Increasing the Accuracy of Low Spatial Resolution Digital Elevation Models using Geostatistical Conflation

Digital Elevation Models (DEMs) are an important data source for a range of scientific and commercial applications, which accuracy depends on the accuracy of the input DEMs. Examples of such applications are hydrological studies, topographic mapping and landscape modelling, among others. Different technologies (e.g. Lidar, radar, photogrammetry) exist for producing accurate high resolution DEMs. However, DEMs produced using these technologies are generally limited to small areas and are expensive. In contrast, low spatial resolution DEMs cover most of the planet and are freely available. Consequently, these DEMs are commonly used in applications with limited resources (Hirt et al. 2010). However, the accuracy of low resolution DEMs that are freely available (e.g. Aster GDEM, SRTM) is undocumented for specific study areas and only an estimate of the global or regional accuracy is provided with them, adding uncertainty to their use. In this paper we recommend the use of geostatistical conflation to reduce the uncertainty associated with the use of low resolution DEMs by increasing their vertical accuracy by means of conflating them with a set of sparsely distributed Ground Control Points (GCPs) using Kriging with an External Drift (KED).