Abstract:
This study assesses the functionality of Sentinel-2 satellite facts to estimate aboveground carbon
density (AGCD) in the Ampara district of Sri Lanka, which is known as a biodiversity hotspot vital
for climate resilience. The look at advanced a linear regression version using Google Earth Engine,
incorporating Sentinel-2 surface reflectance imagery from 2020 to 2021 alongside the World
Conservation Monitoring Centre (WCMC) international carbon dataset. Key predictors comprised
spectral bands, NDVI, and masks for dynamic world land cover to delineate vegetated areas. The
model showed a strong relationship (r = 0.89) between the predicted and actual carbon densities
(tonnes/ha), described by the equation: Predicted Carbon Density = 1.325 × Carbon tonnes per
/ha − 28.774. Systematic errors were observed in low-carbon zones, resulting in implausible
negative estimates. Validation with more than 400 sample points showed a lot of differences in
space: the measured AGCD went from 0.07 to 123.6 tonnes/ha, while the predictions went from
−14.9 to 99.9 tonnes/ha. In dense forests, the measurements were close to each other, but
differences in farming and damaged areas showed that adjustments are needed for varied
landscapes. An RMSE of ±18.2 tonnes/ha showed it was suitable for regional monitoring, but
also pointed out challenges in dealing with detailed ecological details. The study indicates Sentinel
2 demonstrates capability in conducting inexpensive assessments of tropical ecosystem carbon
stocks, which enables policymakers to implement sustainable management tools at different scales.
Future initiatives must incorporate precise biomass measurement techniques like LiDAR for
enhancing accuracy estimates for complex terrain features to support diverse species regions in
climate change initiatives.
