Comparison and fusion of DEM derived from SPOT-5 HRS and SRTM data and estimation of forest heights

There are several ways to produce digital elevation models (DEM) from space- borne satellite data. In this study, two techniques and their results will be compared: DEM derivation with optical stereo data acquired with the French SPOT-5 HRS instrument and DEM derived from C-band and X-band radar data acquired during the SRTM mission. A comparison of the accuracy is given for different landscapes and land use classes using reference DEM from laser scanner data and aerial photography. Both systems show advantages and disadvantages but have similar accuracy values in comparison to a reference DEM, in both cases the resulting DEM is actually a mixture of a digital surface model (DSM) and a digital terrain model (DTM) since the reflection/back scatter results from a mixture of different targets in each resolution cell. Therefore in a forest area neither the forest canopy nor the terrain height itself is modeled. Additionally a DEM fusion, utilizing height error maps for each DEM, is performed. It is shown that a fusion of the DEM derived from optical and radar data leads to higher accuracies in nearly all cases. A comparison of the space borne DEM with the reference DEM shows a higher absolute difference in forest areas than in agricultural regions. The absolute difference gives an estimation of the canopy height of the trees, although the scattering process (in the radar case) and the reflection and matching techniques (in the optical case) cause lower values than the real height of the trees. It is shown whether an estimation of the real mean tree height is possible with the additional knowledge of tree parameters like species and density. Results are compared for both DEM generation techniques to the data received from in-situ measurements.