the added weight. Public author-
senting conditions before and after the gap can ities and insurers may want to define thresholds
be used to estimate the values within the gap. In beyond which damages resulting from extreme
reality, the more variable the weather patterns are conditions become eligible for economic relief.
at a location, the less likely that this assumption
will hold, and consequently the interpolated Design criteria are often expressed in terms of a
values could be in error. return period, which is the mean interval of time
between two occurrences of values equal to or
The seriousness of any error of interpolation is greater than a given value. The return period
related to the use of the data. Conclusions and concept is used to avoid adopting high safety coef-
judgments based on requirements for microscale, ficients that are very costly, but also to prevent
detailed information about a local area will be major damage to equipment and structures from
much more affected by errors than those that are extreme events that are likely to occur during the
based on macroscale, general information for a useful life of the equipment or structures. As such
large area. When estimating data, the sensitivity of equipment can last for years or even centuries,
the results to the use of the data should be consid- accurate estimation of return periods can be a criti-
ered carefully. cal factor in their design. Design criteria may also
be described by the number of expected occur-
Validation is essential whenever spatial interpola- rences of events exceeding a fixed threshold.
tion is performed. Split validation is a simple and
effective technique. A large part of a dataset is used
5.10.1 Return period approach
to develop the estimation procedures and a single,
smaller subset of the dataset is reserved for testing Classical approaches to extreme value analysis
the methodology. The data in the smaller subset represent the behaviour of the sample of extremes
are estimated with the procedures developed from by a probability distribution that fits the observed
the larger portion, and the estimated values are distribution sufficiently well. The extreme value
compared with the observed values. Cross- distributions have assumptions such as stationarity
validation is another simple and effective tool to and independence of data values, as discussed in
compare various assumptions either about the Chapter 4. The three common extreme value distri-
models (such as the type of variogram and its butions are Gumbel, Frechet and Weibull. The
parameters, or the size of a kriging neighbour- generalized extreme value (GEV) distribution
hood) or about the data, using only the information combines these three under a single formulation,
available in the given sample dataset. Cross- which is characterized by a model shape
validation is carried out by removing one parameter.
observation from the data sample, and then esti-
mating the removed value based on the remaining The data that are fitted by an extreme value distri-
observations. This is repeated with the removal of bution model are the maxima (or minima) of
a different observation from the sample, and values observed in a specified time interval. For
repeated again, removing each observation in example, if daily temperatures are observed over a
turn. The residuals between the observed and esti- period of many years, the set of annual maxima
mated values can then be further analysed could be represented by an extreme value distribu-
statistically or can be mapped for visual inspec- tion. Constructing and adequately representing a
tion. Cross-validation offers quantitative insights set of maxima or minima from subintervals of the
into how any estimation method performs. An whole dataset requires that the dataset be large,
analysis of the spatial arrangement of the residuals which may be a strong limitation if the data sample
often suggests further improvements of the esti- covers a limited period. An alternative is to select
mation model. values beyond a given threshold. The generalized
5–14 GUIDE TO CLIMATOLOGICAL PRACTICES

Pareto frequency distribution is usually suitable for (b) Calibration of the model using observations
fitting data beyond a threshold. of storm depth and accompanying atmos-
pheric moisture;
Once a distribution is fitted to an extreme value (c) Use of the calibrated model to estimate what
dataset, return periods are computed. A return would have occurred with maximum observed
period is the mean frequency with which a value is atmospheric moisture;
expected to be equalled or exceeded (such as once (d) Translation of the observed storm character-
in 20 years). Although lengthy return periods for istics from gauged locations to the location
the occurrence of a value can be mathematically where the estimate is required, adjusting for
calculated, the confidence that can be placed in the effects of topography, continentality, and
results may be minimal. As a general rule, confi- similar non-meteorological or non-climato-
dence in a return period decreases rapidly when the logical conditions.
period is more than about twice the length of the
original dataset.

Extreme climate events can have significant impacts 5.11 ROBUST STATISTICS
on both natural and man-made systems, and there-
fore it is important to know if and how climate Robust statistics produce estimators that are not
extremes are changing. Some types of infrastructure unduly affected by small departures from model
currently have little margin to buffer the impacts of assumptions. Statistical inferences are based on
climate change. For example, there are many observations as well as the assumptions of the
communities in low-lying coastal zones through- underlying models (such as randomness, independ-
out the world that are at risk from rising sea levels. ence and model fit). Climatological data often
Adaptation strategies to non-stationary climate violate many of these assumptions because of the
extremes should account for the decadal-scale temporal and spatial dependence of observations,
changes in climate observed in the recent past, as data inhomogeneities, data errors and other factors.
well as for future changes projected by climate
models. Newer statistical models, such as the non- The effect of assumptions on the results of analyses
stationary generalized extreme value, have been should be determined quantitatively if possible, but
developed to try to overcome some of the limita- at least qualitatively, in an assessment of the valid-
tions of the more conventional distributions. As ity of conclusions. The purpose of an analysis is also
models continue to evolve and as their properties important. General conclusions based on large
become better understood, they will likely replace temporal or spatial scale processes with a lot of
the more common approaches to analysing averaging and on a large dataset are often less sensi-
extremes. The Guidelines on Analysis of Extremes in a tive to deviations from assumptions than more
Changing Climate in Support of Informed Decisions for specific conclusions. Robust statistical approaches
Adaptation (WMO/TD-No. 1500) is a publication are often used for regression.
that provides more insight into how one should
account for a changing climate when assessing and If results are sensitive to violations of assumptions,
estimating extremes. the analyst should include this fact when dissemi-
nating the results to users. It may be also be possible
to analyse the data using other methods that are
5.10.2 Probable maximum precipitation
not as sensitive to deviations from assumptions, or
The probable maximum precipitation is defined as that do not make any assumptions about the
the theoretically greatest depth of precipitation for factors causing the sensitivity problems. Since
a given duration that is physically possible over a parametric methods assume more conditions than
storm area of a given size under particular geograph- non-parametric methods, it may be possible to
ical conditions at a specified time of the year. It is reanalyse the data with non-parametric tech-
widely used in the design of dams and other large niques. For example, using the median and
hydraulic systems, for which a very rare event could interquartile range instead of the mean and stand-
have disastrous consequences. ard deviation decreases sensitivity to outliers or to
gross errors in the observational data.
The estimation of probable maximum precipitation
is generally based on heuristic approaches, includ-
ing the following steps:
5.12 STATISTICAL PACKAGES
(a) Use of a conceptual storm model to represent
precipitation processes in terms of physical Since most climatological processing and analyses
elements such as surface dewpoint, depth of are based on universal statistical methods, universal
storm cell, inflow and outflow; statistical packages are convenient computer
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–15

software instruments for the climatologists. Several These points can be interactively excluded from anal-
software products for universal statistical analysis ysis based on a graph of the series, and trend statistics
are available on a variety of computer platforms. can be recalculated automatically. Options are usually
available for analysing and displaying subgroups of
Statistical packages offer numerous data manage- data.
ment, analytical and reporting tools. A chosen
package should have all the capabilities required to
manage, process and analyse data, but not be
burdened with unnecessary tools that lead to inef- 5.13 DATA MINING
ficiencies. Some of the basic tools are often included
in a Climate Data Management System (see Data mining is an analytic process designed to
section 3.3). explore large amounts of data in search of consist-
ent patterns or systematic relationships among
Basic data management tools provide a wide variety elements, and then to validate the findings by
of operations with which to make the data conven- applying the detected patterns to new subsets of
ient for processing and analysis. These operations data. It is often considered a blend of statistics, arti-
include sorting, adding data, subsetting data, trans- ficial intelligence and database research. It is rapidly
posing matrices, arithmetic calculations, and developing into a major field, and important theo-
merging data. Basic statistical processing tools retical and practical advances are being made. Data
include the calculation of sample descriptive statis- mining is fully applicable to climatological prob-
tics, correlations, frequency tables and hypothesis lems when the volume of data available is large,
testing. Analytical tools usually cover many of the and ways to search the significant relationships
needs of climate analysis, such as analysis of vari- among climate elements may not be evident, espe-
ance, regression analysis, discriminant analysis, cially at the early stages of analysis.
cluster analysis, multidimensional analysis and
time series analysis. Calculated results of analyses Data mining is similar to exploratory data analy-
are usually put into resultant datasets and can sis, which is also oriented towards the search for
usually be saved, exported and transformed, and relationships among elements in situations when
thus used for any further analysis and processing. possible relationships are not clear. Data mining is
not concerned with identifying the specific rela-
The graphical tools contained in statistical packages tions among the elements involved. Instead, the
include the creation of two- and three-dimensional focus is on producing a solution that can generate
graphs, the capability to edit the graphs, and the useful predictions. Data mining takes a “black
capability to save the graphs in specific formats of box” approach to data exploration or knowledge
the statistical packages or in standard graphical discovery and uses not only the traditional explor-
formats. Most packages can create scatter plots (two- atory data analysis techniques, but also such
and three-dimensional); bubble plots; line, step and techniques as neural networks, which can gener-
interpolated (smoothed) plots; vertical, horizontal ate valid predictions but are not capable of
and pie charts; box and whisker plots; and three- identifying the specific nature of the interrelations
dimensional surface plots, including contouring of among the elements on which the predictions are
the surface. Some packages contain the tools for based.
displaying values of some element on a map, but
they should not be considered a replacement for a
Geographical Information System (GIS). A
Geographical Information System integrates hard- 5.14 REFERENCES AND ADDITIONAL
ware, software and data for capturing, managing, READING
analysing and displaying all forms of geographically
referenced information. Some GIS programs include
5.14.1 WMO publications
geographical interpolation capabilities such as
cokriging and geographically weighted regression World Meteorological Organization, 1966: Some
tools. Methods in Climatological Analysis (WMO/TN-
No. 81, WMO-No. 199), Geneva.
Interactive analysis tools combine the power of statis- ———, 1981: Selection of Distribution Types for Extremes
tical analysis and the ability to visually manage the of Precipitation (OHR-No. 15, WMO/TN-No. 560),
conditions for any particular statistical analysis. Tools Geneva.
allow the visual selection of values to be included in ———, 1986: Manual for Estimation of Probable
or excluded from analyses, and recalculation based Maximum Precipitation (OHR-No. 1, WMO/TN-
upon these selections. This flexibility is useful, for No. 332), Geneva.
example, in trend calculations when climate data ———, 1990: Extremes and Design Values in Climatology
series contain outliers and other suspicious points. (WCAP-No. 14, WMO/TD-No. 386), Geneva.
5–16 GUIDE TO CLIMATOLOGICAL PRACTICES

———, 1990: On the Statistical Analysis of Series of Dixon, K.W. and M.D. Shulman, 1984: A statistical
Observations (WMO/TN-No. 143,WMO- evaluation of the predictive abilities of climatic
No. 415), Geneva. averages. J. Clim. Appl. Meteorol., 23:1542–1552.
———, 1994: Guide to the Applications of Marine Environmental Systems Research Institute (ESRI),
Climatology (WMO-No. 781), Geneva. 2008: ArcGIS Geographical Information System.
———, 1997: Progress Reports to CCl on Statistical Methods Redlands, ESRI.
(WCDMP-No. 32, WMO/TD-No. 834), Geneva. Fisher, R.A. and L.H.C. Tippet, 1928. Limiting forms
———, 1999: Proceedings of the Second Seminar for of the frequency distribution of the largest or
Homogenization of Surface Climatological Data smallest member of a sample. Proc. Cambridge
(Budapest, Hungary, 9–13 November 1998) Philos. Soc., 24:180–190.
(WCDMP-No. 41, WMO/TD-No. 962), Geneva . Frich, P., L.V. Alexander, P. Della-Marta, B. Gleason,
———, 2002: Guide to the GCOS Surface and Upper- M. Haylock, A.M.G. Klein Tank and T. Peterson,
Air Networks: GSN AND GUAN (GCOS-No. 73, 2002: Global changes in climatic extreme
WMO/TD-No. 1106), Geneva. events during the second half of the
———, 2003: Guidelines on Climate Metadata and 20th century. Climate Res., 19:193–212.
Homogenization (WCDMP-No. 53, WMO/TD- Gabor, D, 1946: Theory of communication. J. IEEE,
No. 1186), Geneva. 93:429–457.
———, 2008: Deficiencies and Constraints in the Gandin, L., 1963: Objective Analysis of Meteorological
DARE Mission over the Eastern Mediterranean: Fields. Leningrad, Hydrometeoizdat, in Russian
MEDARE Workshop Proceedings (WCDMP- (English translation: Israel Program for Scientific
No. 67, WMO/TD-No. 1432), Geneva. Translation, Jerusalem, 1965).
———, 2009: Guidelines on Analysis of Extremes in a Haylock, M.R., N. Hofstra, A.M.G. Klein Tank,
Changing Climate in Support of Informed Decisions E.J. Klok, P.D. Jones and M. New, 2008: A
for Adaptation (WCDMP-No. 72, WMO/TD- European daily high-resolution gridded dataset
No. 1500), Geneva. of surface temperature and precipitation for
1950–2006. J. Geophys. Res., 113: D20119,
DOI:10.1029/2008JD010201.
5.14.2 Additional reading
Heino, R., 1994: Climate in Finland During the Period
Alexandersson, H., 1986: A homogeneity test applied of Meteorological Observations. Contributions,
to precipitation data. Int. J. Climatol., 6:661–675. No. 12. Helsinki, Finnish Meteorological
Angel, J.R., W.R. Easterling and S.W. Kirtsch, 1993: Institute.
Towards defining appropriate averaging periods Intergovernmental Panel on Climate Change
for climate normals. Climatol. Bull., 27:29–44. (IPCC), 1997: An Introduction to Simple Climate
Bénichou, P. and O. Le Breton, 1987: Prise en Models used in the IPCC Second Assessment Report:
compte de la topographie pour la cartographie IPCC Technical Paper II (J.T. Houghton,
des champs pluviométriques statistiques. La L.G. Meira Filho, D.J. Griggs and K. Maskell,
Météorologie, 7(19):23–34. eds). Geneva, IPCC.
Burrough, P.A., 1986: Principles of Geographical International ad hoc Detection and Attribution
Information Systems for Land Resource Assessment. Group (IDAG), 2005: Detecting and attributing
Monographs on Soil and Resources Survey external influences on the climate system: A
No. 12. Oxford, Oxford University Press. review of recent advances. J. Climate,
Cook, E.R., K.R. Briffa and P.D. Jones, 1994: Spatial 18:1291–1314.
regression methods in dendroclimatology: A Jenkinson, A.F., 1955: The frequency distribution
review and comparison of two techniques. Int. of the annual maximum (or minimum) values
J. Climatol., 14:379–402. of meteorological elements. Q. J. R. Meteorol.
Daley, R., 1993: Atmospheric Data Analysis. New Soc., 81:158–171.
York, Cambridge University Press. Jones, P.D., E.B. Horton, C.K. Folland, M. Hulme
Dalgaard, P., 2002: Introductory Statistics with R. New and D.E. Parker, 1999: The use of indices to
York, Springer. identify changes in climatic extremes. Climatic
Daly, C., R.P. Neilson and D.L. Phillips, 1994: A Change, 42:131–149.
statistical-topographic model for mapping Jones, P.D., S.C.B. Raper, B. Santer, B.S.G. Cherry,
climatological precipitation over mountainous C. Goodess, P.M. Kelly, T.M.L. Wigley,
terrain. J. Appl. Meteorol., 33:140–158. R.S. Bradeley and H.F. Diaz, 1985: A Grid Point
Della-Marta, P.M. and H. Wanner, 2006: Method of Surface Air Temperature Data Set for the Northern
homogenizing the extremes and mean of daily Hemisphere, DOE/EV/10098-2. Washington,
temperature measurements. J. Climate, DC, United States Department of Energy.
19:4179–4197. Jureckova, J. and J. Picek, 2006: Robust Statistical Methods
Deutsch, C.V. and A.G. Journel, 1998: GSLIB: with R. Boca Raton, Chapman and Hall/CRC.
Geostatistical Software Library and User’s Guide. Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins,
Second edition. Oxford, Oxford University Press. D. Deavan, M. Iredell, S. Saha, G. White,
 CHAPTER 5. �STATISTICAL METHODS FOR ANALYSING DATASETS 5–17

J. Woolen, Y. Zhu, A. Leetmaa, R. Reynolds, Sen, P.K., 1968: Estimates of the regression coeffi-
M. Chelliah, W. Ebisuzaki, W. Higgins, cient based on Kendall’s tau. J. Amer. Stat. Assoc.,
J. Janowiak, K.C. Mo, C. Ropelewski, J. Wang, 63:1379–1389.
R. Jenne and D. Joseph, 1996: The NCEP/NCAR Sensoy, S., T.C. Peterson, L.V. Alexander and
40-year reanalysis project. Bull. Amer. Meteorol. X. Zhang, 2007: Enhancing Middle East climate
Soc., 77:437–471. change monitoring and indices. Bull. Amer.
Kaplan, A., Y. Kushnir, M. Cane, and M. Blumenthal., Meteorol. Soc., 88:1249–1254.
1997: Reduced space optimal analysis for histori- Shabbar, A. and A.G. Barnston, 1996: Skill of
cal datasets: 136 years of Atlantic sea surface seasonal climate forecasts in Canada using
temperatures. J. Geophys. Res., 102:27835–27860. canonical correlation analysis. Monthly Weather
Karl, T.R., C.N. Williams Jr and P.J. Young, 1986: A Rev., 124(10):2370–2385.
model to estimate the time of observation bias Torrence, C. and G.P. Compo, 1998. A practical
associated with monthly mean maximum, mini- guide to wavelet analysis. Bull. Amer. Meteorol.
mum, and mean temperatures for the United Soc., 79:61–78.
States. J. Clim. Appl. Meteorol., 25:145–160. Ulbrich, U. and M. Christoph, 1999. A shift of the
Kharin, V.V. and F.W. Zwiers, 2005: Estimating NAO and increasing storm track activity over
extremes in transient climate change simula- Europe due to anthropogenic greenhouse gas
tions. J. Climate, 18:1156–1173. forcing. Climate Dynamics, 15:551–559.
Kohler, M.A., 1949: Double-mass analysis for testing Uppala, S.M. P.W. Kållberg, A.J. Simmons,
the consistency of records for making adjust- U. Andrae, V. da Costa Bechtold, M. Fiorino,
ments. Bull. Amer. Meteorol. Soc., 30:188–189. J.K. Gibson, J. Haseler, A. Hernandez, G.A. Kelly,
Mann, M. E., 2004: On smoothing potentially non- X. Li, K. Onogi, S. Saarinen, N. Sokka, R.P, Allan,
stationary climate time series. Geophys. Res. E. Andersson, K. Arpe, M.A. Balmaseda,
Lett., 31: L07214, DOI:10.1029/2004GL019569. A.C.M. Beljaars, L. van de Berg, J. Bidlot,
National Climatic Data Center (NCDC), 2006: N. Bormann, S. Caires, F. Chevallier, A. Dethof,
Second Workshop on Climate Data Homogenization, M. Dragosavac, M. Fisher, M. Fuentes,
National Climatic Data Center (Asheville North S. Hagemann, E. Hólm, B.J. Hoskins, L. Isaksen,
Carolina, 28–30 March 2006). Jointly organized P.A.E.M. Janssen, R. Jenne, A.P. McNally,
by Climate Research Division, Environment J.-F. Mahfouf, J.-J. Morcrette, N.A. Rayner,
Canada, and National Climatic Data Center, R.W. Saunders, P. Simon, A. Sterl, K.E. Trenberth,
NOAA. Asheville, NCDC. A. Untch, D. Vasiljevic, P. Viterbo and
Peterson, T.C., D.R. Easterling, T.R. Karl, P. Groisman, J. Woollen, 2005: The ERA-40 reanalysis.
N. Nicholls, N. Plummer, S. Torok, I. Auer, Q. J. R. Meteorol. Soc., 131:2961–3012.
R. Boehm, D. Gullett, L. Vincent, R. Heino, Vincent, L.A., X. Zhang, B.R. Bonsal, and W.D. Hogg,
H. Tuomenvirta, O. Mestre, T. Szentimrey, 2002: Homogenization of daily temperatures
J. Salinger, E.J. Forland, I. Hanssen-Bauer, over Canada. J. Climate, 15:1322–1334.
H. Alexandersson, P. Jones, and D. Parker, 1998: von Storch, H. and F.W. Zwiers, 1999: Statistical
Homogeneity adjustments of in situ atmospheric Analysis in Climate Research. Cambridge,
climate data: a review. Int. J. Climatol., Cambridge University Press.
18:1493-1517. Wackernagel, H., 1998: Kriging, cokriging and
Peterson, T.C., C. Folland, G. Gruza, W. Hogg, external drift. In: Workshop on Dealing with
A. Mokssit and N. Plummer, 2001: Report on the Spatialisation (B. Gozzini and M. Hims, eds).
Activities of the Working Group on Climate Change Toulouse, European Union COST Action 79.
Detection and Related Rapporteurs 1998–2001. Wahba, G. and J. Wendelberger, 1980: Some new
International CLIVAR Project Office (ICPO) mathematical methods for variational objective
Publication Series, 48. Southampton, United analysis using splines and cross-validation.
Kingdom, ICPO. Monthly Weather Rev., 108:36–57.
Palus, M. and D. Novotna, 2006. Quasi-biennual Wang, X.L. and V. R. Swail, 2001: Changes of
oscillations extracted from the monthly NAO extreme wave heights in Northern Hemisphere
index and temperature records are phase-synchro- oceans and related atmospheric circulation
nized. Nonlin. Process. Geophys., 13:287–296. regimes. J. Climate, 14:2204–2221.
Peterson, T. C. et al., 2002. Recent changes in Wang, X.L., 2008: Accounting for autocorrelation
climate extremes in the Caribbean Region. in detecting mean shifts in climate data series
J. Geophys. Res., 107(D21):4601. using the penalized maximal t or F test. J. Appl.
Philipp, A., P.M. Della-Marta, J. Jacobeit, Meteorol. Climatol., 47:2423–2444.
D.R. Fereday, P.D. Jones, A. Moberg and Wilks, D.S., 2002: Statistical Methods in the
H. Wanner, 2007: Long-term variability of daily Atmospheric Sciences. Second edition. New York,
North Atlantic–European pressure patterns Academic Press.
since 1850 classified by simulated annealing Yue, S., P. Pilon and G. Cavadias, 2002: Power of the
clustering. J. Climate, 20:4065–4095. Mann-Kendall and Spearman’s rho tests for
5–18 GUIDE TO CLIMATOLOGICAL PRACTICES

detecting monotonic trends in hydrological Al-Shabibi, Z. Al-Oulan, Taha Zatari, I. Al Dean

series. J. Hydrol., 259:254–271. Khelet, S. Hammoud, M. Demircan, M. Eken,
Zhang, X., E. Aguilar, S. Sensoy, H. Melkonyan, M. Adiguzel, L. Alexander, T. Peterson and Trevor
U. Tagiyeva, N. Ahmed, N. Kutaladze, Wallis, 2005: Trends in Middle East climate
F. Rahimzadeh, A. Taghipour, T.H. Hantosh, extremes indices during 1930–2003. J. Geophys.
P. Albert, M. Semawi, M. Karam Ali, M. Halal Said Res., 110: D22104, DOI:10.1029/2005JD006181.

.
 CHAPTER 6

SERVICES AND PRODUCTS

6.1 INTRODUCTION They include diverse groups such as media and

public information personnel, farmers, defence
Climate services are the dissemination of climate infor- forces, government departments, business and
mation to the public or a specific user. They involve industry personnel, water and energy managers,
strong partnerships among NMHSs and stakeholders, consumers, tourists, legal professionals, health
including government agencies, private interests and officials, humanitarian and relief organizations,
academia, for the purpose of interpreting and applying and meteorological services. Their needs may be a
past climate information for decision-making, for simple interest in the weather and climate, a school
sustainable development, and for the improvement of project, the design of a building, agricultural
climate information products, predictions and operations, water management, operating an air
outlooks. Public and private sector partnerships conditioning system or a large dam, the planning
encourage the coordination of climate services from of energy production and distribution, or the
local to national to international scales. When the preparation for and response to food or water
roles of each partner are clearly defined, products shortages.
should be provided to users in a timely and efficient
manner. Climate information should be relevant to the The current interest in climate change and in the
social and economic interests of the recipient and impacts of climate change has brought about an
provide the ability to educate consumers more broadly additional need for climate information. In the
on the uses and benefits of climate data and products. past, climate products were mostly limited to infor-
The partnerships also facilitate the coordination of mation about the physical environment of the
applied research, monitoring and prediction across atmosphere near the Earth’s surface. Today, users
national, regional and global levels. may want information about many aspects of the
broader climate and Earth system (solid earth, air,
The three fundamental principles for providing sea, biosphere and land surface, and ice). Data
climate services are: related to climate are now used to describe, repre-
(a) Know the user and understand what is needed, sent and predict both the behaviour of the whole
such as: the climatic elements that are relevant climate system (including the impact of humans on
to the user, how the user wishes to receive climate) and the relationship of climate to other
information, how the user is likely to interpret aspects of the natural world and human society.
the information, for what purpose the infor- These interests have strengthened the need to
mation will be used, the decision process of the monitor and describe the climate in detail in terms
user, and how the information might improve of both space scales and timescales, while placing
the decision-making processes. current events into a range of historical perspec-
(b) Make the information simple, accessible and tives. High-quality baseline climate datasets are
timely: provide products that can be understood being compiled for which complex statistical qual-
and readily applied by the user, along with easy ity assurance techniques have been devised. The
access to follow-up professional advice. compilations include processes to detect and, if
(c) Ensure quality: provide products that have possible, correct data for outliers, and to account
been developed with an understanding of for inhomogeneities such as changes in instrumen-
possible applications and analysis techniques, tation or location of observing stations. Researchers
complete with proper documentation and and other expert users of climate products are also
backed by thorough knowledge of up-to-date keenly interested in metadata to help interpret and
data availability and characteristics. analyse data homogeneity and quality.

Each of these principles, as well as the marketing of The uses of climatological information can be
services, is discussed in more detail in this chapter. classified into two broad categories, strategic and
tactical. Strategic uses refer to products that aid in
the general long-term planning and design of groups
of projects and policies. The types of information
6.2 USERS AND USES OF that are usually required for strategic uses are
CLIMATOLOGICAL INFORMATION probability analyses and risk assessments of
meteorological events for design specifications and
The users of climate services are many and varied, regulations, summaries of historical conditions as
ranging from schoolchildren to global policymakers. background information about past climate
6-2 GUIDE TO CLIMATOLOGICAL PRACTICES

conditions, and climate scenarios as indicators of Climate information can be presented in many
future expectations. An example of a strategic use is ways, such as by data files, time series, tables,
an analysis of climate data for designing a dam. diagrams, statements or maps. It can be delivered via
Tactical uses refer to products and data that aid in a number of methods, ranging from telephone,
solving short-term, singular, immediate problems. facsimile and mail to Internet file transfer protocol
Typical information provided for tactical uses (ftp), e-mail and Website access. Because of the vari-
includes copies of official observations of the ety of presentation and delivery methods, the
occurrence of a meteorological event, summaries of development of a climate service must determine
historical data, and the placement of an event into from the outset what the user of the information
historical context. An example of a tactical use is the needs. The requirements can often be determined by
analysis of recent observational data to assist in surveying the users and by analysing records of user
managing the use of water during a drought. requests. The demands of the users usually dictate
Sometimes the use crosses from one realm to the level of service and nature of products that are
another. For example, the calculation of probabilities developed. Some general considerations for develop-
of the speed and direction of movement of tropical ing climate product and service programmes are:
storms from historical storm data is a strategic use. (a) Providing the best services at the least cost;
But the same information, when used in forecasting (b) Timely delivery of products;
the movement of a current storm, is a tactical use. (c) Enhancing user awareness of the available
climate products and services;
The use of climate information may include, but is (d) Encouragement of user feedback and use of
not limited to: the feedback to improve products;
(a) Monitoring of specific activities that are driven (e) Compliance with quality management
by meteorological conditions (for example, principles;
fuel consumption for heating and cooling, air (f) Information security;
pollution levels crossing thresholds, variabil- (g) Availability of relevant documentation and
ity of sales of goods, drought-related inflation metadata;
in commodity markets); (h) Transparency regarding the reliability and
(b) Prediction of the behaviour of sectoral uncertainty associated with the products;
systems that react to meteorological events (i) Compliance with Resolution 40 of the Twelfth
with a known response time, and in which World Meteorological Congress and Resolu-
the knowledge of recent past weather and tion 25 of the Thirteenth World Meteorologi-
climate allows some forecasting of sectoral cal Congress relating to the exchange of data
impacts (for example, energy production and and products (see section 3.5);
consumption, anticipation of restocking of (j) Compliance with other local, national and
goods, crop production, plant diseases, heat- international laws and treaties.
health warning systems, food security, water
supply and demand); It is important to prepare comprehensive documenta-
(c) Certification for insurance or other purposes tion for the information access system as well as for
that a meteorological event such as a thunder- the products. The time invested in preparing adequate
storm, high winds, frost or drought occurred; documentation will save resources later when answer-
(d) Monitoring to identify the character of a given ing user questions, and it will help minimize
event or period, especially deviations from inappropriate use of products and facilitate regular
normal (for example, intensity of extreme maintenance and updating of the system.
rainfall or dryness);
(e) Design of equipment for which the knowl- Each product should have a statement concerning
edge of the local climatology is critical to the confidence that can be placed in the product
effectiveness and efficiency (such as civil engi- and guidance concerning limitations of the infor-
neering works, air conditioning systems, and mation so that all users, especially those who are
irrigation and drainage networks); less accustomed to climatological products, can
(f) Impact studies, such as knowledge of initial assess the appropriateness of a certain product to
conditions in order to assess the consequences their needs and incorporate the associated uncer-
of installing a power plant or other industrial tainty into their own decision-making process.
enterprise that could affect air quality;
(g) Study of the influence of meteorological
conditions on economic sectors, such as
public transportation and tourism; 6.3 INTERACTION WITH USERS
(h) Planning and risk management for providing
community services to society, such as water, There can be a strong educational and communica-
emergency preparedness and response, and tion aspect to the provision of climate services.
energy. Many people with a need to use climate data have
 CHAPTER 6. �SERVICES AND PRODUCTS 6-3

little understanding of meteorological science and together with information about service standards
related concepts. Thus, they may not know what such as hours of operation or when a return e-mail,
information they really need or how best to use the telephone call or letter can be expected.
information. Many users of digital information
may not even know how to import the information It is important to offer a variety of methods for deliv-
into their own computer systems. ering information. The service providers may have an
up-to-date and powerful system backing their service,
The climate service should ensure that information but many users may not. Modern communications
requested by a user is truly the information that is and technology can now transfer and deliver data and
needed to solve the user’s problem. Often, users products very quickly via the Internet (using e-mail,
simply request products that they know to be avail- ftp or the World Wide Web), but in most countries
able. The service should query the users about their there are users who do not have ready access to these
problems, discuss the kind of information needed facilities. Alternative methods of delivery will be
to solve the problems, and recommend the prod- required in many countries, and the NMHSs
ucts that are most appropriate. concerned must take this into account. While the
systems behind the service may be the most modern
The climate service should ensure that expertise in technology and design available, methods of delivery
the communication, interpretation and use of and the products supplied should be developed with
climate information is available. The service person- sensitivity to the needs of the local users and to the
nel are the link between the technical climatological ways these needs change with time.
aspects of the data, analyses and scenarios, and users
of the information who may not be technically An NMHS often has multiple offices, and inconsisten-
proficient. Climate service staff should be prepared cies in standards, formats and even products can arise
to respond to requests for data with additional infor- among offices. Frequent liaison among offices is vital;
mation about sites, elements and instrumentation, face-to-face meetings at regular intervals are recom-
mathematical definitions of various parameters, the mended, and they should be scheduled at least once a
many aspects of how observations are performed, year. Centralized and commonly applicable opera-
and the science of meteorology and climatology in tional procedures and single reference databases help
particular. Climate service personnel should culti- ensure consistency, as do well-written documenta-
vate a broad range of skills and expertise or have tion, quality standards, instructions and manuals.
access to people with the necessary expertise.
When fees are charged for providing information,
Users sometimes organize meetings and activities to this can be an especially sensitive area for users and
which climate service personnel are invited. A posi- a cause of criticism and dissatisfaction if they are not
tive response to these invitations builds stronger applied consistently or if they are seen as unreason-
relationships and gives the service personnel the ably high. Therefore, it is important to establish a
opportunity to listen and learn about users and their clear and transparent service pricing policy with
problems. It is important and very rewarding to be instructions for implementing the policy. This may
involved in users’ activities as much as possible; feed- involve a formal policy underpinned by a set of prin-
back from the users usually leads to better products, ciples, and a series of practical charges for direct use
new applications, and more efficient and inclusive by those providing the services operationally. The
dissemination of climate information. Continuous or policy should be made available to all who are
frequent communication with users is essential for involved in providing services, as well as to the users.
ensuring that existing products still meet the require-
ments of the users and for determining what changes The interaction with users should be part of a quality
need to be made to products to satisfy the users. management framework that organizes the manage-
ment of a climate service based on the needs of
Customer service personnel need to enjoy dealing customers. Quality management principles adopted
with people. They should be courteous and tactful by the WMO Inter-Commission Task Team on the
and recognize the importance of timely services; Quality Management Framework (WMO, 2008) and
user needs may be urgent for many reasons. Ideally, several NMHSs should be followed.
the climate service should have good basic commu-
nications with technological and training facilities
supporting the customer service personnel.
Personnel providing climate services are the people 6.4 INFORMATION DISSEMINATION
with whom the public directly interacts, the people
on whom the service provider’s reputation depends. How well a user interprets a collection of climatological
Clearly advertised contact points should be availa- information depends very much on how the
ble for a variety of communication methods, such information is presented. Where practical to do so,
as phone, facsimile, e-mail, mail and personal visits, the salient facts should be shown visually with
6-4 GUIDE TO CLIMATOLOGICAL PRACTICES

accompanying text used to qualify, emphasize and conditions is the vision of the World Meteorological
explain. The presentation should be logical, clear, Organization’s Climate Information and Prediction
concise and tailored to suit the user and the aims of Services (CLIPS) Programme, as detailed in the Report of
the presentation. For example, the style used to the Meeting of Experts on Climate Information and
impart climate information to a researcher will be Prediction Services (WMO/TD-No. 680). The Programme
different from that of an article for a newspaper or emphasizes the development and implementation of
popular magazine. applications from the viewpoint of the users. Particular
targets include climate services for water resources
The techniques used to summarize data in a routine management, food production, human health, urban-
monthly climatological bulletin, intended for a wide ized areas, energy production and tourism. The main
range of technical and non-technical users, should be thrusts of the Programme are climate services for
different from those utilized to prepare an interpreta- sustainable development, studies of climate impact
tive report on some specific problem in applied assessments and response strategies to reduce vulnera-
climatology. Technical information should be bility, new frontiers in climate science and prediction,
presented to users with a limited knowledge of atmos- and dedicated observations of the climate system.
pheric sciences in a manner that is simple and
understandable, while remaining scientifically correct. Consistent with these thrusts, the CLIPS Programme
More knowledgeable users will be better able to under- addresses the objectives of:
stand complex information and presentations, but (a) Providing an international framework to
some distillation of technical information is usually enhance and promote economic, environmen-
desirable. The climate service must, therefore, be able tal and social benefits from climate information
to anticipate, investigate and understand the require- and predictions;
ments of government decision-makers, industrial and (b) Facilitating the development of a global
commercial interests, and the general public. It must network of regional and national climate
ensure that the customer service personnel are able to centres, as well as Regional Climate Outlook
understand the issues and respond to questions with Forums, to facilitate consensus and common
current knowledge and professional skill. understanding of regional climate outlooks,
including communications and training, and
Information about the past, current and predicted to act as a focus for providing climate infor-
states of the climate is significant in the develop- mation and prediction services;
ment of national policy and strategies. Providing (c) Demonstrating the value and ultimate social
relevant information for use in preparing and and economic benefits of climate information
implementing national policy and strategies places and prediction services, and the connection
several demands on the information personnel: of those benefits with global observing, moni-
(a) Historical information must be collected, toring, prediction and applications;
subjected to quality control, archived and (d) Encouraging development of operational
made accessible in a timely manner; climate predictions that are directed towards
(b) Assessments of the climate data and infor- user-oriented applications for feasible periods
mation must be related to the needs of the and areas.
decision-makers and those responsible for
implementing decisions; National Meteorological and Hydrological Services
(c) Interpretations and presentations of climatic can develop and maintain their climate informa-
data, information and scenarios, and the degree tion dissemination systems by closely following the
of confidence in the interpretations, must be CLIPS framework, and also by making optimal use
meaningfully communicated to users who may of the available global and regional entities. It is
not be technically knowledgeable about climate; desirable that NMHSs facilitate the development of
(d) Coordination with other public agencies, national climate outlook forums involving the
academic institutions, private interest groups representatives of core climate-sensitive sectors to
and programmes is often necessary to answer interpret the available global and regional informa-
multidisciplinary questions concerning national, tion in the national context, and to enhance
sectoral and community vulnerability related to two-way feedback with the user community.
climate variability and change.

Meeting these demands requires that there be suita-

ble and effective liaison and formal linkages with 6.5 MARKETING OF SERVICES AND
other government departments with climate inter- PRODUCTS
ests, as well as with international climate activities.
Marketing is not simply advertising and selling; it
Utilizing climate information and predictions in order also allows potential users to learn what services
to provide the best possible information about future and products are available, realize the utility of the
 CHAPTER 6. �SERVICES AND PRODUCTS 6-5

services and products, and gain an understanding quality of the services being received, or even
of the value of the information, as discussed in exactly what they are receiving or the appropriateness
Operational Climatology – Climate Applications: On of the information. An important part of marketing
Operational Climate Services and Marketing, involves the identification and understanding of
Information and Publicity (WMO/TD-No. 525). the strengths of the professional personnel
Marketing, public relations, and advertising, underpinning the service and informing users of
promoting and disseminating climate information those strengths.
are essential to the success of most climate services.
Climate information is unlikely to be sufficient in Characteristics of an effective marketing programme
its own right, but is often vital in assessing the include:
viability of food, water and economic projects. It (a) Focusing on user needs by gaining a clear
has social and economic value, and marketing understanding of the user’s problems and
success depends in large measure on how well the requirements and how the climate informa-
promotion and dissemination processes are tion is used;
addressed. The benefits of a product must be sold, (b) Training customer service personnel to
not the product itself. become attuned to customer needs and wants;
(c) Selecting a target market;
The relevance of climate information often is not (d) Promoting the benefits of climate services and
apparent to those who are in a position to benefit products to the target sector;
from the products or to those who approve funds (e) Developing a product or service for a need of
for climate programmes. The benefits and worth of the user and promoting its application to solv-
using the products must be clearly demonstrated in ing the user’s problems;
social and economic terms. Studies are needed to (f) Promoting the professional skills of climate
assess the value of climatological applications. Such service personnel;
studies should not be the sole responsibility of the (g) Deciding on methods of product accessibility
applications climatologist, but rather a shared or delivery and making alternatives available
responsibility with economists, social statisticians to the user;
and specialists in the field of application. One way (h) Evaluating the economics of the products and
to show the effectiveness of a product is by demon- services;
strating the value of the product to the users. An (i) Informing users through promotion and
analysis of the costs of a product or service should public relations;
include not just the direct costs, but also the poten- (j) Monitoring user satisfaction and assessing
tial cost of not applying the product. service performance and marketing efforts;
(k) Ensuring credibility of climate services by
A marketing strategy should gauge user satisfaction. being transparent about the reliability and
User surveys allow an NMHS to evaluate if the serv- limitations of the products and services
ices supplied are those that are needed, and if the offered.
users are engaged in effective application of the
information supplied. Survey results also provide
information on the value, utility and appropriate-
ness of current services; this information allows the 6.6 PRODUCTS
refinement of existing services and the design of
new ones. Survey results form an information base Climate products are information packages that
from which service managers can make financial, include data, summaries, tables, graphs, maps,
investment, policy and programme decisions. User reports and analyses. Spatial distributions may be
surveys should be conducted at least every three shown on maps. More complex products, such as
years. Surveys will require resources both for design climate atlases or analyses, may combine several
of the survey and for analysis of the results, as well kinds of visualization with descriptive text. There
as for development of new products or revisions to also may be databases with software tools that allow
existing products that the surveys indicate are online customers to produce statistics and
needed. New surveys should always be tested on a visualizations according to their own needs.
sample reference group before being distributed to
the full list of recipients or released as a more open
6.6.1 General guidelines
Internet-based survey.
Products and the data upon which they are based
A climate service should market the knowledge and should be of the highest quality possible within the
professional skills of its personnel. Users frequently time constraints for providing the information.
choose to buy a service based on the faith in and There has been a strong and increasing requirement
credibility of the provider’s professional judgment. for climate-related products to be provided as
Users often have limited understanding of the quickly as possible after the aggregating period.
6-6 GUIDE TO CLIMATOLOGICAL PRACTICES

Maintaining the quality standards for such prod- climatological data periodicals are issued on either a
ucts is a concern. The short time between monthly or annual basis. Some services also publish
observation and delivery to a user leaves little or no periodicals for different intervals such as a week or a
time for quality control of the data other than that season, however. Weekly or monthly publications
which can be done automatically. At the very least are issued immediately after the close of the period
some basic checks should be made as the data are in question and usually contain recent data that
received (see Chapter 3). Users must be alerted to have not been subject to complete quality control
possible problems concerning the data, and since procedures. These periodicals contain timely data
these products are usually automatically delivered, that can be of great importance to various economic,
these alerts should be included with the products. A social and environmental sectors, and so publication
proper quality assurance system will provide a is valuable even though the data may contain a few
framework that allows this information to be errors and omissions. Quarterly or seasonal data
handled along with the data. periodicals are often issued to disseminate
summarized seasonal data such as winter snowfall,
Products concerning historical data should be of growing-season precipitation, summer cooling
higher quality than those using very recent data. All degree-days and winter degree-days.
data that contribute to the climate record should be
checked for random and systematic errors, homoge- Most NMHSs issue monthly bulletins containing
neity, spatial representativeness and gaps in time data from a selection of stations within particular
series. For products such as climatic atlases or techni- areas or states or the country as a whole. When
cal regulations, data should be for a standard reference issued a week or two after the end of each month,
period (see section 4.8). Frequent revisions based on these periodicals will usually contain recent data
new periods of record should be avoided. If some that may not have been subject to full quality
content of a product is not stable over a long time control, but if issued a month or more afterwards,
period, there needs to be additional information all data should meet the normal quality control
describing the nature of the variability or change. standards for historical climatological data.
Maximum and minimum temperature and total
The value of historical and statistical climatological precipitation for each day should be listed, as well
data tables can usually be improved by the inclu- as perhaps temperatures at fixed hours, together
sion of a supporting text that serves to interpret the with the associated humidity values. Daily mean
data to the user and to emphasize the more impor- wind speed and prevailing direction, duration of
tant climatological elements. In all publications, bright sunshine, or other locally important data
sufficient information and data must be included (such as heating, cooling and growing degree-days)
regarding the location and elevation of the observ- could be included as well. Monthly averages,
ing stations, the homogeneity of the data from all extremes and other statistical data from all stations
stations, the periods of record used, and the statisti- also should be included when available.
cal or analytical procedures employed.
While most monthly bulletins contain only surface
The display of products should be checked carefully climatological data, some NMHSs include a selec-
before they are made accessible to potential users. tion of basic data from upper-air stations or issue
For example, climatic maps should be well designed separate monthly bulletins containing upper-air
with carefully chosen colours and scales, clear titles data. In such monthly bulletins, daily and monthly
and notations on the map of what is being analysed, mean data are usually published for the standard
identification of the data period of record, and a pressure surfaces. The data usually include altitude
listing of the responsible organizations. There (in geopotential meters), temperature, humidity,
should be reasonable consistency among maps (in and wind speed and direction for one or two sched-
terms of colours, layout and data) to allow for easy uled ascents each day.
comparisons.
Some of the most useful climatological publications
Consultation with all who are affected by environ- are those containing simple tables of monthly and
mental information services is encouraged. Input annual values of mean daily temperature and total
from interested parties should be considered when precipitation. Such tables are prepared by NMHSs and
creating, modifying or discontinuing products and made available either in manuscript or electronic
services. format. The services should publish, at least once a
decade, a comprehensive set of statistical climatologi-
cal data for a selection of representative stations.
6.6.2 Climatological data periodicals

A periodical climatological publication is one that is Periodicals sponsored by WMO include data from
scheduled for preparation and publication on a Member countries. Examples are Monthly Climatic
routine basis over set time intervals. Most Data for the World (data from all CLIMAT stations),
 CHAPTER 6. �SERVICES AND PRODUCTS 6-7

World Weather Records (single-station, historical, usually beneficial to develop a standard product
monthly and annual values of station pressure, sea that can be used by a wide range of users. For exam-
level pressure, temperature and precipitation), and ple, both energy management entities and fruit
Marine Climatological Summaries (monthly, annual growers can make use of a degree-day product.
and decadal climatological statistics and charts for When the content, format and design of a product
the oceans). are carefully chosen, the development costs can be
spread across many users. Such standard products
fill the gap between the climate data periodicals
6.6.3 Occasional publications
and those tailored for individual users. Standard
Unlike climate data periodicals, which are produced products should be locally developed to meet the
to a schedule, occasional publications are produced needs of groups of users.
as the need arises. They are in a form that will
satisfy a large number of users for a considerable Increasingly, products are being requested and
time, so they will not need frequent updating. delivered using the Internet. The user interface to
Occasional publications are designed for those users these systems can be considered another product of
who need information in planning for capital the climate service, and standardizing that inter-
investments or in designing equipment and build- face can be seen as enhancing the quality and
ings to last for decades and centuries; for members utility of the product.
of the general public whose interests are academic
or casual; and for researchers in the atmospheric
6.6.5 Specialized products
and oceanic sciences. They are also designed to
summarize or explain unusual events, such as It is often necessary to develop products that are
extreme weather, and to describe or update an specific to an individual user or sector. The particu-
important predicted event such as a strong El Niño. lar requirements of one user group do not always
The content and format of a specific occasional match the requirements of other groups, so the
publication must reflect the interests and needs of expense of publishing the product for general avail-
the users for whom it is published. ability is not warranted.

Long-term, continuous and homogeneous series of Such applied climatological products are tailored to
data are of great value for comparative climatologi- the needs of a particular user or user group. These
cal studies and for research on climatic fluctuations, products provide a bridge between the observed
trends and changes. Several NMHSs have published data and the specific requirements of a user; they
such series for a selection of stations where observa- transform the observations into a value added
tional practices and the environmental surroundings product that has utility for the particular recipients
have remained essentially unchanged over long of the product. Developing these products involves
periods of time. Data series most commonly avail- analysing the data and presenting the information
able and needed are those of temperature and with a focus on the specifications that will enable
precipitation, although data for wind, pressure, the user to gain optimum benefit from the applica-
bright sunshine, cloudiness and other climatic tion of the information. The use of the product
elements might also be published. Some NMHSs usually dictates the types of analyses and data trans-
include historical climatological data series in year- formations that need to be performed and the
books or other annual bulletins. Monographs on methods by which the product is delivered.
the climate of a country or area are valuable to a
wide range of users and should be published and The climate service should be able to accept requests
updated periodically. It is recommended that the for specialized products and develop the products
publications and data also be available in electronic to the satisfaction of the users, which will require
format for ease of access and exchange. all of the skills of user interaction and marketing
already discussed. Although the product may not
The collection of maps in atlas format is another be published for a general audience, the users will
valuable occasional publication. Legends and expect at least the same level of quality, both in
captions on climatic maps should include precise content and presentation.
information regarding the element mapped, some
indication of the number of stations from which An example of an application-driven product can
data have been obtained, and the period of record be found in the requirement by a fruit grower for
used to generate each map or diagram. daily degree-hour data for pesticide management of
fire blight disease. When only daily maximum and
minimum temperatures are available for the loca-
6.6.4 Standard products
tions of interest, degree-days can be calculated from
Although creating products specifically tailored to the average of the maximum and minimum values.
individual users may be the best for those users, it is Since degree-hours are required but not available,
6-8 GUIDE TO CLIMATOLOGICAL PRACTICES

an analysis is necessary to develop relationships to grasp an understanding of the whole complex of

between degree-days calculated from daily temper- the climate system. For many users it is also difficult
ature extreme data and degree-hours. The to understand the connection between features of the
conditions for which the relationships are valid, global climate system and the current climate condi-
and the degree of error in the relationships, must tions within their own country. Thus, it is appropriate
also be assessed. Once the relationships are estab- for the climate service to process its own data and
lished, the user can be given a product that contains analysis results and, where possible, compile them
degree-hours, even though degree-hours are not together with the material of other agencies into a set
measured directly. of products that can be promptly disseminated with
each agency’s views on current climate conditions.
Flood analysis is another example. Flooding is a
natural feature and varies in scale from water If the necessary data are not available within a
running off a saturated hillside to large rivers burst- given country, the relevant NMHS should obtain
ing their banks. The impacts of floods range from regional or global data and analyses from foreign or
waterlogged fields and blocked roads to widespread international agencies and process the information
inundation of houses and commercial property into a form suitable for local to national use. The
and, occasionally, loss of life. Flood frequency esti- NMHS should, however, add its own views to these
mates are required for the planning and assessment global analyses about the connection between the
of flood defences; the design of structures such as local climate conditions and the large-scale climatic
bridges, culverts and reservoir spillways; and the fields. Monitoring activities require that the climate
preparation of flood risk maps for the planning of service develop expertise in analysing the state of
new developments and for insurance interests. A both past and current climate and global to regional
product that provides the probability of observed teleconnections, and provide summarized informa-
precipitation amounts is a necessary component in tion to both public and private sector users. Good
the development of flood frequency estimates. monitoring products are essential for climate
Developing the precipitation risk information predictions and updates.
involves the value added statistical analysis (see
Chapter 5) of the observed precipitation data that
6.6.7 Indices
are usually presented in standard summaries. If the
resulting risk analyses will be of use to a number of Presentation of historical climate patterns to the
different users, a general publication may be user in a simple and readily understandable form
warranted. may often be accomplished with indices (see section
4.4.6). Climate indices are widely used to character-
ize features of the climate for climate prediction
6.6.6 Climate monitoring products
and to detect climate change. They may apply to
Monitoring climate variability around the world is individual climatological stations or describe some
a goal of the World Climate Data Monitoring aspect of the climate of an area. Indices usually
Programme (WCDMP). Maintenance and accessi- combine several elements into characteristics of,
bility of climate data and information by a climate for example, droughts, continentality, phenologi-
service supports this WCDMP objective. For moni- cal plant phases, heating degree-days, large-scale
toring and diagnosing the climate of a country, it is circulation patterns and teleconnections. When
necessary to understand current climate conditions providing information to users, it is often necessary
in the country as part of the global climate system. for the climate service to interpret the meaning of
In addition to monitoring local climates for an index value, changes in values over time, and
national interests and relating current episodes to sometimes calculation procedures. Examples of
historical patterns, the climate service should aim indices are the El Niño–Southern Oscillation (ENSO)
to place the local variations within a larger regional Index; the North Atlantic Oscillation Index; descrip-
and even global context. tors such as the moisture availability index, used
for deriving crop planning strategies; agrometeoro-
Observation and monitoring of climate may be logical indices such as the Palmer Drought Severity
conducted by more than one agency in a country. Index, aridity index and leaf area index, which are
When an agency publishes observational data, analyt- used for describing and monitoring moisture avail-
ical results and statistical data, the products are usually ability; and the mean monsoon index, which
presented in formats suited for the agency’s own summarizes areas of droughts and floods. The
purposes. Hence, the products may not necessarily be construction and evaluation of indices specific to
appropriate for use by other agencies. In addition, it climate change detection, climate variability and
may not be easy for individual users to choose prod- climate extremes are ongoing processes, as discussed
ucts for their own needs from among the climate in the Report on the Activities of the Working Group on
monitoring products distributed by various agencies, Climate Change Detection and Related Rapporteurs
or to consolidate a disparate set of products in order 1998–2001 (WMO/TD-No. 1071).
 CHAPTER 6. �SERVICES AND PRODUCTS 6-9

6.7 CLIMATE MODELS AND CLIMATE monitoring of indicators such as ENSO and sea
OUTLOOKS surface temperatures in the Indian and Atlantic
oceans. The Update should be considered as comple-
The climate system, its behaviour, its components mentary to more detailed regional and national
and their interactions, and its future development seasonal climate outlooks, such as those produced
and changes can be simulated and studied using by Regional Climate Outlook Forums and NMHSs.
climate models. The growing knowledge about the
climate system and the availability of larger and Operational climate forecasting is practiced by
faster computers have led to the development of NMHSs, GPCs and other international institutions.
highly complex climate models. The methods of climate forecasting can be roughly
classified as empirical-statistical or dynamical. The
Among the simplest models, however, are those empirical-statistical methods use relationships
based on empirical or statistical relationships derived from historical data, while the dynamical
among parts of the climate system. These models methods are numerical predictions using atmos-
are used extensively in producing climate outlooks, pheric general circulation models or coupled
which give the expected average value of a climate ocean–atmosphere general circulation models
element, typically over periods of several months. (GCMs). Because numerical prediction for climate
The most complex models analyse and couple forecasting needs vast computer resources, there are
together the entire climate system for the whole only a small number of climate centres that perform
globe and are used to model climate into the future operational numerical climate predictions. Several
explicitly or under certain assumptions. Regional NMHSs make forecasts for their own countries
climate models concentrate on representing the based on products from GPCs and other interna-
climate on smaller space scales over a limited area. tional institutions, however. Regional Climate
Climate outlooks are derived from the analysis and Centres interpret and downscale global predictions
interpretation of observations and climate model to the regional context, while Regional Climate
outputs. Outlook Forums cooperatively produce consensus-
based climate outlooks in which several countries
6.7.1 Climate outlook products sharing similar climatological characteristics
participate.
Climate outlooks are forecasts of the values of
climate elements averaged over timescales of about Forecast ranges vary widely, from less than one
one month to one year. The climate elements typi- month to more than one year, although three
cally forecast are average surface air temperature months is common. Most forecasts are issued regu-
and total precipitation for a given period. Sunshine larly throughout the year, such as monthly or every
duration, snowfall, the number of occurrences of three months, while some are issued only for
tropical cyclones, and the onset and end of specific seasons, such as before the onset of the
monsoons are also forecast in some centres. Since rainy season. The forecast elements and forecast
the ENSO phenomenon has a significant impact on periods can vary in accordance with the climatic
the climate in many parts of the world, forecasts of features of each country. They also depend on the
the beginning, end and intensity of ENSO events users’ need for climate forecasts.
and forecasts of tropical Pacific Ocean sea surface
temperatures can be regarded as climate forecast Presentation of forecast elements in absolute terms
products. is rare; forecast elements are generally represented
by categories such as above normal, near normal
The WMO El Niño/La Niña Update, a consensus and below normal. Two types of forecast formats
product based on inputs from a worldwide network are possible with the categorical representation: a
of forecasting centres, provides an indication of the categorical forecast gives the most likely category
phase and strength of the El Niño–Southern and a probabilistic forecast gives the chance of
Oscillation. Operational climate forecasting is occurrence of a category. Uncertainty is unavoid­
conducted by NMHSs, Global Producing Centres of able in climate forecasting because of the chaotic
Long-Range Forecasts (GPCs) and other interna- nature of the atmosphere, the shortage of observa-
tional institutions. The Update recognizes that other tional data and the approximations within the
factors in addition to ENSO influence seasonal forecast methods. Given this uncertainty, probabil-
climatic patterns around the world, and that there istic forecasts are generally more robust than
is a need for detailed regional evaluations of prevail- categorical forecasts. Probabilistic forecasts,
ing conditions. Combining expected ENSO however, are more difficult to apply; users need to
influences with those from other geographic regions be made familiar with the merits and limitations of
usually provides the best estimates of the patterns probabilistic forecasts and also with the methods of
of variability to be expected regionally and locally cost–benefit analysis. When forecasts of elements
for the coming months. The effort requires careful are presented as numerical quantities, the forecast
6-10 GUIDE TO CLIMATOLOGICAL PRACTICES

uncertainty can be expressed with confidence limits Intergovernmental Panel on Climate Change (IPCC)
or by attaching verification statistics of past fore- has projected a range of possible temperature
casts. The climate service should consider the increases for the twenty-first century that would
results of past forecast verification experiments to result in the event that the world follows a number
guide the use of probabilistic forecasts. of plausible patterns of population and economic
growth, development of energy technologies, and
Forecast products can be provided directly to emissions. By considering how the resulting changes
specific users, and the climate service is often in atmospheric composition would affect the climate
required to interpret the meaning of the forecasts using different climate models, each with its own
to the user. News media are common means of particular climate sensitivity, the projections of
disseminating forecasts to the public, as is the climate change account for a wide range of reasona-
Internet. Some climate centres provide their fore- ble possibilities of both societal development and
casts only directly to specific users, however. climate behaviour. Projections are neither a predic-
tion nor a forecast of what will or is likely to happen.
For decision-makers, they indicate the likely
6.7.2 Climate predictions and
outcomes resulting in part from the adoption of
projections
specified policy-driven actions.
A climate prediction is a probabilistic statement
about the future climate on timescales ranging
6.7.3 Climate scenarios
from years to decades. It is based on conditions that
are known at present and assumptions about the A major use of global climate models is the genera-
physical processes that will determine future tion of climate scenarios. A climate scenario refers to
changes. A prediction generally assumes that factors a plausible future climate constructed for investigat-
beyond what is explicitly or implicitly included in ing the potential consequences of human-induced
the prediction model will not have a significant climate change, but should also represent future
influence on what is to happen. In this sense, a conditions that account for natural climate variabil-
prediction is most influenced by the current condi- ity. The IPCC reports and publications (for example,
tions that are known through observations (initial IPCC, 2007) provide a good source of information
conditions). For example, a weather prediction that about climate scenarios.
a major snowstorm will develop over the next few
days is mostly determined by the state of the atmos-
6.7.4 Global climate models
phere as observed (and its conditions in the recent
past). The small changes that may occur over the Global climate models are designed mainly for
next few days in other factors that are potentially representing climate processes on a global scale.
influential on longer timescales, such as ocean They provide the essential means to study climate
temperatures or human activities (boundary condi- variability for the past, present and future. They are
tions), are likely to be less important to the weather based upon the physical laws governing the climate
forecast. A prediction is made probabilistic by processes and interactions of all of the components
accounting for various types of uncertainties, for of the climate system, expressed in the form of
example, in the accuracy of observations and in the mathematical equations in three dimensions. The
chaotic state of the atmosphere. For decision- highly non-linear governing equations are solved
makers, what is important is that a prediction is a numerically on a four-dimensional grid of the
statement about the likelihood that something will atmosphere (three space dimensions plus time).
occur, no matter what they do (policymakers Many physical processes such as individual clouds,
cannot change tomorrow’s weather or the amount convection and turbulence take place on much
of rainfall in a coming season). smaller spatial and timescales than can be properly
resolved by the grid. These processes have to be
A climate projection is usually a statement about the included through a simplified representation in
likelihood that something will happen several terms of the large-scale parameters of the model.
decades to centuries in the future if certain influen-
tial conditions develop. In contrast to a prediction, a These models first became practicable in the 1960s,
projection specifically allows for significant changes and since then have undergone rapid development
in the set of boundary conditions, such as an increase and improvement. They have developed in parallel
in greenhouse gases, which might influence the with numerical weather prediction models. Initially,
future climate. As a result, what emerges are condi- GCMs were directed at coupling the atmosphere
tional expectations (if this happens, then that is and ocean; most state-of-the-art GCMs now include
what is expected). For projections extending well out representations of the cryosphere, biosphere, land
into the future, scenarios are developed of what surface and land chemistry in increasingly complex
could happen given various assumptions and judg- integrated models that are sometimes called climate
ments. For example, in its various assessments the system models.
 CHAPTER 6. �SERVICES AND PRODUCTS 6-11

Confidence in GCMs has increased substantially as scales that have been identified in the observed
a result of systematic model intercomparisons, the climate. Features of the large scale, such as averages,
ability shown by some models to reproduce major variability and time dependencies, are functionally
trends in the climate of the twentieth century and related to the smaller regional scale. The most severe
in some paleoclimates, and the improved simula- limitation of statistical downscaling is the necessity
tion of major general circulation features related to for adequate historical observations upon which the
phenomena such as ENSO. In general, over many statistical relationships are based.
parts of the world, GCMs provide credible climate
simulations at subcontinental scales and for Both methods are prone to uncertainties caused by
seasonal to decadal timescales, and are therefore lack of knowledge of the Earth system, model
considered as suitable tools to provide useful future parameter and structure approximations, random-
climate projections. These GCMs have formed the ness, and human actions. Validation and verification
basis for the climate projections in IPCC assess- of the results of a downscaled model are also diffi-
ments and contribute substantially to seasonal cult, especially if there is an inadequate observational
forecasting in climate outlook forums. base. Active research is being carried out to reduce
the uncertainties.
There is considerable interest in refining global
climate modelling in order to simulate climate on
6.7.6 Local climate models
smaller scales, where most impacts are felt and
adaptive capacity exists. Smaller-scale climates are Unlike global and regional climate models, which
determined by an interaction of forcings and circu- seek to model the climate of the entire globe or a
lations on global, large-area and local spatial scales large part of the globe over an extended period of
and on subdaily to multidecadal temporal scales. time, local climate models attempt to simulate
The large-area and local forcings are caused by microscale climate over a limited area of a few
complex topography and land-use characteristics, square metres to a square kilometre for a short time.
features at land–ocean interfaces, area and local Local climate modelling is used for a range of
atmospheric circulations such as sea breezes and purposes, such as planning for potentially hazard-
tropical storms, the distribution of aerosol particles ous industrial plants; planning for odour and noise
and atmospheric gases, and the effects of lakes, emissions from industrial and agricultural plants
snow and sea ice. The climate of an area may also and roads; urban planning of, for example, ventila-
be strongly influenced through teleconnection tion, cold air flow and heat stress in single buildings
processes by forcing anomalies in distant locations. or housing and industrial areas; warning and emer-
The processes are often highly non-linear, making gency services relating to local concerns that may
simulation and prediction difficult. include air quality and pollutant threshold levels;
and managing the spread of hazards.
6.7.5 Downscaling: Regional climate
models The models vary from simple, highly parameterized
but fast methods, such as statistical approaches, to
Models cannot provide direct information for scales complex tools that numerically solve the hydro­
smaller than their own resolution. A process known dynamic equations of motion and include a range
as downscaling relates the properties of a large-scale of other processes. Sometimes the models use only
model to smaller-scale regions. The approach can a very limited number of measuring stations and
be either dynamical or statistical, or a combination data covering periods that range from just several
of the two. weeks on up to a few years. Short-term model runs,
typically of several days’ duration, are repeated for
The dynamical approach involves nesting limited- the specified climate period, set of representative
area high-resolution models within a coarser global weather situations, meteorological elements, and
model. Tools used in this process are known as sometimes chemical elements. This process yields a
regional climate models (RCMs). They typically use sample or ensemble of results that can be inter-
the synoptic and larger-scale information from a preted statistically to provide the required climate
GCM to drive a regional or mesoscale dynamical information.
model. A major problem with these models is how to
relate the coarse-resolution grid cell information from
the GCM through the boundaries into the finer grid
cells of the RCM. Since the RCM is essentially driven 6.8 REANALYSIS PRODUCTS
by the GCM, good performance of the GCM is of
prime importance for the smaller-scale modelling. In operational numerical weather analysis and
prediction, “analysis” refers to the process of creating
Statistical downscaling involves the application of an internally consistent representation of the
statistical relationships between the large and smaller environment on a four-dimensional grid. The
6-12 GUIDE TO CLIMATOLOGICAL PRACTICES

time-critical nature of weather prediction means The assimilation of information from not only atmos-
that the initializing analysis must usually begin pheric sciences, but also oceanography, hydrology
before all observations are available. Reanalysis uses and remote-sensing, is used to create environmental
the same process (and often the same systems), but databases from which systematic changes can be
as it is done weeks or even years later, it is able to use better assessed. Currently, the main global-scale rean-
a more complete set of observations. These reanalysis alysis databases are those of the National Center for
systems generally incorporate a prediction model Atmospheric Research and National Centers for
that provides information on how the environment Environmental Prediction in the United States, the
is changing with time, while maintaining internal European Centre for Medium-Range Weather
consistency. Unlike the “analysis” in operational Forecasts, and the Japan Meteorological Agency. All
weather prediction, in which the models are of these reanalysis efforts have found wide use in
constantly updated to incorporate the latest research climate monitoring, climate variability studies and
advances, the “reanalysis” is performed with a fixed climate change prediction. It is important to assess
modelling system throughout the period of reanalysis the relative skill of the reanalysis techniques in repre-
to prevent the inhomogeneities that generally exist senting the observed features in a given region before
in the operational dataset because of model using their data for further climatological studies.
differences over time.
Greater understanding of the physical, chemical
The output of a reanalysis is on a uniform grid and and biological processes that control the environ-
no data are missing. The result is an integrated ment, combined with data from a range of sources
historical record of the state of the atmospheric that go well beyond the atmospheric sciences,
environment for which all the data have been proc- should promote improvements in reanalysis data-
essed in the same manner. Reanalysis outputs are bases. As the numerical models become more
often used in place of observational data, but this complete, and as computer technology allows for
must be done with care. Although the analysis algo- higher resolution, more accurate and comprehen-
rithms will make good use of observations when sive reanalysis products will emerge.
available, in regions where observations are scarce
the reanalysis grid will be strongly influenced by
the prediction model. For reanalysis projects that
span decades, there is generally a lot of heterogene- 6.9 EXAMPLES OF PRODUCTS AND DATA
ity in the type and coverage of data throughout the DISPLAYS
period, such as between the pre- and post-satellite
periods. Further, the relative influence of the obser- Data can be presented in a number of ways. Figures 6.1
vations and the model is different for different to 6.17 illustrate some of the many simple but effective
climatic variables; certain variables are strongly presentations of information that are possible.
influenced by the observational data used, while
some are purely model-derived. These aspects
should be carefully considered while interpreting
the reanalysis data products. For example, rean-
alyses of dynamical variables are far better than
reanalyses of precipitation, partially because proc-
esses leading to precipitation are not well
represented in the models.

The limitations of reanalysis outputs are most obvi-

ous in the areas with complex orography (in
particular, in mountain regions), as well as in other
areas when the assimilation and processing schemes
are unable, because of smoothing, to reproduce real
atmospheric processes with high spatial and tempo-
ral gradients. Also, there still remains the issue of
localizing to spatial and temporal scales finer than
the reanalysis grid. There are ongoing efforts to
perform “regional reanalysis” using more local
observational data with higher-resolution limited
area models. As with any other analysis technique,
validation of models, quality assurance and indica-
tors of error are necessary to properly interpret the
results. Figure 6.1. Contour map of precipitation
 CHAPTER 6. �SERVICES AND PRODUCTS 6-13

Sunshine

Figure 6.2. Graphical display of daily values of several elements

Energy demand and temperature - Sydney, 1990–1991

Figure 6.3. Scatter diagram of energy demand and temperature with trend line superimposed
6-14 GUIDE TO CLIMATOLOGICAL PRACTICES

Monthly Long-Term Maximum Temperature Averages (1961−1990) for Selected Sites

Degrees Celsius

Figure 6.4. Line graph and table of temperature values for multiple stations

Figure 6.5. Graphical display of growing degree-days throughout a season for multiple stations
 CHAPTER 6. �SERVICES AND PRODUCTS 6-15

Figure 6.6. Map of regional values of rainfall with a table of summary information

Figure 6.7. Symbolic map display of rainfall information

6-16 GUIDE TO CLIMATOLOGICAL PRACTICES

Period of Record: 1 May 1973 to 30 April 2000

Snowfall (in.)

Figure 6.8. Tabular display of a monthly climate summary

Figure 6.9. Composite line and bar chart for multiple elements and stations
 CHAPTER 6. �SERVICES AND PRODUCTS 6-17

Figure 6.10. Visual display of months with complete and missing data

Figure 6.11. Three-dimensional display emphasizing anomalous values

Temperature (Deg F)

1956

Figure 6.12. Three-dimensional display emphasizing time series of monthly values for a given year
6-18 GUIDE TO CLIMATOLOGICAL PRACTICES

Figure 6.13 Line graph of a time series

Figure 6.14. Line graph of a time series emphasizing observed values

No. of years in average calculation

No. of years

Figure 6.15. Composite bar chart and line graph of average temperature values and .
number of years of record
 CHAPTER 6. �SERVICES AND PRODUCTS 6-19

Figure 6.16. Contour map of drought categories and impacts

Figure 6.17. Symbolic map display of wind

6-20 GUIDE TO CLIMATOLOGICAL PRACTICES

6.10 REFERENCES 6.10.2 Additional reading

American Meteorological Society, 1993: Guidelines

6.10.1 WMO publications
for using color to depict meteorological
World Meteorological Organization, 1983: Guide to information. Bull. Amer. Meteor. Soc.,
Climatological Practices. Second edition (WMO- 74:1709–1713.
No. 100), Geneva. Brohan, P., J.J. Kennedy, I. Harris, S.F.B. Tett and
———, 1988: Technical Regulations. Vol. I – General P.D. Jones, 2006: Uncertainty estimates in
Meteorological Standards and Recommended Practices; regional and global observed temperature
Vol. II – Meteorological Service for International Air changes: A new dataset from 1850. J. Geophys.
Navigation; Vol. III – Hydrology; Vol. IV – Quality Res., 111: D12106.
Management (WMO-No. 49), Geneva. Hargreaves, G.H., 1975: Moisture availability and
———, 1991: Manual on the Global Data-processing crop production. Trans. ASAE, 18:980–984.
System. Vol. I – Global Aspects. Supplement Jones, P.D., M. New, D.E. Parker, S. Martin and
No. 10, October 2005 (WMO-No. 485), Geneva. I.G. Rigor, 1999: Surface air temperature and its
———, 1992: Manual on the Global Data-processing variations over the last 150 years. Rev. Geophys.,
System. Vol. II – Regional Aspects. Supplement 37:173–199.
No. 2, August 2003 (WMO-No. 485), Geneva. Khambete, N.N., 1992: Agroclimatic classification
———, 1992: Operational Climatology - Climate for assessment of the crop potential of
Applications: On Operational Climate Services and Karnataka. Mausam, 43(1):91–98.
Marketing, Information and Publicity: Report to the MacCracken, M., 2002: Do the uncertainty ranges
Eleventh Session of the Commission for Climatology in the IPCC and U.S. National Assessments
(Havana, February 1993) (CCl Rapporteurs on account adequately for possibly overlooked
Operational Climatology, J.M. Nicholls, and climatic influences? Climatic Change,
Marketing, Information and Publicity, 52:13–23.
D.W. Phillips). World Climate Applications Palmer, W.C., 1965: Meteorological Drought.
and Services Programme (WMO/TD-No. 525, Weather Bureau Research Paper No. 45.
WCASP-No. 20), Geneva. Washington, DC, United States Department of
———, 1995: Hydrological Forecasts for Hydroelectric Commerce.
Power Production (WMO/TD-No. 118), Geneva. Peterson, T.C., 2005: Climate change indices. WMO
———, 1995: Meeting of Experts on Climate Bulletin, 54:83–86.
Information and Prediction Services (CLIPS): Report Peterson, T.C. and M.J. Manton, 2008: Monitoring
of the Meeting (Melbourne, 28–31 March 1995) changes in climate extremes: A tale of interna-
(WMO/TD-No. 680, WCASP-No. 32), Geneva. tional cooperation. Bull. Amer. Meteor. Soc.,
———, 1996: Economic and Social Benefits of 89:1266–1271.
Climatological Information and Services: A Review of Rayner, N.A., P. Brohan, D.E. Parker, C.K. Folland,
Existing Assessments (J.M. Nicholls) (WMO/TD- J.J. Kennedy, M. Vanicek, T. Ansell and
No. 780, WCASP-No. 38), Geneva. S.F.B. Tett, 2006: Improved analyses of changes
———, 1996: Report of the Second Session of the CCI and uncertainties in marine temperature
Working Group on Operational Use of Climatological measured in situ since the mid-nineteenth
Knowledge (Geneva, 28–31 May 1996) and Report century: The HadSST2 dataset. J. Climate,
of the Meeting of Experts on CLIPS (Geneva, 22–24 19:446–469.
May 1996) (WMO/TD-No. 774, WCASP-No. 37), Sarker, R.P. and B.C. Biswas, 1988: A new approach
Geneva. to agroclimatic classification to find out crop
———, 1999: Report of the Planning Meeting for the potential. Mausam, 39(4):343–358.
Shanghai CLIPS Showcase Project: Heat/Health Tufte, E.R., 1990: Envisioning Information. Cheshire,
Warning System (Shanghai, 6–8 October 1999) Connecticut, Graphic Press.
(WMO/TD-No. 984, WCASP-No. 49), Geneva. United States Global Change Research Program
———, 2007: WMO Statement on the Status of the Office (USGCRP), 2000: Climate Change Impacts
Global Climate in 2006 (WMO-No. 1016), Geneva. on the United States: The Potential Consequences of
———, 2008: Inter-Commission Task Team on the Climate Variability and Change. Washington,
Quality Management Framework, Third Session, DC, USGCRP.
Final Report (Geneva, Switzerland, 28–30 Wang, B. and Z. Fan, 1999: Choice of South Asian
October 2008), Geneva. summer monsoon indices. Bull. Amer. Meteor.
———, 2008: Report on the Activities of the Working Soc., 80:629–638.
Group on Climate Change Detection and Related Wilhelm, W.W., K. Ruwe and M.R. Schlemmer, 2000:
Rapporteurs 1998–2001 (WMO/TD-No. 1071, Comparisons of three Leaf Area Index meters in
WCDMP-No. 47), Geneva. a corn canopy. Crop Science, 40:1179–1183.

_________________________
 ANNEX 1

ACRONYMS

AWS Automated Weather Station

CCl Commission for Climatology

CDMS Climate Data Management System

CLIPS Climate Information and Prediction Services

ENSO El Niño–Southern Oscillation

GAW Global Atmosphere Watch

GCM General Circulation Model

GCOS Global Climate Observing System

GEOSS Global Earth Observation System of Systems

GEWEX Global Energy and Water Cycle Experiment

GOOS Global Ocean Observing System

GPC Global Producing Centres of Long-Range Forecasts

GSN GCOS Surface Network

GUAN GCOS Upper-Air Network

IPCC Intergovernmental Panel on Climate Change

ISO International Organization for Standardization

NMHS National Meteorological and Hydrological Service

RCC Regional Climate Centre

RCM Regional Climate Model

WCASP World Climate Applications and Services Programme

WCDMP World Climate Data and Monitoring Programme

WDC World Data Centre

WMO World Meteorological Organization

.
 ANNEX 2

INTERNATIONAL CLIMATE ACTIVITIES

A2.1 COORDINATION OF CLIMATE as the Food and Agriculture Organization of the

ACTIVITIES United Nations; the World Health Organization;
the United Nations World Tourism Organization;
The Commission for Climatology is responsible for the United Nations Human Settlements Programme;
promoting and facilitating activities relating to the United Nations Environment Programme; the
climate and its relationship with human well-being, United Nations Development Programme; and the
socio-economic activities, natural ecosystems and United Nations Educational, Scientific and Cultural
sustainable development. Specifically, it: Organization. The Commission for Climatology
(a) Coordinates and consolidates general require- also works extensively with non-governmental
ments and standards for observations, data organizations such as the International Federation
collection, archiving and data exchange of Red Cross and Red Crescent Societies, the
for all components of the World Climate International Council for Science, research institu-
Programme; tions, universities, professional associations,
(b) Identifies best practices in management of academia, and development agencies such as the
climate data, including near-real-time data, World Bank.
proxy data, remote-sensing data, and meta-
data (information about data);
(c) Promotes dissemination of data, products and
methods in support of research, applications, A2.2 THE WORLD CLIMATE PROGRAMME
impact assessments and climate system moni-
toring; In the 1970s it became obvious that greater interna-
(d) Supports preparation of authoritative state- tional coordination and effort were needed to tackle
ments on climate, including the Annual State- the deficiencies in the understanding of climate and
ments on the Status of the Global Climate and how to deal with the wide range of climatic influ-
regular El Niño and La Niña Updates; ences on society and the environment, some
(e) Evaluates and reviews development and appli- beneficial and others detrimental. In 1974 the
cation of operational climate predictions; Executive Council of the World Meteorological
(f) Identifies priorities for studying the climates Organization agreed that WMO should initiate an
of natural and managed ecosystems and for international programme on climate and laid the
providing climate information for alleviating foundations of the World Climate Programme. In
problems arising from the influence of human 1978 the Economic and Social Council of the United
activities on local and regional climate; Nations requested that WMO devote particular
(g) Supports capacity-building and technology attention to those aspects of the Programme which
transfer; would provide prompt and effective assistance to
(h) Promotes research and evaluation of the role national planners and decision-makers in formulat-
of climate in key social and economic sectors ing economic and social programmes and activities
in partnership with other WMO Technical in their respective countries.
Commissions, other United Nations agencies
and relevant international and regional insti- The First World Climate Conference in February
tutions; 1979 recognized that the problem of possible
(i) Assesses the potential for application of human influence on climate was a matter of special
seasonal prediction and other climatologi- importance. The Conference Declaration stressed
cal services for social and economic benefit, the urgent need “for the nations of the world to
including reduction of risk to climate-related take full advantage of man’s present knowledge of
hazards and optimal utilization of climate as a climate; to take steps to improve significantly that
resource; knowledge; and to foresee and to prevent potential
(j) Provides advice on issues relating to the access man-made changes in climate that might be adverse
and availability of climatological data and to the well-being of humanity”.
services.
The eighth World Meteorological Congress formally
The Commission promotes and relies on a range of created the World Climate Programme in 1979,
national, regional and global entities involved in recognizing that the cooperation of many other
climate-related matters. Apart from NMHSs, these United Nations agencies and other international
entities include other United Nations agencies such organizations was needed to support the
Ann–2 GUIDE TO CLIMATOLOGICAL PRACTICES

programme. Responsibility for the overall coordi- The Second World Climate Conference in 1990
nation of the programme and for climate data and recognized an urgent need to acquire comprehen-
applications was assumed by WMO, while the sive information on the properties and evolution of
United Nations Environment Programme assumed the Earth’s climate system, to detect climate change,
responsibility for climate impact studies, and WMO to support climatological applications for economic
together with the International Council for Science development, and to develop climate science and
agreed to jointly implement the climate research predictions. The Eleventh World Meteorological
programme. The structure of the World Climate Congress in 1991 decided that a global climate
Programme and the partners cooperating with observing system should be established. This system
WMO have evolved over time. would be based on the coordination and associa-
tion of existing or planned operational and research
There are four main components of the World programmes for observing the global environment
Climate Programme. The World Climate Data and on the further development of those
and Monitoring Programme (WCDMP) facilitates programmes required to ensure continuity of infor-
the effective collection and management of mation over decades.
climate data and the monitoring of the global
climate system, including the detection of climate In 1992, the Global Climate Observing System
variability and change. The World Climate (GCOS) was formally established under a
Applications and Services Programme (WCASP) Memorandum of Understanding among WMO, the
fosters scientific understanding of the role of Intergovernmental Oceanographic Commission of
climate in human activities, effective application the United Nations Educational, Scientific and
of climate knowledge and information for the Cultural Organization, the United Nations
benefit of society, tailored climate services for Environment Programme, and the International
users in various socioeconomic sectors, and the Council for Science. While GCOS does not directly
prediction of significant climate variations (both make observations or generate data products, it
natural and as a result of human activity). The works closely with and builds upon existing and
Climate Information and Prediction Services developing observing systems and provides a frame-
(CLIPS) project was established by WMO as an work for integrating the systems of participating
implementation arm of WCASP. The World countries and organizations. The systems and
Climate Impact Assessment and Response networks included are:
Strategies Programme, implemented by the (a) GCOS Surface Network (GSN) and Upper-Air
United Nations Environment Programme, Network (GUAN) (subsets of the WMO World
assesses the impacts of climate variability and Weather Watch Global Observing System);
changes that could markedly affect economic or (b) WMO Global Atmosphere Watch (GAW);
social activities and advises governments on (c) WCRP/GEWEX Baseline Surface Radiation
these matters; it also contributes to the develop- Network;
ment of a range of socio-economic response (d) Global Ocean Observing System (GOOS),
strategies that could be used by governments and including ocean observing systems such as
communities. The World Climate Research the Global Sea Level Observing System;
Programme (WCRP), jointly sponsored by WMO, (e) Global Terrestrial Observing System, includ-
the International Council for Science and the ing the Global Terrestrial Network for Glaciers
Intergovernmental Oceanographic Commission and the Global Terrestrial Network for Perma-
of the United Nations Educational, Scientific and frost;
Cultural Organization, seeks to improve the basic (f) World Hydrological Cycle Observing System.
scientific understanding of climate processes for
determining the predictability of climate, climate The concept of coordinated and comprehensive
variability and change, and the extent of human Earth observations was promoted at the First Earth
influence on climate, and for developing the Observation Summit in 2003. At the Third Earth
capability for climate prediction. The activities Observation Summit held in 2005, the Global
and projects of WCRP include studies of the Earth Observation System of Systems (GEOSS) was
dynamics and thermodynamics of the Earth’s formally established to build on, and add value to,
atmosphere, the atmosphere’s interactions with existing national, regional and international
the Earth’s surface, and the global water cycle; observation systems by coordinating efforts,
climate variability and predictability; interaction addressing critical gaps, supporting interoperabil-
of dynamical, radiative and chemical processes; ity, sharing information, reaching a common
processes by which the cryosphere interacts with understanding of user requirements, and improv-
the rest of the climate system; and biogeochemi- ing delivery of information to users. GEOSS
cal and physical interactions between the ocean supports GCOS by expanding on the range of
and the atmosphere. climate-related variables identified in the GCOS
 ANNEX 2. INTERNATIONAL CLIMATE ACTIVITIES Ann–3

Implementation Plan and by assisting the parties significant warming of the global climate in the
in meeting their responsibilities under the United next century, as detailed in the Report of the
Nations Framework Convention on Climate International Conference on the Assessment of the Role
Change (see section A2.4). of Carbon Dioxide and of Other Greenhouse Gases in
Climate Variations and Associated Impacts (WMO-
No. 661). They also noted that past climate data
may no longer be a reliable guide for long-term
A2.3 THE CLIMATE AGENDA projects because of expected warming of the global
climate, that climate change and sea level rises are
In the early 1990s, development of increasingly closely linked with other major environmental
close collaboration among climate-related issues, that some warming appears inevitable
programmes of a number of international organi- because of past activities, and that the future rate
zations led to the establishment by WMO of the and degree of warming could be profoundly
Coordinating Committee for the World Climate affected by policies on emissions of greenhouse
Programme. In 1995 a draft interagency document gases. Responding to these concerns, the United
entitled The Climate Agenda – International Climate- Nations Environment Programme, WMO and the
Related Programmes: A Proposal for an Integrating International Council for Science established an
Framework was issued. The four main thrusts were Advisory Group on Greenhouse Gases in 1986 to
new frontiers in climate science and prediction, ensure periodic assessments of the state of scien-
climate services for sustainable development, tific knowledge on climate change and its
dedicated observations of the climate system, and implications.
studies of climate impact assessments and response
strategies to reduce vulnerability. The leaders of In 1987 the Tenth World Meteorological Congress
these thrusts were WMO, the Food and Agriculture recognized the need for objective, balanced and
Organization of the United Nations and the United internationally coordinated scientific assessment of
Nations Environment Programme. the understanding of the effects of increasing
concentrations of greenhouse gases on the Earth’s
The Twelfth World Meteorological Congress climate and the ways in which these changes may
endorsed the Climate Agenda in 1995 and estab- influence social and economic patterns. The United
lished an Interagency Committee on the Climate Nations Environment Programme and WMO agreed
Agenda. While the Committee provided some on an intergovernmental mechanism to provide
initial guidance, the fast pace of developments in scientific assessments of climate change, and in 1988
climate-related matters led WMO at the 2001 the WMO Executive Council, with support from the
session of its Executive Council to begin a process United Nations Environment Programme, estab-
of considering new mechanisms for climate coordi- lished the Intergovernmental Panel on Climate
nation. At its sixty-first session in 2009, the WMO Change to consider the need for:
Executive Council agreed not to revitalize the (a) Identification of uncertainties and gaps in
Interagency Committee on the Climate Agenda, our present knowledge with regard to climate
but to use the United Nations Delivering as One change and its potential impacts, and prepa-
initiative to coordinate climate issues at the United ration of a plan of action over the short term
Nations level. to fill these gaps;
(b) Identification of information needed to evalu-
ate policy implications of climate change and
response strategies;
A2.4 INTERNATIONAL CLIMATE CHANGE (c) Review of current and planned national and
PROGRAMMES international policies related to the green-
house gas issue;
At the First World Climate Conference in 1979, (d) Scientific and environmental assessments of
the international climate community expressed all aspects of the greenhouse gas issue and the
concern that “continued expansion of man’s transfer of these assessments and other rele-
activities on Earth may cause significant extended vant information to governments and inter-
regional and even global changes of climate”, and governmental organizations, to allow them to
called for global cooperation to “explore the possi- be taken into account in policies on social and
ble future course of global climate and to take this economic development and environmental
new understanding into account in planning for programmes.
the future development of human society”. At a
climate conference in Villach, Austria, in 1985 In November 1988 the IPCC established working
scientists from 29 developed and developing coun- groups to prepare assessment reports on the available
tries concluded that the increasing concentrations scientific information on climate change (Working
of greenhouse gases were expected to cause a Group I), on environmental, social and economic
Ann–4 GUIDE TO CLIMATOLOGICAL PRACTICES

impacts of climate change (Working Group II), and of America, is testimony to the remarkable success of
on formulation of response strategies (Working Group the IPCC process in informing the policymakers as
III). The forty-third session of the United Nations well as the general public on the scientific underpin-
General Assembly in 1988 endorsed the action of nings of the climate change issue.
WMO and the United Nations Environment
Programme to establish the IPCC and requested “a
comprehensive review and recommendations with
respect to, inter alia, the state of knowledge of the A2.5 REFERENCES AND ADDITIONAL
science of climate and climatic change; programmes READING
and studies on the social and economic impact of
climate change, including global warming; and possi- Group on Earth Observations (GEO), 2005: Global
ble response strategies to delay, limit, or mitigate the Earth Observation System of Systems (GEOSS):
impacts of adverse climate change”. 10-year Implementation Plan. Reference
Document GEO 1000R/ESA SP-1284. Noordwijk,
The IPCC adopted its first assessment report on 30 European Space Agency Publications Division,
August 1990. Its findings, coupled with the outcomes ESTEC.
of the Second World Climate Conference that same ———, 2007: GEO 2007–2009 Work Plan. Toward
year, spurred governments to create the United Convergence, Geneva.
Nations Framework Convention on Climate Change Intergovernmental Panel on Climate Change
in March 1994. The IPCC Second Assessment Report (IPCC), 2007: The Fourth Assessment Report:
was completed in late 1995, the third in 2001, and Climate Change 2007 (AR4), Vols. 1–4.
the fourth in 2007. The reports issued by the IPCC Cambridge, Cambridge University Press.
are frequently used to inform decisions made under United Nations System Chief Executives Board for
the Framework Convention. They also played a major Coordination, 2008: Acting on Climate Change:
role in the negotiations leading to the Kyoto Protocol, The UN System Delivering as One, New York.
a February 2005 international treaty that builds on World Meteorological Organization, 1986: Report of
the Framework Convention and sets legally binding the International Conference on the Assessment of
targets and timetables for reducing the greenhouse the Role of Carbon Dioxide and of Other Greenhouse
gas emissions of industrialized countries. The award Gases in Climate Variations and Associated
of the 2007 Nobel Peace Prize to the IPCC, along with Impacts, (Villach, Austria, 9–15 October 1985)
Mr Al Gore, Former Vice-President of the United States (WMO-No. 661), Geneva.

.
 Guide to Climatological Practices

GUIDE TO CLIMATOLOGICAL PRACTICES

2011 edition

P-CLW_101264

WMO-No. 100

www.wmo.int WMO-No. 100