Precipitation Uncertainty and its Impacts on Hydrologic Modelling and Flood Prediction: An Investigation in IFloodS Focal Basins

Thursday, 18 December 2014
Huan Wu1, Robert F Adler2, Yudong Tian3 and George John Huffman3, (1)ESSIC/NASA GSFC, College Park, MD, United States, (2)University of Maryland College Park, College Park, MD, United States, (3)NASA Goddard Space Flight Center, Greenbelt, MD, United States
The purpose of this study is to investigate precipitation uncertainty and to understand its impact on flood estimation, toward to further improve the accuracy of a real-time Global Flood Monitoring System (GFMS, http://flood.umd.edu/). We investigate the precipitation uncertainty impacts on streamflow calculations through the Dominant river tracing-Routing Integrated with VIC Environment (DRIVE) model (the core of the GFMS) with a set of real precipitation products. These precipitation products include satellite-based, ground radar based, gauge based and reanalysis (with gauge data) based precipitation estimations. There are significant differences in the mean annual, seasonal, and IFloodS period precipitation estimation among the existing precipitation products. To estimate the bias in each precipitation products, a “reference” of mean annual precipitation was created. The simulated streamflow was compared to the observation, while the upstream basin-area-averaged mean annual precipitation at each location with USGS gauge observation was derived to compare with the “reference”.

The sensitivity of the DRIVE model performance in reproducing streamflow to precipitation inputs, clearly showed that better precipitation inputs (with less bias) tend to result in better streamflow simulations. This indicates better accuracy of the GFMS is expected when improved satellite-based real-time precipitation products are available through the GPM in near future. Real-time Satellite-based precipitation resulted in overall lower model performance scores than the conventional precipitation products and gauge-adjusted remote sensing precipitation estimations. However, the lower scores are mainly attributed to the inconsistency of precipitation estimation in a few extreme events. While precipitation bias mainly contributes to the over/underestimation of the flood magnitude, the DRIVE model tends to lead to faster flood waves for relatively larger floods in downstream part of the Iowa and Cedar River basin. Model calibration or a better representation of the overbank flow (for floodplain) is expected to further improve the flood timing estimation.