MODELLING CHANGES IN RAINFALL INTENSITY-DURATION THRESHOLDS FOR FLASH FLOODS IN A SOUTHERN CALIFORNIA WATERSHED OVER FIVE YEARS OF POST-FIRE RECOVERY

Rainfall intensity-duration thresholds (ID) are a commonly used tool for assessing the potential for flash floods following wildfires. Observations suggest that ID thresholds increase with time due to hydrologic recovery of landscape and potential changes in flash flood generation mechanisms. These changes in ID thresholds, however, are not often quantified, which limits our capability to reliably predict flash floods throughout the post-fire recovery period. In this study, we force the KINEROS2 hydrologic model with radar-derived precipitation to estimate rainfall ID thresholds for post-fire flash floods in a 41 km² watershed in southern California, USA. Prior work in this study area provides constraints on how key hydrologic model parameters change through time over the first five years of recovery. Therefore, we can use the model to explore how rainfall ID thresholds change as a function of time since burning. We also examine a variety of rainfall metrics, specifically averaging rainfall intensity over different temporal scales from 5 minutes to 60 minutes, to assess their ability to serve as reliable thresholds for flash floods. Thresholds based on the average rainfall intensity over relatively long durations (30 or 60 minutes) perform better than those based on the average rainfall intensity over shorter time intervals (5, 10, or 15 minutes). We also find thresholds based on the 3^rd quantile of the radar-derived rainfall intensity over the watershed perform better than threshold based on the 1^st or 2^nd quantile of the rainfall intensity. The model indicates an increase in the magnitude of the ID threshold of more than a factor of 2 from post-fire year 1 to post-fire year 5. Our findings highlight the potential of using hydrologic models to estimate temporal changes in ID thresholds following disturbance and provide insight into which rainfall metrics are likely to be most reliable for assessing the potential for post-fire flash floods. Our study also presents a methodology, which includes the effects of temporal changes in soil hydrologic parameters, that could be used in similar settings to inform post-fire flash flood hazard mitigation and management.