Detecting Change of Aboveground Biomass (AGB) in Kwahu South, Ghana

GEDI L4A specifically provides globally available aboveground biomass density (AGBD) estimates, serving as reliable reference data for training and validating remote sensing models and reducing reliance on labor-intensive field campaigns. Vegetation indices and topographic variables were extracted from each satellite dataset and used as explanatory variables in a Random Forest regression model trained using ArcGIS Pro.

Results showed that rescaled PlanetScope imagery achieved the highest prediction accuracy with a model explaining 84% of variability in observed AGB and the lowest prediction error. Despite its higher spatial resolution, the original Planet model was less accurate compared to Landsat 9. These findings highlight the importance of matching the spatial scale of satellite data to that of reference observations.

They also suggest that increasing resolution does not always improve model performance and that thoughtful data preprocessing can significantly enhance prediction accuracy. This research offers a scalable and cost-effective method for monitoring biomass in regions where ground measurements are limited, contributing to better-informed forest management and carbon accounting efforts.