Blue Carbon Stock Assessment of Mangroves in Barangay Tinib, Casiguran, Aurora, through Remote Sensing and Machine Learning Regression Techniques

Mangroves are important blue carbon ecosystems, yet ongoing degradation threatens their capacity to mitigate climate change.Using the surface reflectance of Sentinel-2 image, normalized difference vegetation index (NDVI) and modified normalized difference water index (MNDWI) information, three machine learning regression algorithms namely, k-nearest neighbor (k-NN), Random Forest (RF), and Gradient Boosting (GB), were applied to derive the blue carbon stocks of the mangrove forest of Barangay Tinib, Casiguran, Aurora via allometric models of the mangrove species and field inventory data.Three coastal proximity-based strata were established to capture spatial variability of blue carbon as affected by the distance to shoreline.Sixty plots, 20 per strata, were established for the collection of field-based aboveground carbon (AGC).Using the mean absolute error (MAE), root mean square error (RMSE), percentage RMSE (PRMSE) and root mean square percentage error (RMSPE) as the evaluation criteria, the results showed that RF has the best performance (MAE: 23.91 t/ha; RMSE: 33.61 t/ha; RMSPE: 22.97%; PRMSE 28.63%) and was used to generate wall-to-wall maps of aboveground biomass (AGB), AGC, and their carbon dioxide (CO 2 ) and oxygen (O 2 ) equivalents.The mean AGB was 112.55 t/ha, AGC was 47.08 t/ha, equivalent to 172.78 t CO 2 /ha and 125.70 t O 2 /ha.Carbon stock decreased with distance to the coast, and ANOVA confirmed significant differences among strata (p < 0.05).These findings demonstrate the effectiveness of integrating remote sensing (RS) and machine learning for mangrove carbon quantification and provide essential data for conservation, restoration, and climate-resilient policy-development.