Evaluation of CMIP6 model performance in simulating the PDO and its future change

Abstract

This study comprehensively assesses the model performance in simulating the variation of the Pacific Decadal Oscillation (PDO) in the winter time (November to March) during the historical period of 1870–2014 using 40 models from phase 6 of the Coupled Model Intercomparison Project (CMIP6). Its future change is also investigated under four Shared Socioeconomic Pathway scenarios on the basis of 17 CMIP6 models. Results show that CMIP6 models can reasonably reproduce the observed sea surface temperature anomaly mode associated with the PDO, with high spatial correlation coefficients of above 0.7. However, CMIP6 models generally perform poorly in simulating the variability of the PDO, with low correlation coefficients for most models, except FGOALS-g3 (0.57), GISS-E2-1-H (0.42), and NESM3 (0.37), which present a relatively greater correlation. The multi-model ensemble (MME) outperforms the individual models and can capture the dominant periodicity of the PDO (∼50 years), but it fails to capture the periodicity of ∼20 years. Additionally, the MME result underestimates the variability of the PDO. In response to future warming, the PDO is expected to transfer from a negative to positive phase around the 2050s. The 50-year periodicity of the PDO is projected to decrease compared to the historical period.摘要本文基于40个CMIP6模式输出结果, 系统评估了模式对历史时期太平洋年代际振荡(PDO)的模拟性能, 并利用其中17个模式的4种共享社会经济路径情景数据预估了PDO未来可能变化趋势. 结果表明, CMIP6模式能够合理再现PDO相关海表温度异常的空间模态, 但模式模拟PDO位相演变的能力普遍较弱. 多模式集合能够合理再现PDO的50年左右周期, 但无法模拟出其20年左右周期, 并且低估了PDO的变化幅度. 在未来变暖情景下, PDO可能在2050左右出现由负位相向正位相的转变, 同时其50年左右周期也将明显缩短.