ASSESSMENT AND REDUCTION OF THE VERTICAL ERROR IN SRTM DEMS USING ICESAT

The successful Shuttle Radar Topography Mission (SRTM) mission provided scientists with land-surface data necessary to study processes on a global scale with highly consistent accuracy. Experience with the SRTM DEM indicates that it does not exactly match the true surface elevation of the ground derived from other remote sensing datasets. SRTM DEMs include various types of errors in the vertical direction, due to motion of 60 m antenna baseline and backscatter characteristics of radar phase from ground features. This study compares elevation values of the C-band SRTM 30-meter DEM with point wise elevations from ICESat laser altimetry to classify errors in the SRTM DEM. The accuracy measure is separately calculated as a function of the land use cover derived from the Landsat-7 image and physiographic characteristics of the land surface. By separating the study area into different physiography and land use covers, it is possible to identify SRTM DEM uncertainties over different surface types. This study addresses Otter Tail County, MN, and examines elevation differences between the SRTM DEM and ICESat elevations. The classified land use cover is categorized into waters, wetlands, forests, urban areas, and bare grounds, while the physiographic characteristics include stagnation moraines, hilly till plains, outwash and pitted outwash plains, the flat Lake Agassiz Plain, and Wadena Drumlin Field. The errors in the C-band SRTM DEM are most commonly associated with the land use types as well as topography. Also, the typical mean vertical difference between SRTM DEM and ICESat was found in each classified land use type to be approximately 1.5 m, with SRTM DEM measuring lower elevations. Forest typically shows much greater RMS errors than the bare rock/soil area of more than 5 m. Over different surface type, the SRTM 30-m DEM can be simply adjusted (bias removal) and improved using highly precise ICESat elevations as ground control points.