Regression and analytical models for estimating mangrove wetland biomass in South China using Radarsat images

Abstract

Mangrove wetlands have been rapidly diminishing because of human pressures worldwide. The Guangdong Province in South China, which has the largest area of mangrove wetlands in the nation, is under severe threat as a result of rapid urbanization and economic development. In this paper, comparisons were made between optical Landsat TM images and Radarsat fine-mode images for estimating wetland biomass. Regression and analytical models were used to establish the relationships between remote sensing data and wetland biomass. The optimal parameter values for the analytical model were determined using genetic algorithms. Experiments indicate that the models using Radarsat fine-mode images have significant accuracy improvement in terms of Root Mean-Square Error (RMSE) whereas the use of the single Normalized Difference Vegetation Index (NDVI) may produce serious errors in biomass estimation. The Radarsat images can obtain more accurate trunk information about mangrove forests because of higher resolution and side-looking geometry. The use of genetic algorithms can help to decompose backscatter into vegetation and soil backscattering, which is very useful for ecological modelling.