基于空间相关性和非平稳性的淮河流域极端气候事件时空特征研究

Extreme climate events happen frequently under the background of climate change, which led to serious impact on ecological environment and socio-economy. Studying the temporal and spatial changes of extreme climate events has become hot issues in the field of hydrometeorology. The project plans to get scientific characterization of the typical extreme climate events based on calculating and evaluating the applicability of various extreme indices in Huaihe River Basin, China. The spatial dependence of climatic extremes is fully considered in this project. The Max-stable process will be established by taking the Generalized Extreme Value Distribution (GEV) as the marginal distributions for extremes of Annual Maxium (AM) time series, and the elevation, longitude, latitude, and distance from coastline as the covariate. For the extremes of Peak Over Threshold (POT), Bayesian hierarchy models are constructed to describe the spatial and temporal changes of climate extremes. For multivariate analysis of extreme climate events, the Mann-Kendall test, CUSUM test, and Sliding F-test are used to test the stationary of the climate extremes. Additionally, we will allow non-stationary modeling of changes in the marginal distribution parameters in Copula. The results are illustrated by calculating the two-dimensional and three-dimensional joint probability of the intensity, frequency and duration. The combinative probability, conditional probability and the corresponding return period also will be provided. By using moving window and GIS technology, the spatial and temporal evolutions of multivariable Copula analysis for the extreme climate events will be revealed. This study will strengthen the understanding in evolution of extreme climate events in the typical basin of the North-South climate transition zone, and provide theoretical basis for regional water resources management as well as disaster prevention and mitigation.

气候变化导致极端气候事件频发，给生态环境和社会经济带来严重影响。极端气候事件的时空变化过程已成为水文气象领域的研究热点。本项目拟对淮河流域极端气候事件指数计算及适用性评价基础上，实现对极端降水和极端气温的科学表征；充分考虑极端气候事件的空间相关性，利用Max-stable过程模型和贝叶斯层次模型探究极端气候事件时空分布特征。同时，将非平稳特性引入到极端气候事件边缘分布函数中，利用Copula函数获取极端气候事件不同特征（频率、强度、持续时间）之间的二维、三维联合分布，计算不同特征的组合概率、条件概率和相应的重现期，揭示极端气候事件Copula多变量时空演变规律。该研究有望深化对南北气候过渡带典型流域极端气候事件演变过程的理解，同时可以为流域水资源管理、区域防灾减灾提供理论支撑。

气候变化导致极端气候事件频发，给生态环境和社会经济带来严重影响。本项目对淮河流域极端气候事件指数计算及适用性评价基础上，充分考虑极端气候事件的空间相关性，利用Max-stable过程模型和贝叶斯层次模型探究极端气候事件时空分布特征。构建非平稳性判别体系，分析极端气候事件非平稳特征，揭示其Copula多变量时空演变规律。结果如下：1）冷极值呈显著下降趋势，暖极值表现为波动上升趋势；日较差呈显著下降趋势，降水极值在全流域未表现出一致上升或下降趋势，且变化趋势在全流域均不显著。 2）充分考虑极端气候事件的空间相关性，以广义极值分布为边际分布，将经度、纬度、海拔、GDP和气候指标为协变量，构建空间贝叶斯层次模型和Max-stable 模型，研究表明所建模型可获得没有观测台站的极端气候重现水平。3）充分考虑空间相关性和边缘分布函数非正态性对计数极值的影响，将Poisson分布作为边缘函数建立贝叶斯层次模型，剖析淮河流域计数极值的空间分布规律，结果表现建立的空间贝叶斯层次模型可以很好的拟合观测数据变化，且增加协变量可以改善模型模拟效果，但过多增加与重现水平相关性较低的协变量将降低模型模拟效率。 4）构建淮河流域极端气候事件非平稳判别体系，采用非平稳GEV和GPD解读极端气候事件的空间分布特征。同时，将EOF时间系数作为GAMLSS模型的分析对象，揭示流域尺度极端气候事件的非平稳性特征。结果表现气温极值为非平稳性变化，且CO2变化是导致极值非平稳的原因。GAMLSS模型可以模拟气候极值序列位置、尺度参数的平稳和非平稳性变化，与EOF结合，可以减少站点尺度非平稳分析的不确定性。5）采用Copula函数对淮河流域极端气候事件进行联合概率特征研究，结果表现GEV和PIII为气候极值的最佳边缘分布函数。最优Copula模型在不同站点和气候极值之间差异较大，不同Copula函数对同现重现期影响较大，而对联合重现期影响相对较小。该研究深化了对南北气候过渡带典型流域极端气候事件演变过程的理解，同时可以为流域水资源管理和区域防灾减灾提供理论支撑。