Evaluation of Quantitative Precipitation Estimations (QPE) and Hydrological Modelling in IFloodS Focal Basins

Abstract:

A hydrology approach based on the intercomparisons of multiple-product-driven hydrological simulations was implemented for reliably evaluating both quantitative precipitation estimations (QPE) and hydrological modelling at river basin and subbasin scales in the IFloodS focal basin, Iowa-Cedar River Basin (ICRB), over a long-term (2002-2013) and a short-term period (Apr. 1-June 30, 2013). A reference precipitation dataset was created for the evaluation first by reversing the mean annual precipitation from independent observed streamflow and satellite-based ET product and then it was disaggregated to 3-hourly time steps based on NLDAS2. The intercomparisons from different perspectives consistently showed the QPE products with less bias leaded to better streamflow simulation. The reference dataset leaded to overall the best model performance, slightly better than original NLDAS2 (biased -4%) which derived better model performance than all other products in the long-term simulations with daily and monthly NSC of 0.81 and 0.88 respectively and MARE of -2% when compared to observed streamflow at the ICRB outlet, while having reasonable water budgets simulation. Other products (CPC-U, StageIV and TMPARP and satellite-only products) derived gradually decreased performance. All products with long-term records showed consistent merit over the IFloodS period, while the Q2 seemed to be the best estimation for the short-term period. Good correlation between the bias in precipitation and streamflow was found at all from annual to daily scales while the relation and the slope depended on seasons, river basin concentration (routing) time and antecedent river basin water storage condition. Precipitation products also showed significant impacts on streamflow and peak timing. Although satellite-only products could derive even better simulations (than conventional products) for some sub-basins from the short-term evaluation, it had less temporal-spatial quality consistency. While the new GPM IMERG data are expected to bring further improvement of the Global Flood Monitoring System (GFMS), such streamflow-precipitation bias relationship can provide GFMS users of uncertainty estimation for ongoing flood events when real-time validation information is not available.