Right on time: evaluating accuracy and timing of burn-severity assessments to support post-fire management

Effective wildfire impact assessment and post-fire management rely on burn-severity maps that accurately characterize fire effects on vegetation and soils. These maps inform time-sensitive activities such as post-fire flood and debris-flow modeling, risk mitigation for homes and communities, and ecosystem recovery planning, but the satellite observations used to generate them are often obstructed or delayed by smoke, clouds, and snow. Across 9,129 large fires (≥ 404 ha) in Canada and the United States from 2016–2024, we quantified post-fire data obstruction and evaluated observational latency – defined as the time required to acquire the first usable scene – for four freely available moderate-resolution remote sensing datasets: Landsat, Sentinel-1, Sentinel-2, and the Harmonized Landsat Sentinel-2 (HLS) product. For multispectral datasets (Landsat, Sentinel-2, and HLS), cloud cover was the primary source of data obstruction, particularly in mesic and coastal regions, whereas Sentinel-1 (synthetic aperture radar) can penetrate clouds. Consequently, Sentinel-1 had the shortest observational latency (median = 4 days), followed by HLS (7 days), Sentinel-2 (10 days), and Landsat (17 days). Likewise, one-month composite imagery from Sentinel-1 (89.6%) and HLS (82.8%) achieved near-complete coverage for most fires, substantially better than Sentinel-2 (70.5%) and Landsat (56.1%). Despite its widespread use in operational burn-severity mapping, Landsat provided the slowest and least complete early post-fire coverage. For 77 fires in forested ecosystems, we also evaluated relationships between early post-fire imagery and field-derived burn-severity measurements collected 0–3 years after fire. All multispectral datasets corresponded well with field data and reliably identified areas of near-total canopy loss using imagery acquired within two months after fire (> 80% balanced accuracy), performing nearly as well as extended assessments using imagery from a full post-fire growing season. Sentinel-1 produced less accurate burn-severity estimates but may remain valuable where persistent cloud cover limits observations, including eastern Canada and the Pacific Northwest. Overall, HLS provided the best combination of latency, spatial coverage, and burn-severity mapping performance, making it a strong candidate for operational use and ecological monitoring across diverse North American ecosystems.