Mapping high poplar growth areas for bioenergy cultivation: A swarm-optimized approach

Abstract

Poplar trees are a renewable energy source, and identifying high-growth areas for their cultivation is essential for successful bioenergy production. This study presents a novel approach to identify these areas using support vector regression (SVR) combined with two swarm intelligence optimization algorithms, i.e., particle swarm optimization (PSO) and gray wolf optimizer (GWO). The approach uses a spatial database of thirteen geoenvironmental variables and the location of 136 poplar farms in a rugged and heterogeneous region in western Iran. The PSO and GWO algorithms optimized the hyperparameters of the base SVR model and generated two swarm-optimized models: SVR-PSO and SVR-GWO. The swarm-optimized models were compared to the standalone SVR model using various evaluation metrics. The results showed that the SVR-GWO model achieved the highest performance (accuracy = 82%, sensitivity = 82%, specificity = 74%, and prediction rate = 92%), followed by the SVR-PSO model (accuracy = 81%, sensitivity = 81%, specificity = 72%, and prediction rate = 90%), while the standalone SVR model had accuracy and stability (accuracy = 75%, sensitivity = 77%, specificity = 68%, and prediction rate = 82%). The results also revealed that elevation, slope, topographic wetness index, and proximity to roads and rivers were strongly associated with the distribution of poplar farms in the study area. This study provided a novel and effective approach for identifying high poplar growth areas that can help decision-makers and stakeholders allocate resources and plan interventions to promote poplar cultivation for renewable energy production.