ECOSTRESS reveals pre-fire vegetation controls on burn severity for Southern California wildfires of 2020

Aim Topography, climate and plant water stress influence the spatial patterns of burn severity (damage to soils and vegetation associated with wildfires); however, the relative importance of these predictors remains contested. We hypothesized that high-resolution pre-fire vegetation water stress, measured using evapotranspiration (ET), evaporative stress index (ESI) and water use efficiency (WUE), are strong predictors of burn severity. Location California, USA. Time period 2019-2020. Major taxa studied Forests, chaparral, oaks and grasslands. Methods We used random forest regression modelling with remotely sensed ECOsystem and Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) plant water stress from the spring-summer and year before the fires, predictors to represent spatial autocorrelation, and topography to predict the spatial patterns of burn severity for six fires in the Sierra Nevada and the Southern California Mountains. Results Plant water stress was the dominant predictor for the Southern California Mountains, and spatial autocorrelation in Sierra Nevada. Elevation was a leading predictor for burn severity for both regions, with greater predictability in higher-elevation and steeper terrain. Model results varied by vegetation type: for the Southern California Mountains, grasslands had the highest predictability [R-2 = .50 (high elevation); R-2 = .58 (steep)], followed by savanna and open shrublands, which are typified by chaparral-type vegetation [R-2 = .43 (savanna, high elevation); R-2 = .40 (open shrubland, high elevation)]. For grasslands, annual ET was the leading predictor (R-2 = .58), consistent with the idea that growing season grassland enhancement increases fuel availability. The predictability of burn severity was also sensitive to the time of day of the ECOSTRESS acquisition and varied by vegetation type. Afternoon ET was the leading predictor for high-elevation grasslands (R-2 = .50), and morning predictors were important for steep-sloped woody savanna (R-2 = .18). Main conclusions Our findings provide insights on how antecedent plant water stress predicts burn severity and can inform pre-fire season monitoring and fuel management.