The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
1350 
ActiveX GIS control) to perform this operation. The result was 
a dBase table which was imported into Microsoft Excel 
environment to perform the statistical analysis. The results of 
the computation are summarized in Table 1 below. The table 
shows that the absolute vertical accuracy of the CGIAR-CSI 
SRTM elevation data for our study site is ± 7.748m and a 
mean difference in elevation between the two data sets is 
3.539m. Figure 2(a) depicts a graphic plot of the GPS 
elevations against the SRTM elevations. The coefficient of 
correlation and F-statistic between the two data sets were 
respectively +0.993094578 and 0.767496217. These results 
indicate a strong positive correlation between the two data sets. 
Since this study also involved determining the vertical accuracy 
of the DEM derived from the 1:50,000 topographic map, this 
statistic was computed by applying the same method described 
for the SRTM data. The results of the statistical computation 
are presented in Table 2. The results show that the absolute 
vertical accuracy of the topo DEM for our test site was 
± 3.926 and a mean value of-0.151m. A graphic plot of the 
GPS elevations against the topo elevations is as depicted in 
Figure 2(b). The coefficient of correlation for the data set was 
+0.998253 with an F-statistic of 0.990632. 
Statistical 
Measure 
SRTM 
Elevation (m) 
GPS 
Elevation(m) 
Ah 
(SRTM-GPS)fm) 
Minimum 
190.000 
184.324 
-19.996 
Maximum 
489.000 
507.168 
17.162 
Mean 
327.308 
323.768 
3.539 
Std Dev. 
± 63.425 
± 65.101 
± 7.748 
Count 
130 
130 
130 
Table 1: Statistical analysis of SRTM and GPS elevation data 
Statistical 
Measure 
Topo 
Elevation (m) 
GPS 
Elevation(m) 
Ah 
(Topo-GPS)(m) 
Minimum 
182.88 
184.324 
-18.985 
Maximum 
502.92 
507.168 
18.542 
Mean 
326.568 
323.768 
-0.151 
Std Dev 
± 66.38 
± 65.101 
± 3.926 
Count 
139 
139 
139 
Table 2 Statistical analysis of Topo and GPS elevation data 
Graph of GPS Elevations against SRTM Elevations 
Graph of GPS Elevations against Topo Elevations 
Figure 2. Graph of GPS elevations against 
(a) SRTM elevations (b) Topo elevations 
3.3.4 Comparison of the SRTM DEM and the 
Topographical DEM surfaces: For contours interpolated from 
the SRTM DEM to be deployed in 1/25,000 topographic 
mapping, the accuracy with which the DEM represents the 
surface morphology must be tested and proved to be of a 
sufficiently high degree. In this study, we employed two 
methods to test the accuracy of the SRTM data for good terrain 
representation. The methods involved the following 
(1) interpolating a DEM from the topographical map 
with the same spatial resolution as the SRTM DEM 
and visually comparing the two surfaces; 
(2) generating profiles along defined transects on the 
two surfaces and visually comparing the plots. 
The above operations required that the two DEMs must have 
the same spatial resolution, georeference and spatial extent. To 
satisfy this requirement, the output DEM from the 
topographical map with a resolution of 90m was trimmed to the 
size of the SRTM DEM. Both grid-based DEMs were converted 
into TIN-based DEMs in the ArcGIS 9.2 environment and 
overlaid with the hydrographic layer. The results of this 
operation are as shown in Figure 3 (a) and Figure 3(b). Figure 
4(a) and Figure 4(b) show the perspective views of the two 
DEMs. 
(a) (b) 
Figure 3 TIN-based DEMs with the hydrographic network 
superimposed (a) Derived from 1/50,000 Topo Map 
(b) Derived from SRTM Grid 
(a) (b) 
Figure 4. Perspective views of derived DEMs with 
hydrographic network superimposed 
(a) Topo DEM (b) SRTM DEM 
Further analysis of the SRTM surface involved a comparison of 
height profiles on the two surfaces. Two transects running from 
SW to NE and NW to SE were created on the two DEMs. To 
extract the elevations for the profile plot, we implemented a 
small Visual Basic 6.0 program to project the transects onto the 
DEM surfaces and to extract elevations from the two surfaces at 
regular intervals. The result of the computer run of the program 
was a dBase table containing the elevations along the transects 
on the two surfaces and the cumulative distances from the 
starting points of the profiles. The table was imported into