Integrative Use of SAR and Optical Data for Forest Mapping in the Congo Basin

This study presents an innovative processing chain and various test results for using both SAR (PALSAR) and optical (such as AVNIR, LANDSAT) data in an integrative way for forest mapping in the Congo Basin. Three activities are described in this context: (i) geometric adjustment of SAR and optical data by automatic image matching; (ii) analysis of various pre-processing steps for SAR data and (iii) a method for efficiently classify SAR data based on an existing optical classification. In part one, the robust and fully automatic matching procedure based on the mutual information maximization principle has proven to be useful to ensure geometrical congruence between optical and SAR data sets. Results show that the RMSE is reduced from over 80 m to less than 10 m without manual interaction. The second analysis covers the wide variety of SAR pre-processing methods and options. From over hundred different options, the best processing steps are selected by using cross-correlation analysis and in addition considering the typical MMU needed for forest monitoring in the tropics. The third development concerns the so-called classification-based trainer. This method allows filling classification gaps caused by clouds or sensor failures in optical data by using SAR data without much manual effort. A first benchmarking test involving AVNIR and PALSAR shows a slight overestimation of 0.8% of non-forest area for the resulting classification compared to the classification on optical data only. Considering the difference in data quality and properties between optical and SAR data, this is a very promising result.