CONTROLS ON POST-WILDFIRE CHANNEL SEDIMENT YIELD

Wildfire disturbance of mountainous areas is being exacerbated globally by climate change and ongoing and historical land-use, increasing the potential for post-wildfire flooding and debris flow hazards. Predicting sediment yield from recently burned areas remains a challenge but is important for effective hazard and resource management as wildfire impacts increase. In this study, we were interested in identifying the dominant controls on postfire sediment yield, their evolution through time, and use this knowledge to develop statistical predictive models. We used a unique dataset of multitemporal channel airborne lidar surveys to quantify sediment flux across >40 km² of shrub-dominated mountains burned by the 2018 Holy Fire in southern California, USA. Using co-located data on geomorphic variables, rainfall, and burn severity, we quantified variable importance for predicting sediment yield from initial events using a data-driven random forest approach. We were able to explain greater than 50% of out-of-sample variance for initial runoff events following wildfire, with geomorphic variables related to sediment supply and watershed average slope being the most important predictors. Sediment loading during a dry weather window was an important factor in supplying sediment to convergent terrain for subsequent runoff-driven erosion. Later in the rainy season, channel sediment yield by runoff was more subdued and in many parts of the landscape, channels acted as temporary sinks of hillslope-derived sediments. This work, along with other previous studies in the region, implies that continued erosion of soils on hillslopes acts as a persistent source of sediment for later season events. In all, this study demonstrates the importance of watershed slope and sediment supply as important controls on the magnitude of initial channel-derived sediment yield from burned catchments subject to intense rainfall and highlights the importance of shallow hillslope erosion in later rainy season events. These findings can guide both future hazard management and monitoring of burned and recovering landscapes.