Mapping of boreal forest biomass using SAR

Knowledge of the distribution and structure of forests can lead to improved estimates of the carbon balance in terms of above ground storage and exchange with the atmosphere from maintenance respiration. As part of the Boreal Ecosystem Atmosphere Study (BOREAS) a study is being made of the use of remote sensing data for estimating land cover and biophysical characteristics of boreal forest study sites in Canada. Earlier work by the authors and others have developed maps of forest cover and above ground biomass. The biomass maps are produced using a regression equation developed from forest measurements and radar backscatter data. Other researchers have made a case for stratifying forest cover and constructing separate biomass relationships for each strata. In this paper the authors examine the use of forest classification maps derived from SIR-C/XSAR images to improve estimates of total biomass. The classification procedure used a supervised Bayesian classifier to map the forest area into several forest and non-forest classes. Forest classes were combined into dry conifer, wet conifer and deciduous. Regression equations of dry biomass and radar backscatter were developed for each forest type and applied to the radar data. A stepwise technique was used to determine the best set or combinations of radar channels for mapping biomass. The best channels for biomass estimation was determined to be LHV and CHV for conifers and a combination of LHH, LHV and CHV worked best for the limited deciduous data available. Analysis of the results indicates that forest type should be considered when mapping biomass in this type of forest. However, the results from a more general equation were adequate