Remote sensing techniques have been demonstrated to be efficient tools for producing reliable and accurate information regarding forest inventory. Airborne Laser Scanning (ALS) can be used to develop Predictive Ecosystem Mapping (PEM) models based on Digital Elevation Model (DEM) derived layers. These PEM models, and the variables that they are based on may be used to generate reliable information to predict forest growth potential using an age-independent approach. This study evaluated the efficacy of using ALS-derived attributes to infer Site Index (SI) in plantation areas up to 30 years old, and compares the results with the current Biogeoclimatic Ecosystem Classification (BEC) estimates. I used a machine learning approach and Random Forest techniques to develop an age-independent SI model (SI_pem) incorporating DEM-derived topographic layers. The resulting SI_pem model produces accurate SI estimates at a fine grain across different forest ecosystems. The model performed significantly better when compared to Site Index by Biogeoclimatic Ecosystem Classification site series (SIBEC) lookup tables. Microclimate variables associated with water process, such as Topographic Positioning Index, Diurnal Anisotropic Heating, Topographic Openness Dominance and Overland Flow Horizontal Distance had significant importance to predict SI across the three areas studied in this work.
