Decision Tree and Random Forest Classification Algorithms for Mangrove Forest Mapping in Sembilang National Park, Indonesia

Sembilang National Park, one of the best and largest mangrove areas in Indonesia, is very vulnerable to disturbance by community activities. Changes dynamic condition forests Park must be quickly easily accompanied monitoring efforts. One way monitor use remote sensing technology. Recently, machine-learning classification techniques have been widely used classify forests. This study aims investigate ability decision tree (DT) random forest (RF) algorithms determine distribution Park. The satellite data are Landsat-7 ETM+ acquired on 30 June 2002 Landsat-8 OLI 9 September 2019, as well supporting such SPOT 6/7 image 2020–2021, MERIT DEM an existing map. pre-processing includes radiometric atmospheric corrections performed using semi-automatic plugin contained Quantum GIS. We applied forest. In DT algorithm, threshold analysis carried out obtain most optimal value distinguishing non-mangrove objects. Here, RF involves several important parameters, namely, normalized difference moisture index (NDMI), soil (NDSI), near-infrared (NIR) band, digital elevation model (DEM) data. results from images show similarities regarding spatial distribution. algorithm with parameter combination NDMI + NDSI effective classifying image, while NIR image. Image (6 bands), number trees = 100, variables predictor (mtry) square root (√k), minimum node sizes 6, provides highest overall accuracy for combining (7 bands) parameters mtry all (k), size 6 higher when (99.12%) instead (92.82%) but it slightly (98.34%) (97.79%) outperforms because provide a producer mapping development method should support rehabilitation programs mangroves more easily, particularly Indonesia.