**1. Introduction**

218 Studies on Environmental and Applied Geomorphology

Holm, Donald A. (1960) Desert Geomorphology in the Arabian Peninsula. *Science, New* 

IUCNR/MEPA (1987) *Saudi Arabia: An assessment of coastal zone management requirements for* 

Kampf, J. and Sadrinasab, M. (2005) The Circulation of the Persian Gulf: A Numerical Study*,* 

Mah, A. (2004) Detecting mangrove changes using Landsat imagery: Tarut Bay Pilot Project.

NCWCD – National Commission for Wildlife Conservation and Development (2004)

PME – Presidency of Meteorology and Environment (2004) Environmental Protection

Rakha, K, Al-Salem, K. and Neelamani, S. (2007) Hydrodynamic Atlas for the Arabian Gulf*.* 

Seber, Dogan, Vallve, Marisa, Sandvol, Eric Steer, David and Barazangi, Muawia. (1997)

http://atlas.geo.cornell.edu/htmls/gsa\_today.html accessed 22 April 2009

Sulayem, M. and Joubert, E. (1994) Management of protected areas in the kingdom of Saudi

Summerfield, Michael A. (1991) *Global Geomorphology*. Pearson Education Limited, ISBN 0-

Viles, Heather and Spencer, Tom (1995) *Coastal Problems – Geomorphology, Ecology and Society* 

NCWCD and its mandate from *NCWCD* website accessed October 2004

Standards. *From Presidency of Meteorology and Environment* website accessed

Middle East Tectonics: Applications of Geographic Information Systems (GIS). *Institute of the Study of the Continents*, Cornell University, Snee Hall, Ithaca, NY.

Arabia. *Unasylva* - No. 176 - Parks and protected areas Vol. 45 - 1994/1. Food and

*at the Coast*. Edward Arnold, ISBN 0-340-53197-5 (Pb), ISBN 0-340-62540-6 (Hb),

Geology, October 2006 V. 34 no. 10 p. 869-872; doi: 10.1130/G22713.1. Haq, Bilal U. and Al-Qahtani, Abdul Motaleb (2005) Phanerozoic Cycles of Sea-level Change

on the Arabian Platform. *GeoArabia*, Vol. 10, No.2 2005.

*Series*, Vol. 132, No. 3437, pp. 1369-1379.

Conservation and Natural Resources.

http://www.populstat.info/Asia/saudiarc.htm

*Ocean Science Discussions.* 2, pp129-164, May 2005.

http://www.ncwcd.gov.sa/english/intro.htm

September 2004. http://www.pme.gov.sa/Env.asp

Saudi Wildlife Commission (2010) Website accessed February 2011

Agriculture Organization of the United Nations (FAO)

http://www.swc.gov.sa/English/default.aspx

582-30156-4, Harlow, Essex, England.

London, UK.

http://www.fao.org/docrep/v2900E/v2900e08.htm

Lahmeyer, J. J. (2006) Population Statistics Webpage Accessed February 2011.

In *EnviroNews*, Issue No. 12, Saudi Aramco publication, p1-3.

*Journal of Coastal Research*, Special Issue 50, pp 550-554, 2007.

Saudi Arabia: Evidence from Shear-wave Splitting and Seismic Anisotropy.

*the Red Sea*, Arabian Gulf,*Saudi Arabia: An assessment of biotopes and coastal zone management requirements for the Arabian Gulf* [Table of contents only], *Saudi Arabia: An assessment of national coastal zone management requirement*s [Table of contents only]. Technical reports No. 3, 5 and 7. Gland, Switzerland: International Union for

> A Digital Elevation Model (DEM) refers to a quantitative model of a part of the earth's surface in digital form (Burrough and McDonnell, 1998). A DEM consists of either (1) a twodimensional array of numbers that represents the spatial distribution of elevations on a regular grid; (2) a set of *x*, *y*, and *z* coordinates for an irregular network of points; or (3) contour strings stored in the form of *x*, *y* coordinate pairs along each contour line of specified elevation (Walker and Willgoose, 1999). Though there are some disadvantages (Gao, 1997), regular grid DEMs are nowadays the most popular due to their computational efficiency. The use of DEM in this paper, therefore, refers to a regular gridded DEM.

> DEMs are useful for many purposes, and are an important precondition for many applications (Kim and Kang, 2001; Vadon, 2003). They are particularly useful in regions that are devoid of detailed topographic maps. DEMs have been found useful in many fields of study such as geomorphometry, as these are primarily related to surface processes such as landslides which can directly be depicted from a DEM (Hengl and Evans, 2009), archaeology as subtle changes due to previous human activity in the sub surface can be inferred on detailed DEMs (Menze *et al*., 2006), (commercial) forestry, e.g. height of trees and relation to preferred tree stem size (Simard *et al*., 2006), hydrology, like deriving drainage network and overland flow areas that contribute to (suspended) sediment loads (Lane *et al*., 1994) and analysis of glaciers (movement of glaciers using multi temporal DEM's) and glaciated terrains (change of glacier thickness by comparison of multi temporal DEM's) (Bishop *et al*., 2001). Thus for a whole range of different studies, typically topics of interest to geomorphologists, DEMs that provide a good representation of the terrain, are of utmost importance as a starting point for further analysis.

> DEMs can be generated using several methods, with varying degrees of accuracy and cost (Flood and Gutelius, 1997). Traditionally, they have been derived from contours that are extracted with photogrammetric techniques from aerial stereo photographs using the

Comparison of SRTM and ASTER Derived Digital Elevation Models

have reported on an absolute accuracy assessment though.

software (Hutchinson, 1988, 1989).

In short, the objectives of this work are to:

**1.1 Objectives** 

In this study, the C-band SRTM - and ASTER – derived DEMs are compared and validated against a reference DEM for two regions in Ghana. Next to an absolute accuracy assessment also a relative assessment has been conducted. For many environmental and hydrological processes, even if the absolute elevation is not correct, relative to its surrounding regions it might for example exhibit the appropriate flow direction which is the basis of many GIS routines to derive a hydrological network. The reference DEM used was generated using data from a 1:50 000 scale contour map with the TOPOGRID function in the ArcInfo™

• Assess how the elevation data from SRTM and ASTER do compare in an absolute manner with respect to a contour based "traditionally derived" elevation model in an

over Two Regions in Ghana – Implications for Hydrological and Environmental Modeling 221

SRTM and ASTER-derived DEMs are two post-processed elevation datasets that are frequently used for a wide range of applications due to their near-global coverage (Nikolakopoulos and Chrysoulakis, 2006). Dozens of researchers have, over the past few years, carried out a series of global and local assessments of these products. Rodriquez et al. (2006) performed a global validation of the SRTM product using globally distributed set of ground control points derived from Kinematic Global Positioning System (KGPS) transects and National Geospatial-Intelligence Agency's (NGA) level 2 Digital Terrain Elevation Data (DTED). They concluded that SRTM has an absolute height error that exceeds the mission's goal of 16m (90%) often by a factor of two. Hofton et al. (2006) also used data from NASA's Laser Vegetated Imaging Sensor (LVIS) to assess SRTM's accuracy at study sites of variable relief and landcover (i.e. vegetated and non-vegetated terrain). They found out that, under "bare earth" conditions, SRTM data are accurate measurements of LVIS data whereas in vegetated areas, SRTM fell above the ground but below the canopy top, indicating an increase in the vertical offset between SRTM and the LVIS ground data. Jarvis et al. (2004) examined relative and absolute differences between SRTM DEM and a cartographically derived (TOPO) DEM using 1:50,000 scale contours digitized for all Honduras. They validated the two datasets using fifty-nine (59) high accuracy GPS points derived from the National Geodetic Survey (NGS) GPS database from the Central America High Accuracy Reference Network project (NGS, 2003). They also computed slope, aspect, curvature and Pearson's Correlation Coefficient. Their analysis revealed that, SRTM DEM has an average error of 8m as opposed to 20m for the TOPO DEM, although some systematic errors were identified in the SRTM data, related to aspect. They, therefore, concluded that SRTM is more accurate than the 1:50,000 scale cartographically derived DEM for Honduras. Slater et al. (2009) conducted an evaluation of the new ASTER Global Digital Elevation Model (GDEM) using 20 sites in 16 countries. Datasets used for comparison (reference data) with the ASTER GDEM are (1) level 1 and 2 DTED data and (2) control points data (GCPs) photogrametrically derived from satellite imagery. Standard DEM-to-DEM comparisons and DEM-to-control-point comparisons as well as detailed visual analyses of the data were conducted. It was found out that, in almost every area, the mean ASTER GDEM elevations are lower than the reference DEM elevations (biased negatively). The mean ASTER elevations are also lower than the GCPs in every area. Most of the assessments conducted

parallax displacement on the overlapping images, in conjunction with the flight parameters, to derive elevation information (X,Y,X, Omega, Phi and Kappa) (Petrie and Kennie, 1990). Currently using digital photogrammetric software, when the 6 orientation parameters are recorded during the flight (using an inertial measurement unit (IMU) in conjunction with GPS), direct georeferencing is possible and the initial DEM extraction can be achieved directly. Further post processing might be required though. DEMs are also generated from optical satellite images using similar elevation extraction methods (Jacobson, 2003). In 1986, SPOT was the first satellite to provide stereoscopic images that allowed the extraction of DEMs (Nikolakopoulos *et al*., 2006). Advances in space technology have now resulted in many more satellites - ASTER, IKONOS, Quickbird and IRS-1C/1D (etc.) - producing stereoscopic images facilitating the extraction of DEMs, both relative, using only orbital information or absolute, integrating also known ground control points. The last procedure is done through provision of actual ground locations in X,Y,Z.

Use of satellite images for DEM generation have a tremendous advantage over traditional methods in that DEMs over large and inaccessible areas can nowadays be easily produced (near) real-time and within a relatively short time and at remarkable cheaper costs. A disadvantage, however, is that optical spectral range requires a cloud free view and appropriate (time of the day) light conditions in order to generate a good quality and high resolution (in accordance with flight parameters) DEM. Radar Interferometry or Interferometric Synthetic Aperture Radar (InSAR) have recently become popular in extracting elevation data. This technique uses two or more Synthetic Aperture Radar (SAR) recordings to generate DEMs, using differences in the phase of the waves returning to the satellite or airborne platform (Rosen *et al*., 2000). Radars have two main advantages over optical techniques (Massonnet and Feigl, 1998): (1) as an active system, they self-transmit and receive electromagnetic waves. This means image acquisition is independent of natural illumination and therefore images can be taken at night. (2) Observations are not affected by cloud cover since the atmospheric absorption at typical radar wavelengths is very low. National Aeronautics and Space Administrations' (NASA) Shuttle Radar Topographic Mission (SRTM), which produced a near-global DEM, followed this methodology. This system applied a so called single overpass technique using a dual antenna setup. Two systems recorded different wavelength bands (the German Space Research Centre, DLR, operated an X band and NASA a C-band InSAR). For each wavelength, the two antennas' on board of the shuttle were displaced by a certain baseline distance (60 m). When using single overpass, dual antenna techniques, images are recorded at the same time. Using different overpasses from different orbits, recording the same area, mostly the dielectric properties at the Earth surface have changed (given the fact that environmental conditions have changed with respect to a next overpass!) the initial interferogram might show low correlations for locations where environmental conditions have changed in the mean time and therefore the actual elevation might not be properly extracted for these locations.

Another method currently popular in the US and Europe is the Lidar or Airborne Laser Scanning. Due to the fact that these types of recordings are not readily available over Western Africa they are not further discussed here.

Irrespective of the method used, generated DEMs are inevitably subjected to errors, mainly due to the methodology followed (see above) or the various post-processing steps the models have to undergo (e. g. interpolation). It is, therefore, imperative that errors are quantified so as to provide users with first hand information on the accuracy of the DEM.

SRTM and ASTER-derived DEMs are two post-processed elevation datasets that are frequently used for a wide range of applications due to their near-global coverage (Nikolakopoulos and Chrysoulakis, 2006). Dozens of researchers have, over the past few years, carried out a series of global and local assessments of these products. Rodriquez et al. (2006) performed a global validation of the SRTM product using globally distributed set of ground control points derived from Kinematic Global Positioning System (KGPS) transects and National Geospatial-Intelligence Agency's (NGA) level 2 Digital Terrain Elevation Data (DTED). They concluded that SRTM has an absolute height error that exceeds the mission's goal of 16m (90%) often by a factor of two. Hofton et al. (2006) also used data from NASA's Laser Vegetated Imaging Sensor (LVIS) to assess SRTM's accuracy at study sites of variable relief and landcover (i.e. vegetated and non-vegetated terrain). They found out that, under "bare earth" conditions, SRTM data are accurate measurements of LVIS data whereas in vegetated areas, SRTM fell above the ground but below the canopy top, indicating an increase in the vertical offset between SRTM and the LVIS ground data. Jarvis et al. (2004) examined relative and absolute differences between SRTM DEM and a cartographically derived (TOPO) DEM using 1:50,000 scale contours digitized for all Honduras. They validated the two datasets using fifty-nine (59) high accuracy GPS points derived from the National Geodetic Survey (NGS) GPS database from the Central America High Accuracy Reference Network project (NGS, 2003). They also computed slope, aspect, curvature and Pearson's Correlation Coefficient. Their analysis revealed that, SRTM DEM has an average error of 8m as opposed to 20m for the TOPO DEM, although some systematic errors were identified in the SRTM data, related to aspect. They, therefore, concluded that SRTM is more accurate than the 1:50,000 scale cartographically derived DEM for Honduras. Slater et al. (2009) conducted an evaluation of the new ASTER Global Digital Elevation Model (GDEM) using 20 sites in 16 countries. Datasets used for comparison (reference data) with the ASTER GDEM are (1) level 1 and 2 DTED data and (2) control points data (GCPs) photogrametrically derived from satellite imagery. Standard DEM-to-DEM comparisons and DEM-to-control-point comparisons as well as detailed visual analyses of the data were conducted. It was found out that, in almost every area, the mean ASTER GDEM elevations are lower than the reference DEM elevations (biased negatively). The mean ASTER elevations are also lower than the GCPs in every area. Most of the assessments conducted have reported on an absolute accuracy assessment though.

In this study, the C-band SRTM - and ASTER – derived DEMs are compared and validated against a reference DEM for two regions in Ghana. Next to an absolute accuracy assessment also a relative assessment has been conducted. For many environmental and hydrological processes, even if the absolute elevation is not correct, relative to its surrounding regions it might for example exhibit the appropriate flow direction which is the basis of many GIS routines to derive a hydrological network. The reference DEM used was generated using data from a 1:50 000 scale contour map with the TOPOGRID function in the ArcInfo™ software (Hutchinson, 1988, 1989).

#### **1.1 Objectives**

220 Studies on Environmental and Applied Geomorphology

parallax displacement on the overlapping images, in conjunction with the flight parameters, to derive elevation information (X,Y,X, Omega, Phi and Kappa) (Petrie and Kennie, 1990). Currently using digital photogrammetric software, when the 6 orientation parameters are recorded during the flight (using an inertial measurement unit (IMU) in conjunction with GPS), direct georeferencing is possible and the initial DEM extraction can be achieved directly. Further post processing might be required though. DEMs are also generated from optical satellite images using similar elevation extraction methods (Jacobson, 2003). In 1986, SPOT was the first satellite to provide stereoscopic images that allowed the extraction of DEMs (Nikolakopoulos *et al*., 2006). Advances in space technology have now resulted in many more satellites - ASTER, IKONOS, Quickbird and IRS-1C/1D (etc.) - producing stereoscopic images facilitating the extraction of DEMs, both relative, using only orbital information or absolute, integrating also known ground control points. The last procedure is

Use of satellite images for DEM generation have a tremendous advantage over traditional methods in that DEMs over large and inaccessible areas can nowadays be easily produced (near) real-time and within a relatively short time and at remarkable cheaper costs. A disadvantage, however, is that optical spectral range requires a cloud free view and appropriate (time of the day) light conditions in order to generate a good quality and high resolution (in accordance with flight parameters) DEM. Radar Interferometry or Interferometric Synthetic Aperture Radar (InSAR) have recently become popular in extracting elevation data. This technique uses two or more Synthetic Aperture Radar (SAR) recordings to generate DEMs, using differences in the phase of the waves returning to the satellite or airborne platform (Rosen *et al*., 2000). Radars have two main advantages over optical techniques (Massonnet and Feigl, 1998): (1) as an active system, they self-transmit and receive electromagnetic waves. This means image acquisition is independent of natural illumination and therefore images can be taken at night. (2) Observations are not affected by cloud cover since the atmospheric absorption at typical radar wavelengths is very low. National Aeronautics and Space Administrations' (NASA) Shuttle Radar Topographic Mission (SRTM), which produced a near-global DEM, followed this methodology. This system applied a so called single overpass technique using a dual antenna setup. Two systems recorded different wavelength bands (the German Space Research Centre, DLR, operated an X band and NASA a C-band InSAR). For each wavelength, the two antennas' on board of the shuttle were displaced by a certain baseline distance (60 m). When using single overpass, dual antenna techniques, images are recorded at the same time. Using different overpasses from different orbits, recording the same area, mostly the dielectric properties at the Earth surface have changed (given the fact that environmental conditions have changed with respect to a next overpass!) the initial interferogram might show low correlations for locations where environmental conditions have changed in the mean time and therefore the actual elevation might not be properly extracted for these locations.

Another method currently popular in the US and Europe is the Lidar or Airborne Laser Scanning. Due to the fact that these types of recordings are not readily available over

Irrespective of the method used, generated DEMs are inevitably subjected to errors, mainly due to the methodology followed (see above) or the various post-processing steps the models have to undergo (e. g. interpolation). It is, therefore, imperative that errors are quantified so as to provide users with first hand information on the accuracy of the DEM.

done through provision of actual ground locations in X,Y,Z.

Western Africa they are not further discussed here.

In short, the objectives of this work are to:

• Assess how the elevation data from SRTM and ASTER do compare in an absolute manner with respect to a contour based "traditionally derived" elevation model in an

Comparison of SRTM and ASTER Derived Digital Elevation Models

correction for individual scenes (Fujisada *et al*., 2005).

data in many operational and research settings.

Inventory Search Tool (WIST - https://wist.echo.nasa.gov/api/)

**2. Data** 

**2.1 Aster GDEM** 

2000; Hirano *et al*. 2003).

**2.2 SRTM DEM** 

over Two Regions in Ghana – Implications for Hydrological and Environmental Modeling 223

The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is an advanced multispectral imager that was launched onboard NASA's Terra spacecraft in December, 1999. ASTER has three spectral bands in the visible near-infrared (VNIR), six bands in the shortwave infrared (SWIR), and five bands in the thermal infrared (TIR) regions, with 15-, 30-, and 90-m ground resolution, respectively (Yamaguchi *et al.* 1998). The VNIR subsystem has one backward-viewing (27.70 off-nadir) instrument for stereoscopic observation in the along-track direction, making imagery acquired by this satellite suitable for DEM generation. Other important properties that make ASTER data suitable for DEM generation are the platform altitude (705 km) and its base-to-height ratio of 0.6 (Abrams

The Ministry of Economy, Trade and Industry of Japan (METI) and the United States NASA recently released a global DEM (ASTER GDEM) derived from ASTER images acquired since its launch (1999) to the end of August, 2008. It covers land surfaces between 830 N and 830 S, comprising of 22 600 10 –by- 10 tiles. The GDEM is provided at a one arcsec resolution (30 m) and referenced to World Geodetic System (WGS) 1984. Elevations are computed with respect to the WGS 84 EGM96 geoid. The vertical accuracy of the DEM data generated from the Level-1A data is 20 m with 95% confidence without ground control point (GCP)

METI and NASA acknowledges that Version 1 of the ASTER GDEM is "research grade" due to the presence of certain residual anomalies and artifacts in the data that may affect the accuracy of the product and hinder its effective utilization for certain applications. For this study, two 10 –by- 10 tiles in Ghana (see figure 1) were download from NASA's Warehouse

The SRTM (Werner, 2001; Rosen *et al*., 2001a), undertaken by NASA and the NGA, collected interferometric radar data which has been used by the Jet Propulsion Laboratory (JPL) to generate a near-global (80% of earth's land mass) DEM for latitudes smaller than 600. SRTM has been the first mission using space-borne interferometric SAR (InSAR). The SRTM mission has been a breakthrough in remote sensing of topography (van Zyl, 2001), producing the most complete, highest resolution DEM of the world (Farr *et al*., 2007). An extensive global assessment revealed that the data meets and exceeds the mission's 16m (90 percent) absolute height accuracy, often by a factor of two (Rodríguez *et al*., 2006). Since its release in 2005, the user community has embraced the availability of SRTM data, using the

SRTM data for this study was downloaded from the website of the Consultative Group on International Agricultural Research Consortium for Spatial Information (CGIAR-CSI http://srtm.csi.cgiar.org). Data available from this site has been upgraded to version 4, which was derived using new interpolation algorithms and better auxiliary DEMs. This

version, thus, represent a significant improvement from previous ones.

absolute comparison. DEMs over different topography and land cover have been selected.


#### **1.2 Study sites**

The study is conducted for two sites in Ghana (Figure 1). These sites fall in different agro climatic zones, have different elevation ranges and differ in landcover. Site 1, which climatologically falls in the Guinea Savannah zone (northern Ghana), has an elevation range of about 400 m. It is fairly flat, with an average slope of 0.90. The major landcover types are deciduous woodland (55 %) and shrubland (36 %). Site 2 lies between two climatic zones – moist semi-deciduous forest and transitional zones. It has an elevation range of about 780 m, with an average slope of 3.30. Though also fairly flat, this site has a range of mountains that borders the Volta Lake. The dominant landcover types are forest (52 %) and woodland/shrubland (34 %).

Fig. 1. Map of the study sites
