The Effect of Storm Direction on Flood Frequency Analysis Using Physically-Based Streamflow Simulations

Abstract:

How will future changes in preferential storm direction affect current patterns of runoff generation and peak flow distributions across the conterminous United States? This is an essential question with practical implications for the management of water resources given that both long-term records and climate models predict significant changes in rainfall magnitude and direction. In this work, we use an extensive set of numerical experiments to systematic explore this question with an initial focus on two typical watersheds within the Midwest United States: the Turkey River basin (medium scale with a drainage area of 4,385 km²) and the Cedar River basin (mesoscale with a drainage area of 20,168 km²). Our approach uses hydrologic simulations driven by rainfall fields generated with stochastic storm transposition (SST), constrained from long-term radar observations. Systematic rotations of the basin domain are used to recreate different storm directions based on the SST framework. Our results show that the peak flow response is affected by the storm direction in both basins. However, storm direction only significantly impacted the peak flow distribution in the mesoscale basin. The lack of sensitivity for the medium-scale basin is attributed to the high diversity of the selected storms in the SST framework and the relative difference between storm scale and basin size. These findings highlight the need to develop new flood frequency methodologies that take into account future climate-driven rainfall pattern changes, and also to re-evaluate the current selection of explanatory variables in regional flood frequency equations, which has ignored the effect of storm direction in peak flow estimations.