The 30-m Seamless DEM Reconstruction Based On Multi-source Data Fusion

With the rapid development of remote sensing and photogrammetry,many global digital elevation model(DEM)products derived from satellite observations have been publicly released.Nevertheless,due to the limitations of the observation technologies and the processing methods,the individual datasets are inevitably affected by quality degradation or space loss,which further restricted DEMs' application in hydrology,geology,meteorology,military and other fields.Although individual DEM dataset has quality degradation problems,DEMs from different sources also have different quality advantages and complementary features.Therefore,multi-source fusion has become one of the effective ways to improve DEM data quality.Based on point-surface fusion method with multi-source data and multi-scale fusion method,we reconstruct a new high-quality seamless DEM dataset,which is refered to as GSDEM-30.The main parts of this paper are as follows.(1)The situation of the global DEM products was analyzed.We analyzed the data source,release and quality of the existing individual DEMs and multi-source fused DEM datasets,and the relations among different DEMs,which layed a foundation for the selection of multi-source DEM datasets about GSDEM-30 reconstruction.(2)GSDEM-30 dataset spanning the land surface between 82°S and 82°N was reconstructed.Considering the characteristics of multi-source datasets at different latitudes,we reconstructed GSDEM-30 construction using DEMs' complementary features in spatial resolution and coverage.For hole data and 50° to 60° latitude data,the 30-m GSDEM-30 was mainly reconstructed from the recently released SRTM1,AW3D30 v1.1,and ASTER GDEM v2 datasets combining the terrain-adaptive point-surface fusion method with multi-source and multi-scale fusion method.(3)The quality of different DEM datasets was comprehensively evaluated.In this thesis,Global ICESat GLAS data and the local National Elevation Dataset(NED)were used as the reference from global and local scales for the vertical accuracy validation,while GlobeLand30 was introduced for the elevation error distribution of different landscape analysis.Furthermore,we employed the maximum slope approach(MSA)to detect the potential artefacts in the DEMs,without the involvement of reference data.The results showed that the ASTER GDEM dataset was seriously affected by noise and artefacts,and the overall accuracies of the GSDEM-30 and AW3D30 DEM were comparable.Moreover,the GSDEM-30 with fewer anomalies than AW3D30 had great potential in geoscience related applications.