水文时间序列变点分析的贝叶斯方法

建立了用于时间序列变点分析的贝叶斯数学模型，以此来研究随机水文时间序列均值的突变。该模型的核心部分是根据观测到的资料，通过蒙特卡洛马尔科夫链随机抽样的方法来估计变点位置的后验概率分布。对应着最大后验概率的位置被认为是发生变点的最可能位置。该方法在长江宜昌站的年径流资料系列上进行了应用。结果发现，在过去120a，长江宜昌站的年最小流量系列和年均流量系列的均值都极有可能存在着突变，而且其变化趋势都是均值明显减少。

Bayesian Method for Detecting Change-points of Hydrological Time Series

A Bayesian model for analyzing change-points of time series is established to study the abrupt change of the mean value of hydrological time series. Given the observed hydrological data, the model can estimate the posterior probability distribution of each location of change-point by using the Monte Carlo Markov Chain (MCMC) sampling method. The location with the largest posterior probability is regarded as the location of the most probable change-point. This Bayesian model has been applied to the annual discharge series of Yangtze River at Yichang hydrological station. The results show that, during the past 120 years, the mean values of both the annual minimum discharge series and the annual average discharge series are very likely to have changed abruptly, and the change tendency for both series is that the mean value has significantly reduced.