Remote Sensing and Machine Learning-based assessment of Forest Biomass Changes in Pench Tiger Reserve, Maharashtra, India

This study evaluates spatio-temporal variations in forest biomass in the Pench Tiger Reserve, Maharashtra, India, using a combination of remote sensing, field-based observations, and machine learning approaches. Fractional Vegetation Cover (FVC) layers were generated using the Linear Spectral Unmixing (LSU) Algorithm applied to multi-temporal LANDSAT data (2001-2022). A Linear Regression Model (LRM) was developed to estimate forest biomass across three forest strata: Very Dense Forest (VDF), Moderate Dense Forest (MDF), and Open Forest (OF), based on field sampling plots. The results revealed a significant decline in carbon stock from 2001 to 2017, followed by an increase in 2022, correlating with FVC trends. The study also highlights that NDVI-based FVC models provided the highest correlation with field biomass data. A degradation matrix was developed to assess forest degradation trends. The findings contribute to policy decisions on land-use management, carbon sequestration enhancement, and climate balance maintenance.