Multidecadal satellite-derived Portuguese Burn Severity Atlas (1984–2022)

Abstract. Long-term burn severity assessment can support better pre- and post-fire management plans. In this study, Portuguese Burn Severity Atlas was created containing historical fires in Portugal from 1984 to 2022. As prerequisites, fire data were gathered and delimited for all years. Due to the availability of satellite images, for different years, different imagery from Landsat sensors (30 m) were applied. Exploratory analysis showed that burn severity estimates are significantly affected by the time lag between the satellite imagery acquisition and the fire date. We explicitly incorporated the effect of time lag in the degradation of burn severity estimates in the selection of the most suitable pre- and post-fire satellite images for each fire. Using Google Earth Engine, burn severity estimates were calculated for fires equal to or larger than 500 ha between 1984 and 2000 and larger than 100 ha for fires from 2001 to 2022 with known start and end dates (valid fires). Different indices were calculated, such as the differenced Normalized Burn Ratio (dNBR), relative dNBR (RdNBR), Relativized Burn Ratio (RBR), and a burn severity index that combines dNBR with enhanced vegetation index (dNBR-EVI). Overall, in Portugal, 4.92 M ha burned over the 38-year period (1984–2022), from which 3.19 million ha were caused by valid fires (64.8 %). Among these, a total area of 3.11 million ha had burn severity estimates via the applied indices (97 % of valid and 63 % of all fires). Results show that Portugal has experienced, on average, “high” burn severity throughout this period, with large percentages of dNBR pixels between 0.419 and 0.66 (29 %) and > 0.66 (20 %). Estimates from different burn severity indices provided a more complete representation of the burn severity impacts. Via the analysis of only three fires throughout the study period, the dNBR-EVI showed potential in differentiating the “unburned” and “regrowth” burn severity while RBR was more prone to signal saturation, i.e., inability to show “high” and “very high” burn severity. However, more in-depth research is needed to fully confirm these properties. This atlas can be accessed at https://doi.org/10.5281/zenodo.12773611 (Jahanianfard et al., 2024) and be used by researchers to have a better understanding of historical fires, their corresponding impacts on vegetation cover, air, soil, and water quality, and identification of the most influential environmental and climatical drivers of burn severity.