Modeling floods in large river basins: Model resolution and storm patterns

Abstract:

Floods in large river basins are not the simple result of heavy rainfall; rather it is the confluence of the spatio-temporal pattern of single to multiple rainfall events, antecedent moisture conditions, and the river network. Increasing the routing model resolution can improve modeled flood peak, volume and duration; however this can be computationally intensive. We use a high-resolution river routing model coupled to the VIC land surface model over the Ohio and Danube Rivers, both of which are large, flood-prone river basins. Using detailed river channel data from across the Ohio River basin to parameterize the river routing model, we demonstrate that channel width parameters can be transferred to the Danube River, providing a potential strategy for implementing high-resolution models in data-poor regions. Spatial and temporal resolutions at which runoff is generated for the flood model has varying impacts for different sub-basins, and we hypothesize that this is tied to the scale of the flood-generating precipitation events. This modeling framework, which is able to accurately simulate peak flood rate and flood volume, then allows us to explore the relative importance of antecedent moisture conditions and the spatio-temporal pattern of rainfall. We find that working back from the annual maximum flood, we have high correlation with the 7-day runoff preceding the flood event for the Ohio. However starting with the annual maximum 7-day runoff does not necessarily result in a flood, pointing to the importance of considering both the rainfall patterns and river network for flooding in large river basins.