hide
Free keywords:
-
Abstract:
Parameter optimization is the effective way to reduce parameter uncertainty of hydrological models, but the traditional way used for lumped hydrological models needs a series of hydrological data, such as data from 10-year or 20 flood events or more, this requirement usually cannot be satisfied in the data-scarce watershed. The distributed hydrological model derives parameters directly from the terrain properties, such as land use/cover type, soil type, but current studies has showed the uncertainty is still high. In modeling the flash flood process in the data-scarce watershed not only faces the challenge of huge computationally load in parameter optimization, but also faces the hydrological scarcity. In this study, a hypothesis is proposed, i.e., the parameters of a distributed hydrological model could be optimized by using data only from one hydrological event, which is based on the precondition that the parameters of these kinds of model have physical meaning, and could be derived from the terrain properties related to the parameters. This hypothesis is tested in several cases in china for flash flood modelling, including data-scarce and data-plenty watersheds. The distributed hydrological model is Liuxihe model that is proposed for watershed flood simulation and prediction, which developed a parameter optimization scheme by using PSO algorithm and is implemented in a super-computer. The study results show that data from one flood event is enough for parameter optimization, and the model performance is reasonable, more data used could not improve model performance.
