Machine learning-based assessment of offshore wind farm impacts on soft-bottom benthic communities in the Shandong Peninsula

Offshore wind power, as a crucial component of clean energy, is rapidly expanding globally; however, the long-term impact mechanisms on marine benthic ecosystems remain unclear. This study focuses on four offshore wind farms (South 3, South 4, Site V, and Site U1) in the southern waters of the Shandong Peninsula. Based on benthic organism survey data and multi-source remote sensing environmental data from 2015 to 2024, a remote sensing-in-situ integrated machine learning prediction framework was constructed to systematically assess the spatiotemporal impact of wind farm construction and operation on soft-bottom benthic communities. The study employed the XGBoost model as the main model and the Generalized Additive Model (GAM) as the baseline model, using SHAP interpretability analysis to reveal key driving factors. The results show that the XGBoost model achieved an R² of 0.742 on the test set, significantly outperforming the GAM model (R²=0.625). The years in operation (YSI) was the most important factor affecting benthic community diversity; after a brief disturbance during the initial construction phase, the community showed a significant recovery trend after 2–4 years of operation. The artificial reef effect caused by the conversion to hard bottom near the pile foundations resulted in an approximately 13% increase in the Shannon diversity index and an approximately 40% increase in species richness compared to the control area. This study provides a reproducible methodological framework for the ecological impact assessment of offshore wind farms, and the findings can provide scientific basis for the environmentally friendly layout planning of offshore wind power in China.