Predicting post-fire hydrological and erosive catchment response during rainfall events. A comparison of OpenLISEM and MOHID Land models

Wildfires are a source of instability for the natural water cycle in forested watersheds, endangering the water quantity and quality reaching downstream water bodies. The faster hydrological response of a burned area leads to increased runoff and transport of sediment and ash particles during and after rainfall events. Therefore, the use of an adequate spatiotemporal resolution in hydrological models is necessary to properly estimate post-fire impacts. Especially when addressing hydrological events such as flash floods and debris flows, which are highly unpredictable and are characterized by short duration and high impact outside the burned area. This study aims to compare the ability of two hydrological models to simulate the hydrological response and sediment transport during the first year after a fire to ultimately understand which one would best serve as a post-fire hydrological predicting tool at event scale. To achieve this goal, OpenLISEM, an event-based hydrological model, and MOHID Land, a continuous model with variable timestep, were compared. Driven by several limitations identified in previous modeling exercises at this scale during the calibration phase, this work performed a parametrization through the variation in boundary conditions characterizing each event. OpenLISEM and MOHID Land models exhibited similar capabilities in simulating runoff during the first postfire year. However, the larger erosion input parameters required by MOHID Land increase the complexity of erosion prediction and increase equifinality. In addition, MOHID Land limited capacity to perform sensitivity and uncertainty analyses emerged as a major disadvantage, hindering the assessment of the reliability of the model's predictions. Despite its limitations for not integrating subsurface flow and base flow, OpenLISEM is the most suitable model for assessing post-fire impacts on runoff and sediment production at the event scale, because of its ease of implementation and its reduced computational requirements.