Comparison Between Traditional Forest Inventory and Remote Sensing with Random Forest for Estimating the Periodic Annual Increment in a Dry Tropical Forest

This study evaluates the effectiveness of combining remote sensing techniques with the Random Forest algorithm for estimating the Periodic Annual Increment (PAI) in a dry tropical forest located within the Caatinga biome in northeastern Brazil. The analysis integrates forest inventory data collected from permanent plots monitored between 2011 and 2019 with Landsat satellite imagery processed through the Google Earth Engine platform. By incorporating surface reflectance and vegetation indices, the approach significantly improved the accuracy of productivity estimates while reducing the costs and efforts associated with traditional field-based methods. The Random Forest model achieved a strong performance (R2 = 0.8867; RMSE = 0.87), and its predictions were further refined using post-processing correction factors. These results demonstrate the potential of data-driven modeling to support forest monitoring and sustainable management practices, especially in ecosystems vulnerable to the impacts of climate change.