✨ TL;DR
This paper presents a global dataset of Sentinel-1 SAR time series tracking offshore wind infrastructure deployment and operations from 2016-2025, with 15,606 time series and expert-annotated benchmarks for monitoring wind farm construction and operational dynamics. The dataset enables independent, high-resolution global monitoring of the rapidly expanding offshore wind sector.
The offshore wind energy sector is expanding rapidly worldwide, but there is a critical gap in independent, high-temporal-resolution monitoring of infrastructure deployment and operational dynamics at global scale. While Earth Observation has matured for spatial localization of offshore wind infrastructure, existing open datasets lack the temporal density and semantic detail needed to track construction phases, operational events, and regional deployment patterns.
The authors leverage Sentinel-1 synthetic aperture radar (SAR) data to create a dense time series corpus spanning 2016Q1 to 2025Q1. They employ an updated object detection workflow to identify infrastructure locations and extract 1D SAR backscatter profiles from each Sentinel-1 acquisition. The dataset includes 15,606 time series with 14,840,637 events, supplemented by a rule-based classifier for semantic labeling and an expert-annotated benchmark of 553 time series with 328,657 event labels for validation and benchmarking.
What the paper shows.
The baseline rule-based classifier achieved a macro F1 score of 0.84 in event-wise evaluation and an area under the collapsed edit similarity-quality threshold curve (AUC) of 0.785, demonstrating temporal coherence. The compiled dataset contains 15,606 time series with 14,840,637 analysis-ready 1D SAR backscatter profiles, and the expert-annotated benchmark comprises 553 time series with 328,657 event labels. The corpus successfully supports global-scale analyses of deployment dynamics, regional pattern differences, and operational events.
The paper does not explicitly discuss limitations, but implicit constraints include: reliance on Sentinel-1 SAR data which may have weather-related acquisition gaps; the rule-based classifier's performance depends on the quality of underlying SAR backscatter features; the expert-annotated benchmark of 553 time series represents a small fraction of the full 15,606 time series; and the approach's applicability may vary across different geographic regions with different oceanographic and atmospheric conditions.
✨ Generated by Claude · Apr 25, 2026 · Read the PDF for authoritative content.