Earth Science Informatics (2021) 14:985–995

https://doi.org/10.1007/s12145-021-00578-6

RESEARCH ARTICLE

Comparing different space-borne sensors and methods

for the retrieval of land surface temperature
Arif UR Rehman 1 & Sami Ullah 2,3 & Qijing Liu 1 & Muhammad Sadiq Khan 4

Received: 4 August 2020 / Accepted: 25 January 2021 / Published online: 20 March 2021
# The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2021

Abstract
The importance of land surface temperature (LST) is increasingly recognized, and various methodologies have been proposed for
the retrieval of LST using space-borne thermal infrared data. However, the selection of LST retrieval from Thermal Infrared
Sensor (TIRS) of Landsat-8 based on different methods and the readily available MODIS LST products is still a challenging topic
for local and global environmental studies. In this study, the potential of three different methods for retrieving LST using Landsat-
8 TIRS data, including Radiative Transfer Equation (RTE), Single Channel (SC), and Split Window (SW) method in comparison
with MODIS MOD11A1 LST product was evaluated. For accuracy assessment, 0 cm ground surface temperature (LSTGST) data
was used. Our results almost showed same accuracy for RTEB10 with RMSE = 0.35 °C, followed by MODIS with RMSE =
0.36 °C, and SCB10 with RMSE = 0.38 °C. Secondly, SCmean (Mean of B10 and B11), and RTEmean (Mean of B10 and B11)
generate nearly the same accuracy with RMSE = 0.53 °C, and RMSE = 0.54 °C, respectively. The other methods viz., SCB11,
RTEB11, and SW method slightly showed lower accuracy with RMSE = 0.87 °C, RMSE = 0.88 °C, and RMSE = 0.91 °C,
respectively. We found all the methods highly accurate and can be used successfully by climatologists, environmentalists,
hydrologists, and urban planners concerning planning, and monitoring of the ever-increasing LST at local and global scale
studies.

Keywords Landsat-8 TIRS . MODIS . Radiative transfer equation . Single Channel method . Split window algorithm

Introduction temporal measurements of LST across the heterogeneous

landscape requires precise ground measurements, which can
Land Surface Temperature (LST) is an essential parameter in only be obtained from ground-based stations. Although these
the physics of different earth surfaces at regional and global ground stations provide point-based measurements which
scales (Kustas and Anderson 2009). It plays an important role cannot represent the spatial pattern of LST over a large geo-
in a wide range of scientific studies, such as ecology, hydrol- graphic area. Remote Sensing can be used successfully for the
ogy, climatology, and other land-use/land-cover changes stud- retrieval of LST over a large geographical area across the
ies (He et al. 2018; Amir Siddique et al. 2020; Sadiq Khan globe in a variety of spatial and temporal scales. Among re-
et al. 2020; Zhang and He 2013). Accurate, updated, and mote sensing, Thermal Infrared sensor is the key source of

Communicated by: H. Babaie

* Qijing Liu 1
College of Forestry, Beijing Forestry University, Beijing 100083,
liuqijing@bjfu.edu.cn China
2
Department of Forestry, Shaheed Benazir Bhutto University
Arif UR Rehman Sheringal, Dir 18050, Pakistan
arifrehman@bjfu.edu.cn 3
GIS & Space Applications in Geosciences (G-SAG) laboratory at the
Sami Ullah NCE in Geology, University of Peshawar, National Center of GIS
sami.ullah@sbbu.edu.pk and Space Applications, Peshawar, Pakistan
4
State Key Laboratory of Urban and Regional Ecology, Research
Muhammad Sadiq Khan Center for Eco-Environmental Sciences, Chinese Academy of
khan_st@rcees.ac.cn Sciences, Beijing 100085, China
986 Earth Sci Inform (2021) 14:985–995

retrieving LST information at regional and global scales as In this study, Guizhou province located in the southwest of
most of energy detected by a sensor at the thermal region of China, with 32 meteorological stations was selected as a study
the electromagnetic spectrum is directly emitted from the land area, to check the potential of three LST retrieval methods
surfaces (Jiménez-Muñoz and Sobrino 2008; Reuter et al. including RTE, SC and SW using both Landsat-8 TIRS bands
2015). Apparently many efforts have been made to develop (B10 and B11). Furthermore, this study validates the derived
accurate methods for retrieving LST from remote sensing data LSTs from different methods and MODIS LST product
and significant progress has been made in the last three de- MOD11A1 with 0 cm ground surface temperature
cades (Du et al. 2015; Jin et al. 2015; Li et al. 2014; Rongali (LSTGST). The 0 cm ground surface temperature is used for
et al. 2018; Yang et al. 2014; Yu et al. 2014). However, eval- the first time in our study as a ground validation which is a
uating the methods for retrieval of LST and selection of read- secondary product generated by the National Meteorological
ily available LST product is still an essential and challenging Information Center (NMIC) of China Meteorological Agency
topic for scientific research. (CMA) based on meteorological observations.
Over past decades many algorithms have been developed
and proposed to treat the characteristics of various satellite-
based sensors for LST retrieval. For instance, Chatterjee et al. Data and methods
(2017) used a single channel radiative transfer algorithm to
retrieve LST from Landsat TM6 and TIRS data using ground- Ground validation data
based inputs. While Jiménez-Muñoz and Sobrino (2010) used
a single channel algorithm for LST retrieval from ASTER China Meteorological Agency (CMA) serves as primary sup-
data. Some studies compared the Radiative Transfer port for climate monitoring in the country comprised of about
Equation (RTE), Mono-Window (MW) algorithm and 2419 ground meteorological stations including 699 national
Single-Channel (SC) method for retrieve LST from benchmarks distributed unevenly, sparse in northern and
Landsat-5 TM data (Sobrino et al. 2004; Vlassova et al. western and dense in the eastern and southern parts of China
2014; Zhou et al. 2012). For instance, Jiménez-Muñoz et al. (Missions 2016). All the stations observed several climate
(2014) Compared the SC method and Split-Window (SW) variables four times a day (02:00, 08:00, 14:00 and 20:00)
algorithm using Landsat Thermal Infrared Sensor (TIRS) da- and send it to the CMA data center (Ying et al. 2014). CMA
ta. While Yu et al. (2014), and García-Santos et al. (2018) continuously updates and screen data for quality control.
compared RTE, SC and SW methods for the retrieval of LST National Meteorological Information Center (NMIC) of
using Landsat 8 TIRS Band 10 (B10) only. The former used CMA, generate secondary datasets from the primary observed
four selected energy balance monitoring sites from the values at all stations. These secondary products include; baro-
Surface Radiation Budget Network (SURFRAD) while the metric pressure, air temperature, relative humidity, precipita-
latter used surface energy budget (SEB) stations available in tion, evaporation, wind direction and speed, sunshine and
the study area for validation of results. In addition to a com- 0 cm ground surface temperature (GST). Where these second-
parison of methods, some operational LST products from ary datasets are the daily average values of four-time observa-
different sensors like Moderate Resolution Imaging tions at ground stations (Song et al. 2007). The GST was
Spectroradiometer (MODIS) (Wan and Dozier 1996), evaluated by the In-Situ ground temperature and the result
Advance Very High-Resolution Radiometer (AVHRR) show that GST is in good agreement with the in-situ observa-
(Kerr et al. 1992), Advanced Along-Track Scanning tion with RSME 1.8 k and Bias 1.4 k. NMIC issued these
Radiometer (AATSR) (Coll et al. 2012b), Spinning products for research purposes on request bases. For this
Enhanced Visible and Infrared Imager (SEVIRI) (Niclòs study, we order the datasets record for the single summer
et al. 2011), and Geostationary Operational Environmental month, August 2018. The datasets naming consist of data set
Satellites (GOES) (Sun and Pinker 2003) are available at code (SURF_CLI_CHN_MUL_DAY-XXX-XXXXX-
different spatial scales. However, in LST retrieval many as- YYYYMM.TXT), climate variable (XXX), variable code
pects like an estimation of surface emissivity (ε) from the (XXXXX), year (YYYY) and month (MM), where SURF
same Landsat-8 scene used for LST retrieval, testing SC means surface meteorological data, CLI means climate,
and RTE methods for TIRS Band 11 (B11), including mean CHN means China, MUL means multi-variables and DAY
LST from TIRS B10 and B11 seem a research gap in the means daily (Xu et al. 2009). For example the GST data set
previous studies. Further, a comparison of LST retrieval code is SURF_CLI_CHN_MUL_DAY-GST-12030-
methods and its evaluation with 0 cm ground surface temper- 201,808.TXT. However, GST is used for validation of
ature (LSTGST) derived from station data and MODIS daily MOD11A1 and LSTs resulted from different algorithms while
LST product (MOD11A1) brings more novelty at our paper relative humidity, pressure and air temperature are used as
which can be useful to the researcher by a selection of suit- input to the MODTRAN model to calculate atmospheric var-
able methods and datasets in their interest area. iables as defined in section 2.6.
Earth Sci Inform (2021) 14:985–995 987

Landsat 8 data were developed, while the Landsat-8 TIRS launch in

2013, has two adjacent thermal bands (B10 and B11) making
Over the last three decades, the Landsat project provides an it suitable for the SW algorithm (Irons et al. 2012). Landsat
opportunity to estimate LST (Roy et al. 2014; Weng and Fu scenes used in this study were selected in order to maximize
2014), as it has moderate spatial resolution and long data the number of available validation stations. First, all ground
record period. The project started with the launch of stations were overlaid over Landsat WRS-2 path/row
Landsat-3 Multispectral Scanner (MSS) in 1978, followed shapefiles (www.landsat.gsfc.nasa.gov), representing
by Landsat-4 and 5 Thematic Mapper (TM) in 1982 and Landsat-8 scenes boundaries, in ArcMap and query was
1984 respectively and Landsat-7 Enhanced Thematic applied to select the Landsat-8 scenes with maximum
Mapper Plus (ETM+) in 1999 (Behrens 2009; Williams ground stations. The query resulted in four adjacent WRS-
et al. 2006; Wulder et al. 2016). Both the TM and ETM+ 2 boundaries with 32 ground stations covering the Guizhou
sensors have only one thermal band (Markham et al. 2004), province located in the southwest of China (Fig. 1).
therefore SC and RTE methods based on single-channel Secondly, all the images of selected four boundaries,

Fig. 1 Study area map

988 Earth Sci Inform (2021) 14:985–995

acquired in august 2018 were investigated and minimum Table 2 Emissivity

values for soil and Es Ev
cloud cover scenes were selected (Table 1). Finally, Both
vegetation
Level 1 and Level 2 products of Landsat-8 were Band 10 0.9668 0.9863
downloaded from the USGS website (www.earthexplorer. Band 11 0.9747 0.9896
usgs.gov). Level 1 TIRS bands used for retrieving LST, as
Level 2 OLI bands were atmospherically corrected before
delivery (USGS 2016a; Young et al. 2017), therefore used NDVI based ε estimation because of its simplicity and good
in the estimation of NDVI and surface emissivity. working in comparison with other methods (Becker and Li
1990; Sobrino et al. 2001). However, the NDVI based method
MODIS data obtains ε values considering three different NDVI cases
(Sobrino et al. 2004; Sobrino et al. 2008).
MODIS onboard Terra and Aqua sensors were launched by
National Aerodynamics and Space Administration (NASA) a) NDVI <0.2
Earth Observing System (EOS) in December 1999 and
May 2002 respectively (Jang et al. 2010). MODIS provides a In this case, pixels represent bare soil and the ε is obtained
sequence of LST products with different spatial and temporal from reflectivity values in the red region.
resolutions. The revised document describes the algorithms and
quality assessment of the Collection-6 (C6) MODIS LST prod- b) NDVI >0.5
ucts and recommended to use C6 MODIS LST product as its
quality is better than C4.1 and C5 LST products (Hulley et al. In this case, pixels represent full vegetation, and constant
2016). C6 is a series of nine LST products, varying in spatial value for ε is assumed, (ε = 0.99).
and temporal resolutions. In this study, we used the
MOD11A1 day time daily LST product of MODIS C6, iii) 0.2 < NDVI <0.5
downloaded from the USGS website (www.earthexplorer.
usgs.gov). MOD11A1 are level 3 products generated by In this case, pixels are a mixture of soil and vegetation, and
MOD11_L2 LST product using a generalized SW algorithm ε is calculated using Eq. (1):
(correction of atmospheric effects based on different absorption
in adjacent thermal bands) (Wan and Dozier 1996), under clear ε ¼ Esð1−PvÞ þ EvPv þ C ð1Þ
sky condition. The MOD11A1 LST product was validated with
where Ev and Es are soil and vegetation ε (Yu et al. 2014),
0 cm ground surface temperature (LSTGST).
respectively (Table 2), Pv is the vegetation proportion, obtain
through Eq. (2), following (Carlson and Ripley 1997) and C
Calculating land surface emissivity (ε) based on NDVI represents the shape factor (Oltra-Carrió et al. 2012; Sobrino
method et al. 2008; Valor and Caselles 1996), and mean value (C =
0.55) is assumed (Becker and Li 1990; Sobrino et al. 2008).
The knowledge of ε is necessary in order to estimate LST, The


ε of different surfaces can differ significantly from unity and ðNDVI−NDVIminÞ 2
vary with land use, moisture, and surface roughness (Avdan Pv ¼ ð2Þ
NDVImax−NDVImin
2016). A variety of methods are available to retrieve ε (Li et al.
2013; Masiello and Serio 2013; Oltra-Carrió et al. 2012;
Sobrino et al. 2008), where mostly LST based studies (Coll Where, NDVImin = 0.2 and NDVImax = 0.5.
et al. 2012b; Jiménez-Muñoz et al. 2014; Niclòs et al. 2011; For, NDVI we used the Landsat-8 level 2 product, visible and
Oltra-Carrió et al. 2012; Sobrino et al. 2008) have used the near-infrared bands. Equation (3) is used for retrieving NDVI:

Table 1 Satellite data overview

Landsat 8 MOD11A1

Date Acquisition Time Path/Rows Resolutions Acquisition Time Resolution

18 Aug 2018 03:15:06 126/41 30 m OLI and 100 m TIRS 09:05:55 1 Km
18 Aug 2018 03:15:30 126/42
26 Aug 2018 03:21:20 127/41
26 Aug 2018 03:21:44 127/42
Earth Sci Inform (2021) 14:985–995 989

temperature, and humidity) used as input data to the

NIR−RED
NDVI ¼ ð3Þ MODTRAN code to predict the atmospheric spectral param-
NIR þ RED eters (Berk et al. 2014; Jiménez-Muñoz and Sobrino 2003). In
Where NIR represents the near-infrared band (band 5), and previous studies several Thermodynamics Initial Guess
RED represents the red band (band 4) of Landsat-8. Retrieval (TIGR) versions (i.e., TIGR61, TIGR1761, and
TIGR2311) (Cristóbal et al. 2009; Cristóbal et al. 2018b;
Estimation of brightness temperature Jiménez-Muñoz and Sobrino 2010; Jiménez-Muñoz et al.
2014), National Center for Environmental Prediction
The development of all algorithms for LST retrieval using (NCEP) reanalysis data (Chatterjee et al. 2017; Coll et al.
thermal bands of Landsat uses the brightness temperature of 2012a; Vlassova et al. 2014; Yu et al. 2014), and the Global
thermal bands at the satellite level. Where the brightness tem- Atmospheric Profiles from Reanalysis Information (GAPRI)
perature from thermal bands includes the estimation of spec- data used for extracting the atmospheric profiles (Cristóbal
tral radiance from the pixels values (DN) using radian scaling et al. 2018a; Jiménez-Muñoz et al. 2014). The GAPRI data-
factors provided in the metadata file. Equation (4) developed base consists of 4714 atmospheric profiles covering tropical,
by the National Aeronautics and Space Administration mid-latitude, subarctic, and arctic conditions (Mattar et al.
(NASA) (USGS 2016b) was used to retrieve the spectral ra- 2015). However, in this study the atmospheric variables (I↑,
diance from the DN value of Landsat data. I↓ and τ) were calculated with the MODTRAN radiative trans-
fer code (Berk et al. 1999; Berk et al. 2014) using the synthetic
Lλ ¼ ML*Qcal þ AL ð4Þ atmospheric profile provided by a web-tool calculator (Barsi
et al. 2003; Barsi et al. 2005) based on NCEP model (Kalnay
Where Lλ is Spectral Radiance (m W cm-2 sr-1 μm-1), ML
et al. 1996).
is Radiance multiplicative scaling factor for the band
The web-tool calculator allows users to retrieve a more
(RADIANCE_MULT_BAND_n in metadata), AL is
synthetic atmospheric profile, using specific date, time, loca-
Radiance Additive scaling factor for the band
tion, and local surface conditions as input. These surface pa-
(RADIANCE_ADD_BAND_n in metadata) and Qcal = DN
rameters include altitude (km), pressure (mb), air temperature
values.
(°C) and humidity (%) and they were provided by the ground
Once the spectral radiance Lλ is calculated, Eq. (5) is used
stations as described in section 2.1. The atmospheric variables
to estimate the brightness temperature at the satellite level.
were calculated for each Landsat scene separately and once
K2 these parameters (I↑, I↓ and τ) for each Landsat scene were
Ts ¼

  −273:15 ð5Þ
K1 obtained, Eqs. (6a, 6b, and 6c) are used to derive other atmo-
ln þ1 spheric functions (Ψ1, Ψ2, and Ψ3).
Lλ
1
Where Ts is the at-satellite brightness temperature in Ψ1 ¼ ð6aÞ
Kelvin, K1 and K2 are constants, and to convert Kelvin to τ

↑
Celsius 273.15 is subtracted, and this is the most popular I
Ψ 2 ¼ I ↓− ð6bÞ
approach suggested by USGS to compute brightness τ
temperature(USGS 2016b; Reuter et al. 2015).
Ψ3 ¼ I↓ ð6cÞ
Obtaining atmospheric parameters using simulated
data Radiative transfer equation (RTE) based method

LST algorithm of the Radiative Transfer Equation uses some For estimating LST, top of atmospheric spectral radiance
atmospheric parameters like downwelling radiance (I↓), up- needs to be corrected to obtain surface spectral radiance be-
welling radiance (I↑) and atmospheric transmission (τ), which cause atmospheric effects are crucial for temperature studies.
were also used in the SC method to derive the atmospheric In this study, we used a standard RTE, following Yu et al.
functions (Ψ1, Ψ2 and Ψ3). In many previous studies, these (2014), Zhou et al. (2012), and Jiménez-Muñoz et al. (2014),
parameters were obtained by a simulation procedure (Coll expressed in Eq. (7).
et al. 2012a), which is also the most practical way to determine  
the atmospheric transmittance using local climatic conditions. LST RTEBi ¼ τiðθÞEiBiðTsÞ þ ð1−EiÞI ↓ þ I ↑ ð7Þ
The MODTRAN radiative code is used worldwide for the
prediction and analysis of optical measurements through the
atmosphere (Barsi et al. 2003). Satellite parameters, spectral τi (θ)
functions, and atmospheric profiles (atmospheric pressure, air
990 Earth Sci Inform (2021) 14:985–995

atmospheric transmission for channel i when view Jiménez-Muñoz and Sobrino 2003; Jiménez-Muñoz et al.
zenith angle is θ. 2009; García-Santos et al. 2018).
Ei surface emissivity of the band i.   
Bi a ground radiance. LST SCBi ¼ γ ðLSE Bi Þ−1 ððΨ 1 LiÞ þ Ψ 2 Þ þ Ψ 3 þ δ ð10Þ
(Ts) " !
 #−1
I↓ Downwelling path radiance. C 2 Li λBi 4 −1
γ¼ Li þ λBi ð11Þ
I↑ Upwelling path radiance. ðT sBi Þ2 C1
According to Plank’s law ground radiance Bi(Ts) can be
expressed as: where C1 and C2 are Plank’s radiation constants, λ represents
the wavelength of the specific bands.
2hc2
BiðTsÞ ¼

  ð8Þ δ ¼ −γ:Li þ T sBi ð12Þ
hc
λBi 5
exp −1
λBi kTs The mean value (B10 and B11) was also calculated using
Eq. (13) to check and analyze the effect of the mean value on
Where c is the speed of light (c = 2.9979 × 108 m/s), h is the single band results.
the Planck constant (h = 6.6261 × 10–34 J.s), k is the
Boltzmann constant (k = 1.3806 × 10–23 J/K), λ represents LST SCB10 þ LST SCB11
LST SCmean ¼ ð13Þ
the wavelength of TIRS bands (B10 = 10.602 and B11 = 2
12.511), and Ts is brightness temperature derived by Eq. (5).
This RTE method is for a single band. Therefore, it was ap-
Split window (SW) algorithm
plied to both TIRS bands (B10 and B11) separately. Further,
the results from Eq. (7) for each band were put in Eq. (9) to
SW method estimates LST using both TIRS bands of Landsat
calculate the mean.
8, while OLI bands (2–5) are also used for calculating Land
LST RTEB10 þ LST RTE B11 Surface Emissivity (ε). The SW method suggested by Yu et al.
LST RTEmean ¼ ð9Þ
2 (2014), and Rongali et al. (2018) expressed in Eq. (14) was
used. This method includes the Brightness Temperature
(TsB10 and TsB11), water vapor content (W = 0.018 g/cm2)
Single Channel (SC) method acquired from stations data, Mean and Difference values of
ε and some coefficients (Vlassova et al. 2014; Rongali et al.
As Landsat 8 TIRS has two adjacent spectral bands while the 2018) (Table 3). Whereas other SW methods are also avail-
SC method utilizes a single band. Therefore, the algorithm able (Wang et al. 2015; Wan and Dozier 1996; Trigo et al.
applied to both bands separately. Equation (10) proposed for 2008; Jin et al. 2015; Du et al. 2015).
the SC method was used in this study (Vlassova et al. 2014;

 
LST SW ¼ T sB10 þ C 1 ðT sB10 −T S B11 Þ þ C 2 ðT sB10 −T sB11 Þ2 þ C 0 þ C 3 þ C 4 W ð1−Meanε Þ þ C 5 þ C 6 W Dif f ε ð14Þ

based LSTs and observed LSTGST from the meteorologi-

cal stations. The absolute and relative root mean square
εB10 þ εB11
Meanε ¼ ð15Þ (RMSE) as shown in Eqs. (17) and (18) were computed
2
using leave-one-out cross-validation (LOOCV), where ev-
Dif f ε ¼ εB10 −εB11 ð16Þ
ery single observation was held out as a testing set, and
the remaining data were used as a training set. The entire
Accuracy assessment

The LSTGST data shown in Table 4 for 32 ground station

stations in the study area were used for accuracy assess- Table 3 Split Window Coefficients
ment of remote sensing based LSTs. The ground station
Constants C0 C1 C2 C3 C4 C5 C6
points overlaid over the derived LSTs scenes, and values
were extracted at locations of each ground station. Single Values −0.268 1.378 0.183 54.30 −3.238 −129.20 16.40
linear regression was applied between the remote sensing
Earth Sci Inform (2021) 14:985–995 991

Table 4 Stations LST in 0C with inverted LSTs in 0C using different methods and MODIS MOD11A1 product

Stations ID LSTGST RTEB10 RTEB11 RTEmean SCB10 SCB11 SCmean SW MOD11A1 Land Cover

56,691 24.50 25.00 21.27 23.14 25.48 21.56 23.52 24.85 24.25 Forest Land
56,792 29.30 28.25 23.48 25.87 28.88 23.85 26.37 28.13 29.01 Rural Land
56,793 26.00 27.20 22.53 24.87 27.82 22.90 25.36 26.65 26.15 Rural Land
57,606 23.00 23.01 19.81 21.41 23.57 20.16 21.86 23.52 23.71 Forest Land
57,625 23.00 23.35 20.18 21.77 23.76 20.42 22.09 23.88 23.47 Grass Land
57,647 13.00 13.38 9.89 11.64 13.68 10.05 11.86 12.85 12.87 Grass Land
57,707 26.40 26.22 22.40 24.31 26.83 22.77 24.80 27.30 26.13 Grass Land
57,710 24.40 24.51 20.96 22.74 24.95 21.22 23.08 23.75 24.31 Forest Land
57,718 22.70 22.83 18.49 20.66 23.34 18.78 21.06 23.27 22.79 Crop Land
57,722 12.00 12.17 9.31 10.74 12.43 9.45 10.94 11.93 12.17 Grass Land
57,729 21.00 21.01 18.58 19.80 21.48 18.88 20.18 22.15 21.33 Forest Land
57,731 14.00 14.23 11.97 12.10 14.55 10.13 12.34 14.02 13.93 Forest Land
57,741 16.00 16.57 13.00 14.79 16.93 13.20 15.06 16.61 16.27 Forest Land
57,803 27.20 27.08 22.06 24.57 27.68 22.41 25.05 26.86 27.43 Grass Land
57,805 22.30 22.45 19.42 20.94 22.85 19.66 21.26 21.62 22.31 Grass Land
57,806 21.90 21.95 18.97 20.46 22.31 19.19 20.75 21.80 22.13 Grass Land
57,816 17.00 16.99 18.19 17.59 17.87 18.42 17.85 17.44 18.17 Forest Land
57,825 23.20 23.59 21.56 22.58 24.01 21.82 22.92 23.40 24.85 Forest Land
57,827 28.40 28.18 25.34 26.76 28.80 25.74 27.27 28.57 28.65 Rural Land
57,832 23.10 23.96 21.77 22.86 24.41 22.05 23.23 23.23 24.43 Rural Land
57,839 25.60 25.52 23.19 24.35 26.00 23.49 24.74 24.16 24.93 Grass Land
57,840 18.30 18.52 15.85 17.19 18.94 16.10 17.52 17.14 17.87 Crop Land
57,902 22.30 22.60 20.13 21.36 23.05 20.41 21.73 23.60 22.41 Crop Land
57,906 18.30 18.59 15.92 17.25 19.01 16.16 17.58 18.21 18.31 Forest Land
57,907 22.90 22.74 20.86 21.80 23.16 21.12 22.14 23.25 22.91 Forest Land
57,910 21.50 20.29 17.83 19.06 20.76 18.12 19.44 21.52 21.43 Forest Land
57,912 23.00 24.68 21.75 23.22 25.14 22.03 23.58 23.72 23.15 Crop Land
57,916 24.80 25.11 22.20 23.66 25.58 22.49 24.04 25.17 24.63 Grass Land
57,922 24.30 24.28 21.88 22.95 24.45 22.14 23.30 23.78 23.95 Rural Land
57,926 30.30 30.52 27.26 28.89 31.23 27.71 29.47 29.27 30.35 Rural Land
57,932 26.20 26.10 24.34 25.72 27.63 24.67 26.15 25.93 25.81 Rural Land
57,936 24.60 24.64 22.60 23.62 25.07 22.87 23.97 23.46 24.81 Forest Land

calculation was carried out in R-statistical software (Team Results

RC 2016).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Our results almost showed same accuracy for RTEB10 with a
1 n  b R2 = 0.99 and RMSE = 1.54%, followed by MOD11A1 with a
RMSE ðLOOCV Þ½m ¼ ∑ yi−yi 2 ð17Þ R2 = 0.99 and RMSE = 1.60%, and SCB10 with R2 = 0.99 and
n i¼1
RMSE = 1.67%. Secondly, SCmean (B10 and B11), and RTEmean
(B10 and B11) show somewhat same accuracy with R2 = 0.97,
ð18Þ
RMSE = 2.34%, and with R2 = 0.96, RMSE = 2.39%, respec-
tively. The final three methods viz., SCB11, RTEB11, and SW
Method slightly showed lower accuracy with R2 = 0.93,
b is the predicted value of
Where yi is the observed values, yi RMSE = 3.89%, R2 = 0.93, RMSE = 3.91%, and with R2 =
LOOCV, i is the mean of the observed values and n is the total 0.93, RMSE = 4.04%, respectively. Overall there is high degree
number of ground sample plots. Mean absolute Error (MAE) of correlation between the predicted and observed LST as
was also reported between the actual remote sensing-based and shown in Fig. 2. More detail about the regression coefficients,
ground LST derived from the meteorological stations. MAE etc. can be found in Table 5.
992 Earth Sci Inform (2021) 14:985–995

Table 5 Statistical results of LST methods (RTE, SC, and SW) and MOD11A1 product

Types of RS Data and Methods Slope Intercept R2 RMSE RMSE % MAE

Radiative Transfer Equation (B10) 1.01*** −0.32 0.99 0.35 1.54 0.35
MOD11A1 1.00*** −0.18 0.99 0.36 1.60 0.36
Single Channel (B10) 0.99*** −0.33 0.99 0.38 1.67 0.38
Single Channel (Mean of B10 and B11) 1.00*** 0.94 0.97 0.53 2.34 0.53
Radiative Transfer Equation (Mean of B10 and B11) 1.02*** 0.96 0.97 0.54 2.39 0.54
Single Channel (B11) 0.99*** 2.80* 0.93 0.87 3.89 0.87
Radiative Transfer Equation (B11) 1.03*** 2.12. 0.93 0.88 3.91 0.88
Split Window Method 1.00*** 0.19 0.93 0.91 4.04 0.91

Note: R2 = coefficients of determination, and (*** p < 0.001, ** p < 0.01, * p < 0.05, and p < 0.1) indicates the level of significance of the t-test

Discussions by MOD11A1, and SCB10 (Table 5). The mean of B10 and B11
derived from SC and RTE methods can be rank second in
The main goal of our study was to compare the potential of terms of accuracy. While the last three methods viz., SCB11,
three different methods for LST retrieval viz., (1) RTE, (2) RTEB11, and SW method slightly showed lower accuracy. B10
SC, and (3) SW methods using Landsat-8 TIRS bands in has shown the highest accuracy in both the cases i.e. RTE and
comparison with MODIS LST product MOD11A1. Our re- SC which conclude the same findings as in the USGS Landsat
sults almost showed the same accuracy for RTEB10, followed handbook that there is significant calibration uncertainty

Fig. 2 Scatter plots showing

predicted LST derived from
different methods, and datasets
along the x-axis, and observed
(LSTGST oC) along the y-axis
Earth Sci Inform (2021) 14:985–995 993

associated with B11, which leads to error in LST estimation estimating LST. Summarizing our findings all the methods
(USGS 2016b). Also, B11 is more sensitive to errors in atmo- are highly accurate and can be used successfully by researchers,
spheric profiles due to water vapor continuum absorption environmentalists, and urban planners.
(Avdan 2016). Yu et al. (2014) retrieved LST from Landsat-
8 TIRS bands using RTE, SC and SW algorithm. Like our
findings, they also found that the RTE method using B10 has Conclusion
the highest accuracy with RMSE lower than 1 K, while the
SW algorithm has moderate accuracy, and the SC method has Accurate and temporal measurements of LST across the het-
the lowest accuracy. However, our study is different from Yu, erogeneous landscape requires precise ground measurements,
Guo and Wu [13] in many aspects. For instance, Yu, Guo and which can only be obtained from ground-based stations.
Wu [13] used Surface Radiation Budget Network Although these ground stations provide point-based measure-
(SURFRAD) sites operated by the National Oceanic and ments which cannot represent the spatial pattern of LST over a
Atmospheric Administration (NOAA) for validation, whereas large geographic area. The estimation of LST from space
we used local meteorological stations data. In addition, they borne imagery for a large geographic area is a challenging
used NCEP data to simulate atmospheric parameters needed and important topic for investigation. We compared the po-
as inputs for the MODTRAN model while we used the syn- tential of RTE, SC, and SW methods using Landsat-8 TIRS
thetic atmospheric profile provided by a web-tool calculator data in comparison with MODIS MOD11A1 LST product.
(Barsi et al. 2003; Barsi et al. 2005) based on NCEP model We found all the methods, and datasets satisfactory and highly
(Kalnay et al. 1996). Also, they used SW algorithm developed accurate and suggesting being used successfully by climatol-
by Mao et al. (2005) whereas we followed Yu et al. (2014), ogists, environmentalists, hydrologists, and urban planners
Rongali et al. (2018), and García-Santos et al. (2018). Another concerning planning, and monitoring of the ever-increasing
study conducted by García-Santos et al. (2018) by comparing LST at local and global scale studies.
the RTE, SC and SW methods for LST retrieval using
Landsat-8 TIRS data. In contrast to our results, they reported Supplementary Information The online version contains supplementary
material available at https://doi.org/10.1007/s12145-021-00578-6.
lowest RMSE for SW (within 1.6-2 K), whereas SCB10 RMSE
ranging within 2.0–2.3 K and RTE with highest RMSE values
Acknowledgments We are thankful to the China Meteorological
2.0–3.6 K; however, they used surface energy budget stations Department for providing the ground stations temperature data for select-
located at height of 1 m above ground level in the study area ed dates, the US Geological Survey (USGS) for providing level 1 and
for validation whereas we used the 0 cm ground surface tem- level 2 Landsat-8 imagery and MODIS products and to the team of GIS &
Space Applications in Geosciences (G-SAG) laboratory at the NCE in
perature (LSTGST). Besides, they used the ASTER global
Geology, University of Peshawar with the partnership of Shaheed
emissivity product while we derived emissivity from same Benazir Bhutto University, National Center of GIS and Space
Landsat-8 data used for LST retrieval. Chatterjee et al. Applications for helping in processing of remote sensing datasets, analy-
(2017), and Jiménez-Muñoz and Sobrino (2010) compared sis, and final writing up.
RTE, SC and Mono Window (MW) methods for LST retriev-
Author contributions Arif UR Rehman is the lead author and was in-
al using Landsat-5 TM data, and both of them reported high
volved in the overall processing, analysis, and writing. Sami Ullah sup-
accuracy for RTE and SC and low accuracy for MW method; ported the processing of Landsat imagery and developing R codes.
however the first one use network of micrometeorological Muhammad Sadiq Khan contributed to the implementation of the meth-
observation sites while the second one used ground based odological approach. The first draft of the manuscript was prepared by
Arif UR Rehman and was further improved by Sami Ullah, Muhammad
experimental data for validation. In addition, MOD11A1
Sadiq Khan, and finally by Qijing Liu.
LST product also showed a very high accuracy compared to
Landsat-8 TIRS data. Zhu et al. (2013) also reported that Funding “This research is part of a master’s degree thesis of the first
MOD11A1 product is a better estimation of daily temperature. author and it was funded by the Chinese Government Scholarship (CSC
In addition, Fu et al. (2011), and Urban et al. (2013) also Number: 2018SLJ021190).
suggested that MOD11A1 had higher accuracy than other
LST products like MYD11A2 and AVHRR. Declarations
In contrast to other studies reported in a similar direction, the
MODTRAN radiative transfer code based on a web-tool calcu- Conflict of interest “The authors declare no conflict of interest.”
lator was used for estimating atmospheric parameters. The
higher accuracy in the case of RTE method also suggesting that
executing the MODTRAN radiative transfer code using a web- References
tool calculator is accurate enough for retrieving atmospheric
Amir Siddique M, Dongyun L, Li P, Rasool U, Ullah Khan T, Javaid Aini
parameters used in LST estimation. Also, the results reflect that Farooqi T, Wang L, Fan B, Rasool MA (2020) Assessment and
MODIS has the same potential as Landsat-8 TIRS for
994 Earth Sci Inform (2021) 14:985–995

simulation of land use and land cover change impacts on the land He G, Zhang Z, Jiao W, Long T, Peng Y, Wang G, Yin R, Wang W,
surface temperature of Chaoyang District in Beijing, China. PeerJ 8: Zhang X, Liu H (2018) Generation of ready to use (RTU) products
e9115–e9115. https://doi.org/10.7717/peerj.9115 over China based on Landsat series data. Big Earth Data 2(1):56–64
Avdan U (2016) Jovanovska G (2016) algorithm for automated mapping Hulley G, Malakar N, Hughes T, Islam T, Hook S (2016) Moderate
of land surface temperature using LANDSAT 8 satellite data. J Sens resolution imaging spectroradiometer (MODIS) MOD21 land sur-
2016:1–8 face temperature and emissivity algorithm theoretical basis docu-
Barsi JA, Barker JL, Schott JR (2003) An atmospheric correction param- ment. Pasadena, CA: jet Propulsion Laboratory, National
eter calculator for a single thermal band earth-sensing instrument. In: Aeronautics and space administration,
IGARSS 2003. 2003 IEEE International Geoscience and Remote Irons JR, Dwyer JL, Barsi JA (2012) The next Landsat satellite: the
Sensing Symposium. Proceedings (IEEE Cat. No. 03CH37477). Landsat data continuity mission. Remote Sens Environ 122:11–21
Toulouse, France, 2003, pp. 3014-3016 vol.5, https://doi.org/10. Jang K, Kang S, Kim J, Lee CB, Kim T, Kim J, Hirata R, Saigusa N
1109/IGARSS.2003.1294665 (2010) Mapping evapotranspiration using MODIS and MM5 four-
Barsi JA, Schott JR, Palluconi FD, Hook SJ (2005) Validation of a web- dimensional data assimilation. Remote Sens Environ 114(3):657–
based atmospheric correction tool for single thermal band instru- 673. https://doi.org/10.1016/j.rse.2009.11.010
ments," Proc. SPIE 5882, Earth Observing Systems X, 58820E Jiménez-Muñoz JC, Sobrino JA (2003) A generalized single-channel
(22 August 2005). https://doi.org/10.1117/12.619990 method for retrieving land surface temperature from remote sensing
Becker F, Li Z-L (1990) Temperature-independent spectral indices in data. J Geophys Res: Atmos 108 (D22)
thermal infrared bands. Remote Sens Environ 32(1):17–33 Jiménez-Muñoz JC, Sobrino JA (2008) Split-window coefficients for
Behrens CE (2009) Landsat and the data continuity mission. In, 2009. land surface temperature retrieval from low-resolution thermal in-
Congressional research service, Library of Congress, frared sensors. IEEE Geosci Remote Sens Lett 5(4):806–809.
Berk A, Anderson G, Acharya P, Chetwynd J, Bernstein L, Shettle E, https://doi.org/10.1109/LGRS.2008.2001636
Matthew M, Adler-Golden S (1999) MODTRAN4 user’s Jiménez-Muñoz JC, Sobrino JA (2010) A single-channel algorithm for
manual.1. Air Force Research Laboratory 1 land-surface temperature retrieval from ASTER data. IEEE Geosci
Berk A, Conforti P, Kennett R, Perkins T, Hawes F, Van Den Bosch J Remote Sens Lett 7(1):176–179
(2014) MODTRAN® 6: A major upgrade of the MODTRAN® Jiménez-Muñoz JC, Cristóbal J, Sobrino JA, Sòria G, Ninyerola M, Pons
radiative transfer code. In: 2014 6th workshop on Hyperspectral X (2009) Revision of the single-channel algorithm for land surface
image and signal processing: evolution in remote sensing temperature retrieval from Landsat thermal-infrared data. IEEE
(WHISPERS). IEEE, pp 1–4 Trans Geosci Remote Sens 47(1):339–349
Carlson TN, Ripley DA (1997) On the relation between NDVI, fractional Jiménez-Muñoz JC, Sobrino JA, Skoković D, Mattar C, Cristóbal J
vegetation cover, and leaf area index. Remote Sens Environ 62(3): (2014) Land surface temperature retrieval methods from Landsat-8
241–252 thermal infrared sensor data. IEEE Geosci Remote Sens Lett 11(10):
Chatterjee R, Singh N, Thapa S, Sharma D, Kumar D (2017) Retrieval of 1840–1843
land surface temperature (LST) from landsat TM6 and TIRS data by Jin M, Li J, Wang C, Shang R (2015) A practical split-window algorithm
single channel radiative transfer algorithm using satellite and for retrieving land surface temperature from Landsat-8 data and a
ground-based inputs. Int J Appl Earth Obs Geoinf 58:264–277 case study of an urban area in China. Remote Sens 7(4):4371–4390
Coll C, Caselles V, Valor E, Niclòs R (2012a) Comparison between Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L,
different sources of atmospheric profiles for land surface tempera- Iredell M, Saha S, White G, Woollen J (1996) The NCEP/NCAR
ture retrieval from single channel thermal infrared data. Remote 40-year reanalysis project. Bull Am Meteorol Soc 77(3):437–472
Sens Environ 117:199–210 Kerr YH, Lagouarde JP, Imbernon J (1992) Accurate land surface tem-
Coll C, Valor E, Galve JM, Mira M, Bisquert M, García-Santos V, perature retrieval from AVHRR data with use of an improved split
Caselles E, Caselles V (2012b) Long-term accuracy assessment of window algorithm. Remote Sens Environ 41(2–3):197–209
land surface temperatures derived from the advanced along-track Kustas W, Anderson M (2009) Advances in thermal infrared remote
scanning radiometer. Remote Sens Environ 116:211–225 sensing for land surface modeling. Agric For Meteorol 149(12):
Cristóbal J, Jiménez-Muñoz J, Sobrino J, Ninyerola M, Pons X (2009) 2071–2081
Improvements in land surface temperature retrieval from the Li H, Sun D, Yu Y, Wang H, Liu Y, Liu Q, Du Y, Wang H, Cao B (2014)
Landsat series thermal band using water vapor and air temperature. Evaluation of the VIIRS and MODIS LST products in an arid area of
J Geophys Res: Atmospheres 114 (D8) Northwest China. Remote Sens Environ 142:111–121
Cristóbal J, Jiménez-Muñoz J, Prakash A, Mattar C, Skoković D, Sobrino Li Z-L, Wu H, Wang N, Qiu S, Sobrino JA, Wan Z, Tang B-H, Yan G
J (2018a) An improved single-channel method to retrieve land sur- (2013) Land surface emissivity retrieval from satellite data. Int J
face temperature from the Landsat-8 thermal band. Remote Sens Remote Sens 34(9–10):3084–3127
10(3):431 Mao K, Qin Z, Shi J, Gong P (2005) A practical split-window algorithm
Cristóbal J, Jiménez-Muñoz JC, Prakash A, Mattar C, Skoković D, for retrieving land-surface temperature from MODIS data. Int J
Sobrino JA (2018b) An improved single-channel method to retrieve Remote Sens 26(15):3181–3204
land surface temperature from the Landsat-8 thermal band. Remote Markham BL, Storey JC, Williams DL, Irons JR (2004) Landsat sensor
Sens 10(3):431 performance: history and current status. IEEE Trans Geosci Remote
Du C, Ren H, Qin Q, Meng J, Zhao S (2015) A practical split-window Sens 42(12):2691–2694
algorithm for estimating land surface temperature from Landsat 8 Masiello G, Serio C (2013) Simultaneous physical retrieval of surface
data. Remote Sens 7(1):647–665 emissivity spectrum and atmospheric parameters from infrared at-
Fu G, Shen Z, Zhang X, Shi P, Zhang Y, Wu J (2011) Estimating air mospheric sounder interferometer spectral radiances. Appl Opt
temperature of an alpine meadow on the northern Tibetan plateau 52(11):2428–2446. https://doi.org/10.1364/AO.52.002428
using MODIS land surface temperature. Acta Ecol Sin 31(1):8–13. Mattar C, Durán-Alarcón C, Jiménez-Muñoz JC, Santamaría-Artigas A,
https://doi.org/10.1016/j.chnaes.2010.11.002 Olivera-Guerra L, Sobrino JA (2015) Global atmospheric profiles
García-Santos V, Cuxart J, Martínez-Villagrasa D, Jiménez MA, Simó G from reanalysis information (GAPRI): a new database for earth sur-
(2018) Comparison of three methods for estimating land surface face temperature retrieval. Int J Remote Sens 36(19–20):5045–5060
temperature from landsat 8-tirs sensor data. Remote Sens 10(9): Missions UL (2016) Using the USGS Landsat8 product. US Department
1450 of the Interior-US Geological Survey–NASA
Earth Sci Inform (2021) 14:985–995 995

Niclòs R, Galve JM, Valiente JA, Estrela MJ, Coll C (2011) Accuracy Vlassova L, Perez-Cabello F, Nieto H, Martín P, Riaño D, de la Riva J
assessment of land surface temperature retrievals from MSG2- (2014) Assessment of methods for land surface temperature retrieval
SEVIRI data. Remote Sens Environ 115(8):2126–2140 from Landsat-5 TM images applicable to multiscale tree-grass eco-
Oltra-Carrió R, Sobrino J, Franch B, Nerry F (2012) Land surface emis- system modeling. Remote Sens 6(5):4345–4368
sivity retrieval from airborne sensor over urban areas. Remote Sens Wan Z, Dozier J (1996) A generalized split-window algorithm for retriev-
Environ 123:298–305 ing land-surface temperature from space. IEEE Trans Geosci
Reuter D, Richardson C, Pellerano F, Irons J, Allen R, Anderson M, Remote Sens 34(4):892–905
Jhabvala M, Lunsford A, Montanaro M, Smith R (2015) The ther- Wang M, He G, Zhang Z, Wang G, Long T (2015) NDVI-based split-
mal infrared sensor (TIRS) on Landsat 8: design overview and pre- window algorithm for precipitable water vapour retrieval from
launch characterization. Remote Sens 7(1):1135–1153 Landsat-8 TIRS data over land area. Remote Sens Lett 6(11):904–
Rongali G, Keshari AK, Gosain AK, Khosa R (2018) Split-window al- 913. https://doi.org/10.1080/2150704X.2015.1089363
gorithm for retrieval of land surface temperature using Landsat 8 Weng Q, Fu P (2014) Modeling annual parameters of clear-sky land
thermal infrared data. J Geovisualiz Spatial Analys 2(2):14 surface temperature variations and evaluating the impact of cloud
Roy DP, Wulder MA, Loveland TR, Woodcock C, Allen RG, Anderson cover using time series of Landsat TIR data. Remote Sens Environ
MC, Helder D, Irons JR, Johnson DM, Kennedy R (2014) Landsat- 140:267–278. https://doi.org/10.1016/j.rse.2013.09.002
8: science and product vision for terrestrial global change research. Williams DL, Goward S, Arvidson T (2006) Landsat: Yesterday Today
Remote Sens Environ 145:154–172 Tom 72(10):1171–1178
Sadiq Khan M, Ullah S, Sun T, Rehman AU, Chen L (2020) Land-use/
Wulder MA, White JC, Loveland TR, Woodcock CE, Belward AS,
land-cover changes and its contribution to Urban Heat Island: a case
Cohen WB, Fosnight EA, Shaw J, Masek JG, Roy DP (2016) The
study of Islamabad, Pakistan. Sustainability 12(9):3861
global Landsat archive: status, consolidation, and direction. Remote
Sobrino J, Raissouni N, Li Z-L (2001) A comparative study of land
Sens Environ 185:271–283
surface emissivity retrieval from NOAA data. Remote Sens
Environ 75(2):256–266 Xu Y, Gao X, Shen Y, Xu C, Shi Y, Giorgi a (2009) A daily temperature
Sobrino JA, Jiménez-Muñoz JC, Paolini L (2004) Land surface temper- dataset over China and its application in validating a RCM simula-
ature retrieval from LANDSAT TM 5. Remote Sens Environ 90(4): tion. Adv Atmos Sci 26 (4):763–772
434–440 Yang L, Cao Y, Zhu X, Zeng S, Yang G, He J, Yang X (2014) Land
Sobrino JA, Jiménez-Muñoz JC, Sòria G, Romaguera M, Guanter L, surface temperature retrieval for arid regions based on Landsat-8
Moreno J, Plaza A, Martínez P (2008) Land surface emissivity re- TIRS data: a case study in Shihezi, Northwest China. J Arid Land
trieval from different VNIR and TIR sensors. IEEE Trans Geosci 6(6):704–716
Remote Sens 46(2):316–327 Ying M, Zhang W, Yu H, Lu X, Feng J, Fan Y, Zhu Y, Chen D (2014) An
Song F, Zhu Q, Wu R, Jiang Y, Xiong A, Wang B, Zhu Y, Li Q (2007) overview of the China Meteorological Administration tropical cy-
Meteorological data set for building thermal environment analysis of clone database. J Atmos Ocean Technol 31(2):287–301
China. Proceedings of the 10th international building performance Young NE, Anderson RS, Chignell SM, Vorster AG, Lawrence R,
simulation association conference and exhibition, Beijing, pp 9–16 Evangelista PH (2017) A survival guide to Landsat preprocessing.
Sun D, Pinker RT (2003) Estimation of land surface temperature from a Ecology 98(4):920–932. https://doi.org/10.1002/ecy.1730
geostationary operational environmental satellite (GOES-8). J Yu X, Guo X, Wu Z (2014) Land surface temperature retrieval from
Geophys Res: Atmos 108 (D11) Landsat 8 TIRS—comparison between radiative transfer equation-
Team RC (2016) R: A language and environment for statistical comput- based method, split window algorithm and single channel method.
ing. R Foundation for Statistical Computing, Vienna Remote Sens 6(10):9829–9852
Trigo IF, Monteiro IT, Olesen F, Kabsch E (2008) An assessment of Zhang Z, He G (2013) Generation of Landsat surface temperature product
remotely sensed land surface temperature. J Geophys Res: for China, 2000–2010. Int J Remote Sens 34(20):7369–7375
Atmospheres 113 (D17) Zhou J, Li J, Zhang L, Hu D, Zhan W (2012) Intercomparison of methods
Urban M, Eberle J, Hüttich C, Schmullius C, Herold M (2013) for estimating land surface temperature from a Landsat-5 TM image
Comparison of satellite-derived land surface temperature and air in an arid region with low water vapour in the atmosphere. Int J
temperature from meteorological stations on the pan-arctic scale. Remote Sens 33(8):2582–2602
Remote Sens 5(5):2348–2367. https://doi.org/10.3390/rs5052348 Zhu W, Lu A, Jia S (2013) Estimation of daily maximum and minimum
USGS (2016a) Department of the Interior U.S. Geological Survey, vol 8. air temperature using MODIS land surface temperature products.
2.0 edn. USGS Remote Sens Environ 130:62–73. https://doi.org/10.1016/j.rse.
USGS (2016b) Landsat-8 data users handbook version 2.0. EROS, Sioux 2012.10.034
Falls, South Dakota
Valor E, Caselles V (1996) Mapping land surface emissivity from NDVI:
application to European, African, and south American areas. Publisher’s note Springer Nature remains neutral with regard to jurisdic-
Remote Sens Environ 57(3):167–184. https://doi.org/10.1016/ tional claims in published maps and institutional affiliations.
0034-4257(96)00039-9