Carabajal, CC, Harding, DJ (2006). SRTM C-band and ICESat laser altimetry elevation comparisons as a function of tree cover and relief. PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING, 72(3), 287-298.

The Geoscience Laser Altimeter System (GLAS) instrument onboard the Ice, Cloud, and land Elevation Satellite (ICESat) provides a globally distributed elevation data set that is well-suited to independently evaluate the accuracy of digital elevation models (DEMs), such as those produced by the Shuttle Radar Topography Mission (SRTM). We document elevation differences between SRTM C-band 1 and 3 arcsecond resolution DEMs and ICESat 1064 nm altimeter channel elevation data acquired in an areas of variable topography and vegetation cover in the South American Amazon Basin, Asian Tibetan Plateau - Himalayan Mountains, East Africa, western Australia, and the western United States. GLAS received waveforms enable the estimation of SRTM radar phase center elevation biases and variability with respect to the highest (canopy top where vegetated), centroid (distance-weighted average), and lowest (ground) elevations detected within ICESat laser footprints. Distributions of ICESat minus SRTM elevation differences are quantified as a function of waveform extent (a measure of within-footprint relief), SRTM roughness (standard deviation of a 3 X 3 array of elevation posts), and percent tree cover as reported in the Vegetation Continuous Field product derived from Moderate Resolution Imaging Spectrometer (MODIS) data. SRTM roughness is linearly correlated with waveform extent for areas where percent tree cover is low. The SRTM phase center elevation is usually located between the ICESat highest and lowest elevations, and on average is closely correlated with the ICESat centroid. In areas of low relief and sparse tree cover, the mean of ICESat centroid minus SRTM phase center elevation differences for the five regions examined vary between -3.9 and 1.0 m, and the corresponding standard deviations are between 3.0 and 3.7-m. With increasing SRTM roughness and/or tree cover, the SRTM elevation remains essentially unbiased with respect to the ICESat centroid but the standard deviations of the differences increase to between 20 and 34 m, depending on the region. For the Australia, Amazon, Africa, United States, and Asia regions, including all tree cover and roughness conditions, 90 percent of the SRTM elevations are within 6.9, 11.5, 12.1, 16.8, and 37.1 m of the ICESat centroid, respectively. In vegetated areas, the SRTM elevation on average is located approximately 40 percent of the distance from the canopy top to the ground. The variability of this result increases significantly with increasing SRTM roughness. The results are generally consistent for the five regions examined, providing a method to estimate for any location the correspondence between SRTM elevations and highest, average, and lowest elevations using the globally-available MODIS-derived estimate of tree cover and the measure of SRTM roughness.