FOREST CARBON STOCKS ESTIMATION USING MULTIRESOLUTION SATELLITE IMAGERY THROUGH SUPPORT VECTOR MACHINE (SVM) APPROACH (A Literature Review)

Abstract

Study to be conducted with the aimed to developing new methods for estimating tropical forest carbon stock using multi-resolution and multi-temporal satellite imagery with support vector machine (SVM) approach. To cover large area, a coarse resolution imagery of MODIS with seven band will be use as an input and high resolution imagery-QUICKBIRD for training data. The main approach in the study will comprises four steps deriving forest cover vegetation (forest type and forest density) from QUICKBIRD imagery using clustering algorithm, constructing predictor variables derived from MODIS satellite imagery such as surface reflectance, normalized difference vegetation index (NDVI), normalized difference soil index (NDSI) and one groundmeasured variable shortwave radiation, Train the SVM using predictor, apply extended domain of SVM for regression approximation and evaluate model performance and accuracy assessment based on coefficient of determination R2 and a root mean square error (RMSE). Kata kunci: forest carbon,stock estimation, satellitei imagery, Support vector machine