Measuring biomass changes due to woody encroachment and deforestation/degradation in a forest-savanna boundary region of central Africa using multi-temporal L-band radar backscatter

Satellite L-band synthetic aperture radar backscatter data from 1996 and 2007 (from JERS-1 and ALOS PALSAR respectively), were used with field data collected in 2007 and a back-calibration method to produce biomass maps of a 15 000 km(2) forest-savanna ecotone region of central Cameroon. The relationship between the radar backscatter and aboveground biomass (AGB) was strong (r(2)=0.86 for ALOS HV to biomass plots, r(2)=0.95 relating ALOS-derived biomass for 40 suspected unchanged regions to JERS-1 HH). The root mean square error (RMSE) associated with AGB estimation varied from similar to 25% for AGB<100 Mg ha(-1) to similar to 40% for AGB>100 Mg ha(-1) for the ALOS HV data. Change detection showed a significant loss of AGB over high biomass forests, due to suspected deforestation and degradation, and significant biomass gains along the forest-savanna boundary, particularly in areas of low population density. Analysis of the errors involved showed that radar data can detect changes in broad AGB class in forest-savanna transition areas with an accuracy >95%. However, quantitative assessment of changes in AGB in Mg ha(-1) at a pixel level will require radar images from sensors with similar characteristics collecting data from the same season over multiple years. (C) 2011 Elsevier Inc. All rights reserved.