With the surge of extreme meteorological events and intensification of urbanization downstream of hydraulic structures, the need for predicting failure of dams and dikes has become of paramount importance for establishing emergency response procedures (Zhong et al., 2021). To this end, numerical models are instrumental tools for simulating the embankment breaching process. Existing models can be classified into three categories. First, statistical (or parametric) models are based solely on regression analysis of data from past events or laboratory campaigns. They describe some breaching parameters (e.g., final breach width, failure duration or maximum breach discharge) as a function of dam or reservoir properties. These simple, computational-efficient models may lack generality because they entirely rely on data from specific cases without considering underlying physics (Chen et al., 2019;De Lorenzo & Macchione, 2014;Lee, 2019). Conversely, distributed physically based models can describe the phenomenon in greater detail, as they solve the flow and sediment governing equations using a computational mesh of the domain. Their results may be accurate but only if reliable data is available and if physical processes are numerically well represented, for example, erosion of non-homogeneous dam material, slope failure or 3D-flow patterns (Cantero-Chinchilla et al., 2019;Onda et al., 2019;Pheulpin et al., 2020;Shimizu et al., 2020). Additionally, the time required to run distributed physically based models can be substantial (ASCE, 2011). Simplified physics-based models offer a good trade-off (Wu, 2013). Without spatially distributing the flow description nor the embankment morphology, they enable simulating hydraulic and dam breach variables (e.g., time-evolution of breach discharge and dimensions) by describing selected physical processes (