Mapping of Teak (Tectona grandis) Growing Stock Volume based on Sentinel-2 Imagery Classification

Abstract

Remote sensing utilization for plantation forests management in Java was not yet widely used, whereas the ability of remote sensing data for land cover monitoring and forest resource mapping has been developed, ranging from low resolution imagery for global areas to moderate and high resolution for small scale areas. Data availability and human resources often become obstacles in the application. The emergence of Sentinel satellite images becomes an alternative because the dataset is free access and has moderate spatial resolution and high temporal resolution. This study aims to map the growing stock volume in UGM Educational and Training Forest using Sentinel-2 imagery. Three kinds of classification method based on machine learning algorithms i.e. Random Forest, K-NN and SVM were compared for land cover classification. An NDVI algorithm was also used for mapping the spectral value distribution. Moreover, a stand age distribution which obtained from KHDTK UGM manager were also map. A stand classification map based on land cover types, NDVI value and stand age distribution was created as a basis of growing stock volume estimation. The analysis show that the growing stock volume can be estimated using these method with RMSE 177,8 m3 and MAPE 21,9 m3.