Hybrid Sentinel-1 and Survey-Based Forest Disturbance Mapping for the Southeastern United States (USDA Forest Service Region 8), 2017-2020

The Sentinel-1 Forest Disturbance Mapping (S1DM) dataset provides high-precision spatial and temporal information on forest disturbances, including wind, fire, insect outbreaks, and drought, across the Southeastern United States (USDA Forest Service Region 8), from 2017-2020. It combines USDA Forest Service Insect and Disease Survey (IDS) records from that period with Sentinel-1 SAR change detections spanning 2016-2021, refining each surveyed disturbance event to its actual radar-detected extent. Geographic coverage: USDA Forest Service Region 8, covering Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia. While Region 8 formally also includes Puerto Rico and the U.S. Virgin Islands, these territories were excluded from this dataset. Bounding box (EPSG:4326): 95.6°W to 76.5°W longitude, 29.0°N to 36.6°N latitude. Temporal coverage: Disturbance events (from IDS survey records): 2017 to 2020 Sentinel-1 observations used for change detection: 2016 to 2021 Coordinate reference system: EPSG:4326 (WGS 84). Forest disturbance information: The dataset covers the five disturbance types wind and bark beetle (~40% each), defoliators and fire (~10% each), and drought (<1%). Value of this dataset It is based on the Insect and Disease Survey (IDS) of the United States Department of Agriculture (USDA), a unique historical dataset with extensive thematic detail, temporal coverage, and spatial extent, but limited spatial and temporal precision (uncertainty of up to ~500 m and ±1 to 2 years per event). S1DM addresses this limitation by enhancing IDS records with structural change information derived from a commercial Sentinel-1 SAR backscatter product, pinpointing the actual spatial extent and timing of each disturbance far more precisely than the original survey data allows, while retaining the disturbance-type attribution from the original survey. This combination is the core value of the dataset: it gives you the structural precision of a commercial satellite product together with the thematic detail of IDS, without requiring access to that commercial product yourself. S1DM is more spatially and temporally precise than IDS alone, and more thematically informative than Sentinel-1 change detection alone, making it particularly suited for training and evaluating machine learning models on forest disturbance detection, and for ecological research on disturbance regimes. Note: you do not need access to the underlying Sentinel-1 processing product to use S1DM. The refinement has already been done; S1DM is delivered as a ready-to-use dataset under an open license (see below). Dataset creation This dataset is the outcome of a research project. By combining historical survey data with modern satellite observations, S1DM makes this valuable historical record usable with modern data-driven methods that require high spatial and temporal accuracy to be effective. The paper describing the methodology of this dataset: Müller, F., Eifler, L., Cremer, F., Beck, P., Camps-Valls, G., and Bastos, A.: Hybrid forest disturbance classification using Sentinel-1 and inventory data: a case-study for Southeastern USA, Nat. Hazards Earth Syst. Sci., 26, 2785-2815, https://doi.org/10.5194/nhess-26-2785-2026, 2026.