Spatial modelling of mangrove AGC stock using vegetation indices and LULC changes in coastal area of Teluknaga and Kosambi, Indonesia

Abstract Mangrove ecosystem along the coast of Indonesia has experienced degradation and loss of carbon stocks due to extensive coastal development and land conversion. This study aims to develop a spatial model for estimating Above-Ground Carbon (AGC) stocks in the mangrove forests of Teluknaga and Kosambi in 2019, 2022, and 2025. Using Sentinel-2 satellite imagery, Land Use and Land Cover (LULC) was classified using the Random Forest (RF) algorithm, resulting in five main classes including water bodies, mangrove, vegetation, bareland, and built-up area. Vegetation indices (NDVI, EVI, SAVI, and ARVI) were specifically calculated for mangrove areas to construct a linear regression model against the reference AGB values obtained from secondary data. Results indicate that ARVI is the best predictor (R 2 = 0.80; RMSE = 40.98). The maximum AGC value recorded was 177.47 MgC ha −1 in 2019, decreasing to 153.36 MgC ha −1 in 2022 and plummeting to 111.62 MgC ha −1 in 2025. This decline is also reflected in the total carbon stock in the study area, which fell from 987,973.4 MgC to 595,583.8 MgC. This is related to the reduction in mangrove area from 368.45 ha in 2019 to 261.16 ha in 2025. Spatially, the conversion of mangrove to bareland and built-up area reached a total of 122.72 ha, with a predominance of conversion to bareland (78.66 ha) and to built-up area (44.06 ha). This study presents novelty through the utilisation of LULC changes as a spatial approach in estimating carbon reserves and provides a scientific basis for the protection of mangrove ecosystems amidst coastal development pressures.