Abstract: The simulation of runoff probability in an area in lack of runoff data is a difficulty in hydrological research. In this article, we try to establish the probability density function of monthly runoff flow by adopting kernal density estimation method, and give the solution by Markov Chain Monte Carlo (MCMC) simulation method based on Adaptive Metropolis (AM) algorithm. Case study shows that the AM-MCMC algorithm model based on kernel density estimation is of high accuracy and good application value. It can be used in areas in lack of data.

Key words: runoff simulation, probability distribution, kernel density estimation, AM-MCMC algorithm, Luo’ao Reservoir