Prediction and Trade-Off Analysis of Forest Ecological Service in Hunan Province on Explainable Deep Learning

Abstract

Ecosystem services play a crucial role in maintaining ecological balance, providing essential functions. This study examines the trade-offs and synergies among five key ecosystem services in ecological forests across different regions of Hunan Province, China. Various machine learning models are compared to predict ecosystem service value (ESV) levels, with the most effective predictive model identified. The SHAP (SHapley Additive exPlanations) analysis is employed to identify key environmental and management factors influencing ecosystem services. Our findings reveal significant regional variations in ecosystem services, with the eastern and western regions showing superior soil conservation and forest nutrient retention. In contrast, the southern and western regions, particularly karst areas, display fewer trade-offs between ecosystem services, likely due to the effectiveness of ecological policies. SHAP analysis further reveals that factors such as precipitation during the warmest quarter, central government compensation funds, and timber harvesting volume strongly influence regional ESV. This study provides valuable insights for improving ecosystem service management and policy-making in rapidly developing regions, underscoring the importance of ecological protection strategies for sustainable development.