Assessing mangrove health index as a basis for degradation mitigation planning

Abstract

This study aimed to detect mangrove cover, analyze spatio-temporal changes, and assess mangrove health conditions in South Banawa District, Donggala Regency, using multi-temporal Sentinel-2A imagery. Mangrove detection was conducted using a machine learning based decision tree algorithm, while mangrove health was evaluated using the Mangrove Health Index (MHI). The variables included spectral bands and multiple spectral indices (NDVI, NDBI, MNDWI, CMRI, NBR, GCI, SIPI, and ARVI). The classification model demonstrated very high performance, with Overall Accuracy, Kappa, and F1-score values exceeding 98%. The results indicated a decline in mangrove area from 123.96 ha to 95.5 ha, equivalent to a loss of 28.46 ha (22.96%) during the observation period. Degradation was spatially concentrated in areas with high accessibility and proximity to shrimp farming activities. Despite this decline, mangrove conditions were predominantly classified as healthy (87.56%), followed by moderate (12.41%) and poor (0.03%) categories. MHI-based mitigation strategies prioritize low-index areas for restoration through hydrological rehabilitation and buffer zone establishment, while healthy areas are primarily focused on conservation and periodic monitoring. This approach supports data-driven conservation planning, restoration prioritization, and sustainable coastal management based on remote sensing and machine learning.