Ibrahim Olatunji RAUFU1A*, Herbert TATA1B, Solihu OLAOSEGBA2

1
Department of Surveying and Geoinformatics, Federal University of Technology, Akure, Nigeria,
raufuibrahimolatunji@[Link], htata@[Link], 1A[Link]
1B
[Link] 1612
2
Department of Surveying and Geoinformatics, Federal School of Surveying Oyo, Nigeria, olaosegba@fss-
[Link], 2[Link]
[Link]

MODELLING LOCAL GEOID UNDULATIONS USING UNMANNED AERIAL

VEHICLES (UAVS): A CASE STUDY OF THE FEDERAL UNIVERSITY
OF TECHNOLOGY, AKURE, NIGERIA

The study was aimed at developing a geoid model using Unmanned Aerial Vehicle (UAV) technology. To accom-
plish this, a UAV was deployed to capture imagery of the study area from a height of 150m, with a ground resolution
of 4.19cm. A total of 3737 images were obtained, covering an area of 725.804 hectares. The existing ellipsoidal and
orthometric heights were used to georeferenced the acquired images. For the analysis, 35 points were utilized, with 20
points designated as ground control points (GCPs) and the remaining 15 points as check points (CPs). Using the
UAV-derived Digital Terrain Models (DTMs), a dataset comprising 18,492 points was generated for both ellipsoidal
(h) and orthometric (H) heights. The differences between these heights, referred to as geoid heights (N), were calcu-
lated as N = h - H for all 18,492 points. These geoid heights were subsequently employed to generate a geoid model,
including contour maps and 3D maps, of the study area. To assess the accuracy of the UAV-derived geoid heights, a
root mean square error (RMSE) analysis was performed by comparing them with the existing geoid heights and was
found to be 0.113 m. The scientific novelty and practical significance are in the development of a local geoid model
of the study area with centimetre-level precision. Thus, the output of this study can be used for a wide range of appli-
cations, including land management, construction, and environmental impact assessments in the study area.
Key words: geoid, UAV, DTM, ellipsoidal height, orthometric height

Introduction 2018]. The geoid plays a crucial role in engineering

by being widely applied to define physical heights
In the early stages of human intellectual devel-
[Sanso et al., 2019] and it serves as a reference sur-
opment, the question of "what is the earth, its size,
face or datum that refines the fundamental equations
and shape" emerged as one of the first issues that
of motion for the Earth's fluid envelopes [Albayrak et
needed to be resolved. Geodesy deals with the meas-
al., 2020]. Acknowledging the value of the geoid as a
urement and monitoring of Earth's dimensions, shape,
true representation of the shape of the earth makes
gravity field, and the precise determination of point
modelling such a shape even more crucial for a vari-
locations on its surface (NOAA, 2021). Thus, the ba-
ety of applications.
sic geodetic surfaces that describe the earth consist of
Geoid modelling development incorporates
its actual surface, the ellipsoid that provides the best geodetic, gravimetric, and astrogeodetic techniques
mathematical approximation of the Earth's shape, and [Albayrak et al., 2020]. For the precise definition
the geoid, which represents an equipotential surface of a geoid, the Global Navigation Satellite Sys-
perpendicular to the direction of gravity at all points tem/levelling (GNSS/levelling) method, a geodetic
[Oluyori et al., 2018]. The physical surface of the technique, can be utilized. This approach involves
earth contains a variety of landforms like plains, val- converting GNSS-derived ellipsoidal heights (h)
leys, mountains, water features, etc. [Prasad, 2015] into orthometric heights (H). Instead of levelling,
but its main problem as a reference surface for posi- well-established geoid models can be employed to
tion, dimension, and shape determination is its irregu- calculate orthometric heights. These geoid models
larity [Agajelu, 2018]. The geoid on the other hand is enable the computation of geoid height (N), which
an equipotential level surface of the earth’s gravity represents the difference between ellipsoidal and
field closely approximated by the mean sea level un- orthometric height values (N = h - H). Subse-
der ideal conditions of ocean waters, and the exten- quently, orthometric heights can be computed by
sion of that surface underneath the continents to make using the geoid heights and known ellipsoidal
it continuous and covers the whole earth [Agajelu, heights [Jekeli et al., 2012].
64 Geodesy, cartography and aerial photography. Issue 98, 2023

Fig. 1. Relationship between earth surface, geoid and ellipsoid (Albayrak et al., 2020).

Several methodologies have been adopted in employs direct georeferenced airborne LiDAR and
modelling geoid globally, regionally, and locally. indirect georeferenced UAV photogrammetry-
The methodologies range from terrestrial to air- derived point clouds to generate DTMs in ellipsoidal
borne and to space-borne. Al-krargy et al. (2014) and geoidal vertical datums. The local geoid model
employed the GPS/levelling technique to model lo- was estimated as the variation between the con-
cal geoid and to evaluate the performance of several structed DTMs with an accuracy of 9.2 cm. In a
Global Geopotential Models (GGMs) in Egypt. The prior study conducted by Raufu & Tata (2021), the
result shows that the best model for local geoid in accuracy of three polynomial geoid models was
the study area was that of the 2nd-order polynomial evaluated in Akure, Nigeria. The best-fitting geoid
and a standard deviation of ±0.050m was obtained model yielded a standard deviation of 14.7 cm. In
which exceeds the precision of all tested GGMs their study, Belay et al. (2021) employed the re-
models over the study area. Erol end Erol (2020) move-compute-restore (RCR) procedure and the
examined four distinct surface interpolation tech- least-squares collocation (LSC) method to formulate
niques aimed at local geoid modelling in the western a gravimetric geoid model for Ethiopia. The accu-
region of Turkey. The methodologies evaluated en- racy of this model was evaluated using geometric
compass multivariable polynomial regression geoid heights, resulting in a precision of 13 cm.
(MPR), least-squares collocation (LSC), bivariate However, due to the required level of accuracy,
(BIVAR) interpolation, and wavelet neural networks global and regional geoid models may not be suit-
(WNN). The result obtained shows that the BIVAR able for local applications such as engineering and
technique demonstrated a better performance with construction. The use of ellipsoidal heights derived
an accuracy of 2.65 cm, surpassing even a gravimet- from GNSS requires the use of a geoid model to
ric geoid model in the study area. In another study transform them into orthometric heights. However,
by Maglione et al. (2018), the accuracy of global Odera & Fakuda (2015) opined that global models
geoid height models in a local area was studied in tend to be overly generalized for localized applica-
the Campania region of Italy. The result shows that tions. Furthermore, the absence of a national geoid
the global geoid height models are often not suitable model in Nigeria, as emphasized by [Raufu & Tata,
for local applications with RMS values of 1.157m, 2021; Oluyori et al., 2018], highlights the signifi-
0.444m, and 0.288m obtained for EGM84, EGM96, cance of developing local geoid models for precise
and EGM2008 respectively. Erol et al. (2020) assess localized applications.
the performance of photogrammetry methods for Over the past few years, the study area has wit-
determining local geoid model in Turkey. The study nessed tremendous structural development. To en-
Geodesy, cartography and aerial photography. Issue 98, 2023 65

sure adequate and sustainable development, the GPS and Geographic Information System (GIS)
establishment of a reliable reference model for techniques [Quaye-Ballard et al., 2020]. The accu-
height determination has become an essential prior- racy of the topographic map can be achieved as
ity. However, the traditional methods of extending low as 5cm [Quaye-Ballard et al., 2020].
orthometric height such as terrestrial survey meth- Considering the cost-effectiveness and suffi-
ods, can be time-consuming, expensive, and tedi- cient accuracy of UAV technology in height deter-
ous. The use of gravimeters is also costly, and it is mination, and recognizing that geoid modelling
only feasible to occupy a limited number of points relies on precise height data (especially when using
for interpolation purposes. In light of these chal- terrestrial methods), we decided to use UAV tech-
lenges, it was essential to adopt an alternative ap- nology for geoidal undulation modelling in this
proach that could effectively model the entire study study.
area in a relatively short period. This necessity led
Aim of the study
to the adoption of Unmanned Aerial Vehicles
(UAVs) for the purpose of modelling geoidal undu- The aim of this study is to develop a local geoid
lation for the study area. undulation model of the Federal University of
The utilization of UAVs presents a promising Technology, Akure, using a UAV technology. The
solution for the acquisition of multi-temporal aerial research was conducted for the purpose of using
stereo photos and high-resolution digital surface the model for important projects that require pre-
models [Chi et al., 2016]. Therefore, by leveraging cise height information, such as urban planning,
UAVs and having a sufficient number of Ground infrastructure design, and flood modelling in the
Control Points (GCPs), it is anticipated that accu- institution.
rate local geoid modelling can be achieved. There-
fore, this research aims to make the most of these Study area
advancements and develop a local geoid model for
The study area chosen for this research is the
the study area.
Federal University of Technology Akure (FUTA),
Unmanned Aerial Vehicle technology located in Akure South Local Government Area of
in locational data acquisition Ondo State, in the southwestern region of Nigeria.
Geographically, it is situated between latitudes 07°
The application of UAVs for remote data collection
18’ to 07° 20’ and longitudes 05° 06’ to 05° 09’.
has witnessed significant advancements in recent
years [Quaye-Ballard et al., 2020] and it has be- The study area covered a total area of 577.97 hec-
come a new surveying technique for acquiring spa- tares as shown in Fig. 2 and the distribution of
tial information [Yeh et al., 2018]. This practice of ground control points is shown in Fig. 3.
employing UAVs in surveying has gained global
Method
popularity in recent times [Christiansen et al.,
2017; Gonzalez et al., 2016; Turner et al., 2016]. The methodology employed in this study in-
While specialized UAVs equipped with high- cludes three main components: GPS observation
accuracy Global Positioning System (GPS) for using South Differential GNSS receivers, precise
mapping purposes are available, they come at a leveling observation using a Leica DNA03 digital
higher cost. However, consumer-grade UAVs, level instrument, and UAV mapping using a DJI
which are more affordable, user-friendly, and read- Phantom 4 Pro UAV. A total of thirty-five (35) con-
ily accessible, are being employed for various ap- trol points were observed for the GNSS and leveling
plications such as topographical mapping and gen- operations, with twenty-nine (29) pre-existing points
eration of precise digital elevation model (DEM) and six (6) newly established points. The acquired
[Polat & Uysal, 2017]. By leveraging GPS and GNSS data was processed using the South GNSS
GCPs, these consumer-grade UAVs have facilitated Processor software to obtain ellipsoidal heights. Pre-
topographical surveying and have reduced both cise leveling data was processed using the Leica
time and cost in acquiring data for inaccessible Geo-Office Software, and a Microsoft Excel pro-
land areas, particularly through integration with gram was developed to adjust the data using the ob-
66 Geodesy, cartography and aerial photography. Issue 98, 2023

servation equation method for leveling network ad- ric heights, was used to calculate the geoidal height
justment, resulting in orthometric heights. The UAV for each point. The coordinates (eastings and nor-
data was processed using DJI Pix4D Enterprise things) and geoidal heights were imported into
software to generate point cloud data, 3D mesh, Surfer software to create a contour-based model.
digital terrain models (DTMs), and orthomosaic. This model allows for obtaining the geoidal height
The point cloud, consisting of at least eighteen thou- of any point within the study area by inputting the
sand (18,000) points with ellipsoidal and orthomet- corresponding easting and northing coordinates.

Fig. 2. Study area

Fig. 3. Distribution of ground control points in the study area

Geodesy, cartography and aerial photography. Issue 98, 2023 67

GNSS observation adjusted using observation equation method of level-

ling network adjustment and corrections was done to
A total of thirty-five (35) GCPs in the study area obtain the final orthometric heights. The accuracy
were observed using South differential Global Posi- obtained for the levelling measurements in this study
tioning System (DGPS). The existing ones (29 in is 1.3 mm, indicating the vertical precision achieved
total) were observed to confirm their position's sta- through this traditional surveying technique.
bility, while six (6) new points were established and
observed to ascertain their geographic coordinates UAV mapping
and ellipsoidal heights. The observations were car-
To collect data on surface properties such as
ried out in static mode, with a minimum duration of
vegetation, elevation, and ground subsidence, we
thirty (30) minutes for each existing station and
used a DJI Phantom 4 Pro UAV and the orthomosaic
maximum of one (1) hours for each of the new sta-
and digital surface model of the study area were
tion with five (5) seconds epoch rate to monitor a
generated. The UAV flight was conducted at a height
sufficient number of satellites, thereby improving
of 150m above ground level, with a forward overlap
the data streaming quality and ensuring greater accu-
of 60% and a camera angle of 90°. For accurate geo-
racy. The obtained accuracy of the GNSS observa-
referencing, a total of 20 GCPs were used. The posi-
tions is 10 mm, demonstrating the vertical precision
tions of these GCPs were collected using a South
achieved through satellite-based positioning.
differential Global Positioning System (DGPS),
Precise levelling observation while their orthometric heights were determined
through precise levelling techniques.
The orthometric heights of the six (6) newly es-
tablished GCPs were determined using a precise UAV data processing
geometric levelling technique with the Leica
The 3,737 images captured by the drone were
DNA03 digital level instrument. To ensure the in-
strument's quality and accuracy, a two-peg test was imported into the DJI Pix4DEnterprise environment.
performed, revealing a collimation error of 0.002 These images underwent photo alignment process-
mm and confirming the instrument's excellent con- ing to enhance the optimization of key points, tie
dition for observations. To achieve higher accuracy, points, and the estimation of photo match points.
a total of six (6) loops were conducted, covering Subsequently, 35 Ground Control Points (GCPs),
twelve (12) lines and seven (7) points. which were acquired using the South DGPS instru-
ment, served as reference points to improve the
camera positions and orientation. Of these GCPs, 20
were utilized for bundle adjustment, while the re-
maining 15 GCPs were designated as checkpoints.
The process of creating a dense point cloud and a
3D mesh commenced with the generation of essen-
tial points required for constructing the terrain
model. The point cloud data were interpolated to
generate a digital terrain model. The final step in-
volved the creation of the orthomosaic depicted in
Fig. 5, which was achieved through the orthorectifi-
cation of the generated digital terrain model.

Generation of DTMs from point cloud data

Fig. 4. The levelling loop
The Digital Terrain Model (DTM) generation proc-
The SVG/1305 benchmark was utilized as the ess commenced with the careful georeferencing of
reference point, as shown in Fig. 4 Each section of UAV-derived images. The georeferencing process
the levelling survey had a maximum distance of involved the incorporation of both ellipsoidal and
sixty (60) meters, with thirty (30) meters measured orthometric heights to ensure accurate horizontal
from the instrument position to both the back and and vertical positioning with the vertical datum set
foresight points. The reduced levels of the new to World Geodetic System 1984 (WGS84). Follow-
points and their change in height from the were ing georeferencing, a pre-processing step was im-
68 Geodesy, cartography and aerial photography. Issue 98, 2023

plemented to filter the UAV-derived point cloud canopy and structures so as to ensure that the re-
data. This filtering process involved selectively sulting DTM accurately reflected the topographical
removing non-ground points such as vegetation surface.

Fig. 5. Orthomosaic imagery of the study area

Fig. 6. DTM-generated contour map from ellipsoidal heights

Fig. 7. DTM-generated contour map from orthometric heights

Geodesy, cartography and aerial photography. Issue 98, 2023 69

In the point cloud data filtering process, the aver- RMSE between the geoidal undulations obtained
age spatial distance of each point to its neighboring from existing data and the predicted geoidal undu-
points is estimated. The distribution of differences lations from UAV-generated data is estimated as:
between calculated distances and their average is as- n 2
sumed to follow a Gaussian distribution (Erol et al.,
RMSE =
å (O - P )
i =1 i
,
i
(2)
2020). Points with distance differences that deviate N
from this distribution are identified as outliers and where Oi is the geoidal undulations obtained
removed. This removal is based on a specified maxi- through existing data, Pi is the predicted geoidal
mum distance criterion, ensuring that points farther undulations from UAV, and N is the number of
than the defined maximum distance are considered points.
blunders and subsequently excluded. Results
Following the removal of blunders, the data
undergo denoising through an elevation threshold- In diverse geographical contexts, researchers have
ing method. Specifically, the threshold wasis set as adopted a range of techniques to model the geoid.
the sum of the mean height and the standard devia- Each technique comes with its own unique merits and
tion of the height. Thus, after filtering the data dif- limitations. In the present study, we explore an inno-
ference from their mean. Five neighboring points vative approach to geoid modelling by harnessing the
were chosen for the application of these filtering capabilities of UAVs. This approach offers the distinct
processes. Subsequently, 27 % of all points were advantage of efficiently covering extensive areas
removed from the point cloud derived from the within a shorter timeframe. The results of the UAV-
UAV mapping. After filtering the data, a regular derived ellipsoidal, orthometric, and geoidal undula-
grid covering the study area was generated and tions for twenty (20) ground control points that were
kriging interpolation was employed to estimate used in creating the model are presented in Table 1.
elevations at grid points for creating ellipsoidal and The geoidal undulation results of the final fifteen (15)
orthometric DTMs using Surfer 20 software. Fig 6.
GCPs, which were determined by incorporating the
and 7 shows the DTM-generated contour map from
coordinate and ellipsoidal height obtained from the
ellipsoidal and orthometric heights.
GPS observation into the model, are shown in Table 2
Computation of the UAV-derived geoid model as a means of validating the geoid model and an
RMSE of 0.113 is obtained.
The geoidal undulation for all the points in the Visual representations of our findings are pre-
study area was determined using the ellipsoidal and sented in Fig. 8 and 9. Fig. 8 portrays a UAV-
orthometric heights model generated from the point generated contour map, providing a visual under-
cloud of over 18,000 points. A program was written standing of the terrain's characteristics. Meanwhile,
in a Microsoft Excel spreadsheet according to the Fig. 9 offers a three-dimensional representation of
equation below to compute the geoidal undulations the geoid model for the study area. The model
with respect to WGS84. shows that the geoidal height values in the study
NUAV = hUAV – HUAV (1) area are all positive ranging between 17.5 m to
In order to test the validity of the method used 32.5 m at 5 m grid interval with the highest values
in this study, fifteen (15) GCPs were set aside as more prominent in the Northern part of the area
check points out of the thirty-five (35) GCPs. So, while the lowest values are prominent in the
the ellipsoidal, orthometric, and geoidal heights of Southwestern part of the area. The geoid undula-
the twenty (20) GCPs over the study area were tion within the study area experiences variations
computed using the Microsoft excel worksheet. influenced by both the gentle slope of the topogra-
phy and the heterogeneous mass distribution of the
Root Mean Square Error
earth's surface. These variations are primarily due
In order to check the accuracy and reliability of to the ongoing developmental changes that have
the geoidal undulations computed in this study, the occurred within the school since its establishment
root mean square error (RMSE) is employed. The in 1981. As the institution has seen increased struc-
RMSE estimates the difference between the ob- tural development over the years, these changes in
served values and the predicted values. Thus, the mass distribution have led to corresponding shifts
70 Geodesy, cartography and aerial photography. Issue 98, 2023

in gravity values at various locations, further influ- known for exhibiting substantial geoid variations
encing the geoid undulation observed in our model. due to the irregular distribution of mass, geological
It is important to note that an area with a complex features, and topographical changes. Thus, the ob-
topography (comprising of lowland and hilly ter- served geoid variation aligns with expectations for
rains), such as the one under investigation, are such terrain types.
Table 1

UAV-derived heights for the ground control points

Geographic Coordinates UAV Acquired Heights (m) UAV-Derived Geoidal
Station Ellipsoidal Orthometric Undulations (m)
Latitude Longitude
Height (m) Height (m)
SVG1301 7.29842 5.13533 375.746 348.637 27.109
SVG1305 7.30342 5.13283 378.862 350.787 28.075
SVG1306 7.30536 5.13273 385.458 356.015 29.443
SVG1307 7.30451 5.13486 392.514 361.592 30.922
SVG1309 7.30354 5.13698 384.519 354.699 29.820
SVG1310 7.30105 5.13601 374.892 347.723 27.169
SVG1412 7.30325 5.12816 374.160 347.625 26.535
SVG1446 7.30810 5.12793 372.356 345.861 26.495
GPS1449 7.29488 5.13544 361.789 336.843 24.946
SVG1454 7.29840 5.14305 360.754 336.483 24.271
SVG1518 7.30831 5.13455 393.490 362.427 31.063
SVG1522 7.30381 5.13895 382.575 353.650 28.925
SVG1524 7.30231 5.13975 377.069 348.960 28.109
SVG1629 7.30103 5.14089 371.818 344.010 27.808
SVG1633 7.29649 5.14465 357.020 333.646 23.373
SVG1637 7.29473 5.14948 361.265 337.817 23.448
SVG1756 7.30354 5.14069 375.458 347.855 27.603
SVG1758 7.30246 5.14311 375.287 347.893 27.394
PT1 7.31078 5.12196 379.106 354.981 24.126
PT4 7.29915 5.12835 361.199 338.008 23.191
Table 2

Geoidal Undulations derived from the model for the check points
Geographic Coordinates Existing Geoidal UAV-Derived Ge- Difference (m)
Station Longitude Undulations (m) oidal Undulations
Latitude
(m)
SVG1308 7.30387 5.13579 30.137 30.219 -0.082
SVG1447 7.30812 5.12844 26.825 26.945 -0.120
SVG1521 7.30514 5.13924 29.171 29.365 -0.194
SVG1525 7.30124 5.13944 27.871 27.761 0.111
SVG1628 7.30219 5.14010 27.640 27.420 0.220
SVG1630 7.29983 5.14203 25.416 25.523 -0.107
SVG1634 7.29638 5.14603 23.545 23.461 0.085
SVG1636 7.29660 5.14952 23.626 23.656 -0.030
SVG1638 7.29306 5.14968 22.827 22.796 0.031
SVG1755 7.30466 5.14011 28.527 28.495 0.031
SVG1757 7.30202 5.14222 26.734 26.838 -0.104
PT2 7.30387 5.11935 20.873 20.694 0.179
PT3 7.31250 5.12503 25.005 25.014 -0.009
PT5 7.29453 5.14587 22.070 22.152 -0.082
PT6 7.30238 5.14817 25.132 25.084 0.048
RMSE = 0.113
Geodesy, cartography and aerial photography. Issue 98, 2023 71

Fig. 8. UAV-generated geoidal map of the study area

Fig. 9. 3D UAV-generated geoidal model of the study area

Hypothesis testing H1 = The mean geoidal undulation estimated

from UAV mapping is not equal to the mean geoi-
In this study, hypothesis testing was conducted
dal undulation estimated from the geodetic method.
on the check point data using t-distribution statis-
The decision rule stated that if tcal > ttab at a
tics in a Microsoft Excel spreadsheet. The objective 0.05 significance level, we reject H0 and accept H1.
of the hypothesis was to examine whether there is a However, since tcal < ttab for both the one-tail and
significant difference between the mean geoidal two-tail tests (specifically, 0.480 < 1.761 in the
undulations obtained from UAV mapping and those one-tail test and 0.961 < 2.145), we accept H0. This
obtained from the geodetic method, with a signifi- indicates that although there are slight variations in
cance level of 5%. The hypotheses tested were as the geoidal heights obtained from the existing and
follows: UAV-generated data in the study area, these varia-
H0 = The mean geoidal undulation estimated tions are not statistically significant. Thus, the ac-
from UAV mapping is equal to the mean geoidal curacy of the UAV-derived geoid model is con-
undulation estimated from the geodetic method. firmed.
72 Geodesy, cartography and aerial photography. Issue 98, 2023

Comparison of geoidal undulation differences: vering the Akure region, reported a standard devia-
UAV vs existing geometric and global geoids tion of 14.7 cm for the best-fitting geoid model.
Herbert and Olatunji (2021) also assessed the accu-
In this study, we compared the geometric geoid
model estimated from UAV data with both an exist- racy of estimating orthometric height in the study
ing polynomial geoid model and the Earth Gravita- area using GNSS and EGM data, reporting standard
tional Model 2008 (EGM08). The statistics derived errors of 1.361 m, 1.365 m, and 1.367 m for the
from these analyses provide insights into the fitting EGM08, EGM96, and EGM84, respectively. There-
performance of the compared models. The polyno- fore, in terms of accuracy the UAV-geoid model es-
mial geoid model, developed by Raufu and Tata timated in this study is considered comparable to the
(2021) using high-accuracy GNSS/levelling data co- existing polynomial and EGM08 geoid models.

Fig. 10. Map of the difference between UAV-generated

and existing geometric geoidal model of the study area

Fig. 11. Map of the differences between UAV-generated

and EGM08 geoidal model of the study area
Geodesy, cartography and aerial photography. Issue 98, 2023 73

Table 3 shows the statistical differences in geoi- Conclusions

dal undulations estimated from the UAV compared
The paper investigates the use of unmanned ae-
to existing geometric and global geoid models at
rial vehicle (UAV) technology for local geoid
selected geodetic control points while Fig. 10 and
11 shows the visual differences respectively. The modelling. Imagery of the study area was cap-
results indicate that the differences in geoidal undu- tured by the UAV at a flying height of 150 m. A
lation between the UAV-geoid and the existing total of thirty-five (35) points with ellipsoidal
polynomial geoid model range from -22 cm to 23 and orthometric height values were utilized to
cm, with a mean value of -1.2 cm and an RMSE georeferenced the UAV images, with twenty (20)
value of 10.9 cm. Furthermore, the differences in points serving as ground control points (GCPs)
geoidal undulation between the UAV-geoid and and fifteen (15) as validation points. Point clouds
EGM08 range from -42.1 dm to 61.5 dm, with a data derived from the UAV were used to generate
mean value of 15.5 dm and an RMSE value of 25.9 DTM for ellipsoidal and orthometric heights. The
dm respectively. The statistical evaluation suggests geoidal undulation was calculated as the differ-
that the fit of the global geoid to the local geoid ence between the UAV-derived ellipsoidal and
model is not superior when compared to the exis-\ orthometric heights, leading to the development
ting polynomial geoid model in the study area. of a local geoid model for the study area. The
Table 3 analysis resulted in a root mean square error
value of 0.113 m when comparing the existing
Statistical differences between
and UAV-derived geoidal heights. Hypothesis
UAV-geoid and existing geometric
testing supported the null hypothesis, indicating
and global geoids (EGM08)
no significant difference between the two models
Geoidal un- Mini- Maxi- at a 5% significance level. Further comparison
Mean RMS
dulation dif- mum mum between the UAV-derived geoid with the existing
(m) E (m)
ferences (m) (m) polynomial geoid and EGM08 model indicated
UAV-geoid -0.220 0.230 -0.012 0.109 that the polynomial geoid fitted better compared
vs existing to the EGM08 with RMSE of 10.9 cm. Notably,
polynomial
DTMs obtained through remote data collection
geoid
with UAV have a great benefit in geoid determi-
UAV-geoid -4.207 6.152 1.545 2.590
nation because they capture all geoid features in
vs EGM08
high detail. This study highlights the suitability
of UAV technology for achieving centimeter-
The scientific novelty
level accuracy in determining geoidal undula-
and practical significance
tions within smaller areas for localized purposes.
The research showcases the scientific importance While this approach facilitates the creation of a
of employing UAV technology for geoid model- local geoid model with enhanced accuracy in re-
ling, offering a cost-effective, high-precision solu- gions characterized by complex topography, it is
tion that can have practical applications in various acknowledged that the obtained accuracy is not
domains. This novel approach not only advances enough for large-scale engineering and mapping
geodetic surveying techniques but also holds prom- applications. To address this limitation, it is rec-
ise for applications in various domains. By leverag- ommended to conduct additional data analysis
ing UAV technology, this research opens new pos- and employ diverse numerical methods for more
sibilities for professionals and researchers in geo- comprehensive geoid modelling, particularly in
spatial sciences and related fields, providing valu- larger areas. Furthermore, future research initia-
able insights for real-world problem-solving and tives should explore the potential enhancement
decision-making. of height accuracies by integrating innovative
74 Geodesy, cartography and aerial photography. Issue 98, 2023

data sources and technologies. Exploring the Erol, S., & Erol, B. (2020). A comparative assessment of
synergistic potential of various remote sensing different interpolation algorithms for prediction of
techniques, including the integration of photo- GNSS/levelling geoid surface using scattered control
grammetric imagery, Lidar data, and advanced data. Measurement, 173, 108623.
methods like Interferometric Synthetic Aperture Gonzalez, L. F., Montes, G. A., Puig, E., Johnson, S.,
Radar (InSAR), along with employing Un- Mengersen, K., & Gaston, K. J. (2016). Unmanned ae-
rial vehicles (UAVs) and artificial intelligence revolu-
manned Aerial Vehicle (UAV) technology, could
tionizing wildlife monitoring and conservation. Sen-
present novel avenues for precise height deter-
sors, 16(1), 97.
mination. Furthermore, the development of tai-
Herbert, T. & Olatunji, R.I. (2021). Determination of or-
lored strategies to harness the strengths of these
thometric height using GNSS and EGM Data: A sce-
combined approaches is crucial for achieving nario of the Federal University of Technology Akure.
superior results. This comprehensive approach is International Journal of Environment and Geoinformat-
expected to significantly advance our under- ics (IJEGEO), 8(1):100-105. doi: 10.30897/
standing and modelling of local geoids, espe- ijegeo.754808
cially when leveraging the capabilities of photo- Jekeli, C., Yang, H. J., & Kwon, J. H. (2012). The offset of
grammetric techniques. the South Korean vertical datum from a global geoid.
KSCE Journal of Civil Engineering, 16(5), 816–
REFERENCES 821. doi:10.1007/s12205-012-1320-3
Agajelu, S.I. (2018). Geodesy: The Basic Theories – Clas- Maglione, P., Parente, C., & Vallario, A. (2018). Accuracy
sical & Contemporary. Enugu, El ‘Demak Publishers, of global geoid height models in local area: Tests on
ISBN 978-978-8436-99-0, pp. 4, 5, 9, 11 – 21 Campania region (Italy). International Journal of Civil
Albayrak, M., Ozlüdemir, M.T., Aref, M.M., & Halicioglu, Engineering and Technology, 9, 1049-1057.
K. (2020). Determination of Istanbul geoid using National Oceanic and Atmospheric Administration,
GNSS/levelling and valley cross levelling data. Geod- NOAA. (2021). Is the Earth round? retrieved from Na-
esy and Geodynamics, 11(3), 163-173. tional Ocean Service website, 2021,
Al-Krargy, E. M., Doma, M. I., & Dawod, G. M. (2014). [Link]
Towards an Accurate Definition of the Local Geoid Odera, P. A., & Fukuda, Y. (2015). Recovery of orthometric
Model in Egypt using GPS / Levelling Data: A Case heights from ellipsoidal heights using offsets method
Study at Rosetta Zone. International Journal of Innova- over Japan. Earth, Planets and Space, 67(1).
tive Science and Modern Engineering (IJISME), 2(11). Oluyori, P. D., Ono, M. N., & Eteje, S. O. (2018). Com-
Belay, E. Y., Godah, W., Szelachowska, M., & Tenzer, R. parison of Two Polynomial Geoid Models of
(2021). ETH–GM21: A new gravimetric geoid model GNSS/Leveling Geoid Development for Orthometric
of Ethiopia developed using the least-squares colloca- Heights in FCT, Abuja. International Journal of Engi-
tion method. Journal of African Earth Sciences, neering Research and Advanced Technology (IJERAT),
183,104313. 4(10), 1-9.
Chi, Y.Y., Lee, Y.F., & Tsai, S.E. (2016). Study on High Polat, N., & Uysal, M. (2017). DTM generation with UAV
Accuracy Topographic Mapping via UAV-based Im- based photogrammetric point cloud. ISPRS - Interna-
ages. IOP Conference Series: Earth and Environmental tional Archives of the Photogrammetry, Remote Sens-
Science, 44, 032006. ing and Spatial Information Sciences. XLII-4/W6,
Christiansen, M. P., Laursen, M. S., Jørgensen, R. N., 77-79.
Skovsen, S., & Gislum, R. (2017). Designing and test- Prasad, S. (2015). Basic Geodesy Unit: II Semester: I Paper
ing a UAV mapping system for agricultural field sur- Code: GIS 05: Name of Paper: Earth Positioning Sys-
veying. Sensors, 17(12), 2703. tem, PG Diploma in RS & GIS.
Erol, S., Özögel, E., Kuçak, R. A., & Erol, B. (2020). Util- Quaye-Ballard, N.L., Asenso-Gyambibi, D., & Quaye-
izing Airborne LiDAR and UAV Photogrammetry Ballard, J. (2020). Unmanned Aerial Vehicle for Topog-
Techniques in Local Geoid Model Determination and raphical Mapping of Inaccessible Land Areas in Ghana:
Validation. ISPRS International Journal of Geo- A Cost-Effective Approach. Presented at the 2020 FIG
Information, 9(9), 528. doi:10.3390/ijgi9090528 Working Week, Amsterdam, Netherlands, May 10– 14.
Geodesy, cartography and aerial photography. Issue 98, 2023 75

Raufu, I. O., & Tata, H. (2021). Accuracy Assessment of Turner, I. L., Harley, M. D., & Drummond, C. D. (2016).
Different Polynomial Geoid Models in Orthometric UAVs for coastal surveying. Coastal Engineering, 114,
Height Determination for Akure, Nigeria. Geodetski 19–24
glasnik, 52, 61-73. Yeh, F.H., Huang, C.J., Han, J.Y., & Ge, L. (2018). Model-
Sansò, F., Reguzzoni, M., & Barzaghi, R. (2019). Geodetic ing Slope Topography Using Unmanned Aerial Vehicle
Heights. Springer Nature, Switzerland, [Link] Image Technique. MATEC Web of Conferences, 147,
10.1007/978-3-030-10454-2. 07002. doi:10.1051/matecconf/201814707002.

Ібрагім Олатунджі RAUFU1A*, Герберт TATA1B, Соліху OLAOSEGBA2

1
Департамент геодезії та геоінформатики, Федеральний технологічний університет, Акуре, Нігерія, raufuibrahimola-
tunji@[Link], htata@[Link], 1A[Link] 1B[Link]
2
Кафедра геодезії та геоінформатики, Федеральна школа геодезії Ойо, Нігерія, olaosegba@[Link],
2
[Link]

МОДЕЛЮВАННЯ ЛОКАЛЬНИХ ГЕОЇДНИХ УНДУЛЯЦІЙ

ЗА ДОПОМОГОЮ БЕЗПІЛОТНИХ ЛІТАЛЬНИХ АПАРАТІВ (БПЛА):
ПРИКЛАД ФЕДЕРАЛЬНОГО ТЕХНОЛОГІЧНОГО УНІВЕРСИТЕТУ, АКУРЕ, НІГЕРІЯ

Дослідження було спрямоване на розробку моделі геоїда з використанням технології безпілотних

літальних апаратів (БПЛА). Для цього використано БПЛА для отримання зображень досліджуваної території
з висоти 150 м із роздільною здатністю на Землі 4,19 см. Всього отримано 3737 зображень, які охоплюють
площу 725,804 га. Існуючі еліпсоїдні та ортометричні висоти були використані для географічної прив’язки
отриманих зображень. Для аналізу використано 35 точок, з яких 20 точок визначено як наземні контрольні
точки (GCP), а решта 15 точок – контрольні точки (CPs). Використовуючи отримані з БПЛА цифрові моделі
рельєфу (DTMs), створено набір даних, що містить 18 492 точки як для еліпсоїдальної (h), так і для
ортометричної (H) висот. Різниці між цими висотами, які називаються висотами геоїда (N), були розраховані
як N = h - H для всіх 18 492 точок. Ці висоти геоїда згодом використані для створення моделі геоїда, включаю-
чи контурні карти та 3D-карти досліджуваної території. Щоб оцінити точність висот геоїда, отриманих за до-
помогою БПЛА, виконано аналіз середньоквадратичної помилки (RMSE) шляхом порівняння їх з існуючими
висотами геоїда, і встановлено, що вона становить 0,113 м. Наукова новизна та практична значущість полягає
в розробці локальної моделі геоїда досліджуваної території з точністю до сантиметра. Таким чином, результа-
ти цього дослідження можуть бути використані для широкого спектру застосувань, включаючи землеустрій,
будівництво та оцінку впливу на навколишнє середовище на території дослідження.
Ключові слова: геоїд, БПЛА, цифрові моделі рельєфу, еліпсоїдальна висота, ортометрична висота

Received 20.10.2023