An estimation of tropical forest biomass with a combination of JERS-1 and Landsat TM data

The use of remote sensing to estimate biomass in tropical forests has met with varying degrees of success. Previous work has demonstrated that reliable estimates of biomass can be made with Artificial Neural Networks, using data from single sites and with a degree of transferability between other tropical forests. Here JERS-1 data for three sites is combined with Landsat TM images, with the aim of improving the spatial transferability of the method. The image data were compared to ground data from sites in Malaysia, Brazil and Thailand using a feed-forward artificial neural network. The SAR data were weakly correlated with biomass, and when combined with optical data led to a marginal increase in the correlation coefficient for each site individually. However, improved relationships were observed when combined SAR/optical data for all sites were presented to the network, thus indicating marginally improved transferability in the method compared to using optical data alone