Airborne lidar and machine learning reveal decreased snow depth in burned forests

Abstract

Wildfires are increasingly burning higher in elevations well into the seasonal snow zone, altering snow accumulation and melt dynamics. However, limited spatially distributed observations throughout the full snow season have constrained our understanding of how these changes vary across space and time. Here, we assess the impacts of fire on snow depth across nine basins in California's Sierra Nevada using a machine learning (ML) algorithm, Extreme Gradient Boosting (XGBoost), trained on 114 airborne lidar snow depth acquisitions with 50 m resolution in partially burned basins. We develop and apply an explainable ML framework by fitting an ML algorithm on snow depth for each flight as a function of spatial attributes including burn status. We then predict snow depth for counterfactual burned and unburned conditions for each flight to assess the ML-derived impact of fire on snow depth. Fire impact on snow depth evolved throughout the season, with slightly deeper predicted snow in burned forests in the accumulation season (56 % of acquisitions), and shallower predicted snow in burned forests in the ablation season (83 % of acquisitions) compared to unburned forests. Post-fire snow depth differences were smaller in the accumulation than in the ablation season. Lower elevations (< 2000 m) consistently exhibited smaller, near-zero changes in post-fire snow depth compared to higher elevations (> 2000 m). South- facing slopes experienced the largest negative post-fire snow depth changes. These results illustrate a new approach to assessing fire impacts on snow using lidar-derived snow depth and provide insights into snowpack dynamics in burned forests that are novel in their spatial extent and resolution, as well as their ability to discern fire impacts throughout the snow season.