GEOMOPHOMETRY OF DEMS: QUALITY ASSESSMENT AND SCALE EFFECTS

Digital elevation models (DEMs) provide a foundation for a variety of geological and geophysical operations. While the United States has free complete, high quality coverage at 10-30 m scale, and increasing coverage at 1-5 m scale, the rest of the world often has no data available, and available data can be expensive. Two satellite missions promise to provide high resolution data worldwide: SRTM at 3” (about 90 m) and ASTER at 30 m. We have calculated a suite of over 35 geomorphic parameters for the United States for seven different DEM series (about 70GB compressed). We have complete coverage of the old 1:250K USGS DEM, the new 1” NED, and 1” SRTM. We have partial coverage of 10 m USGS SDTS DEMs, 30 m level 1 and level 2 USGS SDTS DEMs, and ASTER DEMs. We have created a database of over 450,000 2½’x2½’ tiles covering the United States for each of the seven DEM series. The database includes land cover, climate, and regional geology characteristics that we can relate to geomorphology.

Correlation matrices from a single DEM series confirm that a large proportion of the metrics that have been proposed really measure a single parameter, surface slope or roughness. Variogram, Fourier, and fractal methods have been resistant to automation, because they require special preprocessing, or subjective determination of linearity or trends in noisy periodic data. The most robust, independent, and interpretable parameters appear to be surface slope or roughness, terrain organization (the tendency of ridges and valleys to share a common orientation), and the elevation-relief ratio or hypsometric integral. Curvature moments contain a noisier signal due to DEM error and quantization effects.

Correlations between series enable us to assess the effects of DEM scale and collection method on geomorophometric parametrs. For example, compared to NED, SRTM computes steeper average slopes for flat areas because of the radar speckle, and gentler slopes for steep areas because of the overall smoothing of the radar DEM. While within the United States the choice is obvious (use NED and ignore the many problems with SRTM and ASTER DEMs), study with the ground truth provided by NED will help improve analyses in the rest of the world where SRTM and ASTER are likely to be the best available DEMs for the foreseeable future, and can provide useful geomorphic information.