基于改进AM-MCMC的Nash模型参数不确定性分析-以挠力河流域为例

In order to improve the Adaptive sampling (AM) of the Markov Chain Monte Carlo (MCMC) simulation sample the amount of information, in this paper, Bayesian theory is used to produce improvements on the acceptance probability formula of AM sampling algorithm, to build AM-MCMC algorithm（BAM-MCMC）. Based on Bayesian theory, sample posteriori information can be imported to the differences of simulation sample, to increase the amount of MCMC simulation sample posteriori information. By extracting simulation samples of this algorithm can be analyzed the random distribution characteristics of sample variables. This research regards naoli river basin as example, establishes a BAM - MCMC algorithm is used to study the uncertainty of Nash model two parameters, and obtain posterior probability distribution of Nash model parameters of the survey region. The research results show the Nash model two parameters of naoli river basin conform to logarithmic normal distribution. The probability distribution of parameters can provide a theoretical basis for the uncertainty research of the parameters, also provide the parameter uncertainty information for the uncertainty research of subsequent Nash model forecast results.