Comparative Performance of Sentinel-2 Vegetation Indices for Estimating Aboveground Biomass and Carbon Stock in Forest Areas

This study evaluated the performance of Sentinel-2 vegetation indices for estimating aboveground biomass (AGB) and carbon stock in the Mahasarakham University Forest area, northeastern Thailand.Field data were collected from six 40 × 40 m sample plots, comprising 1,127 trees representing 61 species.AGB was estimated using allometric equations and converted to carbon stock.Six Sentinel-2 vegetation indices-the Normalized Difference Vegetation Index (NDVI), Soil-Adjusted Vegetation Index (SAVI), Modified Soil-Adjusted Vegetation Index 2 (MSAVI2), Ratio Vegetation Index (RVI), Difference Vegetation Index (DVI), and Green Ratio Vegetation Index (GRVI)-were transformed into fractional vegetation cover (FC) and related to field-derived AGB using exponential regression models.A consistent comparative framework was applied to determine how vegetation-index selection affects model fit, prediction error, and agreement with field-derived carbon stock.Among the six indices, MSAVI2 showed the strongest relationship with field-derived AGB (R² = 0.839) and estimated a total carbon stock of 1,752.866tC, closest to the field-derived estimate of 1,807.721tC, with a difference of 3.03%.In contrast, SAVI produced the lowest prediction error (RMSE = 0.901 t ha⁻¹; RMSE% = 1.339%).These results reveal that the best-performing vegetation index depends on the evaluation criterion, as the index with the strongest model fit does not necessarily produce the lowest prediction error.By directly comparing six Sentinel-2 vegetation indices using the same field dataset and modeling framework, this study provides a consistent basis for vegetation-index selection and supports practical applications of Sentinel-2 for forest AGB and carbon stock assessment.