Aboveground Biomass Mapping Using ALOS-2/PALSAR-2 Time-Series Images for Borneo's Forest

Abstract

Global carbon cycle research and climate change mitigation measures require a means for large-scale monitoring of forest aboveground biomass (AGB). L-band synthetic aperture radar (SAR) is one of promising means, although its signal is saturated at 75–150 Mg ha−1 of AGB, making it difficult to monitor high biomass forest. The Advanced Land Observing Satellite-2/Phased Array L-band SAR-2 (PALSAR-2) conducts ScanSAR mode observations that cover the whole tropical region about nine times a year. We investigated whether such time-series data is effective in overcoming the signal saturation issue. In Borneo, we adopted a spaceborne Light Detection and Ranging (LiDAR) Ice, Cloud, and Land Elevation/Geoscience Laser Altimeter System (GLAS)-derived AGB data for training and validation data, then developed an AGB estimation model using the Random Forest algorithm. As a result, we improved the saturation issue, and estimated the AGB up to 280 Mg ha−1 with a root mean square error (RMSE) of 62.8 Mg ha−1. Such an AGB range covers 83% of Borneo's forests. The developed model was applied to create an AGB map of Borneo with a 250-m resolution as of 2016. Total AGB of Borneo was calculated at 12.8 Gt with an average of 173.3 Mg ha−1. This article showed the PALSAR-2 time-series data to be highly useful in the AGB mapping of high carbon stock forests. However, we needed to correct the difference in observation years between GLAS and PALSAR-2 using a simple biomass growth model, but the accuracy will be improved by using future LiDAR and SAR sensors.