Abstract: Synthetic simulation of streamflow series can afford a reliable basis for planning and management of a water resources system. Previous nonparametric disaggregation models cannot preserve a back and forth auto-correlation structure of variables and may lead to boundary problems of generating a large number of negative values of flow. In this paper, we apply the kernel density estimation theory to develop an improved nonparametric disaggregation model of monthly streamflow through a comprehensive consideration of the relationship among annual runoff, previous first-order monthly runoff, and simulated monthly streamflow, using the variable kernel bandwidth method to reduce boundary impact. This model is compared with the traditional relative disaggregation model, and used for simulating the natural monthly runoffs at Lanzhou station in the Yellow basin for the period of 1932 to 2010. The results show it can better preserve the probability distribution and statistical parameters of the original runoff series, overcome the inconsistency in the back and forth auto-correlation structure, and avoid generating negative flows, thus being applicable to stochastic simulations of monthly streamflow.

Key words: monthly runoff, stochastic simulation, kernel density estimation, improved nonparametric disaggregation model, variable kernel, correlation disaggregation model