The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B7. Beijing 2008
1352
study site. The degree of nearness of the two surfaces to reality
can also be seen from the high F-statistic values of the two data
sets (0.767496217 for the SRTM elevation data and 0.990632
for the topo DEM data). These values indicate that the topo
DEM is a more accurate surface than the SRTM DEM, a
finding that informed our using the topo DEM as a reference
surface for the analyses of the SRTM DEM throughout the rest
of the study.
Figure 6. Contours interpolated from SRTM DEM
(a) 5-m vertical interval contours directly interpolated
from the 90-m SRTM DEM showing artefacts in circles
(b) 5-m vertical interval contours interpolated from the 90-
m DEM derived from the 1:50,000 topographic map by
moving surface interpolation
(c) 5-m vertical interval contours interpolated from the
SRTM DEM re-sampled to 45m by cubic interpolation
(d) 5-m vertical interval contours interpolated from the 45-
m DEM derived from the 1:50,000 topographic map by
moving surface interpolation
(e) Super-imposition of the maps from (a) - (d) above
(f) Super-imposition of the maps from (a) - (d) above and
the hydrographic network
4. RESULTS AND DISCUSSION
The computed TIN-based SRTM and topo DEMs overlaid with
the hydrographic network covering the study site (Figure 3(a),
Figure 3(b), Figure 4(a) and Figure 4(b)) revealed striking
characteristics of the two surface representations. The overlay
showed that the hydrographic network matched the two
surfaces well, with the rivers passing through the water courses
in the two DEMs. This is a particularly interesting result noting
that the two surfaces were derived from two disparate sources
(1:50,000 topographic map and SRTM data) with the
hydrographic network layer coming from the topographic map
data source. The terrain profile presented in Figure 5 further
revealed that the two surfaces are generally close, implying that
SRTM elevation data can be used as a substitute for the
topographic map elevation data for 1:25,000 topographical
mapping.
The results of the various tests conducted to assess the
cartographic accuracy of the CGIAR-CSI SRTM elevation data
are as presented in Figures 6(a) - (f). The tests essentially
involved direct interpolation of 5-m vertical interval contours
from the SRTM DEM, recreation of the SRTM surface by re
sampling and point interpolation and super-imposition of the
various maps for visual interpretation. Based on a visual
interpretation of the results, the following facts can be deduced
about the CGIAR-CSI SRTM:
(1) Direct interpolation of contours from the 90-m SRTM DEM
without further processing produces artefacts in the form of
incomplete contour lines, self-intersecting contour lines and
contour lines intersecting other contour lines having
different contour values (see Figure 6 (a) showing artefacts
in encircled areas of the map). This result is a strong
indication that contour maps directly derived from the
SRTM DEM are generally of low cartographic quality and
are not recommended for deployment in 1:25,000
topographic mapping without prior processing.
The various tests conducted in this study to assess the accuracy
of contour interpolation from the 1:50,000 topographic map and
the SRTM elevation data to investigate their suitability for
1/25,000 topographic mapping revealed striking characteristics
about the data sets. The statistical computation for the absolute
vertical accuracy of CGIAR-CSI SRTM elevation data for our
study site gave a value of ± 7.748m. This statistic indicated
that, for our study area, the 90-m CGIAR-CSI elevation data
featured a much greater absolute vertical accuracy than the
absolute vertical accuracy value of ± 16m published in the
SRTM data specification. The statistical tests also revealed that
the absolute vertical accuracy ( ± 3.926m) of the digital
elevation data derived from the 1:50,000 topographic map
covering our study site was much higher than that of the
CGIAR-CSI SRTM DEM. The graphic plots of the GPS
elevations against the SRTM and topo DEM (Figure 2(a) and
Figure 2(b)) revealed that the two data sets showed strong
positive correlations with the GPS data. This suggests that the
two surfaces are significantly close to the real surface for our
(2) Processing the SRTM DEM before interpolating contours
from it produces results with significantly better
cartographic quality than direct interpolation. This
observation can be clearly seen in Figures 6(b) - (d) which
respectively show the results of moving surface
interpolation (90m) of the derived point raster set, the 45-m
re-sampling of the 90-m resolution SRTM DEM and the
moving surface interpolation (45 m) of the derived point
raster set. In all these output contour maps, it can be seen
that the artefacts present in the directly interpolated contour
map are clearly absent, giving a strong indication that
further processing of the 90-m SRTM elevation data is
required to ensure good cartographic quality of the derived
contour maps for 1:25,000 topographic mapping.
(3) The forms of the contour lines emanating from processed
and unprocessed SRTM DEM are essentially the same. This
can be seen from Figure 6(e) depicting a super-imposition
of the contour maps from the three cases cited above. This
result is a good indication that further processing of the