Attribution analysis for the failure of CMIP5 climate models to simulate the recent global warming hiatus
Attribution analysis for the failure of CMIP5 climate models to simulate the recent global warming hiatus
复制标题
DOI：
10.1007/s11430-015-5465-y
复制
发表时间：
2017-02
期刊：
影响因子：
--
通讯作者：
Meng Wei;F. Qiao
中科院分区：
文献类型：
--
作者：
Meng Wei;F. Qiao
耦合模式相互比较项目第五阶段（CMIP 5）包含一组最先进的气候模式，代表了迄今为止最高水平的气候模拟。然而，这些模型显著高估了2006-2014年期间的全球平均地表温度（GMST）。基于集合经验模态分解（EEMD）方法，分析了1850-2014年HadCRUT 4观测到的GMST时间序列的长期变化，并对33个CMIP 5气候模式模拟的GMST进行了评估。气候模型未能预测最近全球变暖中断的可能原因被揭示出来。结果表明：1850-2014年，全球平均气温在百年时间尺度上呈波动上升趋势，并受长期趋势和年代际变率（MDV）的影响。从世纪初开始，长期趋势相对稳定，过去50年的平均变暖率为0.0883°C/十年。虽然MDV（周期约为65年）在1850-2014年期间显示了2.5个数十年波，随着时间的推移而加深和陡峭，但20世纪最后25年令人震惊的变暖是长期变暖趋势和MDV变暖阶段同时发生的结果，两者都占1975-1998年期间温度升高的三分之一。最近，随着MDV接近其第三个峰值，全球变暖的减缓出现，导致变暖速度降低。将HadCRUT 4记录的GMST时间序列与33个CMIP 5模式的输出结果进行对比分析，结果表明模式能够较好地再现1850-2005年历史模拟期的GMST，尤其是对20世纪后25年全球变暖加速的模拟。然而，根据RCP 4.5设想方案预测的2006-2014年期间的全球平均茎干数及其线性趋势明显高于观测值。这是因为CMIP 5模型混淆了MDV和GMST时间序列的长期趋势，导致了快速的长期趋势和具有不规则相位和小幅度的不正确MDV。这意味着大气CO2在全球变暖中的作用可能被高估，而气候系统内部振荡MDV可能被低估，这可能与对关键气候内部动力过程认识不足有关。我们的研究为新一代气候模式提出了一个重要的标准：它们应该能够模拟GMST的长期趋势和MDV。
The Coupled Model Inter-comparison Project Phase 5 (CMIP5) contains a group of state-of-the-art climate models and represents the highest level of climate simulation thus far. However, these models significantly overestimated global mean surface temperature (GMST) during 2006–2014. Based on the ensemble empirical mode decomposition (EEMD) method, the long term change of the observed GMST time series of HadCRUT4 records during 1850–2014 was analyzed, then the simulated GMST by 33 CMIP5 climate models was assessed. The possible reason that climate models failed to project the recent global warming hiatus was revealed. Results show that during 1850–2014 the GMST on a centennial timescale rose with fluctuation, dominated by the secular trend and the multi-decadal variability (MDV). The secular trend was relatively steady beginning in the early 20th century, with an average warming rate of 0.0883°C/decade over the last 50 years. While the MDV (with a ~65-year cycle) showed 2.5 multi-decadal waves during 1850–2014, which deepened and steepened with time, the alarming warming over the last quarter of the 20th century was a result of the concurrence of the secular warming trend and the warming phase of the MDV, both of which accounted one third of the temperature increase during 1975–1998. Recently the slowdown of global warming emerged as the MDV approached its third peak, leading to a reduction in the warming rate. A comparative analysis between the GMST time series derived from HadCRUT4 records and 33 CMIP5 model outputs reveals that the GMSTs during the historical simulation period of 1850–2005 can be reproduced well by models, especially on the accelerated global warming over the last quarter of 20th century. However, the projected GMSTs and their linear trends during 2006–2014 under the RCP4.5 scenario were significantly higher than observed. This is because the CMIP5 models confused the MDV with secular trend underlying the GMST time series, which results in a fast secular trend and an improper MDV with irregular phases and small amplitudes. This implies that the role of atmospheric CO2in global warming may be overestimated, while the MDV which is an interior oscillation of the climate system may be underestimated, which should be related to insufficient understanding of key climatic internal dynamic processes. Our study puts forward an important criterion for the new generation of climate models: they should be able to simulate both the secular trend and the MDV of GMST.