Estimation of forest biomass using L-band backscatter microwave satellite data.

Abstract

Synthetic Aperture Radar (SAR) satellite data can be used to monitor spatial and temporal changes in forest biomass and timber volume. Previous research suggest the SAR L-band backscatter signals saturate at a relatively low stand biomass threshold, making the application limited to thicket stage crops. In this study, new biomass and L-band backscatter regression models were developed using procedures to reduce interference due to radar incidence angle and surface moisture and by applying cross-image calibration using both forest and non-forest plot data to increase the biomass saturation point for stand biomass and volume. Many of the widely published model formulations were found not to provide a suitable model fit because of non-normal distribution and evidence of heteroskedasticity of model residuals. The model re-developed in this study performed better than published models, based on lower Akaike Information Criteria values and no heteroskedasticity of model residuals. The backscatter saturation for the re-developed model occurred at biomass values of c. 100 Mg ha-1, so accurate determination of biomass using this approach may be limited to immature forest stands. However, the L-band backscatter-biomass model may be suitable to detect changes in forest biomass or volume due to disturbance events.