Accuracy Assessment of a Post-processing Differential GPS Device:

A Case Study in Kaman and Hacturul, Central Turkey

Yuichi S. HAYAKAWA and Hiroomi TSUMURA
Tsukuba

Kyoto

ABSTRACT
The accuracy of a differential GPS (DGPS) device is preliminarily examined
prior to using it for topographic survey in Kaman and Hacturul in central Turkey.
Measurement duration is revealed to most dominantly affect the resultant position
accuracy, and a certain minimum time is necessary for each measurement point to
acquire the position accuracy needed for detailed field survey. This is not always
clearly indicated by device manufacturers. Users should therefore be aware of the
characteristics and capability of devices in advance of use in field survey.
Keywords: DGPS, accuracy assessment, field survey, measurement duration

INTRODUCTION
The use of GPS (Global Positioning System) is
fundamental for measuring locations and shapes of
archaeological sites and structures (e.g., Fenwick 2004;
Miyahara 2006). However, commercial GPS devices
often do not correct for GPS signal errors, giving only
limited accuracy (usually tens of meters), and so can be
inappropriate for use in field surveys requiring higher
accuracies. DGPS (Differential GPS) devices that are
capable of applying corrections to the received signals
have better accuracy, on the submeter scale. However,
because the accuracies stated by the manufacturers of
GPS devices are based on measurement testing under
a limited range of conditions, they often do not reflect
the accuracy during use in field surveys. It is therefore
important to examine how a particular device works at
any field site before using it for detailed measurements.
Here we examine the accuracy characteristics of DGPS
devices used at two field sites (Kaman and Hacturul
in central Turkey), prior to using the devices for
measurements in topographic surveys at the sites.
Although we tested only one type of DGPS device

here due to financial concerns, our assessment is not

limited to a specific product, and we aim to show the
importance of assessing the functionality of any GPS
device before using it for field survey.

METHOD
We used a portable DGPS device, MobileMapper
Pro by Magellan Navigation Inc (Fig.1). This DGPS
device is small (16.57.33 cm) and light (220 g), so
can readily be carried around in the field. The device
works continuously for about 8 hours with two AA
cell batteries. The GPS signal data received from the
satellites are recorded onto a SD memory card. The GPS
signals can be received by either built-in or external
antenna; we used the built-in antenna, which is most
practical. This DGPS device is capable of differential
correction of GPS signals by two means: real-time
SBAS (Satellite-Based Augmentation System) correction
and post-processing correction. The SBAS correction
uses another signal comprising correction information,
such as atmospheric conditions, which is transferred

266

Y.S. HAYAKAWA and H. TSUMURA

Fig. 1 DGPS device tested (MobileMapper Pro by Magellan

Navigation Inc.).

from ground-based public measurement stations. GPS

measurement with SBAS correction can be carried out
by a single device. The SBAS in the American region
is named WAAS (Wide Area Augmentation System), in
Europe EGNOS (European Geostationary Navigation
Overlay Service), and in east Asia MSAS (MTSAT-based
Satellite Augmentation System). The manufacturers
stated accuracy of position measurements with SBAS
correction is 23 m. Another method of differential
correction uses two or more devices for measurement,
one of which is used as a base station and others as
mobile stations. The recorded GPS signals in the devices
are collated after measurement to carry out differential
correction, so the method is called post-processing
correction. The manufacturers stated position accuracy
after post-processing correction is less than a meter
(Thales Navigation 2004).
DGPS measurements were carried out at two sites,
Kaman (392042N, 334724E) and Hacturul

AAS XVII

(394220N, 321316E) in central Anatolia, Turkey.

Measurement of DGPS positioning for a relatively
short time duration was undertaken at Kaman, and
measurement for a relatively long duration was
undertaken at Hacturul as a component of topographic
measurement of the tepe. The two sites have similar
conditions of latitude, climate and vegetation, and the
reception status of GPS signals was almost the same
for the two sites, with open sky (almost no shielding by
topographic reliefs, trees and/or buildings), fair weather,
dry air condition and daytime air temperature of ca.
3040C. DGPS measurements were made over a 200- to
300-m diameter area at Kaman and ca. 1-km diameter at
Hacturul.
Three DGPS devices were used for the
measurements, one of which was set as the base station
and the other two as mobile stations. The base station
was set at a fixed location at each site on each day.
Measurement readings with the mobile devices were
taken with differing measurement durations. At Kaman,
the measurements were taken with a relatively short
duration (~15 min); 131 measurements were taken,
all in one day (July 23, 2007). At Hacturul, the
measurements were taken with a relatively long duration
(~ 40 min each); 114 measurements were carried out over
10 days between July 25 and August 8, 2007.
After the field measurements, the DGPS data
were imported into a PC via SD memory card, and the
DGPS log data of the base and mobile devices were
post-processed using bundled software (MobileMapper
Office), giving corrected coordinates of the measured
points in the UTM projection (Zone 36N).

RESULTS AND DISCUSSION

Accuracy by SBAS-based differential correction at base
station
The DGPS logs measured at each base station were
corrected based only on SBAS information, and here we
show daily changes of the SBAS-corrected coordinates
of the base station device at Hacturul during the 10-day
measurement (Fig. 2). The daily average horizontal
coordinates of the base station (XY in the UTM
projection), as well as the average vertical coordinates

2008

Accuracy Assessment of a Post-processing Differential GPS Device

267

Fig. 2 Daily changes of position coordinates of DGPS base station. (A) Horizontal coordinates in UTM
Zone 36N projection. (B) Elevation.

(Z), change by up to 2.8 m and 4.8 m, respectively,

despite the fixed location of the DGPS base device
on the land surface. These values, no better than the
manufacturers stated accuracy (23 m), represent actual
accuracy with the SBAS differential correction at this
site. The overall measurement period each day was 46
hours, so the lower accuracy was not due to insufficient
measurement duration. The fluctuation of the coordinates
on the order of meters therefore represents the ability of
SBAS correction even when the measurement duration is
sufficiently long. At the study sites, the available SBAS
satellites (EGNOS) are located at a low angle in the sky,
and thus the SBAS correction might work relatively
ineffectively.
Accuracies by post-processed differential correction for
mobile stations
Using the post-processing software (MobileMapper
Office) designed for the DGPS devices, the measured
GPS log data, comprising number of satellites captured,
PDOP (position dilution of precision) and measure
duration, are retrieved and horizontal error and
vertical error are automatically computed. Since
these errors rely solely on the log data, they should
be further examined with some other comparable data
having better accuracy, for more robust assessment.
These errors are therefore regarded as an index of the
absolute or true error.

Fig. 3 Relationships between number of satellites, PDOP and

horizontal and vertical errors, based on two DGPS devices used
as mobile stations at Kaman.

268

Y.S. HAYAKAWA and H. TSUMURA

9HUWLFFDOHUURUP

+RUL]RQWDOHUURUP








AAS XVII

















'XUDWLRQRIPHDVXUHPHQWV
'XUDWLRQRIPHDVXUHPHQWV
Fig. 4 Relationships between measurement time and horizontal (A) and vertical (B) errors. Outlined
symbols are the data of Kaman, and black solid ones are the data of Hacturul.

Table 1. A summary of relationships between measurement

duration and the error of post-processed DGPS coordinates
Duration of
measurement (min)

Position errors after post-processing

correction

<2

1 - 7 m (no dependence on duration)

ca. 1 m

10

All data are less than 1 m

20

All data are less than 0.5 m

In general, the accuracy of GPS positioning is better

when the number of satellites captured is larger, or the
PDOP value is smaller. Our data at both Kaman and
Hacturul show that the error value tends to be small,
especially when the number of satellites is more than 6
and/or the PDOP value is less than 3 (Fig. 3), although
the full relationship between errors versus number of
satellites and PDOP is unclear. The error seems to be
much more clearly related to the measurement duration
(Fig. 4), suggesting the dominant effect of measurement
duration on the accuracy of DGPS measurement.
According to these data, measurement duration
should be no shorter than 2 minutes in order to let the
horizontal and vertical errors be less than ca. 1 m, in
turn, to give the accuracy of 1 m. Furthermore, more

than 10-minute duration is necessary to make all the

errors less than 1 m, and more than 20-minute duration
is needed for the errors to be less than 0.5 m. In contrast,
when the duration is shorter than 2 minutes, the errors
do not correlate with the measurement duration and
fluctuate within the range of 1 to 7 m (Fig. 4, Table
1). These values are below the manufacturers stated
accuracy of the DGPS device (<1 m), indicating
that measurement duration shorter than 2 minutes is
insufficient for post-processing correction.

CONCLUSIONS
This paper examined the accuracy of a postprocessing DGPS device for field studies. The accuracy
or errors of DGPS positioning with post-processing
differential correction is dominantly affected by the
duration of the measurement. It was found that 2 minutes
is the minimum necessary time to enable efficient
post-processing differential correction for the DGPS
devices used. The position accuracy of SBAS-based
differential correction is several meters, even when the
measurements are taken over a period as long as 6 hours.

2008

Accuracy Assessment of a Post-processing Differential GPS Device

269

Further assessments of other GPS devices are

necessary to generalize these results, with different
types of GPS in different areas under various conditions.
Nonetheless, when using any kind of device, it is
important to examine and quantify the practical ability
of such devices prior to actual use in field survey. As a
matter of course, this kind of analysis on the accuracy
of devices is often necessary for the users, even if the
vendor provides such information on the specification
and characteristics of the devices.
It should be noted that some troubles in device
operation occurred during the survey: for instance,
the liquid crystal displays of two of the devices were
malfunctioned and the display barely worked. The most
serious trouble was corruption of data written to the SD
memory card in the device, and this frequently caused
loss of some of the DGPS log data. Furthermore, failures
in finalizing the GPS log data occurred when the battery
died or at accidental shut-down of the device. These
troubles may have occurred due partly to the strong
sunlight and heat at the sites, although such troubles
also occurred later in other areas (such as in Japan)
without strong sunlight. The device itself may have
some vulnerability in the display and memory systems,
and their cause should further be investigated by the
provider of the device. We users, however, should also be
careful to avoid accidents in devices we use in the tough
environments of field survey.

Miyahara, K.
2006 G P S T O I S E K I C H O U S A N Y U M O N
(Introduction to GPS and archaeological
excavation), in: T. Uno (ed.), JISSEN KOUKOGAKU GIS (Practice in Archaeological GIS),
NTT Press, pp. 3338 (in Japanese).
Thales Navigation
2004 White Paper: MobileMapper Post-processed
Accuracy, Thales Navigation.
Takahashi, A., T. Oguchi and H. Sugimori
2003 Effects of Digital Elevation Model resolution
on topographic representation: A case study in
the Tama area, western Tokyo, Geographical
Review of Japan 76, pp. 800818.
Tsumura, H., and S. Suzuki
2006 T h e A r c h a e o l o g i c a l I n f o r m a i c s a n d
Spatiotemporal Digital Archive System
(AISDAS) 1. GIS-based investigations in
Kaman-Kalehyk and surrounding areas using
AISDAS, AAS XV, pp. 181189.

BIBLIOGRAPHY

Hiroomi Tsumura

Fenwick, H.
2004 Ancient roads and GPS survey: modelling the
Amarna Plain, Antiquity 78, pp. 880885.
Hayakawa, Y. S., and K. Kashima
2006 To p o g r a p h i c m a p c o n s t r u c t i o n u s i n g a
handheld laser range finder and GIS at KamanKalehyk and Kltepe, AAS XV, pp. 191195.
Hayakawa, Y. S., T. Oguchi, J. Komatsubara, K. Ito, K.
Hori and Y. Nishiaki
2007 R a p i d o n - s i t e t o p o g r a p h i c m a p p i n g
with a handheld laser range finder for
a geoarchaeological survey in Syria,
Geographical Research 45 (1), pp. 95104.

Doshisha University

Yuichi S. Hayakawa

Geoenvironmental Sciences

Graduate School of Life and Environmental Science

University of Tsukuba
Ibaraki 305-8572
Japan

Faculty of Culture and Information Science

Kyoto 610-0394
Japan