We demonstrate a near-real time forest disturbance alerting system for Europe using Sentinel-1 radar data. Sentinel-1 radar can penetrate clouds and offers high spatial (∼20 m) and temporal (3- to 6-day) detail. We directly integrated near-real time ERA5-Land temperature and Copernicus forest type data into the disturbance detection framework to address freezing temperatures and seasonal phenology, both of which influence the Sentinel-1 backscatter signal and thus need to be accounted for. This facilitates year-round monitoring across a range of environmental conditions (sub-zero, wet and dry) and forest types (coniferous, deciduous) throughout the boreal, temperate, and Mediterranean forests of Europe. Validation across Europe showed high accuracy, with a user accuracy of 91.2% (±1.3%) and producer accuracy of 74.5% (±6.0%). User accuracy increased to 99% (±0.4%) when excluding errors in the European-scale forest cover mask primarily caused by local overestimation of forest height and density. Disturbances were detected with a median delay of 27 days relative to the first high-resolution optical Planet reference image, which can further be reduced to 1 day through retrospective event-based correction of late detection bias. Compared to existing annual optical-based products, our method improves the detection of small-scale disturbances such as group fellings in Romania. We generated European-scale estimates of intra-annual disturbance seasonality, capturing variation in forest management practices and disturbance regimes such as winter harvesting in northern Europe, spring sanitation cutting in central Europe, and summer wildfires in southern Europe. Overall, this alerting system provides timely and detailed forest disturbance information in support of sustainable forest management, biodiversity conservation, carbon accounting, and law enforcement efforts across Europe. The alerts are available at https://wurnrt-raddeurope.projects.earthengine.app/view/radd-europe. • Near-real time Sentinel-1 forest disturbance alerting demonstrated across Europe. • Direct integration of ERA5-Land temperature data enabled year-round monitoring. • User accuracy of 91%, up to 99% with forest mask errors excluded; producer accuracy of 75%. • Median relative detection delays of 27 days and 1 day after event-based bias-correction. • First European-scale estimates of intra-annual forest disturbance seasonality.
Woude et al. (Fri,) studied this question.