Geodesy and Cartography

ISSN 2029-6991 / eISSN 2029-7009

2022 Volume 48 Issue 1: 36–45
https://doi.org/10.3846/gac.2022.14266

UDC 528.932

VERTICAL ACCURACY ASSESSMENT OF OPEN ACCESS DIGITAL

ELEVATION MODELS: BUCARAMANGA-COLOMBIA CASE STUDY

Jhonathan APONTE SARAVIA *

Faculty of Natural Sciences and Engineering, Unidades Tecnológicas de Santander, Bucaramanga, Colombia

Received 20 January 2021; accepted 10 March 2022

Abstract. Digital Elevation Models (DEMs), are fundamental data that allow to represent topographic information con-
tinuously. They are widely used in various applications such as geoscience, and in the graphical representation of the
landscape surface. Performing the analysis by using DEMs in which the real shape of the surface is adjusted, this would
contribute significantly in obtaining their results as we would be approaching the actual occurrence of the object of study
in the landscape. Currently, several global DEMs are freely available. However, various investigations have produced differ-
ent results, so there are uncertainties as to which model is more appropriate for some areas. In that sense, the research was
aimed at comparing the vertical accuracy of four DEMs in the city of Bucaramanga using central tendency statistical meth-
ods such as mean analysis, standard deviation and root mean squared error. As a result, the model that showed the best
vertical accuracy was the one generated by the Advanced Land Observation Satellite program – Synthetic Aperture Radar
and X-band Shuttle Radar Topography Mission, with a root mean squared error of 8.22 and 8.55 m respectively. Moreo-
ver, the one that best represented the shape of the landscape was the X-band Shuttle Radar Topography Mission X model.
Keywords: vertical uncertainty, digital surface model, accuracy assessment, Free DEM comparison.

Introduction SRTM data generated in the X-band (SRTM X) were also

published by the German Space Agency (DLR). Also, the
Digital Elevation Models are a dataset that allows the con- DSMs of the Advanced Land Observation Satellite [ALOS]
tinuous representation of land topography using quantita- program launched in 2006 by Japan Aerospace Explora-
tive values (Li et al., 2005). They are one of the most im- tion Agency (JAXA) (Tadono et  al., 2014) are free and
portant attributes for terrain modelling (Yahaya & Azzab, publicly available in digital repositories.
2019), which has multiple applications in the survey and These DSMs have been successfully used in various
analysis of geographic objects that are bounded to the applications, such as: Hydrological modelling (Wilson
Earth’s surface (Florinsky, 2016; Yahaya & Azzab, 2019). & Gallant, 2000), glaciology studies (Mölg et  al., 2017),
They are classified into two groups: Digital Terrain Models geomorphology studies (Florinsky, 2016), environmen-
(DTM), which describe objects at the ground level, and tal studies (Li et al., 2005), civil engineering applications
Digital Surface Models (DSMs), which reflect the Earth’s (Li et al., 2005), soil mapping and prediction (Florinsky,
surface including all man-made objects and natural ob- 2016), spatial distribution of vegetation (Jelaska, 2009),
jects (Martha et al., 2010; Florinsky, 2016). meteorological applications (Emeis & Knoche, 2009),
Currently, several organizations have released digital among others. Therefore, it is essential to know the DSMs
surface model data generated through indirect methods, generation processes and their levels of uncertainty with
some of which have global coverage. In addition, in recent respect to the real shape of the Earth’s surface.
years these data have reportedly improved their perfor- However, DSMs have several errors, which affect the
mance in describing the shape of the land surface. This is representation of the Earth’s surface. One of them are ran-
the case of the SRTM program, which has released DSM dom errors that may be associated with the altitude varia-
data in different periods (version 1, version 2 and ver- ble. In turn, they are categorized by their spatial scale into
sion 3). For example, at the end of 2014 the version 3 of long-wavelength absolute biases (Rodríguez et  al., 2006;
the SRTM model generated in the C-band (SRTM C) was Dowling et al., 2011). Also, the DSMs can be affected by
released by Jet Propulsion Laboratory (2020). Similarly, systematic errors, which are caused by the tree canopies in

*Corresponding author. E-mail: [email protected]

Copyright © 2022 The Author(s). Published by Vilnius Gediminas Technical University

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unre-
stricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Geodesy and Cartography, 2022, 48(1): 36–45 37

forested areas (O’Loughlin et al., 2016). Moreover, there is few investigations have evaluated the performance of
evidence of these errors in areas without vegetation cover– DEMs in Latin America.
especially in areas with a steep slope (González-Moradas
& Viveen, 2020). 1. Materials and methods
When the DSMs are obtained, the metadata generally
indicates the values of precision and accuracy. However, 1.1. Study area
it is understood that these values are a global reference, The study area is located in the city of Bucaramanga,
and in some parts of the Earth the terrain shape is abrupt northeast of Colombia’s capital, on the north-eastern side
and complex, especially in mountainous landscapes. of the Andes Mountains, in the geographic coordinates
Therefore, landscape analyses in areas with complex ter- (7°7′8.79′′ N; 73° 7′21.65′′ W). The study area is made up
rain topography generate results with unknown levels of of a variety of terrain types relief such as: Hills, valleys, de-
uncertainty. In addition, research has shown that vertical pressions, plains, plateaus, and mountain elevations in the
accuracy values vary according to slope (Vaka et al., 2019; most extreme parts. The study area covers approximately
Yahaya & Azzab, 2019), where the Root Mean Squared Er- 263.22 km2, see Figure 1. This whole area has a diverse
ror (RMSE) varies between ±4.22 and ±27.58 m (Sharma topography, where land slopes of less than 12% predomi-
et al., 2010), Moreover, SRTM version 4.1 with a pixel size nate. There are also small flat and semi-flat areas, where
of 90 m for South America shows an RMSE of approxi- the city of Bucaramanga, outlying cities and agricultural
mately 11.2 m (Mukul et al., 2017). Thus, these results are production areas are located. And in the areas where the
still very general. land slope varies between 12% and 25%, there are creeks
On the other hand, it is essential to know which model and small valleys. The areas where the land topography
allows to present with greater accuracy and precision the is higher than 25% correspond to the highest parts of the
continuous description of the data associated with the selected area, especially in the north-east.
altitude variable and its spatial location; it is crucial for
the success of many applications in geoscience, engineer-
1.2. Data
ing and other areas (Vaka et  al., 2019). After all, using
accurate data in landscape modelling processes and their The data used for this study were collected from the SRTM
interactions will provide results that are highly correlated X digital surface model, which was developed by the Na-
with real-world situations. Accordingly, it may be possible tional Aeronautics and Space Administration [NASA] in
to understand why certain complex processes are occur- cooperation with the National Geospatial-Intelligence
ring in the landscape (Wechsler, 2007). Thus, the aim of Agency (NGA), the German Space Agency (DLR) and
the research was to compare the vertical accuracy of four the Italian Space Agency (ASI) (German Space Agency,
open-access DEMs from different sources for the city of 2020b), using the SAR technique, in the X-band, whose
Bucaramanga, the area has different types of surfaces, such wavelength is 3.1 cm (Farr et al., 2007). Additionally, the
as: surfaces with rugged topography and areas where the DSM metadata indicates that it has a relative vertical ac-
surface has been modified by human activity, in addition, curacy of around 6 m and an absolute vertical accuracy of

Figure 1. Shows the location of the study area: a – study area coverage using continuous DEM data; b – the location of the study
area in the territory of Colombia; c – the presentation of the study area in a 3D model
38 J. Aponte Saravia. Vertical accuracy assessment of open access digital elevation models: Bucaramanga-Colombia...

around ±16 m at the 90% confidence level (German Space In addition, a digital surface model developed by the
Agency, 2020a). PALSAR program (ALOS SAR) was used – a project de-
Similarly, the SRTM C version 3 digital surface model veloped by JAXA and the Japan Resource Observation
was used, which was generated using the SAR technique, System (JAROS) using the SAR technique. This data was
in the C-band, wavelength of 5.6 cm (Farr et  al., 2007), polarimetrically obtained in the L-band frequency, gener-
with spatial resolution similar to the SRTM X model. Its ated from a horizontal polarization (HH) and a horizon-
purpose was to obtain data that did not exceed the 16 m tal-vertical polarization (HV) (Advanced Land Observing
vertical accuracy error at 90% confidence level (Mukul Satellite, 2020), see Figure 2.
et al., 2017). This version was released in 2015 (National Furthermore, for the vertical accuracy assessment of
Aeronautics & Space Administration, 2015). digital surface models, 88 high precision checkpoints (ver-
The ALOS digital surface model was also used, which tices of the national geodetic network of Colombia) were
consisted in a project called ALOS World 3D (AW3D) used, configured in the National Geocentric Reference
version 3.1, developed by the JAXA and launched in Janu- Framework, densification of the Geocentric Reference Sys-
ary 2006 (Tadono et  al., 2017). It used more than three tem for the Americas MAGNA SIRGAS (Sanchez, 2004).
million satellite images to create 3D topographic data, ap- The data were configured in the Geo-Col2004 Geoidal
plying the traditional technique of optical stereo matching Model, which represents the direct relationship between
of panchromatic images: Panchromatic Remote Sensing the geoid and the quasigeoid, and is equivalent to the dif-
Instrument for Stereo Mapping (PRISM) (Takaku et  al., ference of orthometric and normal heights (Sánchez Rod-
2014), with a vertical mean squared error of less than 4 m ríguez, 2003). These data are available in the geoportal of
(Tadono et al., 2017). Instituto Geográfico Agustín Codazzi.

Figure 2. Presents the visual comparison of the digital elevation models, the blue pixels indicate the zones of lower altitude and the
red zones show the zones of higher altitude: a – DEM AW3D; b – DEM ALOS SAR; c – DEM SRTM C; d – DEM SRTM X
 Geodesy and Cartography, 2022, 48(1): 36–45 39

Table 1. Presents the descriptive statistical distribution of the altitudes of the digital elevation models of the selected area and the
number of pixels covered by the grid.

Data Height min. Height Max. Mean SD Datum vertical No data cells
AW3D 581.04 1773.03 959.48 209.42 WGS84 0
ALOS SAR 599.96 1784.92 967.06 208.84 EMG96 0
SRTM C 592.65 1774.80 957.87 208.75 EMG96 0
SRTM X 485.33 1813.40 968.68 208.07 WGS84 1206

1.3. Methods geoscience as well as in DEMs (Vaka et  al., 2019; Wes-

sel et al., 2018; González-Moradas & Viveen, 2020) – was
After downloading and organizing data, both in the digital
calculated using the following expression:
surface models and in the checkpoints (geodesic points),
they were converted into flat coordinates, in order to have 1 n
( )
2
data in the same reference system. The DEMs associated =
RMSE ∑ hpoint − hDEM
n i =1
, (3)
with the area of interest were then extracted, with pixel
sizes of 30 m. for each of the DEMs. Subsequently, the de- where n is the number of points used in the sample,
scriptive statistical exploration, as well as the verification ( h point ) is the height of the checkpoints and (hDEM ) is the
of the vertical datum, were performed on these selected height of the digital elevation model.
data as shown in Table 1. Then, assuming that the values of the vertical errors
Similarly, a preliminary analysis of the checkpoints have a normal distribution and the linear error is propor-
(geodetic points) revealed that these data contained geo- tional to the standard error (Grohmann, 2018) and in ac-
metric height and undulation values. Thus, it was possi- cordance with the American Society for Photogrammetry
ble to convert data into ellipsoidal heights, since digital and Remote Sensing (ASPRS) standards for these process-
surface models denote ellipsoidal heights. For such pur- es (ASPRS, 2015; González-Moradas & Viveen, 2020), the
pose, the equation applied by Yahaya and Azzab (2019) confidence intervals were calculated at 95% using the fol-
was used. lowing expression.
The pixel values of DEMs that match the horizontal
95% 1.9600 × SD. (4)
LE=
position of the checkpoints were then extracted. Subse-
quently, statistical assessments of vertical accuracy com- Similarly, the residual distribution was assessed to
parison analysis were performed, using the System for Au- understand how these values are distributed according
tomated Geoscientific Analyses (SAGA), R Package and to the slope, which was also assessed in similar studies
RStudio for data processing and analysis, example the dif- by González-Moradas and Viveen (2020) and Vaka et al.
ferences between the height of the checkpoints (h point ) and (2019). In the same way, linear correlation between the
the height of the DEMs (hDEM ) were calculated (Ibrahim height of the checkpoints and the height of the DEMs was
Yahaya & El Azzab, 2019). The vertical accuracy analysis analysed in order to identify the model with the highest
is based on the analysis of statistical parameters such as: correlation value.
Mean (ME), Standard Deviation (SD) and RMSE (Vaka Then, the digital surface models were compared. For
et  al., 2019; Wessel et  al., 2018), For this purpose, the this process, data referring to the DSM that provides the
mean residual values were calculated using the following most accurate statistical analysis was considered. It was
expression: then compared with the other less precise models, follow-
ing the criteria applied by González-Moradas and Viveen
1 n 1 n (2020), using the following relation:
ME = ∑h point − hDEM = ∑∆hi , (1)
n n
=i 1 =i 1 ∆DEM= DEMref − DEMi , (5)
where ∆hi are the residual error values, and n is the num- where ∆DEM , are the residual values of the models,
ber of checkpoints. These values were converted into ab- DEMref is the reference model, and DEMi is the model
solute values for their analysis. This expression was also selected for comparison. After obtaining the residual val-
applied by González-Moradas and Viveen (2020). In ad- ues of the digital elevation models, the absolute value was
dition, SD (American Society for Photogrammetry and calculated and comparisons were made with the pending
Remote Sensing [ASPRS], 2015) was calculated on each topographic attribute generated from the most accurate
model using the following equation: DEM, in order to observe the statistical and spatial behav-
iour in different landforms.
1 n 2
=SD ∑
n − 1 i =1
( ∆hi − ME ) . (2)
2. Results and discussions
Similarly, the RMSE – which is a widely used metric The following describes the most important results of the
for measuring the accuracy of continuous variables in vertical uncertainty assessment process for the different
40 J. Aponte Saravia. Vertical accuracy assessment of open access digital elevation models: Bucaramanga-Colombia...

Figure 3. Presents the methodological scheme of the research process, indicating the entry,
data analysis and obtaining the results of vertical accuracy

global digital surface models. Also, the presentation and and 7.87 respectively – Alganci et  al. (2018) stated that
comparison of the results was done using the R package the ALOS model showed a greater accuracy than SRTM
software product, RStudio and the presentation of surface C. Whereas, González-Moradas and Viveen (2020) men-
profiles in SAGA. tioned that the SRTM C model has higher vertical ac-
curacy when compared to other models such as: ASTER
Table 2. Presents the descriptive statistical distribution of the GDEM2, AW3D. Other research has shown more accurate
residual errors between the control points and the digital results as described in (Alganci et al., 2018). Therefore, it
surface models
is important to make these comparisons, because vertical
Data Statistics AW3D ALOS SAR SRTM C SRTM X accuracy depends significantly on landscape conditions
and geographic locations.
Absolute maximum On the other hand, when analysing the behaviour of
21.02 20.70 28.75 24.07
residual (m)
the residual errors according to the slope, it can be seen
Absolute minimun
0.30 0.10 0.40 0.35 that the SD increases as the slope increases. This indi-
residual (m)
cates that the vertical accuracy is greater in areas where
Absolute mean
residual (m)
10.13 6.13 11.27 6.32 the slope is less than 2.5%, and that the higher the slope,
the lower the vertical accuracy, see image (a) in Figure 4.
SD 4.63 5.51 7.18 5.79
This criterion is generally met in all four DEMs, which is
RMSE 11.13 8.22 13.34 8.55 consistent with Vaka et al. (2019) findings. Model SRTM
LE 95% 9.07 10.79 14.07 11.34 X errors showed greater vertical accuracy in landscapes
where land topography is less than 2.5% and uncertainty
From the vertical comparison analysis of digital sur- is higher in areas where land topography is greater than
face models with high-precision checkpoints, it is evident 7.5% compared to other models. While ALOS SAR re-
that the ALOS SAR model has the lowest RMSE, followed sidual errors show lower SD in areas where the slope is
by SRTM X. It also has the lowest SD values, except in greater than 20%, this model shows less uncertainty for
AW3D, indicating that the residual error dataset is clus- the study area on various slope types, see Table 2 and Im-
tered in a relatively small range. However, AW3D and age (a) in Figure 4.
SRTM C have the highest mean squared errors, which in- Thus, there are different vertical accuracy behaviours
dicates a greater uncertainty, see Table 2. Similar to Flor- in various types of vegetation cover, see image (b) of
insky et  al. (2018), who also mentioned that the SRTM Figure 3, which shows a greater vertical accuracy in ar-
model is more inaccurate than AW3D–RMSE of 17.91 eas considered as flat and semi-flat areas, where there is
Geodesy and Cartography, 2022, 48(1): 36–45 41

Figure 4. Shows the behavior of residual errors: a – SD in different slope ranges; b – SD in different types of coverage

little vegetation and no urban constructions. This pattern models because the pixel size is large and this does not
is consistent in all models, and uncertainty is greater in allow to generate surfaces with a high level of detail,
mixed areas (constructed areas, areas with vegetation and however, it is appropriate to generate cartographic in-
open spaces); RMSE is higher in areas with dense vegeta- formation with a low level of detail.
tion and abrupt topography, revealing a greater uncertain- Considering that the ALOS SAR digital surface model
ty in these places. It is inferred that dense vegetation and is the one with the highest vertical accuracy, it was used
urban constructions are the components that significantly as a reference model to perform the residual error calcula-
increase the uncertainty of the global DEMs generated by tion on the other models with lower vertical accuracy, in
indirect methods, which is consistent with the findings of order to describe and assess the behaviour of residual er-
Vaka et  al. (2019), Florinsky et  al. (2018), Alganci et  al. rors using the central trend descriptive criteria. The results
(2018). are shown in Figure 6.
From the analysis of the surface profile, the model From the analysis of the residual data (between ALOS
which best describes the shape of the landscape at sur- SAR – SRTM C), it is observed that the central trend of
face level is the SRTM X model, since this model – in data is 2.078, with this value being the closest to zero with
areas with dense vegetation and a slope greater than 20% respect to the other models, see Figure 6. The above shows
– is able to describe the surface profile in greater detail, that most residual data is concentrated between –6.499
including the peaks of the surface profiles in the most and 9.655 m, while the other residual datasets are within
prominent trees, see image (a) and (b), in Figure 5. Thus, greater ranges, despite the fact that SD values are lower.
in urban areas it manages to represent in greater detail the Hence, the higher residual values observed when com-
surface profiles of the buildings and vegetation, followed paring the most accurate model with the others suggest
by the AW3D model, which also represents the shape of a greater uncertainty. Given this criterion, the model that
the surface, but in less detail, as shown in picture (c), (d), presented higher residual values is ALOS SAR – SRTM X.
(g) and (h) in Figure 5. Similarly, in areas with little bush This is because the model SRTM X represents in greater
vegetation and topography of less than 20%, the pixels of detail the landscape surface shape, see Figure 4. On the
the SRTM X model describe in greater detail the shape of other hand, the ALOS SAR model represents the sur-
the landscape at the surface level. face shape in less detail, i.e. the description of the surface
The images (d) and (h) of Figure 5, shows the surface shape is more homogeneous. However, the residual values
profiles of different digital elevation models, where it of ALOS SAR – SRTM C show less uncertainty, probably
allows us to show with better performance the descrip- because the characteristics of both models provide a more
tion of the surface of each model, because this sec- homogeneous description of the landscape shape. In other
tion represents surfaces with abrupt slopes generating words, these models do not represent the surface shape in
a more complex surface edge, see images (c) and (g) much detail. However, this assessment is not statistically
of the same figure, also the length of the profile sec- relevant because there are few data with high residual val-
tion is shorter, achieving to observe the surface pro- ues, see Figure 6.
files with greater detail in the SRTM X model and the
AW3D model. On the other hand, the images (f ) and Conclusions
(j) do not help to visually differentiate the performance
of the model in the process of describing the surface, From the vertical accuracy comparison analysis of the
even though the profile section is located in surfaces models evaluated, the ones that showed the greatest verti-
with smooth, abrupt undulations, where the vegetation cal accuracy were ALOS SAR and SRTM X, whose mean
does not have significant heights, generating confusion. squared error values are 8.22 and 8.55, respectively. On the
In this sense, these data generate inaccurate surface other hand, the models that better described the shape of
42 J. Aponte Saravia. Vertical accuracy assessment of open access digital elevation models: Bucaramanga-Colombia...

Figure 5. Shows the behaviour of the surface profiles in different types of coverings: a – vegetation coverage in topography of land
greater than 20%; b – surface profile behaviour in vegetation zones; c – the image shows urban infrastructure; d – surface profile
of different digital surface models in urban areas; e – shows zones with scarce vegetation and areas of flat and semi-flat slopes; f –
indicates the surface profile in areas of low vegetation; g – the image shows urban infrastructure and in areas with dense vegetation;
h – shows the surface profile in areas of urban structures and in areas with dense vegetation; i – shows the image in areas where the
slope varies between 30% and 52%; j – surface profile in rugged areas
Geodesy and Cartography, 2022, 48(1): 36–45 43

Figure 6. Shows the spatial distribution of the residual errors values of the DEM: a – residual error of the difference between
ALOS SAR and AW3D; b – distribution of the residual data of the difference between ALOS SAR and AW3D; c – residual error of
the difference between ALOS SAR and SRTM C models; d – distribution of residual difference errors of ALOS SAR and SRTM C
models; e – spatial distribution of the difference errors between ALOS SAR and SRTM X; f – difference residual error histogram
of ALOS SAR and SRTM X models
44 J. Aponte Saravia. Vertical accuracy assessment of open access digital elevation models: Bucaramanga-Colombia...

the landscape were SRTM X and AW3D. Specifically, the Oskin, M., Burbank, D., & Alsdorf, D. (2007). The shuttle
SRTM X model tries to represent in greater detail the el- radar topography mission. Reviews of Geophysics, 45(2), 1–33.
evations of urban structures and dominant trees. However, https://doi.org/10.1029/2005RG000183
the models that represent the terrain shape more homo- Florinsky, I. (2016). Digital terrain modeling: A brief historical
overview. In Digital terrain analysis in soil science and geology
geneously were ALOS SAR and SRTM C, because these
(2nd ed.). Elsevier Inc.
models failed to represent the edges of urban structures
https://doi.org/10.1016/B978-0-12-804632-6.00001-8
or dominant trees. Florinsky, I., Skrypitsyna, T., & Luschikova, O. (2018). Com-
Accordingly, it can be deduced that for the study area parative accuracy of the AW3D30 DSM, ASTER GDEM, and
the SRTM X model has a high potential to make adjust- SRTM1 DEM: A case study on the Zaoksky testing ground,
ments for the landscape modelling at surface level. On the Central European Russia. Remote Sensing Letters, 9(7), 706–
other hand, the ALOS SAR model has no potential to de- 714. https://doi.org/10.1080/2150704X.2018.1468098
scribe the shape of the surface even though it presented German Space Agency. (2020a). Shuttle Radar Topography Mis-
the lowest mean square error value. sion SRTM product. Retrieved December 2022, from https://
geoservice.dlr.de/resources/licenses/srtm_xsar/DLR_SRTM_
XSAR_ReadMe.pdf
Acknowledgements German Space Agency. (2020b). The SRTM X-SAR digital eleva-
tion model. Retrieved December 2020, from https://geoser-
My deepest thanks to the research group “Grupo de In-
vice.dlr.de/web/dataguide/srtm/#further_information_mis-
vestigación Medio Ambiente y Territorio – GRIMAT”, for sion
giving me the necessary time to conclude this investiga- González-Moradas, M., & Viveen, W. (2020). Evaluation of AS-
tion and the manuscript. TER GDEM2, SRTMv3.0, ALOS AW3D30 and TanDEM-X
DEMs for the Peruvian Andes against highly accurate GNSS
Author contributions ground control points and geomorphological-hydrological
metrics. Remote Sensing of Environment, 237, 111509.
Jhonathan Aponte Saravia, has contributed to the design, https://doi.org/10.1016/j.rse.2019.111509
data collection, data analysis, interpretation of results and Grohmann, C. H. (2018) Evaluation of TanDEM-X DEMs on
construction of the manuscript. selected Brazilian sites: Comparison with SRTM, ASTER
GDEM and ALOS AW3D30. Remote Sensing of Environment,
212, 121–133. https://doi.org/10.1016/j.rse.2018.04.043
Disclosure statement Ibrahim Yahaya, S., & El Azzab, D. (2019). Vertical accuracy
assessment of global digital elevation models and validation
The author of the manuscript has no interest in financial,
of gravity database heights in Niger. International Journal of
professional or personal issues regarding the publication Remote Sensing, 40(20), 7966–7985.
of this document, but on the contrary the purpose is to https://doi.org/10.1080/01431161.2019.1607982
share the results of the research with the scientific com- Jelaska, S. D. (2009). Vegetation mapping applications. In
munity. T.  Hengl & H.  Reuter (Eds.), Developments in soil science:
Vol.  33. Geomorphometry – concepts, software, aplications
(pp. 481–496). Elsevier.
References
https://doi.org/10.1016/S0166-2481(08)00021-4
Advancend Land Ovserving Satellite. (2020). Palsar phased ar- Jet Propulsion Laboratory. (2021). U.S. releases enhanced shuttle
ray type l-band synthetic aperture radar. Retrieved July 2020, land elevation data. California Institute of Technology. Re-
from https://www.eorc.jaxa.jp/ALOS/en/about/palsar.htm trieved January 2021 from https://www2.jpl.nasa.gov/srtm/
Alganci, U., Besol, B., & Sertel, E. (2018). Accuracy assessment of Li, Z., Zhu, C., & Gold, C. (2005). Digital terrain modeling: Prin-
different digital surface models. ISPRS International Journal of ciples and methodology. CRC Press.
Geo-Information, 7(3), 114. Martha, T. R., Kerle, N., Jetten, V., van Westen, C. J., & Kumar, K. V.
https://doi.org/10.3390/ijgi7030114 (2010). Characterising spectral, spatial and morphometric
American Society for Photogrammetry and Remote Sensing. properties of landslides for semi-automatic detection using
(2015). ASPRS positional accuracy standards for digital geo- object-oriented methods. Geomorphology, 116(1–2), 24–36.
spatial data. Photogrammetric Engineering & Remote Sensing, https://doi.org/10.1016/j.geomorph.2009.10.004
81(3), A1–A26. https://doi.org/10.14358/PERS.81.3.A1-A26 Mölg, N., Ceballos, J. L., Huggel, C., Micheletti, N., Rabatel, A.,
Dowling, T. I., Brooks, M., & Read, A. M. (2011, December). & Zemp, M. (2017). Ten years of monthly mass balance of
Continental hydrologic assessment using the 1 second (30 m) conejeras glacier, colombia, and their evaluation using dif-
resolution Shuttle Radar Topographic Mission DEM of Aus- ferent interpolation methods. Geografiska Annaler: Series A,
tralia. In 19th International Congress on Modelling and Simula- Physical Geography, 99(2), 155–176.
tion. Perth. https://doi.org/10.1080/04353676.2017.1297678
Emeis, S., & Knoche, H. R. (2009). Applications in meteorology. Mukul, M., Srivastava, V., Jade, S., & Mukul, M. (2017). Uncer-
In T. Hengl & H. Reuter (Eds.), Developments in soil science: tainties in the Shuttle Radar Topography Mission (SRTM)
Vol. 33. Geomorphometry – concepts, software, aplications (pp. heights: Insights from the Indian Himalaya and Peninsula.
603–622). Elsevier. Scientific Reports, 7, 41672. https://doi.org/10.1038/srep41672
https://doi.org/10.1016/S0166-2481(08)00026-3 National Aeronautics and Space Administration. (2015). The
Farr, T. G., Rosen, P. A., Caro, E., Crippen, R., Duren, R., Shuttle Radar Topography Mission (SRTM) collection user
Hensley, S., Kobrick, M., Paller, M., Rodriguez, E., Roth, L., guide. Retrieved June 2015, from https://lpdaac.usgs.gov/
Seal,  D., Shaffer, S., Shimada, J., Umland, J., Werner, M., documents/179/SRTM_User_Guide_V3.pdf
Geodesy and Cartography, 2022, 48(1): 36–45 45

O’Loughlin, F. E., Paiva, R. C., Durand, M., Alsdorf, D. E., & face model dataset. In IEEE International Geoscience and Re-
Bates, P. D. (2016). A multi-sensor approach towards a global mote Sensing Symposium (IGARSS) (pp. 5656–5657).
vegetation corrected SRTM DEM product. Remote Sensing of https://doi.org/10.1109/IGARSS.2017.8128290
Environment, 182, 49–59. Takaku, J., Tadono, T., & Tsutsui, K. (2014). Generation of high-
https://doi.org/10.1016/j.rse.2016.04.018 resolution global DSM from ALOS PRISM. ISPRS Annals of
Rodríguez, E., Morris, Ch. S., & Belz, J. E. (2006). A Global as- Photogrammetry, Remote Sensing & Spatial Information Sci-
sessment of the SRTM performance. Photogrammetric Engi- ences, 2(4), 243–248. https://www.aw3d.jp/wp/wp-content/
neering & Remote Sensing, 72(3), 249–260. themes/AW3DEnglish/technology/doc/pdf/technology_02.
https://doi.org/10.14358/PERS.72.3.249 pdf
Sánchez Rodríguez, L. (2003). Determinación de la superficie ver- Vaka, D. S., Kumar, V., Rao, Y., & Deo, R. (2019, July). Com-
tical de referencia para Colombia [Master’s Thesis]. Technis- parison of various DEMs for Height accuracy assessment
che Universitat Dresden. https://www.igac.gov.co/sites/igac. over different terrains of India. In IGARSS 2019 – 2019 IEEE
gov.co/files/modelogeoidalgeocol2004:pdf International Geoscience and Remote Sensing Symposium
Sanchez Rodriguez L. (2004). Aspectos Prácticos de la Adopción (pp. 1998–2001). Yokohama, Japan.
del Marco Geocéntrico Nacional de Referencia MAGNA SIR- https://doi.org/10.1109/IGARSS.2019.8898492
GAS como datum oficial de Colombia (Tech. Rep.). Subdirec- Wechsler, S. P. (2007). Uncertainties associated with digital eleva-
ción de Geografía y Cartografía. Instituto Geografico Agustin tion models for hydrologic applications: A review. Hydrology
Codazzi. https://www.igac.gov.co/sites/igac.gov.co/files/aspec- and Earth System Sciences, 11(4), 1481–1500.
tospracticos.pdf https://doi.org/10.5194/hess-11-1481-2007
Sharma, A., Tiwari, K., & Bhadoria, P. (2010). Vertical accuracy Wessel, B., Huber, M., Wohlfart, C., Marschalk, U., Kosmann, D.,
of digital elevation model from Shuttle Radar Topographic & Roth, A. (2018). Accuracy assessment of the global Tan-
Mission – a case study. Geocarto International, 25(4), 257– DEM-X Digital Elevation Model with GPS data. ISPRS Jour-
267. https://doi.org/10.1080/10106040903302931 nal of Photogrammetry and Remote Sensing, 139, 171–182.
Tadono, T., Ishida, H., Oda, F., Naito, S., Minakawa, K., & Iwa- https://doi.org/10.1016/j.isprsjprs.2018.02.017
moto, H. (2014). Precise global DEM generation by ALOS Wilson, J. P., & Gallant, J. C. (Eds.) (2000). Terrain analysis: Prin-
PRISM. ISPRS Annals of the Photogrammetry, Remote Sensing ciples and applications. Wiley.
and Spatial Information Sciences, 2(4), 71–76. https://www. Yahaya, S. I., & Azzab, D. E. (2019). Vertical accuracy assessment
aw3d.jp/wp/wp-content/themes/AW3DEnglish/technology/ of global digital elevation models and validation of gravity
doc/pdf/technology_03.pdf database heights in Niger. International Journal of Remote
Tadono, T., Takaku, J., Ohgushi, F., Doutsu, M., & Kobayashi, K. Sensing, 40(20), 7966–7985.
(2017). Updates of ‘AW3D30’ 30 M-MESH global digital sur- https://doi.org/10.1080/01431161.2019.1607982
 © 2022. This work is published under
https://creativecommons.org/licenses/by/4.0/ (the “License”).
Notwithstanding the ProQuest Terms and Conditions, you may use this
content in accordance with the terms of the License.