An investigation of post-wildfire changes in hydrologic parameters using data assimilation in a southern California watershed

The Matilija Creek watershed in southern California, USA, is characterized by pronounced vulnerability to post-wildfire debris flow and sediment-laden flood hazards which are challenging to predict since they occur as a result of the confluence of diverse but interconnected physical mechanisms. These events are a cascading hazard, in that wildfire increases susceptibility to mass movements. Southern California is prone to wildfires due to its dry climate during the summer months. The fires in turn cause changes in hydrologic response, including increased runoff and decreased soil cohesion. The region has also experienced severe drought in the mid-2010s. Stationarity is often an assumption of both statistical and physically-based hydrologic models, but in the case of Matilija Creek watershed it is likely that the best hydrologic parameters vary as a result of both drought and fire. Changes in hydrologic response can be detected through a wide variety of statistical analyses, including traditional methods for detecting changes in water yield, double-mass analysis and flow-duration curves. Data assimilation is a promising approach for dynamically capturing post-disturbance changes in hydrologic response over time. This study aims to assess the utility of data assimilation with a physically-based hydrologic model to detect changes in hydrologic parameters during a drought and following a fire. In our previous work on this method, over-parameterization has likely caused inconsistent results between runs. If many parameters are allowed to change too drastically, different parameter shifts can cause the same results. This new analysis will therefore focus on a small set of infiltration-related parameters. Data assimilation is also compared to a statistical method – a performance of a linear model of runoff-ratio over time - for detecting post-wildfire changes in the hydrologic response. In choosing a data assimilation algorithm, this study seeks to provide a more objective and data-driven assessment of the wildfire-driven changes in hydrologic parameters than would be possible with other methods.