Files
Abstract
Industrial plantations are rapidly expanding to counter deforestation and meet urbanization-driven timber demands. This growth requires efficient data-driven management to enhance yields in the face of rising wood consumption, population, illegal logging, and land-use change. Plantations, like the southeastern U.S. loblolly pines, often underperform due to generic management approaches. This underscores the demand for robust data-driven management systems, which rely on the accurate and repetitive measurement of forest attributes. Collecting such data through traditional on-site data collection is not feasible. It necessitates high-resolution remote sensing technologies like LiDAR for effective forest management.To address these challenges, this dissertation first compares the performance of four different modeling methods for predicting various plot-level forest attributes necessary for decision-making using aerial LiDAR data: (1) Least Squares Regression (LSR), (2) Adaptive Least Absolute Shrinkage and Selection Operator (ALASSO), (3) Random Forest (RF), and (4) Generalized Additive Modeling Selection (GAMSEL). The results revealed that no single method exhibits a significant advantage in accuracy over the others. Nevertheless, the ALASSO method exhibited shighltly higher accuracy. Moreover, it is considered less biased and more accessible compared to the other three.
The second part of this dissertation focuses on the development of methods for assessing understory vegetation in plantation forests, particularly in the southeastern coastal plain pine plantations, where the high competition from evergreen understory species constrains pine growth. Our study identifies Terrestrial Laser Scanning (TLS) as a promising tool for understory vegetation assessment, as the model based on variables derived from this sensor demonstrates high accuracy in detecting and quantifying understory biomass. However, implementing this sensor for larger areas is not practical. Therefore, the third part of this dissertation addresses the upscaling of the TLS-derived model to Unmanned Aerial Vehicles and Aerial Laser Scanning-derived variables, enabling its application to larger areas. The result shows that both the UAV and ALS-based methods can assess the understory biomass. However, the ALS-based upscaled models were more accurate and robust, demonstrating efficacy and potential for enhanced forest management and decision-making.