Global Offshore Wind Turbine Mapping in 2025 Using the CPEF Framework and Sentinel-1 SAR

Accurate and up-to-date spatial information on offshore wind turbines (OWTs) is essential for offshore resource assessment, marine spatial planning, and environmental impact analysis. However, few existing datasets offer open access, global consistency, full-year 2025 coverage, and individual turbine-level spatial detail simultaneously. To address these gaps, we propose CPEF, a two-stage framework guided by constant false alarm rate (CFAR) detection that combines PCA-aligned peak profile features with EfficientNet-B0 image features for large-scale OWT detection using Sentinel-1 SAR imagery. We first composited multitemporal SAR observations into annual images to enhance persistent strong scatterers and suppress transient targets. A trimmed mean CFAR (TM-CFAR) was then used to generate high-recall candidate regions, followed by an EfficientNet-B0-based refinement model for candidate-level discrimination. To distinguish true turbines from false targets, PCA-aligned peak profile features were extracted along the principal and orthogonal directions and fused with deep visual features. CPEF detected 16,270 OWTs globally in 2025, with an overall accuracy of 96.33% and an F1-score of 97.25%. Compared with DeepOWT, our dataset identified additional turbines in regions such as Jiangsu, Shanghai, and the Dogger Bank area. Spatial analysis reveals that global OWTs primarily concentrate in nearshore Europe and coastal East Asia. In Asia, most turbines are in waters shallower than 20 m, whereas Europe has higher proportions in the 20–30 m and deeper depth ranges. These results demonstrate that CPEF offers an efficient, interpretable approach for global OWT mapping and provides an updated turbine-level dataset for spatial pattern analysis and offshore wind resource assessment.