**2.3 Reference DEM**

The reference DEM was generated from the hypsographic (contours) and hydrographic (rivers/streams) data in the 1:50,000 topographical map series produced by the Survey Department of Ghana (SDG). The contours used have an interval and vertical accuracy of 50 feet (≈ 15m). This accuracy supersedes that of the 16m and 20m of SRTM and ASTER respectively, hence its use as a reference. Map sheets covering the study sites were merged prior to generating the DEM.

The SDG, in January 2009, adopted the WGS 1984 (UTM) coordinate system (Daily Graphic, 14th October 2008). Prior to this, the department's maps (including the ones used in this study) were produced based on a local datum and the "war office" ellipsoid. Thus, there was the need to transform the data from the old system (Ghana grid) into the new adopted system (WGS 84 UTM). Table 1 shows the transformation parameters, as published by the SDG, used for the conversion.

The TOPOGRID function in the ArcInfo™ software was used to interpolate the transformed contour data into a DEM. There are many interpolation methods used to generate a DEM from contour data (Hengl and Evans, 2009). Comparison of these techniques has been extensively carried out by many researchers (Wise, 2000). Li *et al*. (2005) points out that there is no universal interpolation technique that is clearly superior, and appropriate for all sampling techniques and DEM applications. Despite this conclusion, Hengl and Evans (2009) recommend that, where possible, algorithms that can incorporate secondary information (such as layers representing pits, streams, ridges, scarps or break lines) should be implemented for DEM interpolation. One such method, and the one used in this study, is the ANUDEM algorithm by Hutchinson (1988, 1989), and implemented as the TOPOGRID function in the ArcInfo™ software. ANUDEM is based on the discretized thin-plate spline technique (Wahba, 1990), and is an iterative DEM generation algorithm that produces hydrologically correct DEMs. The algorithm starts with a coarse grid, and then enforces drainage conditions, the spatial resolution is increased, and then drainage enforcement is performed again, and so on, until the desired resolution is reached (Hengl and Evans, 2009).


Source: Daily Graphic of Ghana, 14th October 2008

Table 1. Transformation parameters (War Office to WGS 84)

The grid resolution chosen for the reference DEM was based on equation (1). In cases where a DEM is based on contours, a suitable grid resolution can be estimated from the total length of the contours (Hengl and Evans, 2009). It is important to choose a grid resolution that optimally reflect the complexity of the terrain and represent the majority of the geomorphic features (Smith *et al*., 2006).

$$
\Delta S = \frac{A}{2\sum L} \tag{1}
$$

where "A" is the area of the study site (km2) and "L" is the cumulative length of all contour lines (km).

Table 2 shows the implementation of equation (1) for both study sites and the resultant grid resolution. Based on these results, a cell size of 90 m was chosen in generating the reference DEM. The choice of a 90 m spatial resolution reflects the complexity of terrains for both study sites and allows direct comparison with the 90 m SRTM data.


Table 2. Suitable grid resolution for the Reference DEM
