Integration of Sentinel-1A and Sentinel-2A on Identifying Mangrove Using Random Forest Algorithm in Youtefa Bay, Papua Indonesia

Mangroves have an important role in the carbon storage cycle within coastal ecosystems. Indonesia, home to the largest mangrove forest in the world, has a potential in carbon trading particularly in blue carbon. It is essential to accurately identify the presence of mangrove to realize that potential. Remote Sensing (RS) and Machine Learning (ML) technology can be used to identify the presence of mangroves. In this research, the integration of active, passive sensors, National Digital Elevation Model (DEMNAS) and the used of the Random Forest (RF) algorithm were applied to identify the presence of mangroves. There were 30 independent variables, consisting of 4 independent variables from Sentinel-1A, 25 independent variables from Sentinel-2A and 1 independent variable from DEMNAS. The model was built from 75 sampling plots, 32 cross validation internal plot, and 25 testing plots. The optimal number of trees used for mapping mangrove in Youtefa Bay were 100, 200, and 800 with internal validation model 0.854. The results of this research show that mangrove forest area in Youtefa Bay is 189 ha with total accuracy 91 % and kappa index 0.89. DEMNAS, NDMI and MVI are the best independent variables on identifying mangroves in Youtefa Bay.