Skripsi

DEFANI ARMAN ALFITRIANSYAH. Mangrove Carbon Stock Modeling Based on Sentinel-2A Imagery in the Cipunegara River Estuary, Subang, West Java. Supervised by DEDE KOSASIH and IING NASIHIN.

Mangrove forests play a vital role in mitigating climate change due to their high carbon sequestration capacity. However, traditional field measurements for carbon estimation are time-consuming and limited in spatial coverage. This study aims to develop a locally calibrated Above Ground Carbon (AGC) estimation model for the mangrove ecosystem in the Cipunegara River Estuary, West Java, utilizing Sentinel-2A imagery and python based cloud computing platform. Field data were collected from 25 plots to measure the diameter at breast height (DBH) and calculate AGC using allometric equations. Spatial analysis involved cloud masking using Cloud Score+ and mangrove mapping via the Random Forest classifier. The study evaluated three vegetation indices (NDVI, IRECI, and TRVI) and their combinations through linear regression models.

The classification identified 710.44 ha of mangrove area with an overall accuracy of 95.4%. Field measurements revealed an average AGC of 80.21 ton/ha, dominated by Avicennia genus. The supervised regression analysis demonstrated that the combination of IRECI and TRVI yielded the most accurate predictive model, outperforming single indices like NDVI, which suffered from spectral saturation. The optimal model achieved a Coefficient of Determination (R²) of 0.743 and a Root Mean Square Error (RMSE) of 19.724 ton/ha. The resulting mathematical equation $AGC = 81.6197(IRECI) - 1.0512(TRVI) - 6.5443$, was applied to generate a spatial carbon map, estimating AGC values between 23.21 and 171.16 ton/ha. This study confirms that integrating red-edge and near-infrared bands significantly enhances the accuracy of mangrove carbon modeling.

*Keywords : AGB, Sentinel 2A, Remote Sensing*