Resource information
It is difficult to monitor forests in tropical regions with frequent cloud cover using optical remote-sensing data. Adequate multi-temporal, high-resolution imagery is often not available. Microwave imagery is able to penetrate cloud cover, enabling imagery of the land surface to be recorded more frequently. This study seeks to improve tropical forest mapping by combining optical and microwave imagery, with one of the main objectives being the discrimination of planted and natural forests. First, multi-spectral Advanced Land Observing Satellite (ALOS)/Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2) images were used to create a forest and land-cover classification of the study area. Subsequently, ALOS/Phased Array type L-band Synthetic Aperture Radar (PALSAR) single-polarized and dual-polarized microwave images were used to generate forest and land-cover masks to be used in combination with the ALOS/AVNIR-2 classification. The overall accuracy of the ALOS/AVNIR-2 classification was 77%. When the ALOS/PALSAR masks were used in combination with the ALOS/AVNIR-2 classification, the overall accuracy increased to 88% with higher than 90% accuracy for the main forest classes.
