Mapping forest types with TerraSAR-X imagery

Forest mapping and monitoring surveys are often based on costly and time-consuming field work. Satellite remote sensing data can facilitate these procedures over large forest areas [1]. Therefore, the Forest Research Institute of Baden Württemberg and the Department of Remote Sensing and Landscape Information Systems of the University of Freiburg aim on developing methods to support and to improve traditional monitoring of forest areas. The objective is the implementation of long-term, low-cost, and high-quality monitoring programs. Synthetic aperture radar (SAR) imagery provided by high resolution satellites as TerraSAR-X and RADARSAT-2 are now available for large scale forest mapping and forest monitoring applications. In order to examine TerraSAR-X images to map deciduous and coniferous forest, High Resolution Spot Light (HS) images were acquired for two study sites in southern Germany. A correct interpretation of SAR images requires an adequate pre-processing of the data. The quality of the Digital Elevation Model (DEM) used as reference is fundamental for this purpose. Therefore, the aim of this study was to analyze the influence of the quality of the DEM on mapping accuracy of forest types using SAR data. The use of high resolution (5 m) DEMs derived from Airborne Laser Scanning (ALS) for the orthorectification of SAR images was found to be important for the reduction of errors in pixel location and to increase the classification accuracy of the forest types. No significant difference was found between using an ALS Digital Surface Model (DSM) or a Digital Terrain Model (DTM) for the orthorectification. Analysis of the radar backscatter indicated that plots dominated by coniferous trees tended to have lower radar backscatter (σ) values than plots dominated by deciduous trees. Results of forest type classifications based on images in leaf-off and leaf-on conditions were similar. Nonetheless, the combination of leaf-off and leaf-on images preprocessed with an ALS DTM resulted in the highest overall accuracy with kappa coefficients of 0.41 and 0.49, respectively. The classification of forest types using SAR images pre-processed with the SRTM DEM resulted in poor accuracies and kappa coefficients of less than 0.10.