CMIP5 Climate Multi-model Ensemble Optimization Based on Spatial-temporal Distribution

ZUO Zhengkang1,2, ZHANG Feizhou2, ZHANG Ling1,3, SUN Yiyuan2, ZHANG Ruihua2, YU Tian4, LU Jianzhong1,†

1. State Key laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079; 2. Institute of Remote Sensing and Geographic Information System, Peking University, Beijing 100871; 3. Department of Atmospheric Science, School of Environmental Studies, China University of Geoscience, Wuhan 430078; 4. Information Engineering College, Capital Normal University, Beijing 100048; † Corresponding author, E-mail: lujzhong@whu.edu.cn

Abstract The multi-mode ensemble based on spatiotemporal distribution is constructed to reduce the uncertainty of a single-model and the non-uniform distribution of the traditional model ensembles. The improved genetic algorithm is employed to optimize the multi-model ensemble of CMIP5 global climate data from temporal and spatial scales, and Taylor diagram is used to evaluate its simulation performance. The experimental results show that the multi-mode ensemble based on spatiotemporal distribution is superior to the traditional equal weight multimode ensemble scheme. Key words multi-mode ensemble; optimization; genetic algorithm; spatial-temporal distribution; CMIP5

气候变化是当今国际社会普遍关注的全球性问题[1], 全球气候模式(global climate model, GCM)数据是气候模拟和预估气候未来变化的重要工具[2]。其中, CMIP5模式是迄今为止时间最长且内容最全的气候变化模式资料库, 广泛应用于气温[3]、降水[4]、径流[5]以及海温变化[6]等模拟能力评估和预估。CMIP5模拟的是简化的真实的气候系统, 会造成结果的不确定性[7]。为克服单一模式模拟精度有限、空间和时间上分布不均等问题, 学者们提出多

模式集合。Kharin等[8]和 Kang等[9]提出等权集合方案(EE), 但未考虑集合成员间的不确定性差异, 而是赋予所有成员相同的权重。由于动力气候模式具有系统误差, Peng 等[10]和 Ke等[11]提出先消除模式的系统误差, 再等权集合的方法(Cali-ee)。因不同

[12]集合成员的预测能力不同, Krishnamurti 等 提出基于多元线性回归分析, 赋予集合成员不同权重的多模式集合方法(MLR)。以上 3 种基于经典统计学的多模式集合方案主要用于确定性预测。近几年出

Fig. 1

gi ( X )  X ),其他, i  qh ( X )。i 1 i 0,

R   f

 r

图 1 Poyang Lake Basin[21]