OBJECTIVE RAINFALL INTENSITY-DURATION THRESHOLDS FOR THE INITIATION OF POST-FIRE DEBRIS FLOWS IN SOUTHERN CALIFORNIA

Rainfall intensity-duration thresholds are commonly used to predict the temporal occurrence of debris flows and shallow landslides. Most often, thresholds are subjectively defined as the upper limit of intensities and durations that do not produce debris flows and landslides, or as the lower limit of intensities and durations that do produce debris flows and landslides. In addition, data regarding the precise timing of debris flows within storms are not readily available. Only 5% of 2626 published landslide and debris-flow occurrences were temporally precise to ± 12 hours (Guzzetti et al. 2007). Here, we attempt to improve rainfall intensity-duration thresholds for post-fire debris flows in southern California using a combination of 2) recent data on the precise timing of debris flows relative to rainfall and 2) an objective method to define the threshold intensities. These data are derived from rainfall data from 353 storms, during which 58 debris flows were recorded at 13 different monitoring sites. We objectively defined the thresholds at durations from 5 to 720 minutes using an iterative receiver operating characteristics analysis to maximize the number of correct predictions of debris flow occurrence while maintaining a false alarm rate (threshold exceeding intensities without debris-flow initiation) at or below 10%.

Peak storm intensities were often significantly higher than those that immediately preceded recorded debris flows, and often occurred several minutes to hours after debris flows occurred. As such, peak storm intensities do not necessarily represent the meteorological conditions that contribute to the initiation of post-fire debris flows. Objectively defined thresholds were typically exceeded within minutes of the time a debris flow was recorded at a monitoring site. Correct predictions of debris-flow occurrence were significantly more numerous for the objectively defined thresholds than for the subjectively defined thresholds. False alarm rates were only slightly higher than previously defined upper-limit thresholds, and significantly lower than lower-limit thresholds. We conclude that objectively defined rainfall thresholds derived from precise timing data are better predictors of debris-flow occurrence and timing than those subjectively derived from peak storm intensities.