基于多变量GAMLSS模型的洪水理论分布研究

Research on flood theory distribution based on multivariate GAMLSS model

为探讨水文数据非一致性对洪水频率分析的影响,提出基于位置、尺度、形状的广义可加模型(Generalized additive models for location,scale and shape,GAMLSS),从时间和降水两类因素出发,计算单变量、多变量洪水频率,分析经验点据与理论分位曲线拟合效果,选取不同变化条件下洪水适宜理论分布。以汤旺河流域为例,研究结果表明,单变量洪水频率最优分布选取较稳定,而受时间和降水因素影响,多变量洪水频率最优分布选取均不同。与前者相比,引入协变量使原序列参考时间连续性变化和降水极端信息,改进传统洪水频率计算方法。

In order to explore the influence of non-uniformity of hydrological data on flood frequency analysis,it was proposed to use generalized additive models for location,scale and shape(GAMLSS)based on location,scale and shape.Starting from the same factors,the univariate and multivariate flood frequency calculations were carried out,the fitting effect of the empirical data and the theoretical quantile curve was analyzed,and the appropriate theoretical distribution of floods under changing conditions was selected.Taking the Tangwanghe Basin as an example,the research results showed that the optimal distribution of univariate flood frequency was relatively stable,while the optimal distribution of multivariate flood frequency was different due to the influence of time and precipitation factors.Compared with the former,the introduction of covariates made the original series refer to the continuity of time and the extreme information of precipitation,which improved the traditional flood frequency calculation method.