Identifying the minimum set of parameters for flood forecasting in Iowa using the HLM model

Hydrological model parametrization is a challenging and unsolved issue. There have been multiple attempts to address it; however, these are usually focused on maximizing performance at the outlet of a given watershed, leaving unsolved the problem of regional scale parameterization. In our work, we propose a top-down methodology to address this problem. We search the best combinations for three parameters of the Hillslope Link Model (HLM) using observed hydrographs for multiple small watersheds of Iowa (areas below 1000 km²). The parameters include λ1 and Vo that control the hydrograph, and Tl (top layer depth) which controls the runoff volume. In our methodology, we use two versions of the HLM (Hillslope Link Model): the HLM-RC with fixed runoff coefficient, and the HLM-TP with the inclusion of a top-layer soil. Nine combinations of λ1 and Vo were evaluated at multiple sites and events by varying over time the parameter runoff coefficient (RC) of the HLM-RC model. With the HLM-TP model, we evaluate seven values of Tl to fit the observed RC. In both cases, we applied the concept of consistency to determine the best parameters set up for each station. Then, we interpolate the obtained parameters to the domain and implement them into the HLM-TP model. Finally, we compare our results with observed records of the region, including watershed above 1000 km². Our results suggest the possibility to obtain robust parameters for a region derived from field observations. We use as reference for performance enhancement the simplest model with only 3 parameter values for all of Iowa, and we showed that using a set of 54 groups of parameters for the state increases performance significatively.