长江流域基于空间指标的GCMs模拟结果优选组合
Optimal combination of GCMs simulation results based on spatial metrics in the Yangtze River Basin
雷雅文 1顾磊2
作者信息
- 1. 武汉大学水资源工程与调度全国重点实验室,湖北武汉 430072
- 2. 华中科技大学土木与水利工程学院,湖北武汉 430074
- 折叠
摘要
提出一种针对长江流域全球气候模式(global climate models,GCMs)输出的降水及气温变量的优化组合方法.首先,利用7个空间指标对国际耦合模式比较计划第6阶段(couple model inter-comparison project phase 6,CMIP6)数据集的不同GCMs输出的降水量及最高、最低气温进行综合排序,并采用简单平均法、随机森林法对子流域及全流域的多GCMs组合进行排序,进而建立各变量GCMs优化组合.针对气候模式数据空间评价较少的问题,提出一种高精度的空间优化组合的融合方法,该方法将空间优化组合问题解耦为单模型排序和多GCMs优化组合的问题.在单模型排序后,基于当前排序结果构建一个组合优化问题,通过求解该问题得到模型优化组合.为GCMs提出了一种有效提高空间精度的融合方法,GCMs经随机森林法组合后对长江流域降水及气温空间模拟效果良好.
Abstract
This study proposes an optimized combination method for precipitation and temperature variables from global climate models(GCMs)outputs in the Yangtze River Basin.Firstly,the outputs of precipitation and maximum and minimum temperatures of different GCMs in the couple model inter-comparison project phase 6(CMIP6)data set are comprehensively sorted by using 7 spatial metrics,and multi-combinations of sub-basins and the whole basin are sorted by using simple average method and random forest method,and then GCMs optimization combinations of various variables are established.For the problem of less spatial evaluation of climate model data,a high-precision fusion method of spatial optimization combination is proposed,which decouples the spatial optimization combination problem into the problems of single-model sequencing and multi-GCMs optimization combination.After the sorting of single model,a combinatorial optimization problem is constructed according to the current sort results,and the model optimization combination is obtained by solving the problem.This study proposes an effective fusion method to improve the spatial accuracy of GCMs.The combination of GCMs with random forest method has a good effect on the spatial simulation of precipitation and temperature in the Yangtze River Basin.
关键词
空间指标/CMIP6气候模式/多GCMs组合/随机森林法Key words
spatial metrics/CMIP6 climate model/multi-GCMs combination/random forest method引用本文复制引用
基金项目
国家重点研发计划资助项目(2017YFA0603704)
湖北省自然科学基金资助项目(2020CFA100)
出版年
2024
NETL
NSTL
万方数据