Flood after Fire – Predicting Peak Streamflows using Practical Hydrologic Models
Wildfires can significantly impact land cover, soil properties, and hydrologic processes. These changes in the watershed can lead to an increased risk of debris flows and larger peak streamflows, which pose risks to human lives, infrastructure and assets, and ecosystems. In the last decade, anthropogenic activities such as climate change, human ignited fires, and accumulated fuel loads have contributed to increased severity and occurrence of wildfires in at the wildland-urban interface, increasing hazard risks for communities. For example, the Holy Fire (August 2018) burned about 94 sq-km in Orange County and Riverside County in California (USA) and affected over 1,000 residents. Three months later in California, the Woolsey and Hill Fires (November 2018) burned approximately 392 sq-km in Los Angeles County and Ventura County. These two fires combined resulted in 3 fatalities, 8 non-fatal injuries, and 1,661 destroyed buildings. Quickly following these wildfires (December 2018-Feburary 2019), record amounts of rainfall fell across Southern California, contributing to flooding and debris flows in the burned regions. During this time, the Riverside County Fire Department ordered the evacuation of 300 homes. Additionally in Los Angeles County and Ventura County, over 300,000 people were evacuated from their homes; despite these great efforts, flooding and debris flows killed at least three people. Furthermore in January 2018, following the December 2017 Thomas Fire, communities subsequently experienced debris flows that killed at least 23 people and destroyed over 100 homes in Montecito, California. In Southern California, Rowe, Countryman, and Storey (RCS), an empirical method based on rainfall-runoff data collected prior to 1949, is often used to rapidly estimate post-fire peak streamflows. It was hypothesized that the 1949 RCS method does not reflect current hydrologic and geomorphic conditions and may not adequately predict post-fire runoff events. This research evaluates RCS for 28 small watersheds in Southern California to assess model performance. Observed streamflow data are compared to RCS predictions before and after wildfire. Where data are available, RCS is used to predict peak flows up to 70 years after fire, indicating a full recovery of the watershed and a return to pre-fire conditions. Results show that RCS, without the use of any bulking factors, generally under predict post-fire peak streamflow events during the wildfire recovery period. Results suggest that RCS is overgeneralizing watershed processes and may not represent the spatial and temporal variability present in systems affected by wildfire. Local physiography, land cover, geology, slope, aspect, rainfall intensity, and soil burn severity were incorporated into a calibrated random forest algorithm to characterize model performance across different watersheds. This characterization of important parameters in post-fire peak flow is compared to the structures of existing process-based hydrologic models including Hydrologic Engineering Center- Hydrologic Modeling System (HEC-HMS) and Automated Geospatial Watershed Assessment Tool (AGWA). This work also serves as a basis for improving post-fire watershed predictions and ultimately decision support tools for post-fire hazard evaluation and management in Southern California.