Abstract: Flood forecasting for small and medium-sized catchments with an area smaller than 3000 km2 is a weak spot in the national flood forecasting system. For such catchments, it is imperative to improve accuracy and reduce uncertainty in flood peak forecasting. This paper examines the effects of distributed rainfall input and distributed routing simulation on accuracy and uncertainty in peak flow estimation through a case study of the Dongwan catchment of the upper Yiluo River. Based on the Xinanjiang model and Monte Carlo sampling method, we adopt four methods: lumped simulation, lumped rainfall-distributed routing simulation, distributed rainfall-lumped routing simulation, and distributed rainfall-distributed routing simulation. The results show that a distributed rainfall input improves overall accuracy in peak flow estimation, but it increases estimation errors greatly for rainfall events with extremely uneven spatial distribution is inputted. A distributed rainfall input contains parameters independent of each other and will raise uncertainty in flood peak flow calculations. Distributed routing simulations reduce overall uncertainty in flood peak calculation. The results are useful to flood forecasting for small and medium-sized catchments.

Key words: flood control, rainfall, routing, distributed, flood peak, accuracy, uncertainty