项目名称: 空间极值模型的贝叶斯推断及其在气候变化政策中的应用研究
项目编号: No.41301421
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 天文学、地球科学
项目作者: 曾惠芳
作者单位: 湖南科技大学
项目金额: 25万元
中文摘要: 本项目将利用贝叶斯方法和MCMC抽样技术提出一类新的空间相依性极值模型,包括贝叶斯区间空间极值模型,贝叶斯门限空间极值模型,以及贝叶斯时空极值模型,实现对空间极值相依性和空间边际效用的有效刻画。利用Dirichlet 先验核构建具有极大稳定性的有限维高斯极值过程,解决空间极大稳定过程似然函数不存在的问题,可以实现对空间极值模型的贝叶斯推断。利用近似贝叶斯计算方法构建综合DIC信息准则以及RJMCMC算法实现对模型的选择。考虑到模型参数的不确定性,利用状态空间理论和粒子滤波算法实现对空间极值模型的贝叶斯预测,提高模型的预测精度。进一步,利用贝叶斯空间极值模型研究气候变化政策问题。
中文关键词: 空间极值;贝叶斯推断;气候变化;不确定性;
英文摘要: In this proposal, we will build some novel spatial extreme models based on Bayesian method and MCMC algorithm to modeling the dependence of extremes and their marginal distribution in time and space model.These models include Bayesian block maxima spatial extreme models,threshold spatial extreme models and spatio-temperal extreme models.An approximation to the Gaussian extreme value process(GEVP)based on Dirichlet kernal is proposed, which can solve the problem that the related likelihoods are unavailable in parametric spatial extreme model.Moreover,we employ appproximate Bayesian computing method to build composite DIC information criteria and RJMCMC algorithm to accomplish the model selection.In addition,we carry out the model forecasting based on state space theory and particle filter method.Since the method consider the parameters is a random variable,which can improve the forecast accuracy.Finally, we employ the spatial extreme model to analyze the climate change policy.
英文关键词: Spatial extreme values;Bayesian inference;Climate change;Uncertainty;