Agroforestry Land Use Land Cover Area Classification Using Decision Tree Algorithm

Abstract

Monitoring the location and extent of agroforestry land use land cover (LULC) in Lampung Province is critical for effective policy development and sustainable agroforestry management. However, existing monitoring efforts have been limited to small regions. This study addressed this gap by employing threshold values from five distinct vegetation indices (ARVI, EVI, GDVI, NDVI, and SAVI) derived from Landsat 9 OLI imagery to accurately identify and estimate agroforestry LULC across the Lampung Province. The data collection activities were carried out using a combination of Landsat 9 OLI satellite imagery acquisition, and ground truth validation on 7 classes of different land use (forest, agroforestry, dry land farming, ricefield, settlements, bare land, and water bodies) within 5,600 points of interest (POI) inside 5 regencies as an area of interest (AOI). This study aimed to predict agroforestry area based on vegetation indices (VIs) threshold using the decision tree (DT) algorithm. The research process involved a series of systematic steps, beginning with satellite image data acquisition and preprocessing, VIs values extraction, and DT sequential for agroforestry areas. The DT computation incorporated the value of each LULC type on the 5 VIs. The result showed that the overall accuracy reached 91.59% with a Kappa coefficient of 0.89, indicating a high level of accuracy for land cover identification. The DT algorithm calculation showed that the agroforestry in Lampung Province estimated spanned for 734,739.61 ha, determined only by NDVI and ARVI. The findings have significant implications for both policy development and agroforestry management. Accurate LULC classification enhances decision-making processes by providing reliable data on land use patterns, which can guide sustainable land management practices and support the creation of region-specific agroforestry policies. This research directly informs policymakers on the extent and distribution of agroforestry areas, offering a foundation for crafting strategies aimed at promoting sustainable land use while mitigating environmental degradation. The methodology also provides a scalable approach for other regions facing similar agroforestry and land management challenges.