B53E-0616
Mapping tropical forests and rubber plantations on Hainan Island using PALSAR and multi-temporal Landsat data
Friday, 18 December 2015
Poster Hall (Moscone South)
Bangqian Chen1, Xiangming Xiao2, Jinwei Dong2, Xiangping Li3, Weili Kou4 and Zhixiang Wu1, (1)Rubber Research Institute, Ecology, Danzhou, China, (2)University of Oklahoma Norman Campus, Norman, OK, United States, (3)Fudan University, Shanghai, China, (4)University of Oklahoma, Norman, OK, United States
Abstract:
Updated and accurate maps for tropical forests and plantations such as natural rubber (Hevea brasiliensis) are of vital importance for ecological study and optimal forest management practices. However, the existing optical-based efforts are limited by the coverage of clouds and shadows, which is prevent in the tropical regions. In this study, we combined PALSAR 25-m mosaic and 30-m multi-temporal Landsat (TM/ETM+) images to map tropical forests and rubber plantation in Hainan Island, China. Based on PALSAR imagery and maximum NDVI, we first identified tropical forests in 2010 using a decision tree algorithm. The resultant forest map has high accuracy (producer/user accuracy > 96%) and can serve as a reliable base map for rubber plantation mapping. Deciduous rubber plantation subsequently extracted from the PALSAR/Landsat-based forest map according to its high NDVI in peak growing season, and low LSWI during the rapid defoliation and foliation stage. The rubber map had good accuracy when validated by ground truth-based Region of Interests (ROIs) and three farm-scale land use maps. This study has clearly demonstrated the advantage of integrating PALSAR and Landsat images and phenology-based algorithm to map tropical forests and rubber plantation.
