Prediction of plot-level forest variables using TerraSAR-X stereo SAR data

Abstract

Promising results have been obtained in recent years in the use of high-resolution X-band stereo SAR satellite images (with the spatial resolution being in order of meters) in the extraction of elevation data. In the case of forested areas, the extracted elevation values appear to be somewhere between the ground surface and the top of the canopy, depending on the forest characteristics. If the ground surface elevations are known by using a Digital Terrain Model derived from Airborne Laser Scanning surveys, it is possible to obtain information related to forest resources. To the best of our knowledge, this paper, presents the first attempt to obtain forest variables at plot level based on TerraSAR-X stereo SAR images (non-interferometric data). The study set consisted of 109 circular test plots for which forest variables were observed by performing tree-specific measurements. The statistical features were calculated for each test plot from the elevation values extracted from stereo SAR data. This was followed by predicting field-observed plot-level forest variables from the features derived from stereo SAR data using the Nearest Neighbors approach which employs the Random Forest technique in selection of the nearest neighbors. The relative errors (RMSE%) for predicting the stem volume, basal area, mean forest canopy height, and mean diameter values were 34.0%, 29.0%, 14.0%, and 19.7%, respectively. The results indicate that there was no clear saturation level in stem volume estimation. In this case study, stem volumes were predicted up to about 400 m 3/ha. In the light of these results, stereo SAR data appears to be an interesting remote-sensing technique for future forest inventories. For example, stereo SAR data could have high potential in forest inventories as the SAR-based features can be adapted to the methods currently used in inventories.