Progress in the Attribution of Climate Warming in China for the 20th Century

Letters Article ID: 1673-1719 (2007) Suppl.-0082-05 Progress in the Attribution of Climate Warming in China for the 20th Century Zhou Tianjun 1, Zhao Zongci 2 1 LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China; 2 National Climate Center, China Meteorological Administration, Beijing 100081, China Abstract: Progress in the attribution of climate warming in China for the 20th century is summarized. Three sets of climate model experiments including both coupled and uncoupled runs have been used in the attribution analyses. Comparison of climate model results with the observations proves that in the 20th century, especially in the recent half century, climate warming in China is closely related to the increasing of the anthropogenic emissions of greenhouse gases, while sulfate aerosol should also have contributions. When both external forcing and natural forcing agents are prescribed, coupled climate models have better results in producing the observed variation of temperature in China. The role of oceanic forcing is also emphasized in the attribution analyses. The observed climate warming of China in the 1920s could not be reproduced in any set of climate model simulations. Key words: 20th century; surface air temperature; China; attribution analysis; climate warming Introduction Global warming has large impacts on environment. Climate disasters such as flood, drought and frequent extreme climate events have led to increasingly serious economic losses. As noted in the IPCC Third Assessment Report (TAR) [1], the 20th century, especially recent 50 years, global warming might be related to increase in the anthropogenic emissions of greenhouse gases. In the past decades, much effort has been devoted by Chinese climate research community to the reconstruction of regional average temperature time series, and the diagnosis and numerical model simulation of past climate change [2]. However, mechanisms responsible for the change remain to be open questions. The main motivation of this paper is to summarize some progresses in numerical simulations of the 20th century Chinese climate, and to make attribution analyses of the warming trend by comparing different sets of model results. 1 Model experiments The outputs of three sets of climate model experiments were used in the comparison. The first set of experiments (E1) came from the IPCC TAR [3]. Impacts of anthropogenic greenhouse gases (GHGs) and sulfate aerosols were included, and major global ocean atmosphere coupled land sea-ice models were involved in E1. In addition to global coupled models, regional climate models were also used in E1. The focus of our discussion on E1 was the simulation of 20th century climate. The second set of experiments (E2) came from The 20th Century Climate in Coupled Models (20C3M) project of IPCC Fouth Assessment Report (AR4), which was set up to address the factors governing the evolution of 20th century climate [4]. In 20C3M, the 20th century climate was simulated with various combinations of forcings including GHGs, sulfate aerosols, ozone, volcanic aerosols and solar variability. These experiments were performed by using Received: December 13, 2006; revised: February 16, 2007 Corresponding to Zhou Tianjun (E-mail: zhoutj@lasg.iap.ac.cn) 2007 National Climate Commission of China. All right reserved 82

global atmosphere-land-ocean-sea ice coupled system model. The change of land use was also included in several models. The integration of these models most started from 1860, and the focus of E2 was the century climate covering 1880J1999. Results from the nineteen coupled models of 20C3M were used in the analyses. The third set of experiments (E3) was an Atmospheric Models Intercomparison Project (AMIP) type simulation covering 1872J2002. In E3, the monthly sea surface temperature and sea ice data from Hadley Center [5] were used to drive the NCAR CAM2 model [6]. An ensemble simulation with 12 realizations was finished under different initial conditions. The E3 was used to reveal the contribution of sea surface temperature and sea ice to the 20th century evolution of surface air temperature (SAT) over China. In addition, for the purpose of model validations, the observed 20th century SAT averaged over China was used, which was reconstructed by Wang et al. [2]. 2 Attribution analyses 2.1 The results of E1 Only forcings of GHGs and sulfate aerosols were prescribed in E1, the simulated annual SAT anomaly averaged over China during the 20th century are shown in Fig.1. All models show reasonable performances in producing the recent half century warming trend. The results of multi-model ensemble mean are closer to the observations, and the ensemble mean 20th century SAT of about 40 models has a correlation coefficient of 0.47 with the observations. The linear trends of the ensemble mean and the observations are 0.71 /100 a, and 0.90 /50 a, respectively. This correspondence indicates that the climate warming in China in the past half century is closely related to increase in GHGs. In the mean time, we should note that hardly any model could reproduce the warming during the 1920sJ1940s. 2.2 The results of E2 The evolutions of the 20th century annual mean SAT anomalies (relative to 1961J1990) simulated by nineteen 20C3M coupled models are shown in Fig. 2. There were large spreads among the simulations of the models before the 1970s, and hardly any model could produce the warming peak between the 1920s and the 1930s, however after the 1970s all simulations converged toward the observations. Calculated correlation coefficients between simulations and observations for different models indicate that simulation results are significantly positive correlated with the observation in terms of annual mean temperature in China [4]. However, among the nineteen models, the correlation coefficient for UKMO-HadCM3 model is low, and both FGOALS_g1.0 and CSIRO-Mk3.0 models have negative correlations. The unacceptable results of UKMO-HadCM3 and CSIRO-Mk3.0 models are attributed to the absence of some forcing agents such as aerosols. The bias of FGOALS_g1.0 model is caused by the poor performance of the model in simulating sea ice. The ensemble mean of the other 12 models has a correlation coefficient of 0.56 with the observations, which is lower than the global/ 3 2 1 0 J1 J2 J3 J4 1900 1920 1940 1960 1980 2000 Year (1900J2002) HADL-GG CCSR-GG IAP2-GG HADL-GS CCSR-GS IAP2-GS CCSRA1 CSIROA2 GFDLA2 GCM7-GG NCCIAPB2 GFDL-GG CCC-GG RCN-GG GFDL-GS CCC-GS RCN-GS CCSRA2 CSIROB2 GFDLB2 GCM7-GS OBSWG DKRZ-GG NCAR-GG IAP1-GG DKRZ-GS NCAR-GS CCCA2 CCSRB1 ECHA2 HADLA2 GCMA2B214 GCMA2 CSIRO-GG NCC95GG YONU-GG CSIRO-GS NCC95GS CCCB2 CCSRB2 ECHB2 HADLB2 NCCIAPA2 GCMB2 Fig. 1 Temporal evolutions of the annual mean temperature anomalies in China for the 20th century by E1 (about 40 climate models) and the observations (black) (relative to 1961J1990) (After Zhao et al. [3] ) 83

Zhou Tianjun et al.: Progress in the Attribution of Climate Warming in China for the 20th Century 2.0 1.5 1.0 CGCM3.1(T47) CCSM3 CNRM-CM3 CSIRO-MK3.0 ECHAM5/MPI-OM GFDL-CM2.0 GFDL-CM2.1 GISS_AOM GISS_ER UKMO-HadCM3 FGOALS_g1.0 INM_CM3.0 IPSL-CM4 MRI-CGCM2.3.2 MIROC3.2(medres) BCC-CM1 BCCR_BCM2.0 CGCM3.1(T63) UKMO-HadGEM1 Observation Ensemble 0.5 0.0 J0.5 J1.0 J1.5 J2.0 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 Year (1880J1999) Fig. 2 Temporal evolutions of the annual mean temperature anomalies in China for the 20th century by E2 (the IPCC AR4 20C3M) and the observations (black) (relative to 1961J1990) (After Zhou et al. [4] ) Northern Hemisphere mean temperature (0.87/0.82), respectively, but still statistically significant at the 95% confidence level. The correlation coefficient between regional mean SAT simulations and observations for North China (102.5 J122.5 E, 35 J43 N) is 0.39, which is higher than 0.33 for South China (102.5 J122.5 E, 22.5 J 35 N). The multi-model ensemble mean is clearly better than the result of any individual model, although the amplitude of anomaly series from the ensemble mean is obviously weakened than those from individual models. Analyses of variance indicate that the external forcing of model including both natural and anthropogenic forcing agents explains 37.6% of the total variance, while the internal noise explains another 62.4% [4]. Variances explained by the external forcing and internal noise are 66.9% and 33.1% respectively, for the global mean temperature, and 60.0% and 40.0% for the Northern Hemispheric average, respectively. Hence for the annual mean temperature in China, the spread among the models is larger than those for the global/northern Hemisphere mean conditions [4]. For the warming trend of the 20th century (1880J1999), the result of E2 ensemble mean is 0.70 /100 a, which is close to the observational value of 0.71 /100 a. The change of solar constant in the past century is small. There are debates on the role of solar variability in addressing climate change in the international climate research community. The contribution of solar variability to regional climate change is more ambiguous. The 1880J 1999 periods are divided in this paper into 1880J1940 part and 1941J1999 part and it is found that for the annual mean SAT in China, models with the inclusion of solar variability, e.g. the CCSM3, CNRM-CM3, GFDL-CM2.0, GFDL- CM2.1, and MRI-CGCM2.3.2 models, generally have significant positive correlations with the observations for 1880J1940. This indicates that the climate change in China in the first-half part of 20th century was at least partly affected by natural forcings. 2.3 The results of E3 The results of NCAR CAM2 model forced by the observational sea surface temperature and sea ice from Hadley Center are shown in Fig. 3. Although there are discrepancies between simulations and observations, the model results correspond to observations, such as the early stage cooling, the 1940s warming, the 1970s cooling and the recent warming starting from the 1980s. Among the 12 realizations, each individual simulation has significant positive correlation with the observations at the 95% confidence level. Employing ensemble mean clearly improves the simulation. The ensemble mean of 12 realizations has a correlation coefficient of 0.50 with the observations, which is comparable to the multi-model ensemble mean of 20C3M experiments. Hence a large part of the observational variation of SAT in China comes from oceanic forcing. The source of this oceanic forcing should also be solar variability and GHGs. However, the linear warming trend of 0.33 /100 a over 1880J1999 for the 84

Table 1 Correlation coefficients between three ensemble experiments (E1, E2, E3) and the observations for the annual mean temperature of the 20th century in China and their linear trends [3] Experiment Correlation coefficient Linear trend 20th century recent 50 years 1960J1970 1980J1999 E1 0.47 0.71 /100 a 0.90 /50 a J0.07 /10 a 0.38 /10 a E2 0.55 0.70 /100 a 0.85 /50 a J0.06 /10 a 0.24 /10 a E3 0.50 0.33 /100 a 0.01 /50 a J0.12 /10 a 0.21 /10 a Observation 0.4J0.8 /100 a 0.5J0.9 /50 a J0.55 /10 a 0.53 /10 a G =The correlation coefficient of E2 is for the ensemble mean of 16 coupled models, while the linear trend is for the ensemble mean of 19 models 1.2 1.0 0.8 0.6 0.4 0.2 0.0 J0.2 J0.4 J0.6 J0.8 J1.0 J1.2 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Year (1872J2002) Fig. 3 Temporal evolutions of the annual mean temperature anomalies in China for the 20th century by E3 (different members) and the ensemble mean (black) (relative to 1961J1990) (After Zhou et al. [4] ) ensemble mean of E3 is weaker than both the observational value of 0.71 /100 a and the E2 result of 0.70 /100 a, and the recent warming trend during 1949J1999 is also weak. Analyses of variance show that SST forcing explains 37.5% of the total variance, while the contribution of internal noise is 62.5% [4]. The observed warming during the 1920s can not be found in any of individual realizations, indicating that this warming might not be explained in terms of internal noise. Note that even when forced by the combination of natural forcing agents and anthropogenic GHGs, none of the 19 coupled models in E2 could reproduce this warming period, the attribution of cause for the 1920s warming is still an open question. Correlation coefficients between three sets of experiments and the observations for the annual mean SAT in China for the 20th century, and simulated and observed linear trends are shown in Table 1. Except the general warming trend, the most prominent feature of the annual SAT in China during the 20th century was a cooling during 1960J1970 and an accelerated warming starting from the year 1980. The 1960J1970 cooling is very weak in E1. The E2 experiments reproduce a 1960J1970 cooling with a linear trend (J0.06 /10 a), which is still much weaker than the observed value of J0.55 /10 a. Among three sets of model experiments, the trend of E3 (J0.12 /10 a) is relatively close to the observations. The decadal scale cooling/warming oscillations of temperature in China starting from the year 1880 is apparent in E3, but with a delay of 10J15 years in their phases, and the huge thermal 85

Zhou Tianjun et al.: Progress in the Attribution of Climate Warming in China for the 20th Century inertia of oceans should be responsible for this delay. 3 Concluding remarks (1) When the anthropogenic GHGs and sulfate aerosol forcing agents were prescribed in climate models, the 20th century warming of China can be reproduced reasonably, and the simulated warming trend of recent half century is comparable to the observations. (2) When forced by natural factors such as variations of solar constant and anthropogenic factors such as GHGs, many coupled climate models show reasonable performances in reproducing the 20th century variation of temperature in China. (3) There is evidence indicating that natural forcing agents contributed to the variations of temperature in China during the first half of 20th century, since models with the inclusion of solar constant and volcanic aerosol variations generally reproduced major characters of annual mean temperature in the first half century. The recent half century warming of China is mainly resulted from increase in GHGs, for all models with the inclusion of GHGs have significant positive correlations with the observations. The impact of GHGs on annual mean temperature in northern China is larger than that in southern China. The outputs of 20C3M experiments closely resemble the observations in the warming trends of either the 20th century or recent half century. (4) When forced by observed SST, the NCAR CAM2 model could partly produce the observed 20th century variation of temperature in China, and the ensemble mean of 12 realizations is comparable to that of coupled models. The 20th century warming trend of the SST-forced simulations is, however, obviously weaker than the observations, and the warming trend for recent half century is also very weak. Nevertheless, the model has reasonable performances in producing the cooling of 1960J1970. (5) Both CAM2 simulation with prescribed SST and coupled model simulations with the inclusion of natural and anthropogenic forcing agents failed in producing the 1920s warming of China. Acknowledgements The international modeling groups is acknowledged for providing their data for analysis, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and archiving the model data, and Mr. Wen Xinyu for performing the SST-forced CAM2 integration on the Lenovo Deep Comp 6800 supercomputer at the Supercomputing Center of the Chinese Academy of Sciences. This work was jointly supported by the International Partnership Creative Group entitled The Climate System Model Development and Application Studies, and the National Natural Science Foundation of China (40375029). References [1] [2] [3] [4] [5] [6] Houghton J T, Ding Y H, Griggs D G, et al. Climate Change 2001: The Scientific Basis. Contribution of Working Group 1 to the Third Assessment Report of the Intergovernmental Panel on Climate Change [M]. Cambridge, UK: Cambridge University Press, 2001 Wang S, Gong D. Enhancement of the warming trend in China [J]. Geophys. Res. Lett., 2000, 27 (16): 2581J2584 Zhao Zongci, Luo Yong, Gao Xuejie, et al. Enlightenment from detection of climate change in China for the 20th century [J]. Climate Change Newsletter (2003/2004), 2004: 20J22 Zhou Tianjun, Yu Rucong. 20th century surface air temperature over China and the globe simulated by coupled climate models [J]. J. Climate, 2006, 19 (22): 5843J5858 Rayner N A, Parker D E, Horton E B, et al. Global analyses of SST, sea ice and night marine air temperature since the late nineteenth century [J]. J. Geophys. Res., 2003, 108 (D14), 4407, doi:10.1029/2002jd002670 Collins W D, Hack J J, Boville B A, et al. Description of the NCAR Community Atmosphere Model (CAM2) [M]. Boulder, Colorado, USA: NCAR, 2003: 171 86