Effect of Rainfall Spatial Distribution on Flood Forecasting in Complex Terrain

Abstract:

This study examines the impact of spatially distributed versus lumped rainfall input data on flood forecasting. For this purpose, MRMS (Multi-Radar/Multi-Sensor) data is used to provide spatially distributed rainfall data and MPE (Mean Precipitation Estimation) data is used to provide spatially lumped rainfall data to force the ModClark rainfall-runoff model, developed by the Hydrologic Engineering Center (US Army Corps of Engineers). ModClark is an event-oriented semi-distributed rainfall-runoff model which simulates surface runoff from each grid and routes these to the outlet point of the basin using a time-lag approach. Also, ModClark uses a soil curve number approach to simulate direct runoff. The study is conducted in the Napa River basin, California. The streamflow gage at St Helena (USGS 11456000, drainage area 78 sq. mi.), located in the upper reaches of the basin, is used as a control gage site as it is has minimal influences by reservoirs and diversions.
There are two main objectives for this study. First, we investigate the effect of spatially distributed rainfall input data on parameter estimation for the ModClark model. For this purpose, parameters are estimated by using MRMS and MPE data for a set of rainfall events by comparing their hydrograph results. To assess the impact of each rainfall data set on ModClark simulated stream flow, the derived parameters are used to simulate hydrographs corresponding to an independent set of rainfall input data which are different from the rainfall data used for estimating parameters. Second, this study examines the uncertainty arising from rainfall error due to the spatial distribution of the rainfall field and the accuracy of quantitative rainfall amount. To perform this analysis, we attempt to find the relationship between rainfall error and runoff error and to compare both runoff results from MPE and MRMS using ensemble rainfall input data derived from Monte Carlo simulations as an estimate of rainfall error. Preliminary results show that there is difference in the ModClark parameters resulting from the spatially distributed versus lumped rainfall input data. The difference in parameters also affected the rainfall-runoff simulation results. It is also shown that the rainfall error arising from the MRMS and MPE inputs becomes amplified the runoff error differently.