This article was downloaded by: [Deakin University Library]

On: 12 August 2015, At: 14:21

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: 5 Howick Place, London, SW1P 1WG

Papers in Applied Geography

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/rpag20

Horizontal Accuracy Assessment of

Global Positioning System Data from
Common Smartphones
a a a
Tyler W. Jones , Luke Marzen & Art Chappelka
a
Auburn University
Published online: 22 Apr 2015.

Click for updates

To cite this article: Tyler W. Jones, Luke Marzen & Art Chappelka (2015) Horizontal Accuracy
Assessment of Global Positioning System Data from Common Smartphones, Papers in Applied
Geography, 1:1, 59-64, DOI: 10.1080/23754931.2015.1009304

To link to this article: http://dx.doi.org/10.1080/23754931.2015.1009304

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
 Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [Deakin University Library] at 14:21 12 August 2015
 Horizontal Accuracy Assessment of Global Positioning System Data
from Common Smartphones

Tyler W. Jones, Luke Marzen, and Art Chappelka

Auburn University

Smartphones have become ever-present in our lives since the launch of the Apple iPhone in 2007. Since then, the number of
smartphones in use has climbed to over 1 billion worldwide. Many users are attracted to the myriad of apps that these devices
offer, including location-based services (LBS) that allow users to track their current location. In this study we seek to establish
some preliminary results concerning the horizontal accuracy of several common smartphones. Many of the devices used in this
study represent several generations of the same device with developmental and technological upgrades differentiating them from
one another. Location coordinate data were collected using volunteer students and their smartphones in the study and compared
to RTK corrected benchmarks to assess horizontal accuracy. Each benchmark represented different types of local obstruction
that have plagued traditional Global Positioning System (GPS) receivers for years. The objective is to create some preliminary
results of smartphone LBS accuracies that can be used as a baseline in future studies. Keywords: A-GPS, GPS, horizontal
accuracy, smartphone.
Downloaded by [Deakin University Library] at 14:21 12 August 2015

T he objective of this study is to generate some

preliminary results that quantify the horizontal
accuracy of commonly available smartphones in real-
European GALILEO and the Chinese COMPASS
(under development). For this reason GPS has
emerged as the leading technology to provide location
world scenarios. According to Strategy Analytics, a information to location-based services (LBS) such as
global research and consulting firm, there were an navigation, tracking, and emergency services (Zand-
estimated 1.038 billion smartphones in use by October bergen 2009). These services have become a mainstay
2012 (MobiThinking 2014). That is up from the esti- of modern commerce and entertainment, as evidenced
mated 708 million smartphones in use at the end of by the more than 1.1 million applications, many of
2011, which is a staggering growth rate of 46.6 per- which use location, available in the Apple App Store as
cent. The widespread availability of these devices and of April 2014 (148Apps 2014).
their potential for contribution to both academic and GPS receivers work well in ideal conditions such as
commercial endeavors make determining their relative clear, precipitation-free weather with an adequate
accuracy an important goal. Currently, many develop- view of the sky, but in many cases it can be difficult for
ers and researchers are unaware of the real-world spa- a GPS receiver to accurately track its position due to
tial accuracy of the devices they already possess. overhead obstructions such as tall buildings that
Providing these data could give these individuals the weaken the signal (< –172 dBW). This has led many
knowledge that expensive stand-alone equipment is manufacturers to use high-sensitivity GPS (HSGPS).
not necessary when doing field work, but that the In addition to this, many smartphones use assisted
smartphones in their pockets will suffice just fine. GPS (A-GPS) messages to reduce the time-to-first-fix
Clearly the smartphone has become a ubiquitous (TTFF; Zandbergen 2009). Stand-alone GPS
part of our everyday lives with all signs pointing to the receivers can take up to several minutes to acquire a
devices only becoming more prevalent in the coming position so manufacturers use other available data to
years. Currently the two most dominant operating sys- speed up the process. For instance, many A-GPS sys-
tems are the Android and iOS developed by Google tems use the already available GPS-synced atomic
and Apple, Inc., respectively (Watzdorf and Micha- time employed by a Code Division Multiple Access
helles 2010; MobiThinking 2014). Many of the appli- (CDMA) cellular network. Also, because cellular car-
cations that are run on these smartphones provide the riers maintain very precise locations of their cellular
user with information tied to the device’s current loca- towers, an A-GPS system can take measurements of
tion. These current location data can be gathered in signal strength for all the towers within range and
several different ways ranging from the use of an inter- trilaterate the area where the signals all overlap.
nal GPS chipset to communicate with satellites in This extra type of positional information is gener-
orbit, using cell-tower based triangulation, and wi-fi ally limited to urban areas with sufficient cell tower
positioning (Watzdorf and Michahelles 2010). density, however. Wi-fi signals have also been used
The first Global Positioning System (GPS) satellite in conjunction with cellular networks to gather
was launched in 1978 by the United States, and in the position data and were first introduced by Apple in
following decades, became the dominant global navi- an upgrade to the original iPhone’s operating sys-
gation satellite system (GNSS). Since then, only the tem (OS) in early 2008. This was likely a tempo-
Russian GLONASS constellation has become opera- rary measure on Apple’s part because the original
tional as a competitor with others such as the iPhone did not have an onboard GPS receiver,

Papers in Applied Geography, 1(1) 2015, pages 59–64 © Copyright 2015 Applied Geography Conferences.
Published by Taylor & Francis Group, LLC.
 60 Jones et al.

which did not make its debut until the release of A-GPS overcomes these signal limitations by incor-
the iPhone 3G (Fleishman 2011). porating external data and resources, such as an assis-
After a brief review of A-GPS technology, this study tance server and reference network, to assist in
investigates the horizontal accuracy of location data determining position (Zhang et al. 2010). An assis-
from several commonly available smartphones loca- tance server has the advantage of being able to access
tion collected over survey grade benchmarks. The data on a reference network along with far superior
objective of this article is to establish some preliminary processing capabilities compared to a GPS receiver.
findings for the horizontal location of several models The communication between the GPS receiver and
of smartphones and cellular-enabled tablets and offer the assistance network occurs via a wireless link. This
possible explanations for differences in accuracy. combination of efforts will produce results that are
Overall, the goal is to help determine which models of more accurate than that of a stand-alone GPS. There
these devices might be the most useful for in situ field are three types of information the assistance server
data collection in research projects across varying will provide to the GPS receiver: (1) GPS satellite
disciplines. orbital metrics and clock information; (2) initial posi-
tion and time estimate; and (3) satellite selection,
range, and range-rate information (LaMance,
Literature Review DeSalas, and Jarvinen 2002). The data collected are
Downloaded by [Deakin University Library] at 14:21 12 August 2015

processed for a position solution by the assistance

Several previous studies on the horizontal accuracy of server instead of the GPS receiver, which makes the
GPS-enabled smartphones have been conducted. receiver’s sole function to collect range information.
Most have found that the onboard sensors, such as The received GPS signals are frequency shifted due to
that of the Apple iPhone, offer consistently poorer the difference in relative motion of the GPS satellite
horizontal positioning than dedicated navigational and the receiver, a phenomenon known as the Dopp-
GPS receivers (Zandbergen 2009; Klimaszewski-Pat- ler effect (Parkinson and Spilkner 1996).
terson 2010). These devices are also susceptible to This Doppler shifted frequency must be locked
local interference, such as where the receiver is located on to by the receiver, but without the ancillary data
on the body or the orientation of the smartphone in a supplied by the assistance server, all possible fre-
person’s hand, as this tends to vary in a real-world sce- quencies would have to be searched individually.
nario (Blunck, Kjærgaard, and Toftegaard 2011). With A-GPS, this time-consuming step can be sig-
Given that modern smartphones’ A-GPS chipsets use nificantly shortened by incorporating known satel-
cell-based triangulation, a device’s positional accuracy lite position and velocity data along with initial
can also be greatly affected by whether it is being used receiver position data from cell tower trilateration
in an urban setting with higher cellular base station to limit the number of frequencies that need to be
density or in a rural setting with a much lower base searched. These techniques have the effect of
station density (Weiss 2003; Lin and Juang 2005; reducing the TTFF, increasing battery life by
Mohr, Edwards, and McCarthy 2008). Another such reducing the processing load required by the
study was able to collect more than 1,000 observation receiver, and increasing receiver sensitivity by
points over a large area; however, the accuracy analysis greatly reducing the number of frequency bins that
relied solely on the accuracy being reported by the need to be searched, allowing the receiver more
device itself rather than ground reference measure- time to dwell on each bin (LaMance, DeSalas, and
ments (Watzdorf & Michahelles 2010). Jarvinen 2002). The increased sensitivity due to
GPS receivers have two basic operating modes. The additional dwell time allows receivers to make range
first occurs when a receiver is powered on and begins measurements at signal strengths that are below
to acquire a signal and the second occurs once a posi- conventional thresholds, effectively allowing the uti-
tional lock has been made, which continues to track lization of data that would otherwise not have been
the receiver as it moves. Once a receiver is in tracking available.
mode, it will remain so until the unit is powered off or Cellular networks also provide positional data
becomes unable to maintain enough satellite signals to within an A-GPS architecture. Cellular service areas
determine position. GPS satellites only deliver a 27W are spatially broken up by the base station nearest and
signal from orbital altitudes of over 12,000 miles and supplying the strongest signal strength to that area.
by the time this signal reaches the receiver it has been Positional tracking of an active cellular receiver is
reduced to no more than –153 dBW. Signal strength achieved by using the two-way radio connecting the
is broken into three categories: (1) strong signal receiver to the network. In its simplest form, a rough
strength > –172 dBW, (2) weak signal strength positional approximation can be achieved using cell
between –172 dBW and –180 dBW, and (3) very weak identification to place the cellular receiver within the
signal strength between –180 dBW and –190 dBW service area of the nearest base station. This method
(Global Positioning Systems Wing 2010). Conven- only allows for a device to be located to within a par-
tional GPS receivers will work well if they have a ticular service area and the positional accuracy is
strong signal, but have problems tracking and acquir- directly linked to the size of any particular service cell.
ing a position at all with weaker signals. Other techniques can be employed if, for instance, a
 Horizontal Accuracy Assessment of GPS Data from Common Smartphones 61
cellular service area is divided into sectors using direc- Methods
tional base station antennae. Another example would
be to use the received signal strength data for The methodology consists of collecting several GPS
increased positional accuracy, though signal strength coordinate pairs over previously established bench-
can vary considerably due to topography, buildings, marks using several different smartphone models that
and fading. These types of techniques are referred to use different cellular networks. Thirty-three different
as enhanced cell identification (Zandbergen 2009). benchmarks were established across the Auburn Uni-
When a device is within the range of several base versity campus (Figure 1) using a Leica GS14 RTK
stations time distance of arrival (TDOA) calcula- (Leica Geosystems, Norcross, GA) corrected GPS
tions may be employed. The accuracy of these receiver capable of subcentimeter horizontal accuracy.
methods is dependent on the spatial distribution and The benchmarks were collected in the East Alabama
density of a network’s base stations as well as the State Plane projected coordinate system in U.S. feet
reliability of the signal measurements (Bensky using the North American Datum of 1983 (NAD83).
2008). Base station density has the greatest impact Each benchmark was installed in locations with partic-
on positional accuracy because it differs so greatly ular characteristics in regard to local interference
from urban to rural settings. These differences in (open, next to large buildings, under canopy cover) to
horizontal accuracy have been observed to be incorporate the most and least ideal conditions. This
Downloaded by [Deakin University Library] at 14:21 12 August 2015

between fifty to several hundred meters in urban occurs when GPS signals encounter interference
areas and several hundred meters to several kilo- between the user’s receiver and the satellite in orbit.
meters in rural areas (Weiss 2003; Lin and Juang In urban environments this can result in a phenome-
2005; Mohr, Edwards, and McCarthy 2008). Signal non called multipath, where the signal has bounced
reliability is the other major factor in determining from the reflective surfaces of nearby buildings and
horizontal accuracy. This depends on the cellular confused the receiver as to its actual horizontal posi-
network’s bandwidth which would make the Code tion. Other occurrences such as heavy canopy cover
Division Multiple Access (CDMA) signal less reli- can result in the satellite’s signal being weakened to
able than the Global System for Mobile Communi- the point of being undetectable by the user’s receiver.
cations (GSM) signal due to its less reliable For this project, six of the points were placed under
bandwidth (Zandbergen 2009). This is due to the moderately dense canopy cover south of the main
differing ways each technology manipulates and campus, with another point placed underneath light
broadcasts signals of similar frequencies. canopy cover in central campus. Twenty-one points

Figure 1 Map of RTK benchmarks.

 62 Jones et al.

were located near large buildings with the view of the longitude and latitude of student data points compared
sky obstructed in every cardinal direction represented. to that point’s respective RTK point were determined
The final five points were located in an open field and and squared to normalize negative values. These
on an open street with no obstructions nearby. A por- squared differences were then added together with the
table wi-fi hotspot device was used to connect the square root of the resulting number representing that
GPS sensor to the Internet and enable RTK correc- data point’s horizontal accuracy. Average distance
tion. Once established over a benchmark, the sensor error by device could then be calculated by taking the
used multipoint averaging to achieve a horizontal sum of each observation’s error and dividing by the
accuracy of not less than .08 feet. number of total observations.
With the benchmarks installed, twenty-five univer-
sity students were recruited to gather positional data
from some of the thirty-three benchmark locations Results and Discussion
using their personal smartphones. Due to time and
scheduling constraints, not every student was able to Results from averaging the horizontal accuracy of each
visit every benchmark within the study. Several smart- device are shown in Table 2. The A-GPS values range
phone models were used in the study (Table 1), with from a low of 28.91 feet for the Apple iPhone 4 to a
each having either an Android OS developed by Goo- high of 438.81 feet for the Android Razr. As this was a
Downloaded by [Deakin University Library] at 14:21 12 August 2015

gle or an iOS US developed by Apple. Each device volunteer project subject to whatever model of smart-
was installed with the One Touch Location GPS phone students already possessed, there are not an
application developed by Creative Workline. This is a equal number of observations per device. In this case
simple, free app that displays the device’s current lati- the number of Apple iPhone 4 observations is twenty-
tude and longitude using decimal degrees in the seven compared to the Android Razr, which had only
World Geographic Survey (WGS) of 1984 geographic four observations. This likely explains some of the dif-
coordinate system. Each student was accompanied by ference in horizontal accuracy.
a member of the research team and was instructed to As the points were represented visually on a map, it
activate the app, thereby activating the onboard GPS was noticed that although the smartphone application
chipset, and to disable their wi-fi connections several is designed to display in decimal degrees up to six deci-
minutes prior to arriving at the first benchmark. Dis- mals many of the device’s chipsets do not have the sen-
connection from the local wi-fi network was required sitivity to record at this level of precision, resulting in
because the network did not cover all of the installed truncation. Many appear to only have a spatial sensi-
benchmarks and would introduce error into the tivity of one to two meters, which corresponds to five
results. decimals. This created a gridding effect noticeable in
All of the students’ coordinate data were collected in Figure 2. This gridding effect also explains why it
the form of Microsoft Excel spreadsheets. The data appears that less than the twenty-seven A-GPS obser-
were checked for proper formatting, combined into a vations that were actually taken at this benchmark are
single spreadsheet, and imported into ESRI’s ArcMap shown. Several of the observations lie in the exact
software as XY data. The data were then converted same position due to the onboard GPS chipset only
from their geographic coordinate system into the East registering measurements in the same spatial intervals.
Alabama State Plane projected coordinate system in Overall the average positional error for all eleven
U.S. feet using ArcGIS Desktop’s Project (Data Man- models of smartphone was 67.47 feet. Average posi-
agement) tool. Once projected into matching coordi- tional error for all tested products using the Apple
nate systems, the data were analyzed for horizontal iOS was 44.79 feet, whereas the average positional
accuracy as compared to the RTK benchmarks. To error for all tested products using the Android OS
determine this, a standard horizontal accuracy equa- was 207.25 feet. Again, much of this can likely be
tion was employed where the difference between both attributed to Apple products having over five times as

Table 1 List of devices Table 2 Average error by device

Device Operating system Device Average distance of error (feet)

HTC One X Android iPhone 4 28.91

Droid Android iPhone 5 29.70
Droid Ultra Android iPhone 5s 30.15
Droid Razr Android Android HTC One X 33.62
Droid X2 Android iPad 35.80
Galaxy 3 Android Galaxy S4 Active 61.10
Galaxy S4 Active Android iPhone 4s 120.35
iPad (Third generation) Apple iOS Droid X2 182.43
iPhone 4 Apple iOS Motorola Droid 243.43
iPhone 4s Apple iOS Droid Ultra 286.44
iPhone 5 Apple iOS Samsung Galaxy III 290.88
iPhone 5s Apple iOS Android Razor 438.81
 Horizontal Accuracy Assessment of GPS Data from Common Smartphones 63
Downloaded by [Deakin University Library] at 14:21 12 August 2015

Figure 2 A-GPS observations near benchmark.

many observations in this study compared to products Conclusions

using the Android OS. Results from averaging the
horizontal accuracy of each device by cellular pro- The objective of this project was to explore hori-
vider are shown in Table 3. Of the three cellular pro- zontal accuracy of commonly used A-GPS-enabled
viders (Verizon Wireless, AT&T Wireless, T-Mobile devices and offer some preliminary results to assist
Wireless) represented in this study, the highest aver- in further, more comprehensive analysis and inves-
age positional accuracy was T-Mobile, with 36.54 tigation. Results from this project show that many
feet; the lowest average positional accuracy was devices are capable of achieving horizontal accura-
117.97 feet. Verizon Wireless was the most popular cies of approximately 30 feet when tested using
choice of provider among this study’s participants, varying degrees of geographic obstruction.
representing twelve of the twenty-five that partici- Although further investigation is needed, the data
pated and had an average positional accuracy of 81.22 also suggest that horizontal accuracy of any particu-
feet. These findings could be a function of each pro- lar device could be impacted by its cellular network
vider’s particular cellular tower density within the provider. Many devices in this study have de facto
study area. spatial precision of less than the six decimal places
logged using the One Touch Location app, result-
Table 3 Average error by provider ing in a visual gridding effect of the data when
overlaid on a map. Also, it does not appear that
Device AT&T T-Mobile Verizon any significant improvements in A-GPS chipset’s
horizontal accuracy have been developed, although
Droid Ultra 286.44 TTFF as well as overall user experience and satis-
Droid X2 264.14
Galaxy S4 Active 61.10 faction are beyond the scope of this study.
iPad 35.80 A-GPS receivers built into common smartphone
iPhone 4 28.91 models are useful in mass market applications to
iPhone 4s 146.63 43.52
iPhone 5 25.75 34.62 allow users to use LBS. These services’ ability to
iPhone 5s 28.59 45.51 supply information and entertainment has become
Motorola Droid 243.43 an integral and expected portion of the user’s expe-
Samsung Galaxy III 290.88
Total 117.97 36.54 81.22 rience that developers and manufacturers continue
to improve on.
 64 Jones et al.

Acknowledgements MobiThinking. 2014. Mobile marketing: Global mobile

statistics 2014 Part A: Mobile subscribers; handset market
The authors wish to thank Mateus Ismael for help in share; mobile operators. http://mobithinking.com/mobile-
data collection and D. Hartel and E. Kuehler (U.S. marketing-tools/latest-mobile-stats/a. Accessed June 10,
Department of Agriculture Forest Service, Urban For- 2014.
estry South, Athens, GA) for providing technical Mohr, M., C. Edwards, and B. McCarthy. 2008. A study of
LBS accuracy in the UK and a novel approach to inferring
assistance.
the positioning technology employed. Computer
Communications 31(6): 1148–59.
Funding Parkinson, B., and J. Spilkner. 1996. Global Positioning System:
Theory and applications. Vol. 1. Reston, VA: American
Partial support for this project has been received from Institute of Aeronautics and Astronautics.
the U.S. Department of Agriculture Forest Service Watzdorf, S., and F. Michahelles. 2010. Accuracy of
under Challenge Cost Share agreement 13-CS- positioning data on smartphones. In Proceedings of the 3rd
I1330144–060 (“Urban Tree Canopy Comparison International Workshop on Location and the Web, 2–5, E.
Wilde, S. Boll, and J. Schoning, eds. Tokyo, Japan.
Study”) as part of the Urban and Community Forestry
New York, NY: ACM.
Program (S&PF). Weiss, A. 2003. On the accuracy of a cellular location system
based on RSS measurements. IEEE Transactions on
Downloaded by [Deakin University Library] at 14:21 12 August 2015

References Vehicular Technology 52(6): 1508–18.

Zandbergen, P. 2009. Accuracy of iPhone locations: A
148Apps. 2014. App Store etrics. Apple Store applications comparison of assisted GPS, WiFi and cellular positioning.
summary. http://148apps.biz/app-store-metrics/. Accessed Transactions in GIS 13(s1): 5–26.
May 31, 2014. Zhang, L., B. Tiwana, R. P. Dick, Z. M. Mao, Z. Wang, and
Bensky, A. 2008. Wireless positioning technologies and L. Yang. 2010. Accurate online power estimation and
applications (GNSS technology and applications). Norwood, automatic battery behavior based power model generation
MA: Artech House. for smartphones. Proceedings of the Eighth IEEE/ACM/IFIP
Blunck, H., M. B. Kjærgaard, and T. S. Toftegaard. 2011. International Conference 8: 105–14.
Sensing and classifying impairments of GPS reception on
mobile devices. Pervasive Computing: Lecture Notes in
Computer Science 6696: 350–67. TYLER W. JONES is a PhD candidate in the School of For-
Fleishman, G. 2011. How the iPhone knows where you are. estry & Wildlife Sciences, Auburn University, 602 Duncan
MacWorld. http://www.macworld.com/article/1159528/ Drive, Auburn, AL 36849. E-mail: [email protected]. His
how_iphone_location_works.html. Accessed May 20, 2014. research interests include remote sensing, object based image
Global Positioning Systems Wing. 2010. Navstar GPS space analysis, and geographic information systems.
segment/navigation user interfaces. El Segundo, CA: Science
Applications International Corporation.
Klimaszewski-Patterson, A. 2010. Smartphones in the field: LUKE MARZEN is Professor in the Department of Geol-
Preliminary study comparing GPS capabilities between a ogy & Geography, Auburn University, 210 Petrie Hall,
smartphone and dedicated GPS device. Papers of the Applied Auburn, AL 36849. E-mail: [email protected]. His
Geography Conference 33: 270–79. research interests include remote sensing and geographic
LaMance, J., J. DeSalas, and J. Jarvinen. 2002. Assisted information systems and land use/land cover change.
GPS, a low-infrastructure approach. GPS World 13(3):
46–51. http://gpsworld.com/innovation-assisted-gps-a- ART CHAPPELKA is Professor in the School of Forestry &
low-infrastructure-approach/. Accessed May 11, 2014. Wildlife Sciences, Auburn University, 602 Duncan Drive,
Lin, D., and R. Juang. 2005. Mobile location estimation based Auburn, AL 36849. E-mail: [email protected]. His
on differences of signal attenuations for GSM systems. research interests include physiological and edaphic effects of
IEEE Transactions on Vehicular Technology 54(4): 1447–54. several atmospheric contaminants.