Model-based compensation of topography effects for improved stem volume retrieval from CARABAS-II VHF-band SAR images

Artikel i vetenskaplig tidskrift, 2009

A limiting factor that has been identified for stem-volume retrieval in coniferous forests using VHF synthetic aperture radar is that the backscatter varies depending on ground topography. On sloping ground, the backscatter from a forest is reduced, since the dominant groundtrunk double-bounce scattering mechanism is changed. This leads to underestimation of stem volume, and the variations caused by topography can obscure real variations in stem volume. By using multiple images acquired with different flight headings and combining the image information with ground-topography data in a model-based inversion method, we are able to compensate for the ground-topography influence on the backscatter. The inversion method is based on image segmentation and the optimal estimation method. Using four or more images from the CARABAS-II system and a coarse digital elevation model with 50-m horizontal grid, the stem volume can be retrieved with an average root-mean-square error (rmse) of less than 60 $\hbox{m}^{3}\hbox{ha}^{-1}$ for stem volumes in range of 80700 $\hbox{m}^{3}\hbox{ha}^{-1}$ (in terms of above-ground biomass, this is equivalent to an rmse of less than 40 $\hbox{ton} \cdot\hbox{ha}^{-1}$ over the range of 50400 $\hbox{ton} \cdot\hbox{ha}^{-1}$). The retrieval accuracy is similar to that previously obtained for similar forests standing on flat and horizontal ground.