Mapping Aboveground Biomass in Northern Japanese Forests Using the ALOS PRISM Digital Surface Model

Abstract

We tested the performance of the stereo observations of the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) onboard the Advanced Land Observing Satellite (ALOS) in the mapping of forest aboveground biomass (AGB) in Japan. Digital canopy height models (DCHMs), which are differences between PRISM digital surface models and surveying-based digital terrain models (DTMs), were compared to in situ AGB measurements of several forest types (number of stands: 28; average stand size: 0.54 ha; stand size range: 0.25-3.00 ha). DCHM values exhibited a significant correlation with AGB (r = 0.66-0.87; five different DCHMs), and the root-mean-square error and bias of the regression model evaluated by the leave-one-out cross-validation were 37.2-57.8t/ha(22.1%-32.6%) and-0.11-1.89 t/ha, respectively. There was no saturation in the relationship between DCHM and AGB (AGB range: 19-332 t/ha). The correlations between DCHM and mean canopy height (r = 0.17-0.52) and between DCHM and Lorey's height (r = 0.26-0.66) were weaker than the correlation between DCHM and AGB. The PRISM AGB distribution estimated by the regression model was consistent with a tree density map produced from aerial photos. Comparison to Phased Array-type L-band Synthetic Aperture Radar (PALSAR) data showed that the PRISM DCHMs can estimate high AGB over the saturation level of PALSAR backscattering coefficient, i.e., 100-200 t/ha. The results described here demonstrate that the PRISM DCHMs are capable of wall-to-wall AGB estimation at 50-m resolution. This approach will be useful for improving the performance of satellite-based AGB estimation when an accurate DTM is available.