Evaluation of the Impacts of Precipitation Uncertainties on Flood Magnitude and Timing

Precipitation is arguably the major uncertainty source in flood modelling. The quantitative precipitation estimation (QPE) products when used as input for hydrologic modelling can cause significant difference in model performance because of the large variations in their estimation of precipitation intensity, frequency, duration, and spatial extent. Objectively evaluating QPE and deriving the best estimation of precipitation at river basin scale, represent a long-standing problem faced by the hydrometeorological community, despite they are crucial for many researches including flood simulation, such as the Global Flood Monitoring System (GFMS) using aDominant river tracing-Routing Integrated with VIC Environment (DRIVE) model with real-time satellite precipitation as input (Wu et al.,2014). Recently we developed a Multiple-product-driven hydrological Modeling Framework (MMF) for objective evaluation of QPE products using hydrological approaches and hydrological models (Wuet al., 2017). In this presentation, we will share the recent results we derived with the MMF, specifically on (1) how precipitation bias causes flood magnitude differences among different QPEs and how the spatial distribution of P-Q bias relation appears as a function of various hydroclimate factors; and (2) how the different precipitation spatial patterns delineated in QPE products are translated into different flood peak timing, through a linear routing scheme which is employed to decompose the contribution of flood peak during rain-flood events. The results and analysis help better understand the real-time global flood prediction performance and their usage, such as the GFMS with satellite precipitation as input, particularly in wide remote areas where there is no ground observation available.