Abstract:
A digital elevation model (DEM) is a 3-D representation of the earth’s topography.
Vertical accuracy of a global digital elevation model (GDEM) is necessary for optimal
application of satellite-based terrain elevation datasets. This study carries out an
assessment of the vertical accuracy of Classical, ASTER (30 m) and SRTM (90 m)
digital elevation models (DEMs) which are normally used for reconnaissance
surveys, hydrological analysis, biomass estimation and geoid modelling among
others. Classical DEMs are drawn from regional topographical maps while ASTER and
SRTM DEMs are obtained from satellite-based remote sensing missions. The
assessment is carried out by comparing orthometric heights from precise levelling
at 18 points and heights derived from the DEMs over Nairobi County and its
environs. The study found that the mean and standard deviation of the direct
differences between precisely levelled heights and DEM heights are: 3.97 m and
±7.76 m respectively for classical DEM; 16.36 m and ±7.79 m respectively for ASTER
DEM and -0.25 m and ±4.00 m respectively for SRTM DEM. The results indicate that
SRTM DEM is the most accurate followed by the classical and ASTER DEMs in that
order. We then modelled the differences between the DEM heights and the
orthometric heights using a second order surface polynomial at 12 points; the
polynomial was then applied to 6 test points in a cross -validation manner. The
results from the polynomial improved accuracy of height determination in SRTM
DEM but degraded accuracies in the classical and ASTER DEMs.
