Forests play a vital role in carbon sequestration, biodiversity conservation, and climate regulation. However, changing environmental conditions, including variations in precipitation, temperature, and soil properties, significantly impact tree physiology and forest dynamics. This study aims to evaluate the potential of machine learning models in predicting canopy height, a key indicator of forest structure, using remote sensing data. Four models Random Forest, Support Vector Machine, and Gradient Boosting Trees (GBT) were tested under multiple scenarios incorporating Sentinel-2, Landsat 8, and Landsat 9 imagery. Predictor variables included Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), precipitation, land surface temperature (LST), soil moisture, population density, elevation, slope, aspect, albedo, and CO concentration. The study area, located south of Shaver Lake in Fresno, California, represents a dynamic forest ecosystem vulnerable to climatic variability. Results indicate that RF and GBT models achieved the highest predictive accuracy (R² = 0.66 and 0.63, respectively) with Sentinel-2 data, while CART and SVM performed less effectively, especially with Landsat 9 data. These findings emphasize the importance of selecting appropriate remote sensing datasets for tree physiology assessments. Future research should explore multi-source data fusion and advanced hybrid modeling approaches to improve forest monitoring and sustainable management.
Uyar et al. (Fri,) studied this question.