Aulia Lanudia Fathah, Bambang Semedi, Fitri Candra Wardana, Andik Isdianto
Mangrove ecosystems are critical blue carbon reservoirs. Accurate estimation of Above-Ground Carbon (AGC) is essential for climate change mitigation and coastal management. This study applies Sentinel-2 multispectral imagery and vegetation indices with a Random Forest model to estimate mangrove AGC in fragmented coastal ecosystems of Ujungpangkah, Indonesia. Field-based AGC from 15 plots, derived using species-specific allometry, was integrated with Google Earth Engine processing of spectral bands and indices. The classification mask achieved high accuracy (OA = 0.946; κ = 0.810). Among predictors, the Shortwave Infrared Bands (SWIR; B11–B12) and the Transformed Vegetation Index (TRVI) were the most effective, outperforming traditional red-edge indices. External validation showed reliable performance (R2 = 0.65; RMSE ≈ 1.05 tons C/ha), while the mapped mangroves averaged 25.3 tons C/ha, totaling ~2,036.5 tons C. SWIR was negatively correlated with AGC due to canopy water and structure, whereas TRVI captured canopy variability and reduced saturation. These findings highlight the potential of integrating Sentinel-2 imagery and vegetation indices with machine learning to provide cost-effective and replicable blue carbon monitoring for mangrove management. © (2025), (Dr D. Pylarinos). All rights reserved.
Master Program of Environmental Management & Development, Brawijaya University, Indonesia; Department of Marine Science, Brawijaya University, Indonesia
