Topographic mapping of forested areas in the western Great Lakes region using spaceborne synthetic aperture radar interferometry.

Jonathan Ward Chipman - 2001
[UMI Proquest Full Citation]

Synthetic aperture radar (SAR) interferometry represents a new method for topographic mapping. Experimental L-band (24 cm wavelength) interferometric SAR images from the second Shuttle Imaging Radar-C (SIR-C) mission in October 1994 were used to produce a digital elevation model (DEM) covering 2100 km2 in northwestern Wisconsin. In non-forested areas this DEM has an accuracy of 3.5 m RMSE. In forested areas, there is a bias in the interferometric elevation estimates, such that the interferometric phase centers appear to be above (or in some cases, below) the ground surface. These biases are dependent upon forest type and structural characteristics. Models were derived to predict the interferometric elevation biases in red pine plantations based on mean stem diameter, for thinned and non-thinned stands (R2 of 0.88 and 0.72 respectively). In addition, the relationship between interferometric elevation biases and image-based characteristics such as SAR image backscatter magnitude, image texture, and normalized difference vegetation index (NDVI) were examined. While L-band backscatter is loosely correlated with interferometric elevation biases, the other variables were not found to be good predictors. To ensure that end-users are aware of the presence of spatially variable biases in interferometric DEMs, it is recommended that such DEMs be accompanied by pixel-level maps showing expected error magnitudes, taking into account differences in accuracy between forested and non-forested areas.