See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/364623012

POSITIONAL ACCURACY ASSESSMENT OF HISTORICAL GOOGLE EARTH

IMAGERY

Preprint · October 2022

CITATIONS READS

0 78

9 authors, including:

Chukwuma John Okolie Ikenna Arungwa

University of Cape Town Federal University of Technology Owerri
141 PUBLICATIONS   343 CITATIONS    18 PUBLICATIONS   35 CITATIONS   

SEE PROFILE SEE PROFILE

Olagoke Daramola Michael Orji

University of Lagos University of Lagos
57 PUBLICATIONS   133 CITATIONS    24 PUBLICATIONS   43 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

COVID-19 Research View project

Satellite Derived Bathymetry View project

All content following this page was uploaded by Ikenna Arungwa on 22 October 2022.

The user has requested enhancement of the downloaded file.

Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

POSITIONAL ACCURACY ASSESSMENT OF HISTORICAL GOOGLE

EARTH IMAGERY
Peter C. Nwilo1., Chukwuma J. Okolie1, Johanson C. Onyegbula1, Ikenna D. Arungwa2*,
Owolabi Q. Ayoade1,3, Olagoke E. Daramola1, Michael J. Orji1, Ikechukwu D. Maduako4 and
Imeime I. Uyo1
1
Department of Surveying and Geoinformatics, Faculty of Engineering, University of Lagos,
Nigeria
2
Department of Surveying and Geoinformatics, Faculty of Engineering, Abia State University,
Nigeria
3
Bureau of Lands and Survey, Abeokuta, Ogun State, Nigeria
4
Department of Geinformatics and Surveying, Faculty of Environmental Studies, University of
Nigeria, Nsukka, Nigeria
*Corresponding author email: arungwaikenna@gmail.com

ABSTRACT
Google Earth (GE) is the most popular virtual globe in use today. Given its popularity and
usefulness, most users do not pay close attention to the positional accuracy of the imagery, and
there is limited information on the subject. This study therefore evaluates the horizontal accuracy
of historical GE imagery at four epochs between year 2000 and 2018, and the vertical accuracy of
its elevation data within Lagos State in Nigeria, West Africa. The horizontal accuracies of the
images were evaluated by comparison with a very high resolution (VHR) digital orthophoto while
the vertical accuracy was assessed by comparison with a network of 558 ground control points.
The GE elevations were also compared to elevation data from two readily available 30-metre
digital elevation models (DEMs) – the Shuttle Radar Topography Mission (SRTM) v3.0 and the
Advanced Land Observing Satellite World 3D (AW3D) DEM v2.1. In terms of the horizontal
accuracy, the root mean square errors (RMSEs) are as follows – year 2000 (29.369m), year 2008
(28.391m), year 2012 (10.615m) and year 2018 (10.603m). The most recent GE imagery (year
2018) was the most accurate while year 2000 was the least accurate. This shows a continuous
enhancement in the accuracy and reliability of satellite imagery data sources which form the source
of Google Earth data. Results also portray that the GE images have a tendency to be skewed
towards the western and north-western directions, indicative of systematic error. In terms of the
vertical accuracy, GE elevation data had the highest RMSE of 6.213m followed by AW3D with
an RMSE of 4.388m and SRTM with an RMSE of 3.682m. Although the vertical accuracy of
SRTM and AW3D are superior, Google Earth still presents clear advantages in terms of its ease-
of-use and contextual awareness.

Keywords: Google Earth Imagery, Positional Accuracy, Global Positioning System, Shuttle Radar
Topography Mission

1
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

1. INTRODUCTION
The integration of spatial technologies with the world wide web has led to the evolution of
virtual globes which provide worldwide access to geospatial data (Elvidge and Tuttle, 2008). Allen
(2008) defines a virtual globe as “a 3D software model of the Earth (or other planet) that provides
a user interactivity and freedom to view the globe from different viewing angles, positions, and
overlays of actual or abstract geographic data.” The ease of use of digital virtual globes and their
capacity for display and visualisation of spatial information make them a powerful
communication tool for researchers, decision makers and the general public (Aurambout et al.,
2008). Virtual globes present a simpler alternative to technocratic and costly Geographic
Information System (GIS) software, and this facilitates sharing of spatial data at a global scale (Yu
and Gong, 2012). Virtual globes can be viewed as technological realisations of the Digital Earth
(DE) concept introduced by former United States Vice President Al Gore (Gore, 1998; Liang et
al., 2018); and have led to new paradigms in the concept of Digital Earth (Goodchild et al., 2012;
Pulighe et al., 2016). Digital Earth has been described as “a multiresolution and three-
dimensional visual representation of Earth that would help humankind take advantage of geo-
referenced information on physical and social environments, linked to an interconnected web of
digital libraries” (Gore, 1999 in Liu et al., 2020).

Examples of free and publicly available virtual globes/image services include Google Earth,
Google Maps, NASA World Wind, Microsoft Bing Maps and Apple Maps (Pulighe et al., 2016;
Goudarzi and Landry, 2017). Among these examples, Google Earth (GE) is the most popular and
versatile. It renders a three-dimensional (3D) representation of Earth by the superimposition of
images obtained from satellite imagery with worldwide coverage, aerial photography from local
or national mapping agencies, near-orthophoto collections in GeoPortals and GIS 3D globe.
Google Earth can show various kinds of images overlaid on the surface of the earth and is also a
Web Map Service client. The core technology behind Google Earth was originally developed at
Intrinsic Graphics in the late 1990s. In version 5.0, Google introduced “Historical Imagery”,
allowing users to view images of a region at different epochs and to observe an area's changes over
time (see Figure 1). 3D coverage of cities by Google Earth began in 2012 (Ubukawa, 2013). By
early 2016, it had been expanded from 21 cities in 4 countries to hundreds of cities in over 40
countries, including every US state and encompassing every continent except Antarctica. The very
high resolution (VHR) satellite images on Google Earth have a spatial resolution finer than 5m
(Lesiv et al., 2018). However, the spatial resolution of the images depends on the characteristics
of the satellite such as the altitude and type of instruments (Buka et al., 2015). In reality, GE images
are not spatiotemporally continuous or homogenous but are mosaicked using multiple images from
different periods, different spatial resolutions ranging from 15m to 10cm, and from different
imagery providers (Lesiv et al., 2018). The images are compiled from a wide variety of sources
such as: SPOT 5, Rapid Eye, Earth Resource Observation Satellites (EROS), Meteosat 2, Geoeye
1, and Digital Globe World View 2 satellite (Buka et al., 2015). Since Google Earth images are
sourced from multiple sources, they do not have identical positional accuracy or spatial resolution
(Goudarzi and Landry, 2017). The satellite images are sometimes supplemented with aerial
photographs which have a higher resolution. In places where high resolution imagery is
unavailable, GE defaults to Landsat imagery (Potere, 2008). On the frequency of updates, Google
aims to update satellite imagery of places that undergo frequent changes, once a year for big cities,
every two years for medium-sized cities and up to every three years for smaller cities (Schottenfels,
2020).

2
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 1: Historical Google Earth imagery over a part of Beijing city in China at three periods –
(a) 8 November, 2002 (b) 29 March, 2012 (c) 28 August, 2020. The Historical Imagery slider is
visible at the top left corner of the images

3
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

There is some ambiguity on the source of Google Earth elevation data (Goudarzi and Landry,
2017). It is possibly derived from the Shuttle Radar Topography Mission (SRTM) DEM, the
Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global DEM, and
from Light Detection and Ranging (LiDAR) (Goudarzi and Landry, 2017, Chigbu et al., 2019;
MES Innovation Sdn Bhd, 2020). The recent introduction of elevation data sourced from LiDAR,
interestingly makes it possible for height accuracy of about 5-25cm. Ironically, the location of such
places where LiDAR data covers are not known or revealed by GE (MES Innovation Sdn Bhd,
2020). It is expected that errors inherent in the elevation data sources would naturally propagate
into GE elevation data.

Since the launch of Google Earth in 2005, it has enjoyed ever-increasing popularity from map
makers, pathfinders, navigators, planners, application developers, etc as a free data source
providing a realistic view of the world through satellite images, maps, digital terrain, 3D buildings,
land use information, identification of monuments and locational data. Google Earth imagery has
found wide applications in health geography research (Curtis et al., 2006), land use/land cover
mapping (Hu et al., 2013; Malarvizhi et al., 2016), land conversion studies (Jacobson et al., 2015),
mapping of lakes (Shen et al., 2006), internet GIS (Henry, 2009), urban household surveys (Ngom
Vougat et al., 2019), real estate (Hwang, 2008), and relief/humanitarian efforts (Nourbakhsh et al.,
2006). GE Historical Imagery provides images taken at different periods and this has wide
applications in land use change detection studies (Malarvizhi et al., 2016). Generally, the use of
GE in research projects have been summarised into the following categories: visualisation, data
collection, validation, data integration, communication/dissemination of research results,
modelling, data exploration and decision support (Yu and Gong, 2012). In the scientific
community, its use pertains to earth surface processes, habitat availability, health and surveillance
systems, biology, land use/land cover (LULC), agriculture, landscape etc. (Pulighe et al., 2016).
Comprehensive reviews of earth science applications of Google Earth are provided in Yu and
Gong (2012) and Liang et al. (2018).

Google Earth presents a new paradigm in Digital Earth and in the quest by man to understand the
environment and effectively manage its resources. It also presents a clear advantage to achieving
the United Nations 2030 Agenda for sustainable development. As a virtual globe, Google Earth
connects all parts of the world in a virtual environment with free access to geospatial data to
support global partnerships in attaining the sustainable development goals (SDGs). More so, policy
and decision making at every level (local, national, regional or global) are dependent on up-to-date
geospatial data. Globally, there is a continuous drive by policy makers to deliver sustainable
development within, and in accordance with the templates provided by the SDGs. The SDGs are
earth-centred and driven by geospatial data. For example, without geospatial data in place, the idea
of location-based services would to a large extent remain a mirage. Google Earth is therefore
relevant for achieving SDG 2 (zero hunger), SDG 6 (clean water and sanitation), SDG 9 (industry,
innovation and infrastructure), SDG 11 (sustainable cities and communities), SDG 13 (climate
action), SDG 14 (life below water), SDG 15 (life on land) and SDG 7 (affordable and clean
energy). As a virtual globe, Google Earth is also a crucial tool for bridging the global North –
South divide in terms of access to geospatial data for international partnerships and collaborations.

Given the popularity of Google Earth, users tend to assume that it is a highly accurate source of
information with no doubt on its positional accuracy (Flanagin and Metzger, 2008). However,

4
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

there are questions surrounding the reliability of GE imagery, since very little is known about its
metadata including the sensors, imagery resolutions, and overlay/mosaicking techniques (Pulighe
et al., 2016). According to Paredes-Hernández et al. (2013), Google geographic data products are
only approximations without officially documented accuracies. Wang et al. (2017) note that
Google has been unwilling to release comprehensive information on the accuracy of the GE
archive. It is also mentioned that GE images are also not orthorectified and lack photogrammetric
accuracy (Goudarzo and Landry, 2017). The uncertainty surrounding the horizontal accuracy of
the imagery could lead to feature misrepresentations, and incorrect inferences (McRoberts 2010 in
Pulighe et al., 2016). There are also errors in image alignment manifesting at the transition zones
between mosaicked images on Google Earth (Potere, 2008) (e.g., disjoint shorelines and roads
shown in Figure 2). This presents some uncertainty on the usability of GE imagery for sensitive
applications requiring very high accuracy such as high-precision engineering surveys and
autonomous navigation. The practice of reporting coordinates with a precision that does not match
its accuracy misleads users to believe that it is an accurate source of information (Goodchild et al.,
2012). Moreover, Benker et al. (2011) noted that Google’s representatives stated that the
coordinates provided by Google and the data available in their geographic products are only
approximations and that Google makes no claim to the accuracy of their geographic information
products. A quick check of the GE historical images at some locations (Figure 3) shows that the
magnitude of these horizontal shifts varies with time. In some cases, the positional errors are not
consistent when viewed at different periods with the Historical Imagery slider (see Figure 3).
Another limitation is that little is known about the volume of historical imagery in Google Earth’s
archive and where it can be found (Lesiv et al., 2018).

Positional accuracy is traditionally divided into two classes: horizontal accuracy and vertical
accuracy (Goudarzi and Landry, 2017). Becek et al. (2011) identified the flaws associated with the
positional identities of some known points from Global Elevation Data Testing Facility (GEDTF)
and their corresponding points on GE imagery. A remarkable error of more than 1.5km was noticed
in some cases after measuring the discrepancies using some tools and basic statistics. In Paredes-
Hernández et al (2013), geo-registration and large horizontal errors were shown to occur in GE
imagery. However, the authors suggested the possibility of GE imagery satisfying the horizontal
accuracy requirements of the American Society for Photogrammetry and Remote Sensing
(ASPRS), assessed in terms of root mean square error (RMSE) for x, y and z coordinates, for the
production of “Class 1” 1: 20,000 maps, if a large number of well-defined points are extracted
from areas of high-resolution imagery over rural areas. Mulu and Derib (2019) evaluated the
accuracy of GE imagery in Khartoum, Sudan and showed that the horizontal RMSE was suitable
for producing a Class 1 map of 1:20,000 scale (as recommended by ASPRS, 1990). However, they
pointed out that the resolution of the acquired Google Earth imagery was a major factor affecting
the accuracy of the GE dataset, as coarser resolutions appeared to have higher RMSE values
probably due to less accurate location of control points on such coarse resolutions. Goudarzi and
Landry (2017) assessed the horizontal accuracy of GE in the city of Montreal, Canada using ten
Global Positioning System (GPS) reference points. In their results, the positional accuracy varied
between ~0.1m in the south to ~2.7 m in the north of the city.

5
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 2: Region of image misalignment (within the red circles) on Google Earth imagery in
Lagos, Nigeria - (a) disjoint shoreline, imagery date - December 2018 (b) disjoint roads along
Carter Bridge, imagery date – March 2005

According to El-Ashmawy (2016), the accuracies of DEMs prepared from GE data are only
suitable for certain engineering applications but inadequate for very precise engineering studies. It
might satisfy the vertical accuracy requirements of the ASPRS (1993) standards for the production
of “Class III” contour maps. Other applications of GE elevation data include the preparation of
large-area cadastral, city planning, or land classification maps. In Aba metropolis of south-eastern
Nigeria, Chigbu et al (2019) assessed GE elevation data using a 10.16km elevation profile data

6
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

obtained by means of ground survey as reference. They reported a mean error of 1.65m, RMSE of
2.79m, standard deviation of 2.27m and median absolute deviation of 1.72m for the GE elevations.
However, on the strength of further incisive statistical tests (Mann-Whitney U Test of group and
the t-Test), they concluded that GE elevation data was unfit for any form of levelling operation
that would eventually lead to engineering construction.

Figure 3: Horizontal shifts in positions of features on Google Earth shown at three locations -
Pont au Change bridge on the Seine River in Paris France (a) 2002 (b) 2005; National Stadium in
Abuja Nigeria (c) 2003 (d) 2018; Yonghongcun in Wuhan China (e) 2004 (f) 2019

The issue of GE’s positional accuracy has received little interest from researchers around the
world. Most of the studies discovered in the literature survey, focused only on the horizontal
accuracy and there was little interest in the vertical accuracy of its elevation data. Moreover, a
literature search did not reveal any studies dealing with the issue of horizontal error in historical

7
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

GE imagery. Errors in the geo-registration of GE images could limit the scientific value of the
archive (Potere, 2008). Hence, the present study investigated the horizontal accuracy of historical
GE imagery at four periods between 2000 and 2018, and the vertical accuracy of its elevation data
within Lagos State in Nigeria, West Africa. The horizontal accuracies of the images were assessed
by comparison with a highly accurate digital orthophoto while the vertical accuracy was assessed
by comparison with a network of ground control points. The GE elevations were also compared to
elevation data from two publicly available 30-metre DEMs – the Shuttle Radar Topography
Mission (SRTM) v3.0 DEM and the Advanced Land Observing Satellite World 3D (AW3D) DEM
v2.1. To our knowledge, this is the first study to assess the horizontal accuracy of Google Earth
“historical imagery”. The findings are important to inform users of the reliability of GE imagery
for use in change detection studies and other analyses that involve spatio-temporal variability. It
also provides a critical knowledge base to inform end-users on the quality and reliability of the
data for a myriad of applications.

2. MATERIALS AND METHODS

The study involved acquisition of various datasets relevant to the evaluation of the positional
accuracy. The steps involved in realising this are as outlined in sub-sections below.

2.1 Study Area

The study area is Lagos State in Nigeria, West Africa. Lagos State is located in the south-western
part of Nigeria (Figure 4) and is bounded in the south by the Atlantic Ocean, in the west by the
Republic of Benin and in the north and east by Ogun State. The state was once the administrative
capital of Nigeria between 1976 and 1991. It is currently the commercial capital, a beehive of
commercial and industrial activities, and has an estimated population of over 24 million (Lagos
Digest of Statistics, 2017). It has a total area of about 3,577.28km2, of which 2,792.72km2 is
covered by land and 779.56km2 is water (BudgIT, 2018). The state is geographically located
between longitudes 2o41’55’E – 4o22’00’E and latitudes 6o22’22”N – 6o43’20”N. It has a
generally low-lying terrain with the Lagos and Lekki lagoons as its major water bodies. There are
two major climatic seasons: the rainy season and the dry season. The mangrove swamp forest and
freshwater swamp constitute some of its most dominant vegetation types. Temperature ranges from
20oC - 32oC and the mean annual rainfall exceeds 1,700mm (Nwilo et al., 2020). For the horizontal
accuracy assessment in this study, a digital orthophoto of the University of Lagos was acquired.
The University of Lagos is one of the federal universities in Nigeria, situated within metropolitan
Lagos. It is located between longitudes 3o23’00”E – 3o24’30”E and latitudes 6o30’00”N –
6o31’30”N. As an institution for learning and research, it is surrounded by research infrastructure,
buildings and commercial activities, and also bounded eastwards by the Lagos Lagoon.

8
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 4: Map showing location of the study area

2.2 Description of Datasets

The datasets used are discussed below.

2.2.1 Ground control points

The rectification of the orthophotos was done using highly accurate ground control points (GCPs)
within the University of Lagos. The GCPs were surveyed with the Trimble R8 dual frequency
Global Navigation Satellite System (GNSS). The GNSS field procedure and data processing is
well explained in Okolie (2019) and Gbopa et al. (2021). Essentially, fourteen GCPs (shown in
Figure 5) were signalised on the ground. Ground control signalisation is the selection of ground
control identification style or pattern. The signalisation was done with cross markings using white
emulsion paint to ensure their visibility from a high altitude during the Unmanned Aerial Vehicle
(UAV) survey. The cross markings were approximately 80 – 100cm in length and 15 – 20cm in
breadth. Figure 6 shows two of the signalised GCPs within the university, YTT 28/186 and XST
347. The GNSS observation was then carried out on the GCPs in static mode with about 30 – 40
minutes occupation time on each point. After completion of the survey, the data was downloaded
from the GNSS receivers and post-processed to derive the final coordinates. For the vertical

9
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

accuracy assessment, 558 geodetic GCPs in Lagos State were acquired from the Lagos State
Surveyor General’s Office.

Figure 5: Spatial distribution of the signalised GCPs

Figure 6: Some of the signalised GCPs within the University of Lagos – (a) YTT 28/186 and (b)
XST 347

10
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

2.2.2 Digital orthophoto

The reference dataset for the horizontal accuracy assessment was a digital orthophoto which was
derived from overlapping images of the University of Lagos main campus captured during a UAV
survey conducted in July 2019. A detailed discussion of the UAV flight and orthophoto processing
is discussed in Okolie (2019) and Gbopa et al. (2021). The DJI Phantom 4 Professional UAV with
a camera lens of 84° field of view (FOV) and a focal length of 8.8 mm/24 mm (35 mm format
equivalent) was used for the survey. The gimbal has a controllable range of −90° to +30° and a
maximum controllable angular speed of 90°/s (Gbopa et al., 2021). The flight was conducted at an
altitude of 90m, a UAV speed limit of 15m/s, flight direction of 126°, and with generous overlaps
of 75% fore and 65% side. Figure 7 shows the UAV flight in progress at two locations within the
University of Lagos while Figure 8 shows some images captured from the UAV survey. After the
flight, the raw photos were downloaded and imported into Pix4D Mapper software environment.
The processing with Pix4D Mapper addressed the image alignment, geo-rectification with GCPs,
the processing of point cloud and mesh, and orthophoto generation. The orthophoto was produced
at a ground sampling distance (GSD) of 4.36cm. Gbopa et al. (2021) evaluated the accuracy
(RMSE) of the orthophoto using 7 GCPs of high fidelity and obtained a horizontal and vertical
accuracy of 0.183m and 0.157m respectively. This shows a high level of accuracy adequate for
use as a reference dataset to evaluate the horizontal accuracy of the GE images. Figure 9 shows
the final orthophoto map with 45 points initially selected as reference points for the horizontal
accuracy assessment.

Figure 7: UAV flight in progress within the University of Lagos main campus

11
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 8: Some images from the UAV survey (a) Wetlands to the right of the University of
Lagos Central Mosque (b) Staff Quarters (c) GTBank building (centre) and Erastus Akingbola
Hall (bottom centre) (d) High-Rise Complex (far-right) and a canal running along the southern
boundary of the campus

12
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 9: Orthophoto map of University of Lagos showing the spatial distribution of the initial
45 reference points selected for horizontal accuracy assessment (27 points were retained in the
final selection)

2.2.3 Google earth historical imagery

The Historical Imagery tool in Google Earth Pro v.7.3 enables the viewing of images at different
epochs, and allows for observation of an area's changes over time. Historical Google Earth images
between the years 2000 and 2018 were considered. The criteria for selecting a particular year

13
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

included the image clarity and degree of cloud cover. After a meticulous inspection, the following
periods were selected: 13th December 2000, 7th October 2008, 4th June 2012 and 28th December
2018. Google Earth images are referenced to the WGS84 datum which is based on the GRS80
ellipsoid (Buka et al., 2015). Figure 10 shows the points selected on the historical images to
coincide with the reference points from the orthophoto.

Figure 10: Coincident points on Google Earth imagery for comparison with the reference pints
on the orthophoto (a) 13th December, 2000 (b) 7th October, 2008 (c) 4th June 2012 (d) 28th
December, 2018

2.2.4 Google earth elevation data, SRTM v3.0 and AW3D30 v2.1
Elevation data were extracted from Google Earth at points coincident with the 558 GCPs in Lagos
State. The SRTM v3.0 DEM covering Lagos State was downloaded from the United States
Geological Survey (USGS) Earth Explorer website (https://earthexplorer.usgs.gov/). The SRTM
project was jointly executed by the National Aeronautics and Space Administration (NASA) and
the National Imagery and Mapping Agency (NIMA) of the US Department of Defense. The
mission was designed to use a single-pass radar interferometer to produce a digital elevation model

14
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

(DEM) of the Earth's land surface between 600N and 560S latitude (Farr and Kobrick, 2000). The
SRTM DEM of 1 arc-second resolution which corresponds to 30m spatial resolution at the equator
is distributed in 1° × 1° tiles, and has higher accuracy than the earlier 90m SRTM DEM product
(Mukul et al., 2017). The SRTM mission goal of LE90 error of 16m (RMSE ∼10 m) was assessed
worldwide and validated using dual-frequency, Real Time Kinematic (RTK) GPS data (Üstün et
al., 2016). In the published global assessment report of SRTM DEM, it is stated that the vertical
accuracy meets and exceeds the performance requirements of the mission by a factor of nearly two
in comparison to ground-truth data such as kinematic GPS trajectories on road networks (Mukul
et al., 2015). The EGM96 geoid provides the vertical datum for SRTM.

In 2015, the Advanced Land Observing Satellite (ALOS) World 3D - 30m (AW3D30) DEM was
made available as an Earth topography elevation data product
(https://www.eorc.jaxa.jp/ALOS/en/aw3d30/index.htm). AW3D30 was photogrammetrically
developed using optical imagery collected during the ALOS mission (Tadono et al., 2016). The
ALOS elevation maps were produced at a spatial resolution of 5m with an accuracy of 5m
(standard deviation). In 2016, JAXA released a 30m product, the AW3D30 dataset which was
generated from the earlier 5m product (JAXA 2017). From the accuracy assessments by Caglar et
al. (2018), the AW3D30 surpassed the accuracies of SRTM and ASTER GDEM Version 2. Figure
11 presents the spatial distribution of the 558 GCPs used for vertical accuracy assessment with
SRTM v3.0 and AW3D30 v2.1 shown as the backdrop.

15
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 11: Overlay of 558 geodetic GCPs for vertical accuracy assessment on elevation map of
(a) SRTMv3.0 and (b) AW3D30 v2.1

2.3 Data Exploration and Extraction

Forty-five points (#0 to #44) were initially marked as reference points on the orthophoto for use
in the horizontal accuracy assessment. The points were carefully placed in easily identifiable
locations such as road intersections, roundabouts, and corner points of buildings. The dates were
set on Google Earth using the historical imagery time slider and the easting and northing
coordinates of corresponding points on Google Earth were extracted. After inspecting the positions
of all the points on the orthophoto and Google Earth images, it was observed that some points on
the Google Earth images were located in problematic places such as cloud-covered areas etc. The
problematic points were eliminated thus reducing the final selection to 27 points. To extract the
Google Earth elevations for vertical accuracy assessment, the 558 GCPs were first converted to
keyhole markup language (kml/kmz) format and imported into Google Earth. Next, each point
was zoomed to and the corresponding elevation on Google Earth was copied from the status bar
and pasted directly in a Microsoft Excel sheet. The corresponding elevations from SRTM and

16
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

AW3D were extracted in the ArcGIS 10.1 environment using the ‘extract values to points’ tool in
Spatial Analyst.

2.4 Quantitative Analysis

The coordinate differences between the orthophoto reference points and the coincident points on
the Google Earth images were used as a basic metric in assessing the horizontal accuracy.
Coordinates of points on the orthophoto were subtracted from those of the various epochs for the
GE historical imagery. The differences in eastings and northings are given as follows:

ΔE = EGE – EOP (1)

ΔN =NGE – NOP (2)

Where,
ΔE and ΔN are the differences in Eastings and Northings respectively for the selected points. EGE
and EOP are Eastings of selected points on the GE historical imagery and orthophoto respectively.
NGE and NOP are Northings of selected points on the GE historical imagery and orthophoto
respectively. The linear separation between the position of a point on the orthophoto and on Google
Earth is the horizontal error/shift (S) (Figure 12).

Figure 12: Illustration of horizontal error/shift between a point on the reference orthophoto and
the corresponding position on Google Earth

The assessment of the horizontal accuracy is based on the ASPRS Positional Accuracy Standards
for Digital Geospatial Data version 1.0 (ASPRS, 2014). Horizontal accuracy was assessed using
root mean square error (RMSE) statistics in the horizontal plane (i.e., RMSEx, RMSEy and

17
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

RMSEr), while the vertical accuracy was assessed in the z-dimension. In addition, the standard
deviation (SD) and standard error of the mean (SEM) of the coordinate differences were also
calculated. Figure 13 shows an overlay of orthophoto reference points and GE points in two areas
within the University of Lagos. ASPRS (2014) defines positional accuracy as “the accuracy of the
position of features, including horizontal and vertical positions, with respect to horizontal and
vertical datums.” Based on the ASPRS recommendations for imagery at an estimated pixel size of
60cm, the Google Earth images used in this study are expected to meet the ASPRS Accuracy
Standards of 120cm RMSEx and RMSEy Horizontal Accuracy Class for Standard Mapping and
GIS work. The absolute horizontal accuracy should be less than or equal to 2.448 * 120cm at the
95% confidence level. The corresponding estimates of horizontal accuracy at the 95% Confidence
level were computed using methodologies documented in the National Standard for Spatial Data
Accuracy (NSSDA). The ASPRS standards for vertical accuracy specify absolute vertical accuracy
measures for various classes of elevation data based on their spatial resolution (ranging from 1cm
to 333.3cm). However, due to the uncertainty surrounding the source of GE elevations including
its resolution characteristics, the vertical accuracy assessment is based solely on the reference
GCPs. It was computed using the RMSE statistics for non-vegetated terrain (most of the 558 GCPs
are located in non-vegetated terrain).

The horizontal RMSE (RMSEr ) is given as:

RMSEr = √(RMSEX )2 + (RMSEY )2

= √∑[ΔE 2 + ΔN2 ]⁄n (3)

The horizontal linear RMSE in the X direction (Easting) is given below:
RMSE𝑋 = √∑ (EGE − EOP )2⁄𝑛 (4)
The horizontal linear RMSE in the Y direction (Northing) is given below:
RMSE𝑌 = √∑ (NGE − NOP )2⁄𝑛 (5)
Where:
n - the number of check points

NSSDA Horizontal Accuracy at 95% Confidence Level = RMSEr × 1.7308 (6)

NSSDA Vertical Accuracy = RMSEz (7)

𝑅𝑀𝑆𝐸𝑍 = √∑ (𝑍𝑖 − 𝑍𝑗 )2 ⁄𝑛 (8)

Zi = GCP elevation value
Zj = Google Earth, SRTM or AW3D elevation value.

18
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 13: An overlay of orthophoto reference points and GE points in two areas within the
University of Lagos, (a) north, and (b) west. The reference points are displayed as red circles,
and corresonding GE points are displayed as crosses (2000 - pink, 2008 - white, 2012 – blue and
2018 -yellow)

3. RESULTS AND DISCUSSION

The result of the findings for this study are analysed and discussed for both horizontal and vertical
accuracy.

19
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

3.1 Analysis of Horizontal Accuracy

Table 1 shows the coordinates of the orthophoto reference points and the Google Earth (GE)
coincident points. The coordinates are presented in the Universal Transverse Mercator (UTM)
system – Eastings and Northings. Table 2 presents the coordinate differences between the
reference points and the GE points. The difference in the eastings of the orthophoto reference
points and GE coincident points ranged from 0 – 11.99m in the year 2000, 2.98 – 18.99m in the
year 2008, 0.04 – 15.98m in the year 2012 and 1.14 – 7.99m in the year 2018. In the northings, the
differences ranged from 0 – 24.92m in the year 2000, 0.11 – 10.32m in the year 2008, 0.09 –
10.32m in the year 2012 and 0.04 – 11.92m in the year 2018. The SDs and RMSEs of the
differences in eastings were derived as follows: year 2000 (SD: 4.97m; RMSEx: 9.52m), year 2008
(SD: 4.34m; RMSEx: 15.97m), year 2012 (SD: 4.36m; RMSEx: 4.93m) and year 2018 (SD: 1.48m;
RMSEx: 4.39m). The SDs and RMSEs of the differences in northings were derived as follows:
year 2000 (SD: 4.83m; RMSEy: 14.05m), year 2008 (SD: 3.71m; RMSEy: 3.72m), year 2012 (SD:
3.54m; RMSEy: 3.65m) and year 2018 (SD: 3.66m; RMSEy: 4.28m).

The differences in eastings and northings of the orthophoto reference points and GE coincident
points are represented by stacked columns in Figures 14a and 14b respectively. Consistent ΔE
negative values in the years 2000, 2008, 2018 are suggestive of a systematic error that displaces
the Google Earth imagery westwards from the true positions. The largest easting differences are
observed in the 2008 imagery with some exceeding 15m while the least differences are observed
in the 2012 imagery. The highest horizontal linear RMSE in the x direction (RMSEx) is observed
in the 2008 imagery (15.975m), whereas the least horizontal linear RMSE is observed in the 2018
imagery (4.385m). It can be seen that the horizontal differences (Figure 14a) oscillate gently about
the mean differences, with spiked differences between points 22 and 27 for the earliest 3 years.
Surprisingly, the year 2000 imagery seems to perform better than 2008 in terms of the magnitude
of its differences relative to the source of groundtruth (orthophotos). In 2012, the easting
coordinates are generally better fitted with absolute differences closer to zero than in prior years.
However, it is interesting to note the spikes at some points such as #25 which indicate anomalies
in the image planimetry. In 2018, there is a generally better fit for all easting coordinates, with
differences very close to zero and the elimination of the spike, which had probably been rectified
in this recent version of the imagery.

The differences in northings (ΔN) are shown in Figure 14b. For years 2000 and 2018, the
differences are mostly positive and fairly consistent, whereas the differences for years 2008 and
2012 are mostly negative. The trend suggests a systematic or consistent shift in the positions of
the GE images generally occurring in the north-south axis; towards the south for years 2000 and
2018, and northwards for years 2008 and 2012. However, the magnitude of the shift is relatively
large northwards but minimal southwards. In terms of the horizontal linear error in the y direction
RMSE (RMSEy), year 2000 imagery has the highest RMSE of 14.046m while year 2012 imagery
has the least RMSE of 3.646m. The northing differences from year 2000 are generally worse off
by a significant measure than all other periods. Its absolute differences generally range from 10 to
25m, with the exception of point 30 whose absolute difference is exactly zero. Spikes at points 19
and 20 are seen at the images for 2008 to 2018. Years 2008 and 2012 images generally yield
northing differences that are closest to zero.

20
 Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Table 1: Coordinates of the orthophoto reference points and GE points

ID Orthophoto GE - 2000 GE - 2008 GE - 2012 GE - 2018
Easting Northing Easting Northin Easting Northin Easting Northin Easting Northin
(m) (m) (m) g (m) (m) g (m) (m) g (m) (m) g (m)
1 543316.50 720678.04 543309 720690 543301 720676 543317 720676 543313 720678
3 543527.11 720608.96 543517 720621 543510 720606 543530 720606 543522 720610
6 543859.21 720597.77 543851 720609 543843 720596 543859 720596 543855 720598
7 544045.31 720641.96 544037 720652 544030 720640 544046 720640 544041 720643
8 544153.99 720572.17 544143 720587 544136 720576 544157 720571 544151 720575
9 543947.95 720682.63 543936 720699 543934 720680 543953 720680 543945 720687
10 543844.04 720691.11 543835 720705 543826 720691 543844 720690 543840 720691
11 543827.14 720670.66 543818 720683 543809 720670 543828 720670 543823 720671
12 543922.67 720561.41 543913 720574 543906 720559 543924 720560 543918 720563
13 543980.75 720490.21 543972 720503 543965 720489 543981 720489 543976 720491
14 544122.65 720547.91 544113 720560 544104 720546 544122 720548 544118 720548
16 544183.70 720504.00 544174 720516 544166 720503 544183 720503 544178 720506
19 544162.74 720246.68 544151 720268 544144 720257 544160 720257 544156 720258
20 544125.86 720197.08 544114 720222 544109 720207 544125 720206 544120 720209
22 544004.85 720190.64 543993 720206 543988 720190 544009 720189 544001 720195
25 543897.02 719983.64 543908 719992 543900 719980 543913 719978 543894 719986
26 543925.06 719973.52 543927 719984 543917 719977 543932 719976 543922 719975
27 543243.01 719891.69 543236 719903 543225 719890 543241 719890 543238 719890
29 543241.67 719747.52 543233 719760 543226 719748 543244 719748 543238 719751
30 543244.93 719734.02 543244 719734 543229 719732 543248 719731 543240 719735
32 543214.77 719927.15 543206 719940 543198 719925 543216 719924 543211 719928
35 542994.99 720368.45 542983 720375 542976 720366 542991 720366 542987 720363
37 542803.65 720336.67 542795 720347 542787 720335 542803 720336 542799 720336
41 542773.38 720456.37 542764 720470 542758 720454 542777 720454 542770 720458
42 543884.70 719262.76 543875 719281 543872 719258 543894 719258 543883 719268
43 543869.93 719167.54 543859 719187 543856 719163 543879 719163 543868 719174
44 543859.14 719099.52 543849 719119 543846 719095 543867 719095 543858 719106

21
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Table 2: Coordinate differences between the orthophoto reference points and the GE points
ID ∆E (m) ID ∆N (m)
Year Year Year Year Year Year Year Year
2000 2008 2012 2018 2000 2008 2012 2018
1 -7.50 -15.50 0.50 -3.50 1 11.96 -2.04 -2.04 -0.04
3 -10.11 -17.11 2.89 -5.11 3 12.04 -2.96 -2.96 1.04
6 -8.21 -16.21 -0.21 -4.21 6 11.23 -1.77 -1.77 0.23
7 -8.31 -15.31 0.69 -4.31 7 10.04 -1.96 -1.96 1.04
8 -10.99 -17.99 3.01 -2.99 8 14.83 3.83 -1.17 2.83
9 -11.95 -13.95 5.05 -2.95 9 16.37 -2.63 -2.63 4.37
10 -9.04 -18.04 -0.04 -4.04 10 13.89 -0.11 -1.11 -0.11
11 -9.14 -18.14 0.86 -4.14 11 12.34 -0.66 -0.66 0.34
12 -9.67 -16.67 1.33 -4.67 12 12.59 -2.41 -1.41 1.59
13 -8.75 -15.75 0.25 -4.75 13 12.79 -1.21 -1.21 0.79
14 -9.65 -18.65 -0.65 -4.65 14 12.09 -1.91 0.09 0.09
16 -9.70 -17.70 -0.70 -5.70 16 12.00 -1.00 -1.00 2.00
19 -11.74 -18.74 -2.74 -6.74 19 21.32 10.32 10.32 11.32
20 -11.86 -16.86 -0.86 -5.86 20 24.92 9.92 8.92 11.92
22 -11.85 -16.85 4.15 -3.85 22 15.36 -0.64 -1.64 4.36
25 10.98 2.98 15.98 -3.02 25 8.36 -3.64 -5.64 2.36
26 1.94 -8.06 6.94 -3.06 26 10.48 3.48 2.48 1.48
27 -7.01 -18.01 -2.01 -5.01 27 11.31 -1.69 -1.69 -1.69
29 -8.67 -15.67 2.33 -3.67 29 12.48 0.48 0.48 3.48
30 0.00 -15.93 3.07 -4.93 30 0.00 -2.02 -3.02 0.98
32 -8.77 -16.77 1.23 -3.77 32 12.85 -2.15 -3.15 0.85
35 -11.99 -18.99 -3.99 -7.99 35 6.55 -2.45 -2.45 -5.45
37 -8.65 -16.65 -0.65 -4.65 37 10.33 -1.67 -0.67 -0.67
41 -9.38 -15.38 3.62 -3.38 41 13.63 -2.37 -2.37 1.63
42 -9.70 -12.70 9.30 -1.70 42 18.24 -4.76 -4.76 5.24
43 -10.93 -13.93 9.07 -1.93 43 19.46 -4.54 -4.54 6.46
44 -10.14 -13.14 7.86 -1.14 44 19.48 -4.52 -4.52 6.48
Mean -8.18 -15.40 2.45 -4.14 21.56 13.22 -0.78 -1.11 2.33
SD 4.97 4.34 4.36 1.48 13.22 4.83 3.71 3.54 3.66
RMSEx 9.523 15.975 4.931 4.385 RMSEy 14.046 3.724 3.646 4.278

Figure 14c shows the horizontal errors/shifts in Google Earth images. Table 3 shows the horizontal
errors in the GE points including the horizontal linear RMSE in the radial direction that includes
both x- and y- coordinate errors (RMSEr). The errors ranged from 0 – 27.6m in year 2000, 4.71 –
21.39m in year 2008, 0.65 – 16.95m in year 2012 and 3.4 – 13.29m in year 2018. The SD and
RMSE of the horizontal errors were derived as follows: year 2000 (SD: 5.05m; RMSEr: 16.969m),
year 2008 (SD: 3.29m; RMSEr: 16.404m), year 2012 (SD: 4.01m; RMSEr: 6.133m), and year
2018 (SD: 2.54m; RMSEr: 6.127m). It can be seen that the positional accuracy degrades starting
from year 2000 onwards, and it is most accurate in year 2018. Year 2000 shows the widest range
of errors, with a vastly undulating trend that represents a very poor fit of points, with the highest
shift at 27.6m. There is a slight improvement in year 2008 where RMSEr drops to 16.404m from
16.969 in year 2000. Years 2012 and 2018 display the closest fit to the orthophoto. The SDs
generally reduce for later epochs for both easting and northing coordinates, with very slight
deviations as shown in the increases in SD of easting coordinate differences from 2008 to 2012,
and northing coordinate differences from 2012 to 2018. The trend is slightly different for the

22
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

RMSEs, where there is a significant increase in RMSEx from 2000 to 2008, before generally
dropping off. While for RMSEy, it decreases from 2000 till 2012 before picking up again in 2018.

Following equation 6, the NSSDA horizontal accuracies at 95% Confidence Level were calculated
as follows – year 2000 (29.369m), year 2008 (28.391m), year 2012 (10.615m) and year 2018
(10.603m). The most recent imagery (year 2018) is the most accurate followed by the year 2012
imagery. However, the difference in accuracy between 2012 and 2018 is negligible. The images
for 2000 and 2008 are the worst in terms of horizontal accuracy.

Table 3: Horizontal errors/shifts (S) in the Google Earth images

ID Horizontal error/shift
Year 2000 (m) Year 2008 (m) Year 2012 (m) Year 2018 (m)
1 14.12 15.63 2.10 3.50
3 15.72 17.36 4.14 5.21
6 13.91 16.30 1.78 4.22
7 13.03 15.44 2.08 4.43
8 18.46 18.39 3.23 4.12
9 20.27 14.19 5.70 5.27
10 16.57 18.04 1.11 4.04
11 15.36 18.16 1.08 4.16
12 15.87 16.84 1.94 4.93
13 15.50 15.80 1.23 4.82
14 15.47 18.75 0.65 4.65
16 15.43 17.73 1.22 6.04
19 24.33 21.39 10.67 13.17
20 27.60 19.57 8.96 13.29
22 19.40 16.87 4.46 5.82
25 13.80 4.71 16.95 3.83
26 10.66 8.78 7.37 3.40
27 13.30 18.09 2.63 5.29
29 15.20 15.68 2.38 5.06
30 0.00 16.05 4.31 5.02
32 15.56 16.91 3.38 3.87
35 13.67 19.15 4.68 9.67
37 13.47 16.73 0.93 4.70
41 16.54 15.56 4.33 3.75
42 20.65 13.56 10.45 5.50
43 22.32 14.66 10.14 6.74
44 21.96 13.89 9.07 6.57
RMSEr 16.969 16.404 6.133 6.127
RMSEr (95% C.L) 29.369 28.391 10.615 10.603

23
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 14: Stacked columns showing (a) differences in eastings of the orthophoto reference
points and GE coincident points (b) differences in northings of the orthophoto reference points
and GE coincident points, and (c) horizontal errors/shifts in Google Earth images

24
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

3.2 Magnitude and Direction of Horizontal Error

Table 4 shows the directions and whole circle bearings (WCBs) of the horizontal errors/shifts. In
general, the maximum shifts in the four periods of study are in the NW direction: 2000 (Mag:
27.60m; Dir: 334.55⁰), 2008 (Mag: 21.39m; Dir: 300.47⁰), 2012 (Mag: 16.95m; Dir: 354.49⁰) and
2018 (Mag: 13.28m; Dir: 350.02⁰). The mean quadrants of the shifts are shown in Table 5. With
the exception of year 2012, the mean shifts are generally skewed towards the western direction:
2000 (Mag: 16.26m; Dir: 300.70⁰), 2008 (Mag: 16.08m; Dir: 262.59⁰), 2018 (Mag: 5.60m; Dir:
294.88⁰). In the 2000 imagery, the general shift is in the northern direction, more notably NW with
mean and maximum displacement of 16.26m and 27.60 respectively in that direction (see Table
5). The mean shifts for the 2000 imagery (16.264m) and 2008 imagery (16.082m) are within a
similar range but there is a sharp drop to 4.703m in year 2012 and in 2018 it rises to 12.142m.
Although considerable similarity exists between the mean and maximum shifts in 2000 and 2008,
their directions are opposite. For year 2012, the mean shift is 4.70m in the SE direction. A
remarkable feature of the 2018 imagery is a general shift that is skewed to the western axis. This
is indicated by a mean value of the eastern displacement component (– 2.909m) and a range of
values between -1.140m and -7.990m along the eastern/western axis. It is further corroborated by
the quadrant/angle of the minimum, maximum and mean shifts. There is a general decrease in the
magnitude of the mean shifts from the earliest year to the latest year, with 2018 having the least
horizontal error along that axis.

Ninety percent of the displacement in year 2000 occurred around the NW quadrant (Table 4), no
displacement exceeds ±13m along the EW axis, and displacement along the NS axis occurred only
in the northern quadrant (Figure 15). In the year 2008, about 80% of the displacements occurred
in the SW quadrant, with a significant amount of shift occurring within -10m to -20m range on the
EW axis. There is however little error between –10m to +5m as one moves eastward along the
EW axis. This may be a strong indication of systematic error that skews position westward away
from their true value. The case is slightly different when one considers displacement along the NS
axis. Displacement/shift along this axis generally occurs within ±5m from the origin. In the year
2012, the shifts are generally skewed towards the SE quadrant. Along the NS axis displacement
generally occur within ±5m. A similar feature is also noticeable in the shift along the EW axis. In
the year 2018, the shifts generally occurred towards the NW quadrant. Interestingly no
displacement occurred as one moves eastward. Only at three locations were displacement within -
2m to -1m observed. A significant number occurred within -6m and -1m along the EW axis. Along
the NS axis, displacement majorly occurs within -2m to +6m.

25
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Table 4: Directions and Whole Circle Bearings (WCBs) of the horizontal errors/shifts
2000 2008 2012 2018
ID Dir (°) WCB (°) Dir (°) WCB (°) Dir (°) WCB (°) Dir (°) WCB (°)
1 -32.09 327.91 82.50 262.50 -13.77 166.23 89.35 269.35
3 -40.02 319.98 80.19 260.19 -44.31 135.69 -78.50 281.50
6 -36.17 323.83 83.77 263.77 6.77 186.77 -86.87 273.13
7 -39.61 320.39 82.70 262.70 -19.39 160.61 -76.43 283.57
8 -36.54 323.46 -77.98 282.02 -68.76 111.24 -46.57 313.43
9 -36.13 323.87 79.32 259.32 -62.49 117.51 -34.02 325.98
10 -33.06 326.94 89.65 269.65 2.06 182.06 88.44 268.44
11 -36.53 323.47 87.92 267.92 -52.50 127.50 -85.31 274.69
12 -37.53 322.47 81.77 261.77 -43.33 136.67 -71.20 288.80
13 -34.38 325.62 85.61 265.61 -11.67 168.33 -80.56 279.44
14 -38.60 321.40 84.15 264.15 -82.12 277.88 -88.89 271.11
16 -38.95 321.05 86.77 266.77 34.99 214.99 -70.67 289.33
19 -28.84 331.16 -61.16 298.84 -14.87 345.13 -30.77 329.23
20 -25.45 334.55 -59.53 300.47 -5.51 354.49 -26.18 333.82
22 -37.65 322.35 87.83 267.83 -68.44 111.56 -41.45 318.55
25 52.72 52.72 -39.31 140.69 -70.56 109.44 -51.99 308.01
26 10.49 10.49 -66.65 293.35 70.34 70.34 -64.19 295.81
27 -31.79 328.21 84.64 264.64 49.94 229.94 71.36 251.36
29 -34.79 325.21 -88.25 271.75 78.36 78.36 -46.52 313.48
30 88.77 268.77 82.77 262.77 -45.47 134.53 -78.76 281.24
32 -34.31 325.69 82.69 262.69 -21.33 158.67 -77.29 282.71
35 -61.35 298.65 82.65 262.65 58.45 238.45 55.70 235.70
37 -39.94 320.06 84.27 264.27 44.13 224.13 81.80 261.80
41 -34.54 325.46 81.24 261.24 -56.79 123.21 -64.25 295.75
42 -28.00 332.00 69.45 249.45 -62.90 117.10 -17.97 342.03
43 -29.32 330.68 71.95 251.95 -63.41 116.59 -16.63 343.37
44 -27.50 332.50 71.02 251.02 -60.10 119.90 -9.98 350.02

26
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Table 5: Summary statistics of the magnitudes and directions of horizontal error

Year Shift ΔE (m) ΔN (m) Magnitude Direction Quadrant
(m) (⁰)

2000 Max. 10.98 24.92 27.598 334.549 NW

Mean -8.212 13.219 16.264 300.699 NW
Min. -11.990 -0.020 0.930 10.488 NE
2008 Max. 2.980 10.320 21.394 300.471 NW
Mean -15.397 -0.781 16.082 262.593 SW
Min. -18.990 -4.760 4.704 140.693 SE
2012 Max. 15.980 10.320 16.946 354.493 NW
Mean 2.455 -1.114 4.703 167.309 SE
Min. -3.990 -5.640 0.656 70.336 NE
2018 Max. -1.140 11.920 13.283 350.022 NW
Mean -2.909 5.163 12.142 294.875 NW
Min. -7.990 -5.450 3.399 235.702 SW

Figure 15: Magnitudes and directions of horizontal error in the Google earth historical images -
(a) year 2000 (b) year 2008 (c) year 2012 and (d) year 2018

27
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

3.3 Analysis of Vertical Accuracy

This section presents the analysis of the vertical accuracy of GE heights (HGE), including the
comparison with the AW3D-30 v2.1 (HAW3D) and SRTM v3.0 (HSRTM) DEMs. The basis of the
assessment was the heights (elevations) of 558 GCPs spread around Lagos State. The heights from
all three datasets fell within the general range of 0 – 73m. The mean heights are as follows: GCP
(16.410m), Google Earth (16.588m) AW3D (18.482m) and SRTM (18.188m). Table 6 presents
the SDs, SEMs and RMSEs of the height differences. The lowest RMSE of 3.682m and highest
vertical accuracy was observed in SRTM followed by AW3D with an RMSE of 4.388m. The
highest RMSE of 6.213m and lowest vertical accuracy occurred in Google Earth. The mean height
differences in GE are the least at 0.178m. However, the range of GE elevations is the highest
(64.152m) and this shows that the elevations have a wider spread relative to the GCPs. These
results show the low reliability of estimated elevations from GE relative to AW3D and SRTM
DEM. This conclusion is also supported by its high values for RMSE and SD, indicating a much
larger spread of elevations about the mean, relative to other DEMs in comparison.

Statistics from the SRTM DEM suggest it as the most reliable of all three height estimation
products, as it has the least range of height differences, a relatively modest mean elevation error
and the least standard deviation, which implies a low spread of elevations about its mean value.
Hence, we can safely conclude it is a more reliable product for estimating heights of objects or
points than AW3D and GE historical imagery. Finally, AW3D has a slightly lower elevation range
than GE, the highest mean elevation and an SD slightly higher than that of SRTM. GE elevations
are the least reliable while SRTM is the most reliable.

Table 6: SD and RMSE of the height differences

Range Minimum Maximum Mean SEM SD RMSEz
∆HGE-GCP 64.152 -28.460 35.692 0.178 0.263 6.216 6.213
∆HAW3D-GCP 58.850 -15.981 42.869 2.072 0.164 3.872 4.388
∆HSRTM-GCP 35.569 -22.981 12.587 1.778 0.137 3.228 3.682

Figure 16 shows histograms of height differences in the three datasets. SRTM DEMs have the least
spread of height differences, with significant height differences ranging from just less than -10m
to over 10m. Also, the relative frequency of various height differences has a near-normal
distribution, suggesting reliability and predictability. The most commonly occurring height
differences are near the zero mark, suggesting a mean close to zero. The GE histogram has a similar
bell shape for its height differences. The most commonly occurring differences and mean also
hover about the zero point, with nearly evenly distributed differences, although slightly skewed to
the left. Height differences for AW3D are significantly skewed to the right, with its most
commonly occurring elevation differences close to zero. However, its obvious range for
differences is much lower than that of GE which is by far the greatest and probably the least
reliable.

28
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Figure 16: Histogram of height differences (a) Google Earth elevations (b) AW3D (c) SRTM

4. DISCUSSION
The difference between 2018 and 2012 images are quite negligible, hence they can be said to be
of similar accuracy. The images for 2000 and 2008 are the worst in terms of horizontal accuracy.
The accuracy is lowest at year 2000 and gradually improves over time till 2018. This suggests that
Google has made efforts at improving the quality of its satellite imagery over the years. Also, it
shows there is a continuous enhancement in the accuracy and reliability of satellite imagery data
sources which form the source of Google Earth data. In terms of the vertical accuracy, Google
Earth elevation data had the highest RMSE of 6.213m followed by AW3D with an RMSE of
4.388m and SRTM with an RMSE of 3.682m. A crucial feature worthy of note are the tendency
of significant variability in accuracy of GE images within a relatively small area of study. The
noticeable spikes at certain points within the area of study (e.g. 19 and 20) supports this fact. It
would therefore be safe to note that accuracy may vary greatly between points within the same
study area. Another fact that becomes clear from this study is thus: GE images are significantly
skewed westward within the study area. With the exception of the 2012 imagery which may be
regarded as having a mean SW skewness, all other images tend to be skewed towards the NW
direction. This skewness towards certain direction which strongly indicates the prersence of a

29
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

systematic error, must be taken into account when using the imagery. More research is needed in
other locations to determine if this skewness is a general trend with GE imagery.

Although the vertical accuracy of SRTM and AW3D are superior, Google Earth still presents clear
advantages in terms of its ease-of-use and contextual awareness. For example, a wide variety of
end-users and software developers including those not allied to spatial sciences are able to easily
interact with GE data and utilise it for their applications. Google Earth APIs (application
programming interfaces) enable developers to embed Google Earth imagery into web applications.
Moreover, Google Earth is ubiquitous, and it has a very user-friendly interface that appeals to users
from all spheres of life. SRTM and AW3D are distributed via online data archives in which the
processes for search and retrieval are not straightforward for citizens without some background
knowledge or exposure to spatial information sciences.

5. CONCLUSION AND RECOMMENDATIONS

This study assessed the positional (horizontal and vertical) accuracy of Google Earth historical
imagery in Lagos Nigeria. The horizontal accuracy was assessed in terms of the RMSE using
historical images of the University of Lagos acquired at four epochs – 2000, 2008, 2012 and 2018.
In accordance with the ASPRS recommendations for images at an estimated pixel size of 60cm,
the Google Earth images assessed in this study are expected to meet the ASPRS Accuracy
Standards of 120cm RMSEx and RMSEy Horizontal Accuracy Class for Standard Mapping and
GIS work. Also, the absolute horizontal accuracy should be less than or equal to 2.448 * 120cm at
the 95% confidence level. The horizontal accuracies were estimated as follows – year 2000
(29.369m), year 2008 (28.391m), year 2012 (10.615m) and year 2018 (10.603m). It is evident that
the absolute horizontal accuracies of the GE historical images in the four years considered fell
short of the ASPRS standard. In terms of the vertical accuracy, Google Earth elevation data had
the highest RMSE of 6.213m and thus the lowest accuracy. The SRTM DEM had the highest
vertical accuracy with an RMSE of 3.682m followed by AW3D with an RMSE of 4.388m. These
findings strenghten the chain of evidence on the positional accuracies of Google Earth imagery in
previous sudies (e.g. Ubukawa, 2013 and Paredes-Hernández et al., 2013).

The main contribution of this study lies in the relevance of the findings to end-users of Google
Earth images and to the broad Digital Earth community. It presents an informed perspective
through quantitative analysis, on the limitations of historical archive of Google Earth’s imagery
in terms of the horizontal accuracy and also, the vertical accuracy of the elevation data. It is clear
from literature that variations exist in map and satellite information thereby making the services
rendered as error-prone. Users of these services are advised to use GE map and satellite
information with caution as stated in GE terms of service. It is also important to point out that the
positional accuracy of GE can be enhanced by carrying out image geometric registration on the
imagery using established geodetic control points. By carrying out this enhancement, the
application of GE imagery becomes extensive. The trends for GE imagery in this study have shown
improvement in positional accuracy in more recent imagery. It might be advisable for users to go
for the most recent imagery as the trend suggests higher accuracy in recent times. Also, users
should not rely on just GE imagery but instead acquire supporting data for their analysis. It is
reasonable to augment GE acquired data with other relevant and more accurate sources.

30
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

ACKNOWLEDGEMENTS
We thank Google LLC for free access to Goolge Earth imagery used in this research, the
Japan Aerospace Exploration Agency (JAXA) for free access to AW3D-30m DEM and the
National Aeronautics and Space Administration (NASA)/United States Geological Surveys
(USGS) for free access to SRTM data. The University of Lagos management is also appreciated
for permitting us to conduct the UAV survey within the campus. We are also grateful to the Lagos
State Surveyor General’s Office for provision of ground control points. The DJI Phantom 4
Professional UAV and accessories were made available by Geospatial Research Limited Nigeria.
The following individuals who assisted with the GPS observation, UAV survey and orthophoto
processing are acknowledged – Kayode Omolaye and Olumide Awe (
Geospatial Research Limited Nigeria), Adefemi Alabi Geosys Nigeria Limited, Shittu Ibrahim
(Federal School of Surveying, Nigeria), Tochi Nwaoru, Oluwaseyi Isaac and Adepo Rahmatullahi
(Department of Surveying and Geoinformatics, University of Lagos, Nigeria).

FUNDING
This research did not receive any specific funding from the public, private or not-for-profit sectors.

CONFLICTS OF INTEREST
There is no conflict of interest regarding the publication of this research.

REFERENCES
Allen, T.R. (2008). Digital Terrain Visualization and Virtual Globes for Teaching
Geomorphology. Journal of Geography, 106:6, 253-266, DOI: 10.1080/00221340701863766

American Society for Photogrammetry and Remote Sensing (ASPRS) (1990). ASPRS Accuracy
Standards for Large-Scale Maps, URL:
http://www.asprs.org/a/society/committees/standards/1990_ jul_1068-1070.pdf (last date
accessed: 22 January 2015)

American Society for Photogrammetry and Remote Sensing (ASPRS) (2014). ASPRS Positional
Accuracy Standards for Digital Geospatial Data - Edition 1, Version 1.0. Photogrammetric
Engineering & Remote Sensing. Vol. 81, No. 3, March 2015, pp. A1–A26. DOI:
10.14358/PERS.81.3.A1-A26

ASPRS, American Society for Photogrammetry and Remote Sensing. (1993). ASPRS accuracy
standards for large-scale maps. Bethesda (MD): ASPRS, USA.

ASPRS. AMERICAN SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING.

ASPRS accuracy standards for large-scale maps. Photogrammetric Engineering & Remote
Sensing, 56(7), 1068-1070, 1990.

Aurambout JP., Pettit C., Lewis H. (2008). Virtual Globes: the Next GIS?. In: Pettit C., Cartwright
W., Bishop I., Lowell K., Pullar D., Duncan D. (eds) Landscape Analysis and Visualisation.
Lecture Notes in Geoinformation and Cartography. Springer, Berlin, Heidelberg.
https://doi.org/10.1007/978-3-540-69168-6_25.

31
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Becek, K., Ibrahim, K., & Barussalam, B. (2011). On the positional accuracy of the Google Earth
imagery. Proceedings of FIG Working Week on Bridging the Gap between Cultures, Marrakech,
Morocco, 8 pp.

Benker, S. C., Langford, R. P., and Pavlis, T. L. (2011). Positional accuracy of the Google Earth
terrain model derived from stratigraphic unconformities in the Big Bend region, Texas, USA.
Geocarto International, 26(4), 291-301.

BudgIT (2018). What we know about Lagos State Finances. Available at

https://yourbudgit.com/wp-content/uploads/2018/05/LAGOS-STATE-DATA-BOOK.pdf

Buka, L., Maruziva, R., and Nenhowe, P. (2015). A comparison of Google Earth extracted points
with GPS surveyed points. Ethiopian Journal of Environmental Studies & Management 8(5): 484
– 493, 2015. doi: http://dx.doi.org/10.4314/ejesm.v8i5.2

Caglar, B., Becek, K., Mekik, C., and Ozendi, M. (2018). On the vertical accuracy of the ALOS
world 3D-30m digital elevation model. Remote Sensing Letters, 9(6), 607–615.

Chigbu, N., Okezie, M., Arungwa, I.D. and Ogba, C. (2019). Comparative Analysis of Google
Earth Derived Elevation with In-Situ Total Station Method for Engineering Constructions, 17p.
FIG Working Week 2019 Geospatial information for a smarter life and environmental resilience
Hanoi, Vietnam, April 22–26, 2019

Curtis, A.J., Mills, J.W., Leitner, M., (2006). Spatial confidentiality and GIS: re-engineering
mortality locations from published maps about Hurricane Katrina. Int. J. Health Geographics 5.
https://doi.org/10.1186/1476-072X-5-44.

El-Ashmawy, K. L. A. (2016). Investigation of the Accuracy of Google Earth Elevation Data.

Artificial Satellites, 51(3). https://doi.org/10.1515/arsa-2016-0008

El-Ashmawy, K.L.A. (2002). Investigation of the accuracy of Google Earth Elevation Data.
Artificial Satellites, Vol. 51, No. 3, 89-97. DOI: 10.1515/arsa-2016-0008

Elvidge, C.D. and Tuttle, B.T. (2008). How Virtual Globes Are Revolutionizing Earth Observation
Data Access and Integration, 137 – 140. The International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B6a. Beijing 2008

Farr, T. G., and Kobrick, M. (2000). Shuttle radar topography mission produces a wealth of
data, Eos Trans. AGU, 81( 48), 583– 585, doi:10.1029/EO081i048p00583.

Flanagin, A. J., and Metzger, M. J. (2008). The credibility of volunteered geographic information.
GeoJournal (72), 137-148.

Gbopa, A.O., Ayodele, E.G., Okolie, C.J., Ajayi, A.O., and Iheaturu, C.J. (2021). Unmanned
Aerial Vehicles for Three‑dimensional Mapping and Change Detection Analysis. Geomatics and
Environmental Engineering, 15(1), 41-61. https://doi.org/10.7494/geom.2021.15.1.41.

32
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Goodchild, M. F., Guo, H., Annoni, A., Bian, L., De Bie, K., Campbell, F., et al. (2012).
Nextgeneration Digital Earth. Proceedings of the National Academy of Sciences of the United
States of America, 109 (28), 11088–11094.

Gore A (1999) The Digital Earth: understanding our planet in the 21st Century. Photogramm Eng
Remote Sens 65(5):528–530

Gore, Al, 1998. The digital earth: understanding our planet in the 21st Century. Austr. Surveyor
43 (2), 89–91. https://doi.org/10.1080/00050326.1998.10441850.

Goudarzi, M. A. and Landry, R. (2017). Assessing Horizontal Positional Accuracy of Google Earth
Imagery in the city of Montreal, Canada. GEODESY AND CARTOGRAPHY Taylor & Francis
Group ISSN 2029-6991 / eISSN 2029-7009 2017 Volume 43(2): 56–65
https://doi.org/10.3846/20296991.2017.1330767

Henry, A. (2009). Using Google Earth for Internet GIS. M.Sc thesis. Institute of Geography,
School of GeoSciences, University of Edinburgh

Hu, Q., Wu, W., Xia, T., Yu, Q., Yang, P., Li, Z., & Song, Q. (2013). Exploring the Use of Google
Earth Imagery and Object-Based Methods in Land Use/Cover Mapping. Remote Sensing, 5(11),
6026–6042. MDPI AG. Retrieved from http://dx.doi.org/10.3390/rs5116026

Hwang, J.T. (2008). An Embedded Google Earth/Maps application on Real Estate Database
Inquiry And Display. 785 – 790. The International Archives of the Photogrammetry, Remote
Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008

Jacobson, A., Dhanota, J., Godfrey, J., Jacobson, H., Rossman, Z., Stanish, A., Walker, H. and
Riggi, J. (2015). A novel approach to mapping land conversion using Google Earth with an
application to East Africa. Environmental Modelling & Software 72 (2015) 1e9.
http://dx.doi.org/10.1016/j.envsoft.2015.06.011

JAXA (2017). ALOS Global Digital Surface Model (DSM). ALOS World 3D-30m. (AW3D30)
Dataset. Product Format Description, Version 1.1.
http://www:eorc:jaxa:jp/ALOS/en/aw3d30/aw3d30v11 format e:pdf

Lagos Digest of Statistics (2017). Lagos Digest of Statistics 2017. Avaiable at

http://lagosstate.gov.ng/Digest_of_Statistics_2017.pdf

Lesiv, M.; See, L.; Laso Bayas, J.C.; Sturn, T.; Schepaschenko, D.; Karner, M.; Moorthy, I.;
McCallum, I.; Fritz, S. Characterizing the Spatial and Temporal Availability of Very High
Resolution Satellite Imagery in Google Earth and Microsoft Bing Maps as a Source of Reference
Data. Land 2018, 7, 118. https://doi.org/10.3390/land7040118

Liang, J., Gong, J., and Li, W. (2018). Applications and impacts of Google Earth: A decadal review
(2006–2016). ISPRS Journal of Photogrammetry and Remote Sensing 146 (2018) 91–107.
https://doi.org/10.1016/j.isprsjprs.2018.08.019

33
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Liu Z., Foresman T., van Genderen J., Wang L. (2020) Understanding Digital Earth. In: Guo H.,
Goodchild M.F., Annoni A. (eds) Manual of Digital Earth. Springer, Singapore.
https://doi.org/10.1007/978-981-32-9915-3_1

Malarvizhi, K., VasanthaKumar, S., and Porchelvan, P. (2016). Use of High Resolution Google
Earth Satellite Imagery in Landuse Map Preparation for Urban Related Applications. Procedia
Technology, Volume 24, 2016, Pages 1835-1842. https://doi.org/10.1016/j.protcy.2016.05.231

McRoberts, R. E. 2010. “The Effects of Rectification and Global Positioning System Errors on
Satellite Image-based Estimates of Forest Area.” Remote Sensing of Environment 114: 1710–1717.
http://dx.doi.org/10.1016/j.rse.2010.03.001.

MES Innovation Sdn Bhd (2018, July 16) How accurate is Google Earth Elevation? Retrieved
from Blog (mes100.com/blog/how-accurate-is-google-earth-elevation/), MiTS 2, Technical
Documentation. Accessed 8/11/2020.

Mukul, M., Srivastava, V., & Mukul, M. (2015). Analysis of the accuracy of Shuttle Radar
Topography Mission (SRTM) height models using International Global Navigation Satellite
System Service (IGS) Network. Journal of Earth Syst Sci, 124(6), 1343-1357.

Mukul, M., Srivastava, V., Jade, S., & Mukul, M. (2017). Uncertainties in the shuttle radar
topography mission (SRTM) Heights: Insights from the indian Himalaya and Peninsula. Scientific
reports, 7, 41672.

Mulu, Y.A. and Derib, S.D. (2019). Positional accuracy evaluation of Google Earth in Addis
Ababa. Artificial Satellites, 54(2), 43-56. DOI: 10.2478/arsa-2019-0005.

Nourbakhsh, I., Sargent, R., Wright, A., Cramer, K., McClendon, B., Jones, M., (2006). Mapping
disaster zones. Nature 439 (7078), 787–788. https://doi.org/10.1038/439787a.

Nwilo, P.C., Okolie, C.J., Onyegbula, J.C., Abolaji, O.E., Orji, M.J., Daramola, O.E., and
Arungwa, I.D. (2020). Vertical Accuracy Assessment of 20-metre SPOT DEM using Ground
Control Points from Lagos and FCT, Nigeria. Journal of Engineering Research, Vol. 25, No. 2,
June 2020.

Okolie, C.J. (2019). Orthophoto Mapping of the University of Lagos Main Campus and Environs.
SURCON Project Report.

Paredes-Hernández C. U., Salinas-Castillo W. E., Guevara-Cortina, F., and Martínez-Becerra, X.,

(2013), Horizontal Positional Accuracy of Google Earth’s Imagery Over Rural Areas: A Study
Case in Tamaulipas, Mexico. BCG - Boletim de Ciências Geodésicas - On-Line version, ISSN
1982-2170 http://dx.doi.org/10.1590/S1982-21702013000400005

Potere, D. (2008). Horizontal Positional Accuracy of Google Earth’s High-Resolution Imagery

Archive. Sensors. ( 8), 7973-7981.

34
Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

Pulighe, G., Baiocchi, V. and Lupia, F. (2016) Horizontal accuracy assessment of very high
resolution Google Earth images in the city of Rome, Italy, International Journal of Digital Earth,
9:4, 342-362, DOI: 10.1080/17538947.2015.1031716

Schottenfels, P (2020). Ask a Techspert: How do satellite images work?

https://blog.google/products/maps/how-do-satellite-images-work/ (Date accessed: 6th May, 2020).

Shen, C., Fan, J., Pi, L., Li, F., (2006). Delineating lakes and enclosed islands in satellite imagery
by geodesic active contour model. Int. J. Remote Sens. 27 (23), 5253–5268.
https://doi.org/10.1080/01431160600857444.

Tadono, T., H. Nagai, H. Ishida, F. Oda, S. Naito, K. Minakawa, and H. Iwamoto. 2016.
“Generation of the 30 m-Mesh Global Digital Surface Model by ALOS PRISM.” International
Archives of the Photogrammetry, Remote Sensing & Spatial Information Sciences 41.

Thomas R. Allen (2008) Digital Terrain Visualization and Virtual Globes for Teaching
Geomorphology. Journal of Geography, 106:6, 253-266, DOI: 10.1080/00221340701863766

Ubakawa, T. (2013). An Evaluation of the Horizontal Positional Accuracy of Google and Bing
Satellite Imagery and Three Roads Data Sets Based on High Resolution Satellite Imagery.
Available at
http://www.ciesin.columbia.edu/confluence/download/attachments/19726351/Ubukawa_Position
alAccuracy_march2013.pdf?version=1&modificationDate=1363795265000

Ubukawa, T. (2013). An Evaluation of the Horizontal Positional Accuracy of Google and Bing
Satellite Imagery and Three Roads Data Sets Based on High Resolution Satellite Imagery. Center
for International Earth Science Information Network (CIESIN), The Earth Institute at Columbia
University.

Üstün, A., Abbak, R. A., & Öztürk, Z. E. (2016). Height biases of SRTM DEM related to EGM96:
from a global perspective to regional practice. Survey Review, 1-9.

Wang Y, Zou Y, Henrickson K, Wang Y, Tang J, Park B-J (2017) Google Earth elevation data
extraction and accuracy assessment for transportation applications. PLoS ONE 12(4): e0175756.
https://doi.org/10.1371/journal.pone.0175756

Yu, L. and Gong, P. (2012) Google Earth as a virtual globe tool for Earth science applications at
the global scale: progress and perspectives, International Journal of Remote Sensing, 33:12, 3966-
3986, DOI: 10.1080/01431161.2011.636081

35
 Nwilo P.C. et al. - Positional Accuracy Assessment of Historical Google Earth Imagery (PREPRINT)

36

View publication stats