|国家预印本平台
首页|基于贝叶斯模型平均理论的水文模型合成预报研究

基于贝叶斯模型平均理论的水文模型合成预报研究

Study on the forecast combination of hydrological models based on Bayesian model averaging theory

中文摘要英文摘要

通过亚高斯模型分别对实测和水文模型预报的洪水序列进行正态分位数变换,并建立变换后的实测与预报时间序列的线性关系;然后根据贝叶斯模型平均理论,以实测序列隶属于某一水文模型的后验概率为权重,对各模型预报变量的条件概率密度函数进行加权,得到预报变量的概率密度函数,即高斯混合模型,从而实现了不同水文模型预报的合成及概率预报;最后,采用期望最大化算法估计高斯混合模型的参数。以密赛流域为实例,对本文的方法进行了验证。结果表明,基于贝叶斯模型平均的水文模型的合成预报不仅可以提供精度较高的均值预报,而且可以通过置信区间估计,定量评价模型预报的不确定性。

he meta-gaussian approach is used to make the flood time serious transformed into normally distributed variables for both the observation and hydrological model forecast, and the linear relationship between these two kinds of variables is established. According to the theory of Bayesian model averaging (BMA), the probability density function of forecast variable or the gaussian mixture model is built up by weighing the conditional probability distribution of individual hydrological model with the posterior probability that observations are belong to this specific hydrological model. Therefore, the forecast combination for hydrological models is obtained, and it is in a form of probabilistic forecast. The Expectation Maximum algorithm is adopted to estimate parameters in the gaussian mixture model. As an example, the proposed model is applied to the flood forecasting for Misai basins, Zhejiang Province. It indicates that the model can provide not only mean forecasting with higher precision, but also the estimation of confidence interval for the forecast by which the uncertainty of the forecasting can be quantitatively assessed.

余钟波、戴荣、梁忠民、王军

水利工程基础科学

贝叶斯模型平均水文模型亚高斯模型髙斯混合模型期望最大化算法预报不确定性

Bayesian model averaginghydrological modelmeta-gaussiangaussian mixture modelExpectation Maximum algorithmforecasting uncertainty

余钟波,戴荣,梁忠民,王军.基于贝叶斯模型平均理论的水文模型合成预报研究[EB/OL].(2008-10-08)[2025-08-23].http://www.paper.edu.cn/releasepaper/content/200810-128.点此复制

评论