Forest Type Classification and Above-Ground Biomass Estimation in the Petawawa Research Forest

This project used Sentinel-2 satellite imagery and LiDAR-derived canopy structure data to classify forest types and estimate above-ground biomass in the Petawawa Research Forest, Ontario, Canada. The workflow combined multi-seasonal optical imagery, forest inventory data, LiDAR canopy height metrics, field plot biomass measurements, and an independent EFI biomass product. A Random Forest classifier was used to separate evergreen and broadleaved forest, while a Random Forest regression model estimated biomass across the study area. The classification reached 78.6% overall accuracy, and the biomass model performed strongly with an R² of 0.794 and an RMSE of 37.7 t/ha. The final results showed that Sentinel-2 and LiDAR complement each other well for forest mapping, carbon assessment, and spatial biomass analysis.