• Annan, J. D., D. J. Lunt, J. C. Hargreaves, and P. J. Valdes, 2005: Parameter estimation in an atmospheric GCM using the ensemble Kalman filter. Nonlinear Processes Geophys., 12, 363371, doi:10.5194/npg-12-363-2005.

    • Search Google Scholar
    • Export Citation
  • Chen, W., Z. Jiang, and L. Li, 2011: Probabilistic projections of climate change over China under the SRES A1B scenario using 28 AOGCMs. J. Climate, 24, 47414756, doi:10.1175/2011JCLI4102.1.

    • Search Google Scholar
    • Export Citation
  • Ding, Y., G. Ren, Z. Zhao, Y. Xu, Y. Luo, Q. Li, and J. Zhang, 2007: Detection, causes and projection of climate change over China: An overview of recent progress. Adv. Atmos. Sci., 24, 954971, doi:10.1007/s00376-007-0954-4.

    • Search Google Scholar
    • Export Citation
  • Fan, Y., and H. van den Dool, 2008: A global monthly land surface air temperature analysis for 1948–present. J. Geophys. Res., 113, D01103, doi:10.1029/2007JD008470.

    • Search Google Scholar
    • Export Citation
  • Gao, J., V. Masson-Delmotte, T. Yao, L. Tian, C. Risi, and G. Hoffmann, 2011: Precipitation water stable isotopes in the south Tibetan plateau: Observations and modeling. J. Climate, 24, 31613178, doi:10.1175/2010JCLI3736.1.

    • Search Google Scholar
    • Export Citation
  • Giorgi, F., X. Bi, and J. S. Pal, 2004: Mean, interannual variability and trends in a regional climate change experiment over Europe. I. Present-day climate (1961–1990). Climate Dyn., 22, 733756, doi:10.1007/s00382-004-0409-x.

    • Search Google Scholar
    • Export Citation
  • Gong, D., and S. Wang, 1999: 1998: The warmest year on record of the century in China (in Chinese). Meteor. Monogr., 25, 35.

  • Hansen, J., R. Ruedy, M. Sato, and K. Lo, 2010: Global surface temperature change. Rev. Geophys., 48, RG4004, doi:10.1029/2010RG000345.

  • Harris, I., P. D. Jones, T. J. Osborn, and D. H. Lister, 2013: Updated high-resolution grids of monthly climatic observations—The CRU TS3.10 dataset. Int. J. Climatol., 34, 623–642, doi:10.1002/joc.3711.

    • Search Google Scholar
    • Export Citation
  • Hu, Z.-Z., S. Yang, and R. Wu, 2003: Long-term climate variations in China and global warming signals. J. Geophys. Res., 108, 4614, doi:10.1029/2003JD003651.

    • Search Google Scholar
    • Export Citation
  • Hua, W., H. Chen, and S. Sun, 2014: Uncertainty in land surface temperature simulation over China by CMIP3/CMIP5 models. Theor. Appl. Climatol., doi:10.1007/s00704-013-1020-z, in press.

    • Search Google Scholar
    • Export Citation
  • Jacob, D., and Coauthors, 2007: An inter-comparison of regional climate models for Europe: Model performance in present-day climate. Climatic Change, 81, 3152, doi:10.1007/s10584-006-9213-4.

    • Search Google Scholar
    • Export Citation
  • Ji, Z., and S. Kang, 2013: Double-nested dynamical downscaling experiments over the Tibetan Plateau and their projection of climate change under two RCP scenarios. J. Atmos. Sci., 70, 12781290, doi:10.1175/JAS-D-12-0155.1.

    • Search Google Scholar
    • Export Citation
  • Kharin, V. V., F. W. Zwiers, X. Zhang, and G. C. Hegerl, 2007: Changes in temperature and precipitation extremes in the IPCC ensemble of global coupled model simulations. J. Climate, 20, 14191444, doi:10.1175/JCLI4066.1.

    • Search Google Scholar
    • Export Citation
  • Knutti, R., and J. Sedlacek, 2013: Robustness and uncertainties in the new CMIP5 climate model projections. Nat. Climate Change, 3, 369373, doi:10.1038/nclimate1716.

    • Search Google Scholar
    • Export Citation
  • Kumar, S., V. Merwade, J. L. Kinter, and D. Niyogi, 2013: Evaluation of temperature and precipitation trends and long-term persistence in CMIP5 20th century climate simulations. J. Climate, 26, 41684185, doi:10.1175/JCLI-D-12-00259.1.

    • Search Google Scholar
    • Export Citation
  • Legates, D. R., and C. J. Willmott, 1990a: Mean seasonal and spatial variability in global surface air temperature. Theor. Appl. Climatol., 41, 1121, doi:10.1007/BF00866198.

    • Search Google Scholar
    • Export Citation
  • Legates, D. R., and C. J. Willmott, 1990b: Mean seasonal and spatial variability in gauge-corrected, global precipitation. Int. J. Climatol., 10, 111127, doi:10.1002/joc.3370100202.

    • Search Google Scholar
    • Export Citation
  • Li, H., L. Feng, and T. Zhou, 2011: Multi-model projection of July–August climate extreme changes over China under CO2 doubling. Part II: Temperature. Adv. Atmos. Sci., 28, 448463, doi:10.1007/s00376-010-0052-x.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., and Coauthors, 2007: THE WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 13831394, doi:10.1175/BAMS-88-9-1383.

    • Search Google Scholar
    • Export Citation
  • Miao, C., Q. Duan, L. Yang, and A. G. L. Borthwick, 2012: On the applicability of temperature and precipitation data from CMIP3 for China. PLoS ONE, 7, e44659, doi:10.1371/journal.pone.0044659.

    • Search Google Scholar
    • Export Citation
  • Qian, W., and A. Qin, 2006: Spatial-temporal characteristics of temperature variation in China. Meteor. Atmos. Phys., 93, 116, doi:10.1007/s00703-005-0163-6.

    • Search Google Scholar
    • Export Citation
  • Ren, G., Y. Ding, Z. Zhao, J. Zheng, T. Wu, G. Tang, and Y. Xu, 2012: Recent progress in studies of climate change in China. Adv. Atmos. Sci., 29, 958977, doi:10.1007/s00376-012-1200-2.

    • Search Google Scholar
    • Export Citation
  • Riahi, K., and Coauthors, 2011: RCP 8.5—A scenario of comparatively high greenhouse gas emissions. Climatic Change, 109, 3357, doi:10.1007/s10584-011-0149-y.

    • Search Google Scholar
    • Export Citation
  • Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K. Averyt, M. M. B. Tignor, and H. L. Miller Jr., Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.

  • Su, F., X. Duan, D. Chen, Z. Hao, and L. Cuo, 2013: Evaluation of the global climate models in the CMIP5 over the Tibetan Plateau. J. Climate, 26, 31873208, doi:10.1175/JCLI-D-12-00321.1.

    • Search Google Scholar
    • Export Citation
  • Tang, G., and G. Ren, 2005: Reanalysis of surface air temperature change of the last 100 years over China (in Chinese). Climatic Environ. Res., 10, 791798.

    • Search Google Scholar
    • Export Citation
  • Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485498, doi:10.1175/BAMS-D-11-00094.1.

    • Search Google Scholar
    • Export Citation
  • Thomson, A., and Coauthors, 2011: RCP4.5: A pathway for stabilization of radiative forcing by 2100. Climatic Change, 109, 7794, doi:10.1007/s10584-011-0151-4.

    • Search Google Scholar
    • Export Citation
  • van Vuuren, D. P., and Coauthors, 2008: Temperature increase of twenty-first century mitigation scenarios. Proc. Natl. Acad. Sci. USA, 105, 15 25815 262, doi:10.1073/pnas.0711129105.

    • Search Google Scholar
    • Export Citation
  • van Vuuren, D. P., and Coauthors, 2011: The representative concentration pathways: An overview. Climatic Change, 109, 531, doi:10.1007/s10584-011-0148-z.

    • Search Google Scholar
    • Export Citation
  • Wang, G., S. Sun, and R. Mei, 2011: Vegetation dynamics contributes to the multi-decadal variability of precipitation in the Amazon region. Geophys. Res. Lett., 38, L19703, doi:10.1029/2011GL049017.

    • Search Google Scholar
    • Export Citation
  • Wang, L., and W. Chen, 2014: A CMIP5 multimodel projection of future temperature, precipitation, and climatological drought in China. Int. J. Climatol., doi:10.1002/joc.3822, in press.

    • Search Google Scholar
    • Export Citation
  • Wang, S., and D. Gong, 2000: Enhancement of the warming trend in China. Geophys. Res. Lett., 27, 25812584, doi:10.1029/1999GL010825.

  • Wang, S., J. Ye, D. Gong, J. Zhu, and T. Yao, 1998: Construction of mean annual temperature series for the last one hundred years in China (in Chinese). Quart. J. Appl. Meteor., 9, 392401.

    • Search Google Scholar
    • Export Citation
  • Wang, S., D. Gong, and J. Zhu, 2001: Twentieth-century climatic warming in China in the context of the Holocene. Holocene, 11, 313321, doi:10.1191/095968301673172698.

    • Search Google Scholar
    • Export Citation
  • Wang, X., F. Chen, E. Hasi, and J. Li, 2008: Desertification in China: An assessment. Earth Sci. Rev., 88, 188206, doi:10.1016/j.earscirev.2008.02.001.

    • Search Google Scholar
    • Export Citation
  • Wild, M., 2009: How well do IPCC-AR4/CMIP3 climate models simulate global dimming/brightening and twentieth-century daytime and nighttime warming? J. Geophys. Res., 114, D00D11, doi:10.1029/2008JD011372.

    • Search Google Scholar
    • Export Citation
  • Willmott, C. J., and S. M. Robeson, 1995: Climatologically aided interpolation (CAI) of terrestrial air temperature. Int. J. Climatol., 15, 221229, doi:10.1002/joc.3370150207.

    • Search Google Scholar
    • Export Citation
  • Xu, Y., X. Gao, Y. Shen, C. Xu, Y. Shi, and F. Giorgi, 2009: A daily temperature dataset over China and its application in validating a RCM simulation. Adv. Atmos. Sci., 26, 763772, doi:10.1007/s00376-009-9029-z.

    • Search Google Scholar
    • Export Citation
  • Yao, T., J. Pu, A. Lu, Y. Wang, and W. Yu, 2007: Recent glacial retreat and its impact on hydrological processes on the Tibetan Plateau, China, and surrounding regions. Arct. Antarct. Alp. Res., 39, 642650, doi:10.1657/1523-0430(07-510)[YAO]2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Yu, R., B. Wang, and T. Zhou, 2004: Climate effects of the deep continental stratus clouds generated by the Tibetan Plateau. J. Climate, 17, 27022713, doi:10.1175/1520-0442(2004)017<2702:CEOTDC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zhao, Z., S. Wang, Y. Xu, G. Ren, Y. Luo, and X. Gao, 2005: Attribution of the 20th century climate warming in China (in Chinese). Climatic Environ. Res., 10, 808817.

    • Search Google Scholar
    • Export Citation
  • Zhou, T., and Z. Li, 2002: Simulation of the East Asian summer monsoon using a variable resolution atmospheric GCM. Climate Dyn., 19, 167180, doi:10.1007/s00382-001-0214-8.

    • Search Google Scholar
    • Export Citation
  • Zhou, T., and R. Yu, 2006: Twentieth-century surface air temperature over China and the globe simulated by coupled climate models. J. Climate, 19, 58435858, doi:10.1175/JCLI3952.1.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Time series of average annual temperatures derived from CRU, GISS, UDEL, GHCN, WANG, and the CMIP3 and CMIP5 multimodel ensemble averages for China. The gray lines correspond to the individual CMIP5 GCMs.

  • View in gallery

    Annual and seasonal temperature biases of CMIP3 and CMIP5 relative to CRU observations. (top) The difference between CMIP3 and CRU and (bottom) the difference between CMIP5 and CRU for (left) annual, (center) summer [June–August (JJA)], and (right) winter [December–February (DJF)] temperature. Stippled regions indicate statistically significant differences (95% level).

  • View in gallery

    Trends in average annual temperatures over the twentieth century derived from (a) CRU and (b) UDEL observations, and (c) CMIP3 and (d) CMIP5 multimodel ensemble averages. Stippled regions indicate statistically significant differences (95% level).

  • View in gallery

    As Fig. 3, but for the second half of the twentieth century (1950–99).

  • View in gallery

    Comparison between CRU observations and the individual CMIP5 GCMs along the x-axis for (top row) annual and (next rows) seasonal temperatures: (left) temperature bias from observations and (right) correlations between observations and individual models. Error bars (red) indicate standard deviation of temperature biases; stars (green) indicate significant (95% level) correlations.

  • View in gallery

    Trends in seasonal temperatures over the twentieth century for China derived from CRU (red line) and UDEL (blue line) observations, and CMIP3 (dashed line) and CMIP5 (solid line) multimodel ensemble averages. The vertical narrow color bars indicate temperature trends derived from the individual CMIP5 models.

  • View in gallery

    Time series of historical (black line) and projected temperature for China from different CMIP5 experiments during 1900–2100: RCP 8.5 (red line), 4.5 (yellow line), and 2.6 (green line).

  • View in gallery

    (left) Annual temperature trends for the twenty-first century for three emission scenarios and (right) changes in annual temperature (2070–99 vs 1961–90) based on CMIP5 projections: RCP (top) 2.6, (middle) 4.5, and (bottom) 8.5. Stippled regions indicate statistically significant trends/differences (95% level).

  • View in gallery

    Trends in annual (blue) and summer (green) and winter (red) temperatures over the twenty-first century derived from individual GCMs for the three experiments: RCP (top) 2.6, (middle) 4.5, and (bottom) 8.5.

  • View in gallery

    Scatterplot of historical temperature trends during the twentieth century and the projected temperature trend under the RCP 4.5 scenario.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1029 622 53
PDF Downloads 465 279 20

Surface Air Temperature Changes over the Twentieth and Twenty-First Centuries in China Simulated by 20 CMIP5 Models

View More View Less
  • 1 Department of Geography, Texas A&M University, College Station, Texas
© Get Permissions
Full access

Abstract

Historical temperature variability over China during the twentieth century and projected changes under three emission scenarios for the twenty-first century are evaluated on the basis of a multimodel ensemble of 20 GCMs from phase 5 of the Coupled Model Intercomparison Project (CMIP5) and two observational datasets. Changes relative to phase 3 of the Coupled Model Intercomparison Project (CMIP3) are assessed, and the performance of individual GCMs is also quantified. Compared with observations, GCMs have substantial cold biases over the Tibetan Plateau, especially in the cold season. The timing and location of these biases also correspond to the greatest disagreement among the individual models, indicating GCMs’ limitations in reproducing climatic features in this complex terrain. The CMIP5 multimodel ensemble shows better agreement with observations than CMIP3 in terms of the temperature biases. Both CMIP3 and CMIP5 capture the climatic warming over the twentieth century. However, the magnitude of the annual mean temperature trends is underestimated. There is also limited agreement in the spatial and seasonal patterns of temperature trends over China. Based on six statistical measures, four individual models—the Max Planck Institute Earth System Model, low resolution (MPI-ESM-LR), Second Generation Canadian Earth System Model (CanESM2), Model for Interdisciplinary Research on Climate, Earth System Model (MIROC-ESM), and Community Climate System Model, version 4 (CCSM4)—best represent surface air temperature variability over China. The future temperature projections indicate that the representative concentration pathway (RCP) 8.5 and RCP 4.5 scenarios exhibit a gradual increase in annual temperature during the twenty-first century at a rate of 0.60° and 0.27°C (10 yr)−1, respectively. As the lowest-emission mitigation scenario, RCP 2.6 projects the lowest rate of temperature increase [0.10°C (10 yr)−1]. By the end of the twenty-first century, temperature is projected to increase by 1.7°–5.7°C, with larger warming over northern China and the Tibetan Plateau.

Corresponding author address: Liang Chen, Department of Geography, Texas A&M University, College Station, TX 77843. E-mail: chenliang08@neo.tamu.edu

Abstract

Historical temperature variability over China during the twentieth century and projected changes under three emission scenarios for the twenty-first century are evaluated on the basis of a multimodel ensemble of 20 GCMs from phase 5 of the Coupled Model Intercomparison Project (CMIP5) and two observational datasets. Changes relative to phase 3 of the Coupled Model Intercomparison Project (CMIP3) are assessed, and the performance of individual GCMs is also quantified. Compared with observations, GCMs have substantial cold biases over the Tibetan Plateau, especially in the cold season. The timing and location of these biases also correspond to the greatest disagreement among the individual models, indicating GCMs’ limitations in reproducing climatic features in this complex terrain. The CMIP5 multimodel ensemble shows better agreement with observations than CMIP3 in terms of the temperature biases. Both CMIP3 and CMIP5 capture the climatic warming over the twentieth century. However, the magnitude of the annual mean temperature trends is underestimated. There is also limited agreement in the spatial and seasonal patterns of temperature trends over China. Based on six statistical measures, four individual models—the Max Planck Institute Earth System Model, low resolution (MPI-ESM-LR), Second Generation Canadian Earth System Model (CanESM2), Model for Interdisciplinary Research on Climate, Earth System Model (MIROC-ESM), and Community Climate System Model, version 4 (CCSM4)—best represent surface air temperature variability over China. The future temperature projections indicate that the representative concentration pathway (RCP) 8.5 and RCP 4.5 scenarios exhibit a gradual increase in annual temperature during the twenty-first century at a rate of 0.60° and 0.27°C (10 yr)−1, respectively. As the lowest-emission mitigation scenario, RCP 2.6 projects the lowest rate of temperature increase [0.10°C (10 yr)−1]. By the end of the twenty-first century, temperature is projected to increase by 1.7°–5.7°C, with larger warming over northern China and the Tibetan Plateau.

Corresponding author address: Liang Chen, Department of Geography, Texas A&M University, College Station, TX 77843. E-mail: chenliang08@neo.tamu.edu

1. Introduction

The global average surface temperature has increased by 0.74° ± 0.18°C during 1906–2005 (Solomon et al. 2007). This warming trend, in particular the warming since the mid-twentieth century, is very likely due to the increased level of anthropogenic greenhouse gas concentrations (Wang et al. 2011). China has also experienced significant temperature increases concurrent with global warming. In fact, previous studies suggest that the central-northern continent of eastern Asia was one of the most rapidly warming regions in the world during the last two decades of the twentieth century (Wang and Gong 2000). For China as a whole, Ding et al. (2007) imply that the annual mean surface air temperature has increased by 0.5°–0.8°C during the twentieth century, with an accelerated warming of 1.1°C during the second half of the century, which is slightly higher than the global temperature trend for the same period. Based on 486 stations during the period 1960–2000, Qian and Qin (2006) suggest that temperature increased at a rate of 0.2°–0.3°C (10 yr)−1 in northern China and less than 0.1°C (10 yr)−1 in southern China. Seasonally, the greatest warming occurred in winter, and a cooling took place in the Yangtze River and Yellow River basins in spring and summer.

Accurate prediction of future climate change, especially in regard to global warming, has been one of the most important scientific and societal challenges in current climate research efforts. Simulations from coupled ocean–atmosphere general circulation models (GCMs) forced with projected greenhouse gas and aerosol emissions are the primary tools for estimating trends and variability of future climate (Kharin et al. 2007). The establishment of the third phase of the Coupled Model Intercomparison Project (CMIP3) by the World Climate Research Programme (WCRP) provided scientists outside of the modeling community an opportunity to conduct comprehensive analyses on climate variability and change at both global and regional scales, based on the output from multiple climate models (Meehl et al. 2007). The Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) estimates a warming of about 0.2°C (10 yr)−1 for a range of emission scenarios over the next two decades. By the end of the twenty-first century, global mean surface temperature is projected to increase by 1.1°–6.4°C over the 1990 level (van Vuuren et al. 2008). Detailed temperature projections at a regional scale are also crucial for local development and decision-making at a country level, such as in China. Under varied emission scenarios, the annual mean temperature in China based on CMIP3 simulations is projected to increase by 3.9°–6.0°C by 2100 relative to the 1961–90 average (Ding et al. 2007). However, limitations exist in CMIP3. Wild (2009) indicates that CMIP3 models underestimate the decadal variations in global land surface warming during the twentieth century. Using 19 coupled climate models driven by historical natural and anthropogenic forcings in CMIP3, Zhou and Yu (2006) found that the robustness of temperature estimates averaged over China is lower than that of the global and hemispheric average, and discrepancies exist between the observed and simulated spatial patterns of temperature trends. By comparing output from 24 models with observational data in China, Miao et al. (2012) evaluated the performance of the CMIP3 GCMs in simulating temperature, and found that 18 models underestimate the annual mean temperature of China. Based on 24 CMIP3 models, Li et al. (2011) indicated that in simulating both the daily maximum and minimum July–August temperatures, the largest cold bias is found over western China.

The IPCC Fifth Assessment Report (AR5) promoted the development of the fifth phase of the Coupled Model Intercomparison Project (CMIP5), which comprises state-of-the-art GCMs. Compared with CMIP3, CMIP5 uses new representative concentration pathways (RCPs) and includes a larger number of GCMs that are more complex and have higher spatial resolution (Taylor et al. 2012). By comparing results from CMIP3 and CMIP5, Knutti and Sedlacek (2013) suggest that the projected global temperature change from CMIP5 is remarkably similar to that from CMIP3 after accounting for the different underlying scenarios. Based on 19 CMIP5 models, Kumar et al. (2013) found that the multimodel ensemble average of the global land average temperature trend [0.07°C (10 yr)−1] agrees well with the observed trend [0.08°C (10 yr)−1] during the twentieth century, but large uncertainties exist in the simulation of regional-scale temperature trends in CMIP5. Therefore, it is important to evaluate the performance of the state-of-the-art climate models in CMIP5 for different regions. Based on 24 CMIP3 and 10 CMIP5 models, Hua et al. (2014) found that all models capture the climatological pattern of land surface temperatures for the period 1960–2000, but a large spread exists among the models in simulation the climatology, interannual variation, and temperature trends. Over the Tibetan Plateau, it is found that cold biases exist in the majority of the CMIP5 models, with a mean underestimation of 1.1°–2.5°C for the months of December–May, and less than 1°C for June–October (Su et al. 2013). Future temperature projections suggest that there will be 1.9°–3.3°C and 3.4°–5°C increases in 2070–99 under RCP4.5 and RCP8.5, respectively (Wang and Chen 2014). However, to date there is a lack of a comprehensive evaluation of CMIP5 temperature simulations and their possible improvement compared with CMIP3 over the entire region of China. China’s extremely complex terrain and confluence of circulation regimes (such as the various monsoons) likely represent a particularly challenging environment for GCMs.

Therefore, the goal of this investigation is to objectively and comprehensively evaluate the performance of CMIP5 surface air temperature simulations over China. Because of the continuing rapid development and industrialization of China and the contribution of this growth to regional and global change, accurate projections for this part of the world are of particular concern. Over the coming years, until the release of the IPCC’s Sixth Assessment Report in approximately 2020, these CMIP5 simulations will likely be used by various subfields of the scientific community, ranging from physical to social sciences, as the basis for future climate projections. CMIP5 simulations for China in particular will see increased usage, as the rapidly emerging science and technology community in China will also increasingly rely on these state-of-the-science projections for their region of the world. It is thus important to assess the accuracy and skill of CMIP5 simulations for China, as well as the degree to which they have improved or otherwise changed with respect to CMIP3. Such evaluations are then also useful for identifying shortcomings and necessary improvements for future modeling efforts.

Our evaluation includes an assessment of the CMIP5 multimodel ensemble average, which is commonly used by researchers because the multimodel-average approach reduces noise in the predictions. This spatial and temporal (annual and seasonal scale) evaluation is conducted with respect to observational data, as well as with CMIP3 to identify potential CMIP5 improvements. The quality of the CMIP5 multimodel average obviously depends on the quality of the individual models. Therefore, we also evaluate and identify the most (and least) suitable individual CMIP5 models for this part of the world to minimize the uncertainty in multimodel ensemble averages and to improve future temperature projections. Finally, we provide a comprehensive overview of CMIP5 temperature projections for China over the twenty-first century. This paper is organized as follows. Section 2 describes the datasets and analysis methods. Section 3 presents a comparison between observations with the simulations from both CMIP3 and CMIP5 in temperature variability during the twentieth century. Section 4 shows the projected temperature trends in China for the twenty-first century. Sections 5 and 6 provide a discussion and a summary, respectively.

2. Data and methods

a. Data

We obtained monthly surface air temperature output from 20 GCMs in the CMIP5 archive (Table 1; expansions of all model names are provided in the appendix) Four sets of experiments were used: one historical experiment for the twentieth century, and three future emission scenarios for the twenty-first century. The historical experiment (HIST) provides simulations of surface temperature for 1850–2005 on the basis of observed natural and anthropogenic forcings. Future climate projections for 2006–2300 include three RCPs adopted by the IPCC AR5, including RCP 8.5, RCP 4.5, and RCP 2.6, which correspond to higher, medium, and lower greenhouse gas emissions, respectively. More specifically, the RCP 8.5 scenario assumes high population growth and high energy demand without climate change policies. It thus results in the pathway with the highest greenhouse gas emissions, brought about by a radiative forcing of 8.5 W m−2 in 2100 (Riahi et al. 2011). Under RCP 4.5, radiative forcing stabilizes at 4.5 W m−2 in 2100 without exceeding that value (Thomson et al. 2011), which results in a medium stabilization scenario. Finally, RCP 2.6 has a peak radiative forcing of ~3 W m−2 before 2100 and then declines to 2.6 W m−2 by 2100, making it a low emission scenario (van Vuuren et al. 2011). To assess the potential improvements in CMIP5 compared with the earlier version, historical simulations of surface air temperature from 22 GCMs in the CMIP3 archive were also used. Monthly surface air temperature output from CMIP5 was obtained online (from http://cmip-pcmdi.llnl.gov/index.html). Output from the CMIP3 twentieth-century climate in coupled models experiment (20C3M) was obtained from the CMIP3 archive (http://esg.llnl.gov:8080/).

Table 1.

Twenty CMIP5 GCMs used in our study (see the appendix for expansions of model names) and their forcings used for the historical simulations (Y indicates the forcing was used in the simulation). Nat: natural forcing (e.g., solar and volcanic), Ant: anthropogenic forcing (e.g., well-mixed greenhouse gases, aerosols, ozone, and land-use changes), GHG: well-mixed greenhouse gases, SA: anthropogenic sulfate aerosol direct and indirect effects, SD: anthropogenic sulfate aerosol, accounting only for direct effects, Oz: ozone, LU: land-use change, Sl: solar irradiance, Vl: volcanic aerosol, SS: sea salt, Ds: dust, BC: black carbon, OC: organic carbon, MD: mineral dust, AA: anthropogenic aerosols. (All forcing information was gathered from the CMIP5 website: http://cmip-pcmdi.llnl.gov/index.html.)

Table 1.

To ensure our study is not biased by the choice of observational datasets, two surface air temperature products were used to evaluate model performance. The Climatic Research Unit (CRU) time series (TS) 3.10 dataset from the University of East Anglia provides monthly mean temperature for the global land surface at 0.5° × 0.5° resolution for the period 1901–2009 (Harris et al. 2013). This dataset is widely used for assessing climate variability and validating climate models at regional scales (Giorgi et al. 2004; Jacob et al. 2007; Xu et al. 2009). The second observational dataset is the 1900–2010 gridded air temperature time series data version 3.01 (referred to as UDEL hereafter), developed by (Willmott and Robeson 1995). This data archive contains terrestrial monthly air temperature with 0.5° × 0.5° resolution, which is derived from the Global Historical Climatology Network (GHCN) version 2 and other mean monthly surface air temperature station records (Legates and Willmott 1990a,b). Other globally gridded air temperature products are also available, such as the National Aeronautics and Space Administration (NASA) Goddard Institute for Space Studies (GISS) surface temperature analysis, and GHCN itself (part of UDEL). GISS provides global surface temperature anomalies (2° × 2°) since 1880 (Hansen et al. 2010), whereas GHCN combines two individual observational datasets and contains global land surface temperatures with a high resolution (0.5° × 0.5°) since 1948 (Fan and van den Dool 2008). However, GISS temperatures are not spatially continuous in the early part of the twentieth century, and the GHCN product only begins in 1948. However, with the exception of the 1915–30 period, these products all agree reasonably well (Fig. 1). We therefore focus on CRU and UDEL because they are both spatially and temporally complete from 1901 onward. Additionally, observational time series of annual mean surface air temperature over China were used to further validate the reliability of the two global observational temperature datasets at a regional scale. This dataset (hereafter referred to as WANG) was constructed based on temperature observations, documentary records, ice core data, and tree ring data over China for the period 1880–2009 (Wang et al. 1998; Gong and Wang 1999; Wang and Gong 2000; Wang et al. 2001). However, it is only available for surface air temperatures at the mean annual scale, so it cannot be used for the seasonal analysis.

Fig. 1.
Fig. 1.

Time series of average annual temperatures derived from CRU, GISS, UDEL, GHCN, WANG, and the CMIP3 and CMIP5 multimodel ensemble averages for China. The gray lines correspond to the individual CMIP5 GCMs.

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

b. Methods

Based on the temporal coverage of observations and the CMIP5 output, the period 1901–2005 was extracted from the HIST simulations, and the period 2006–2100 was extracted from the three RCP experiments. Because the CRU observations start from 1901 and most of the CMIP3 20C3M simulations end in 1999, therefore, our twentieth-century analysis focuses on the period 1901–99, and the second half of twentieth-century analysis focuses on the period 1950–99. Because of the different spatial resolutions adopted by different GCMs (shown in Table 1), we used bilinear interpolation to regrid all GCM output into a uniform resolution (2.5° × 2.5°). Some modeling groups provide multiple realizations for each experiment (e.g., there are six runs for the HIST experiment in the CCSM4 output). To avoid potentially biasing our results with respect to one randomly selected realization, all available realizations were used and ensemble averages were calculated for each model. The multimodel ensemble averages of CMIP3 and CMIP5 were then created. From here on, the terms CMIP3 and CMIP5 refer to these multimodel ensembles. CMIP3 and CMIP5 were compared with the two observational products to assess the performance of CMIP5, as well as CMIP5’s potential improvements over CMIP3. Furthermore, the HIST simulation of each CMIP5 GCM was compared individually with observations to investigate the agreement among different models.

Linear trends were calculated to estimate the long-term trends in monthly, seasonal, and annual temperature for China during the twentieth and twenty-first centuries. For the evaluation and comparison of GCMs, several statistical measures were used in this paper including mean error (ME), mean absolute error (MAE), standard deviation of the error (SDE), root-mean-square error (RMSE), and Pearson’s correlation coefficients (R). Temperature over the whole domain of China was calculated by averaging the grid cell values within the political boundaries (land areas) of China. To further assess the regional historical and future temperatures, four regions of China were selected: the north (35°–43°N, 102.5°–122.5°E), south (22°–35°N, 102.5°–122.5°E), northwest (38°–48°N, 73°–102.5°E), and the Tibetan Plateau (28°–38°N, 80°–102.5°E). Regional temperature trends were calculated as the area average for each region. Because of the likely shortcomings in the observations prior to the 1950s, model performance was also assessed for the second half the twentieth century separately.

3. Temperature in the twentieth century

Figure 1 compares annual temperature anomalies (departures from the 1961–90 baseline) derived from observations and CMIP3 and CMIP5 multimodel ensemble averages for all of China. The CRU and UDEL observations exhibit good agreement after the 1930s (R = 0.98, p < 0.01). However, when the station network was sparse before 1930, discrepancies exist between the observations. Also included here are the GISS, GHCN, and WANG observations for comparison. CRU shows better agreement with the other two observational products than UDEL. The 1901–2000 correlations of CRU and UDEL with GISS are 0.97 and 0.95, respectively. The CRU and UDEL correlations with GHCN are both 0.99 for the period 1951–2000, although this slightly higher correlation is likely due to the greatly enhanced station coverage during 1951–2000. Compared with the WANG observations, both CRU and UDEL show large discrepancies before 1950. The 1901–50 correlation of CRU and UDEL with WANG is 0.49 and 0.45, respectively. In the WANG time series, temperature increased from the 1910s with a peak in the 1940s. This peak was also reported in other observed time series of surface air temperature over China (Tang and Ren 2005). Time series from all global datasets also indicate a warming peak in the 1940s, however, with a smaller magnitude. After 1950, both CRU and UDEL show good agreement with WANG, with 1951–2000 correlations of 0.98 and 0.97. Therefore, this time series comparison (Fig. 1) between all the observational data products supports our choice of CRU and UDEL for subsequent evaluations. However, it is obvious that large uncertainties in temperature variability exist prior to 1950 in the currently available observational datasets. A separate model evaluation for 1951–2000 is therefore needed, in addition to the twentieth-century assessment.

Evaluating the model simulations, both CMIP3 and CMIP5 illustrate an overall warming trend throughout twentieth century. However, interannual variability in temperature is obvious for the individual models, but is suppressed by the multimodel ensemble averages. For instance, the multimodel ensemble average did not reproduce the highest annual temperature in 1998 as indicated by the observations. Interdecadal variability also appears muted in CMIP3 and CMIP5. For instance, neither the CMIP3 nor CMIP5 models capture the observed warming during 1920–30. The accelerated warming trend evident in the observations since the 1970s is also underestimated in the multimodel ensembles (i.e., observed warming in China outpaced model simulations). The annual mean temperature climatology for the 1961–90 baseline was 5.40°C in CMIP5 and 5.15°C in CMIP3. Comparing CMIP3 and CMIP5 shows that CMIP3 has a consistent cold bias in China relative to CMIP5 before the 1960s.

We compared the CMIP3 and CMIP5 multimodel ensemble averages of annual and seasonal air temperature with CRU and UDEL observations in the twentieth century (Table 2). Generally, the mean annual temperatures from both CMIP3 and CMIP5 are significantly lower than both observational products. These cold biases also exist in the seasonal temperatures. Compared with CMIP3, CMIP5 shows less bias relative to observations throughout the four seasons. Both CMIP3 and CMIP5 are significantly correlated with the two observational annual temperature averages (Table 2). The correlation of CMIP3 and CMIP5 with CRU is 0.70 and 0.65, respectively, and their correlation with UDEL is 0.54 and 0.57, respectively. This suggests that only 29%–49% of the observational temperature variability is captured by the multimodel ensembles. For seasonal temperature, CMIP5 exhibits better agreement with observations in summer and autumn, whereas CMIP3 is better in spring and winter. These seasonal correlations, although mostly statistically significant, are even weaker than the annual relationships and account for, at most, 31% of observed temperature variability. In summer, CMIP3 has no significant correlation with UDEL.

Table 2.

Annual and seasonal temperature means and trends for China during the twentieth century derived from UDEL and CRU observations, and CMIP3 and CMIP5 multimodel ensemble averages. Statistically significant differences or trends (95% level) are shown in boldface.

Table 2.

Linear trend analysis reveals that there are significant warming trends in annual temperatures during the twentieth century from both observations and model simulations (Table 2). The CRU observations exhibit the largest warming rate of 0.79°C (100 yr)−1 for China, which is in line with rates of global temperature increases. The UDEL trend is only 0.48°C (100 yr)−1, which is due to the ~1915–30 period when UDEL was anomalously warm compared to other observational products [except WANG, which shows a twentieth-century warming trend of only 0.38°C (100 yr)−1]. CMIP3 and CMIP5 show warming rates of 0.64° and 0.40°C (100 yr)−1, respectively. The greater CMIP3 trend is driven by its cold bias (relative to CMIP5) prior to ~1960. Seasonally, both the CRU and UDEL observations suggest that the largest warming trends in China occurred in winter and spring. Except for CRU (autumn), there is no significant increase in temperature in summer and autumn (i.e., the annual warming in China is driven by changes in winter and spring). However, this seasonality in temperature trends was not well captured by CMIP3 or CMIP5. CMIP3 exhibits consistently larger warming trends than CMIP5, and these trends are statistically significant throughout all four seasons. The seasonal variability of temperature increases is much less than that in observations, although both CMIP3 and CMIP5 do also exhibit the most warming in winter. The winter warming in the models is only about half of the observed warming.

Given the denser observational station network and hence more reliable trend estimates after ~1950, temperature biases and trends during the second half of the twentieth century were also calculated separately (Table 3). Both the observations and model simulations indicate increasing mean annual temperature. CMIP5 shows fewer cold biases than CMIP3 in both annual and seasonal temperatures. However, CMIP3 has higher correlations with the observations (including GHCN, GISS, and WANG) except for summer temperature. For annual temperature, CRU and UDEL show an accelerated warming of 0.20° and 0.16°C (10 yr)−1, respectively, with the largest warming occurring in winter at a rate of 0.43° and 0.35°C (10 yr)−1. Annual temperature trends were also calculated from GHCN and WANG, showing a warming rate of 0.20° and 0.16°C (10 yr)−1, respectively. However, there is no statistically significant trend during summer for either of the observational products. Similar to the period of 1901–99, both CMIP3 and CMIP5 underestimate the annual temperature trends and exhibit less seasonal variability.

Table 3.

As in Table 2, but for the second half of the twentieth century (1950–99).

Table 3.

To obtain a better sense of model variability within China, we also investigate the spatial variability of climatological annual mean temperature over the twentieth century. We provide this assessment relative to the CRU observations only, because the UDEL spatial patterns are similar (not shown). We calculated the difference in annual and seasonal temperatures between CRU and CMIP3 (Figs. 2a–c) and CMIP5 (Figs. 2d–f). CMIP3 and CMIP5 show a similar spatial pattern of temperature bias relative to the CRU observations. Generally, cold biases exist over most parts of China, including the Tibetan Plateau, the central and southern region, and most of the north. Warm biases exist in the northeast and northwest. In CMIP3, there is a consistent warm bias along the southern edge of the Tibetan Plateau, which does not appear in CMIP5. For both annual and seasonal temperatures, CMIP5 has a lower bias than CMIP3, especially over eastern China. In CMIP5, the largest cold bias occurs over the high elevations of the Tibetan Plateau in winter, while the largest warm bias occurs in the desert regions in northwestern China during summer. This suggests that in CMIP5, the cold regions are too cold during winter, while the warm regions are too warm during summer. We also calculated the intermodel standard deviations in annual and seasonal temperatures (not shown), and they exhibit a similar spatial pattern as the temperature bias. The regions with high intermodel standard deviations correspond to the regions of largest bias. The temperature biases for the second half of the twentieth century show the same spatial pattern and are therefore not discussed here.

Fig. 2.
Fig. 2.

Annual and seasonal temperature biases of CMIP3 and CMIP5 relative to CRU observations. (top) The difference between CMIP3 and CRU and (bottom) the difference between CMIP5 and CRU for (left) annual, (center) summer [June–August (JJA)], and (right) winter [December–February (DJF)] temperature. Stippled regions indicate statistically significant differences (95% level).

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

Figure 3 shows the spatial annual temperature trends over the twentieth century based on observations and multimodel ensemble averages. Both CRU and UDEL show that the largest warming trend occurred in the northwest and northeast of China. However, in the UDEL observations, there is no significant warming trend over the Tibetan Plateau, but a significant cooling trend in central and southern China that is not captured by CRU. Compared with the observations, both CMIP3 and CMIP5 underestimated temperature trends over most regions of China. Spatial patterns of temperature trends in CMIP3 exhibit substantial disagreement with the observations. The largest warming trend occurred over the Tibetan Plateau, and there is significant warming in central and southern China. The spatial trend pattern is slightly improved in CMIP5; however, substantial discrepancies remain. Recognizing the inherent limitations in the observations in the early part of the twentieth century, temperature trends for the second half of the twentieth century were calculated (Fig. 4). Over the past half century, there is accelerated warming over most regions of China. The two observational products show that the largest warming occurred in the northwest and northeast. There are significant warming trends over the Tibetan Plateau. In central China, especially over the Sichuan province, temperature tends to decrease. Disagreement in temperature trends also still exist over 1950–99 between the two observational products. Over the south and southeast, CRU shows significant warming trends while UDEL indicates cooling. As for the entire twentieth century, both CMIP3 and CMIP5 underestimate the warming trends over the north, and cannot reproduce the cooling trends in central and southern China.

Fig. 3.
Fig. 3.

Trends in average annual temperatures over the twentieth century derived from (a) CRU and (b) UDEL observations, and (c) CMIP3 and (d) CMIP5 multimodel ensemble averages. Stippled regions indicate statistically significant differences (95% level).

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

Fig. 4.
Fig. 4.

As Fig. 3, but for the second half of the twentieth century (1950–99).

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

Annual and seasonal temperature trends were calculated for individual regions within China (Table 4). CRU shows statistically significant increases in annual temperatures in all four regions, with greater warming trends over the northeast and the northwest [0.88° and 0.94°C (100 yr)−1, respectively]. In UDEL, the temperature increase is only statistically significant over the northwest at a rate of 0.74°C (100 yr)−1. Compared with the CRU observations, both CMIP3 and CMIP5 overestimate the warming trend over the Tibetan Plateau, and underestimate the warming trend over the northwest and northeast. Based on the two observational products, there is no significant seasonal warming trend in the northeast in summer, in the south in autumn and winter, and in the Tibetan Plateau in summer and autumn. However, both CMIP3 and CMIP5 exhibit significant warming trends in all four regions throughout the year. The temperature trends over the second half of the twentieth century indicate accelerated warming in annual mean temperature based on both observations and simulations (Table 5). The observed accelerated warming in the north is mainly attributed to greater temperature trends in winter and spring, and in the northwest it is attributed to temperature increases in winter and autumn. In the south, temperature increased in winter and autumn, but decreased in summer. Over the Tibetan Plateau, there is no significant warming except in winter. However, neither CMIP3 nor CMIP5 reproduced the seasonality of the temperature trends in these regions. Model simulations show significant warming throughout the four seasons, with the greatest warming occurring in autumn, while the observations indicate that winter warming is the greatest.

Table 4.

Historical surface air temperature trends [°C (100 yr)−1] in different subregions of China during the twentieth century. Statistically significant trends (95% level) are shown in boldface.

Table 4.
Table 5.

As in Table 4, but for the second half of the twentieth century (1950–99).

Table 5.

Using multimodel ensemble averages is a common approach because it is thought that noise in the predictions is thereby reduced. However, as illustrated for CMIP5 in Fig. 1 (gray lines), there is a large amount of variability among the individual model ensembles that comprise a multimodel average. CMIP5 likely includes models that are well suited for capturing the temperature variability across China, in addition to potentially underperforming models. We therefore assess the individual GCMs relative to the CRU observations for annual temperature over the twentieth century (Table 6), to identify the most suitable models for our particular region. The assessment is based on the strength of the correlation, as well as on various error measures. Comparisons with UDEL are similar to CRU, and are therefore not shown. Several models exhibit better agreement (higher correlations and smaller errors/biases) with the observations: MPI-ESM-LR, CanESM2, MIROC-ESM, and CCSM4. Even though IPSL-CM5A-LR and EC-EARTH show the highest correlations with the observations (0.68 and 0.66, respectively), large cold biases and RMSEs exist in these models. FGOALS-g2, INM-CM4, and IPSL-CM5A-LR have the largest cold biases, and MIROC5 is the only model with a warm bias. Among the 20 models, ACCESS1.3, CMCC-CESM, GFDL-CM3, HadGEM2-CC, and MIROC5 show no trend throughout the twentieth century. The individual models’ performance was also compared with observations for the four seasons (Fig. 5). The agreement between models and observations varies through the seasons. Most models show good agreement with CRU in summer. However, only six models exhibit significant correlations with the observations in winter. Individual models were also evaluated by comparing simulated surface air temperatures relative to CRU, UDEL, and WANG observations for the second half the twentieth century (not shown). The models exhibit the same agreement as over the entire twentieth century.

Table 6.

Evaluation of individual CMIP5 GCMs relative to annual CRU temperatures based on Pearson’s correlation coefficients (R), mean error (ME), mean absolute error (MAE), standard deviation of the error (SDE), and root-mean-square error (RMSE). Included also are the twentieth-century trends. Statistically significant differences, correlations, and trends (95% level) are shown in boldface.

Table 6.
Fig. 5.
Fig. 5.

Comparison between CRU observations and the individual CMIP5 GCMs along the x-axis for (top row) annual and (next rows) seasonal temperatures: (left) temperature bias from observations and (right) correlations between observations and individual models. Error bars (red) indicate standard deviation of temperature biases; stars (green) indicate significant (95% level) correlations.

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

To investigate the seasonal variability in temperature trends during the twentieth century, we calculate the monthly temperature trends from CRU and UDEL observations, CMIP3 and CMIP5 multimodel ensemble averages, and the individual GCMs (Fig. 6). Between the two observational products, CRU consistently has a greater warming trend than UDEL throughout the year. As discussed before, the two observational products exhibit a clear seasonality in temperature trends for China. Larger warming trends occurred in the cold season, and less warming or even cooling trends occurred in the warm season. However, this seasonality was not captured by CMIP3, and CMIP5 also only shows a very weak seasonal pattern in temperature trends. As was the case for the annual and seasonal trends, CMIP5 trends are lower than CMIP3 for all individual months of the year. Based on the monthly temperature trends in the individual GCMs, some models show cooling trends during most parts of the year, including ACCESS1.3, GFDL-CM3, HadGEM2-CC, and MIROC5. Several individual models did capture the seasonality shown by the observations, including BCC, CCSM4, CMCC-CESM, CSIRO-Mk3.6, GISS-E2-H, and MIROC-ESM. However, despite capturing the seasonality of the trends, the magnitude is generally much lower. For the period 1950–99, both CRU and UDEL exhibit temperature increases for all months, with the greatest warming in February, December, and January, and the least warming in August, July, and June. As for the entire twentieth century, both CMIP3 and CMIP5 show a much weaker annual cycle. Some models do not capture the accelerated warming in winter months, such as HadGEM2-CC, GFDL-CM3, and MIROC5.

Fig. 6.
Fig. 6.

Trends in seasonal temperatures over the twentieth century for China derived from CRU (red line) and UDEL (blue line) observations, and CMIP3 (dashed line) and CMIP5 (solid line) multimodel ensemble averages. The vertical narrow color bars indicate temperature trends derived from the individual CMIP5 models.

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

4. Temperatures in the twenty-first century

The projected future temperatures for China as a whole under three emission scenarios are shown in Fig. 7. RCP 8.5 and RCP 4.5 exhibit a gradual increase in annual temperature during the twenty-first century at a rate of 0.60° and 0.27°C (10 yr)−1, respectively. As the lowest-emission mitigation scenario, the RCP 2.6 experiment projects the lowest rate of temperature increase [0.10°C (10 yr)−1]. By the end of the twenty-first century, temperature will increase by 1.78°–5.66°C over China (Table 7). Under the RCP 2.6 scenario, temperature will increase until 2040, and then remain stable or even decrease slightly. This indicates the effectiveness of anticipated climate mitigation strategies, while largely reflecting the design of the RCP scenarios in terms of the radiative forcing (see section 2a). Additionally, in the near term (before the 2030s) the temperature increase under RCP 2.6 is greater than under RCP 4.5, even though its radiative forcing is lower than RCP 4.5.

Fig. 7.
Fig. 7.

Time series of historical (black line) and projected temperature for China from different CMIP5 experiments during 1900–2100: RCP 8.5 (red line), 4.5 (yellow line), and 2.6 (green line).

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

Table 7.

Projected warming in different subregions of China during the twenty-first century and the temperature change for 2090–99 relative to 1980–99. All trends and changes are statistically significant (95% level).

Table 7.

Figure 8 illustrates the spatial pattern of annual temperature change over China during the twenty-first century under the three emission scenarios. Interestingly, there is not a common spatial pattern in temperature trends among the three scenarios (Figs. 8a–c). RCP 2.6 shows the greatest warming across eastern China, as well as in some isolated regions in the Himalayas and northwestern China. However, the greatest warming under the RCP 4.5 scenario will occur on the western Tibetan Plateau, while drastic countrywide warming is projected for RCP 8.5. We also calculated the temperature difference between the periods 2070–99 and 1961–90 (Figs. 8d,e) to determine where changes will be greatest relative to the twentieth century. We find significantly higher temperature in the late twenty-first century compared to the three decades in the late twentieth century. Unlike the trend patterns, the temperature differences exhibit a somewhat consistent spatial pattern among the three scenarios. For RCP 4.5 and 8.5, the greatest temperature increase will occur over the Tibetan Plateau, the northwest, and the northeast, with smaller increases over eastern China. While the temperature changes are of course smallest under the RCP 2.6 scenario, the Tibetan Plateau would experience relatively smaller temperature changes than the northwest and northeast, with east-central China also being part of this greater warming region. Table 7 shows projected temperature changes in the different subregions of China under the three emission scenarios. For RCP 2.6, there are similar temperature trends in the four subregions, with a slightly greater increase over the northwest and northeast (~1.9°C warming by the end of the twenty-first century). With the increase in radiative forcing in future projections, the northwest and the Tibetan Plateau will have a larger warming trend, and the northwest will experience the greatest temperature increase, 3.1° and 6.2°C for RCP 4.5 and RCP 8.5, respectively.

Fig. 8.
Fig. 8.

(left) Annual temperature trends for the twenty-first century for three emission scenarios and (right) changes in annual temperature (2070–99 vs 1961–90) based on CMIP5 projections: RCP (top) 2.6, (middle) 4.5, and (bottom) 8.5. Stippled regions indicate statistically significant trends/differences (95% level).

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

Figure 9 shows the annual as well as the winter and summer temperature trends in the individual CMIP5 GCMs for the twenty-first century. The variability of temperature trends among the 20 models shows virtually the same pattern under the three scenarios. Several models project greater rates of temperature increase than others, such as CMCC-CESM, CSIRO-Mk3.6, GFDL-CM3, HadGEM2-CC, and MIROC-ESM. It is unclear which of these twenty-first century trend projections may be the most reliable, given that those models that agreed best with observations in terms of the mean climate states (see Table 6; e.g., MPI-ESM-LR, CanESM2, MIROC-ESM, and CCSM4) nonetheless did not capture the observed twentieth-century trends. There is also limited agreement in the seasonal patterns of future temperature trends among the 20 GCMs. For instance, CMCC-CESM, HadGEM2-CC, and NorESM1-M exhibit much higher winter warming while some other models show no obvious difference in temperature trends between the two seasons. This disagreement is evident for all three scenarios. For MIROC-ESM, there is much greater warming in winter than in summer for RCP 8.5. However, under RCP 4.5 and RCP 2.6, the warming trend is greater in summer relative to winter.

Fig. 9.
Fig. 9.

Trends in annual (blue) and summer (green) and winter (red) temperatures over the twenty-first century derived from individual GCMs for the three experiments: RCP (top) 2.6, (middle) 4.5, and (bottom) 8.5.

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

An additional interesting observation regarding the modeled trends is that, when compared to trends in the HIST experiment, the models with the lower temperature trends during the twentieth century usually exhibit greater warming trends during the twenty-first century. Figure 10 illustrates the significant negative correlation between the historical and RCP 4.5 projected temperature trends (R = −0.70, p < 0.01). The negative relationship also exists for the RCP 2.6 and RCP 8.5 experiments, but is not as strong as for RCP 4.5.

Fig. 10.
Fig. 10.

Scatterplot of historical temperature trends during the twentieth century and the projected temperature trend under the RCP 4.5 scenario.

Citation: Journal of Climate 27, 11; 10.1175/JCLI-D-13-00465.1

5. Discussion

The objective of this study is to evaluate CMIP5’s performance in simulating surface air temperature during the twentieth century, relative to two observational datasets. Compared with UDEL, CRU shows a stronger warming trend during the twentieth century of 0.79°C (100 yr)−1, while UDEL only shows ~60% of that warming trend [0.48°C (100 yr)−1]. This discrepancy between the two observational products can be attributed to two factors. First, as mentioned in section 2, different data sources were assimilated into the respective databases. Second, these two datasets were generated based on different interpolation methods. Based on four different observational datasets, Zhao et al. (2005) suggested that the country-averaged annual mean surface air temperature has increased at a rate of 0.2°–0.8°C (100 yr)−1 from 1901 to 1999. Greater warming trends, 0.3°–1.2°C (100 yr)−1, were detected for the period 1906–2005 (Ren et al. 2012). Based on previous studies (e.g., Hu et al. 2003), there was a pronounced warming from 1951 to 2000 in the entire country in winter, spring, and autumn, particularly in the north. There was also a summer cooling trend reported for central China. These conclusions agree well with the results from CRU and UDEL. Therefore, CRU and UDEL provide reasonable observational temperature trends, and are likely valid for our model evaluation.

In our results, both CMIP3 and CMIP5 show a smaller warming trend than CRU observations. CMIP3 shows a more consistently linear warming trend during the whole century than CMIP5. This is also found by Knutti and Sedlacek (2013), who explain that the twentieth-century historical simulations in CMIP5 included more diverse and complete radiative forcings (shown in Table 1), but some models in CMIP3 did not consider solar and volcanic forcings, or aerosol effects. Zhou and Yu (2006) found that there is a significant correlation between CMIP3 simulations and observed annual temperatures over all of China. However, most of the CMIP3 models failed to reproduce the summertime cooling trend in the middle part of eastern China, and many models underestimate the winter warming trend. Our CMIP3 results agree well with these findings. However, our findings for CMIP5 indicate little improvement in simulating the spatial and seasonal patterns of temperature trends.

Compared with observations, there is a significant cold bias over the Tibetan Plateau where we also observe the largest intermodel variation. This disagreement implies that a common deficiency among the CMIP5 models still exists for reproducing atmospheric processes in such a highly spatially heterogeneous and complex terrain. These cold biases have also been reported in previous studies. Substantial cold biases were found over high plateaus, especially the Tibetan Plateau, in GCM simulations (Annan et al. 2005; Gao et al. 2011; Ji and Kang 2013; Su et al. 2013). Seasonally, this cold bias is largest during the cold season and smallest in the warm season, implying that models may fail to represent snow–albedo feedbacks over this mountain region. Additionally, previous studies suggest that model deficiencies in simulating cloud properties over the plateau may introduce insufficient plateau heating, therefore resulting in temperature biases over the Tibetan Plateau (Zhou and Li 2002; Yu et al. 2004).

Based on our model evaluation, several CMIP5 models were identified as the most suitable models for China, including MPI-ESM-LR, CanESM2, MIROC-ESM, and CCSM4. Before CMIP5 had been completed, (Chen et al. 2011) evaluated 28 atmosphere–ocean GCMs and five models were identified with better performance over China, including ECHAM4, the Hadley Centre Coupled Model, version 3 (HadCM3), the Commonwealth Scientific and Industrial Research Organisation Mark, version 3.5 (CSIRO Mk3.5), the National Center for Atmospheric Research (NCAR) Community Climate System Model, version 3 (CCSM3), and the Model for Interdisciplinary Research on Climate, version 3.2 (MIROC3.2). Except for CanESM2, MPI-ESM-LR, MIROC-ESM, and CCSM4 still show the same good results as their earlier versions, indicating a consistently better performance of these models (MPI-ESM and MIROC-ESM are new Earth system models incorporating ECHAM and MIROC, respectively). MPI-ESM-LR and CanESM2 were also ranked in the top five of the best models in simulating temperature over the Tibetan Plateau by Su et al. (2013).

Future temperature projections show that there will be continued warming over China. The greatest warming trend will occur over northern China and the Tibetan Plateau. However, these spatial trend patterns appear to be the least reliable statistics calculated from the CMIP5 archive. Severe temperature increases such as the ones projected as part of RCP 8.5 will probably aggravate environmental degradation in the northern China, such as drought and desertification, which have been documented to be serious problems already (Wang et al. 2008). Over the Tibetan Plateau, which is sometimes called the “third pole” of the earth, glacier retreat has occurred since the 1960s and has intensified in the past 10 years (Yao et al. 2007). Since the Tibetan Plateau is the source region for many of the major rivers of China, further warming may generate substantial hydrological impacts over China.

6. Conclusions

This study evaluated the performance of 20 CMIP5 GCMs in simulating surface air temperature variability over China during the twentieth century with respect to two observational datasets. For seasonal and annual mean temperatures, GCMs show substantial cold biases over the Tibetan Plateau, especially in the cold season. These cold biases over the Tibetan Plateau are also characterized by the greatest disagreement among the individual models, indicating GCMs’ deficiencies in reproducing climatic features in this complex, high-elevation terrain. CMIP5 shows slightly better agreement with observations than CMIP3 in terms of temperature biases. Both CMIP3 and CMIP5 exhibit climatic warming over the twentieth century with an accelerated warming during the second half of the century. However, annual mean temperature trends are underestimated, and the seasonal trends are poorly simulated. The spatial pattern of temperature trends over China is also not simulated well.

Based on six statistical measures, four CMIP5 models better simulate historical surface air temperature variability over China: MPI-ESM-LR, CanESM2, MIROC-ESM, and CCSM4. The two observational products both exhibit clear seasonality in temperature trends during the twentieth century: larger warming trends occurred in cold season, with less warming trend or cooling during the warm season. However, the multimodel ensembles (both CMIP3 and CMIP5) as well as most individual GCMs did not capture this seasonal pattern, in particular ACCESS1.3, GFDL-CM3, HadGEM2-CC, and MIROC5.

The future temperature projections for China indicate that the RCP 8.5 and RCP 4.5 scenarios exhibit a consistent increase in annual temperature during the twenty-first century at a rate of 0.60° and 0.27°C (10 yr)−1, respectively. The lowest-emission mitigation scenario, RCP 2.6, produces the lowest rate of warming [0.10°C (10 yr)−1], all of which is projected to occur by approximately 2040. By the end of the twenty-first century, temperature is projected to increase by 1.7°–5.7°C, with the larger warming over northern China and the Tibetan Plateau.

Acknowledgments

We acknowledge the organizations and individuals who provided the model simulations and observations for our study. We also thank the editor and reviewers for their constructive and thoughtful comments, which helped us to improve this manuscript.

APPENDIX

Expansions of CMIP5 Models Used in this Study

ACCESS1.3 Australian Community Climate and Earth-System Simulator, version 1.3

BCC-CSM1.1 Beijing Climate Center, Climate System Model, version 1.1

CanESM2 Second Generation Canadian Earth System Model

CCSM4 Community Climate System Model, version 4

CESM1(CAM5) Community Earth System Model, version 1 (Community Atmosphere Model, version 5)

CMCC-CESM Centro Euro-Mediterraneo sui Cambiamenti Climatici Community Earth System Model

CNRM-CM5 Centre National de Recherches Météorologiques Coupled Global Climate Model, version 5

CSIRO-Mk3.6.0 Commonwealth Scientific and Industrial Research Organisation Mark, version 3.6.0

EC-EARTH EC-Earth Consortium

FGOALS-g2 Flexible Global Ocean–Atmosphere–Land System Model gridpoint, version 2

GFDL CM3 Geophysical Fluid Dynamics Laboratory Climate Model, version 3

GISS-E2-H Goddard Institute for Space Studies Model E2, coupled with the Hybrid Coordinate Ocean Model (HYCOM)

HadGEM2-CC Hadley Centre Global Environment Model, version 2–Carbon Cycle

INM-CM4 Institute of Numerical Mathematics Coupled Model, version 4.0

IPSL-CM5A-LR L’Institut Pierre-Simon Laplace Coupled Model, version 5A, low resolution

MIROC-ESM Model for Interdisciplinary Research on Climate, Earth System Model

MIROC5 Model for Interdisciplinary Research on Climate, version 5

MPI-ESM-LR Max Planck Institute Earth System Model, low resolution

MRI-CGCM Meteorological Research Institute Coupled Atmosphere–Ocean General Circulation Model

NorESM1-M Norwegian Earth System Model, version 1 (intermediate resolution)

REFERENCES

  • Annan, J. D., D. J. Lunt, J. C. Hargreaves, and P. J. Valdes, 2005: Parameter estimation in an atmospheric GCM using the ensemble Kalman filter. Nonlinear Processes Geophys., 12, 363371, doi:10.5194/npg-12-363-2005.

    • Search Google Scholar
    • Export Citation
  • Chen, W., Z. Jiang, and L. Li, 2011: Probabilistic projections of climate change over China under the SRES A1B scenario using 28 AOGCMs. J. Climate, 24, 47414756, doi:10.1175/2011JCLI4102.1.

    • Search Google Scholar
    • Export Citation
  • Ding, Y., G. Ren, Z. Zhao, Y. Xu, Y. Luo, Q. Li, and J. Zhang, 2007: Detection, causes and projection of climate change over China: An overview of recent progress. Adv. Atmos. Sci., 24, 954971, doi:10.1007/s00376-007-0954-4.

    • Search Google Scholar
    • Export Citation
  • Fan, Y., and H. van den Dool, 2008: A global monthly land surface air temperature analysis for 1948–present. J. Geophys. Res., 113, D01103, doi:10.1029/2007JD008470.

    • Search Google Scholar
    • Export Citation
  • Gao, J., V. Masson-Delmotte, T. Yao, L. Tian, C. Risi, and G. Hoffmann, 2011: Precipitation water stable isotopes in the south Tibetan plateau: Observations and modeling. J. Climate, 24, 31613178, doi:10.1175/2010JCLI3736.1.

    • Search Google Scholar
    • Export Citation
  • Giorgi, F., X. Bi, and J. S. Pal, 2004: Mean, interannual variability and trends in a regional climate change experiment over Europe. I. Present-day climate (1961–1990). Climate Dyn., 22, 733756, doi:10.1007/s00382-004-0409-x.

    • Search Google Scholar
    • Export Citation
  • Gong, D., and S. Wang, 1999: 1998: The warmest year on record of the century in China (in Chinese). Meteor. Monogr., 25, 35.

  • Hansen, J., R. Ruedy, M. Sato, and K. Lo, 2010: Global surface temperature change. Rev. Geophys., 48, RG4004, doi:10.1029/2010RG000345.

  • Harris, I., P. D. Jones, T. J. Osborn, and D. H. Lister, 2013: Updated high-resolution grids of monthly climatic observations—The CRU TS3.10 dataset. Int. J. Climatol., 34, 623–642, doi:10.1002/joc.3711.

    • Search Google Scholar
    • Export Citation
  • Hu, Z.-Z., S. Yang, and R. Wu, 2003: Long-term climate variations in China and global warming signals. J. Geophys. Res., 108, 4614, doi:10.1029/2003JD003651.

    • Search Google Scholar
    • Export Citation
  • Hua, W., H. Chen, and S. Sun, 2014: Uncertainty in land surface temperature simulation over China by CMIP3/CMIP5 models. Theor. Appl. Climatol., doi:10.1007/s00704-013-1020-z, in press.

    • Search Google Scholar
    • Export Citation
  • Jacob, D., and Coauthors, 2007: An inter-comparison of regional climate models for Europe: Model performance in present-day climate. Climatic Change, 81, 3152, doi:10.1007/s10584-006-9213-4.

    • Search Google Scholar
    • Export Citation
  • Ji, Z., and S. Kang, 2013: Double-nested dynamical downscaling experiments over the Tibetan Plateau and their projection of climate change under two RCP scenarios. J. Atmos. Sci., 70, 12781290, doi:10.1175/JAS-D-12-0155.1.

    • Search Google Scholar
    • Export Citation
  • Kharin, V. V., F. W. Zwiers, X. Zhang, and G. C. Hegerl, 2007: Changes in temperature and precipitation extremes in the IPCC ensemble of global coupled model simulations. J. Climate, 20, 14191444, doi:10.1175/JCLI4066.1.

    • Search Google Scholar
    • Export Citation
  • Knutti, R., and J. Sedlacek, 2013: Robustness and uncertainties in the new CMIP5 climate model projections. Nat. Climate Change, 3, 369373, doi:10.1038/nclimate1716.

    • Search Google Scholar
    • Export Citation
  • Kumar, S., V. Merwade, J. L. Kinter, and D. Niyogi, 2013: Evaluation of temperature and precipitation trends and long-term persistence in CMIP5 20th century climate simulations. J. Climate, 26, 41684185, doi:10.1175/JCLI-D-12-00259.1.

    • Search Google Scholar
    • Export Citation
  • Legates, D. R., and C. J. Willmott, 1990a: Mean seasonal and spatial variability in global surface air temperature. Theor. Appl. Climatol., 41, 1121, doi:10.1007/BF00866198.

    • Search Google Scholar
    • Export Citation
  • Legates, D. R., and C. J. Willmott, 1990b: Mean seasonal and spatial variability in gauge-corrected, global precipitation. Int. J. Climatol., 10, 111127, doi:10.1002/joc.3370100202.

    • Search Google Scholar
    • Export Citation
  • Li, H., L. Feng, and T. Zhou, 2011: Multi-model projection of July–August climate extreme changes over China under CO2 doubling. Part II: Temperature. Adv. Atmos. Sci., 28, 448463, doi:10.1007/s00376-010-0052-x.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., and Coauthors, 2007: THE WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 13831394, doi:10.1175/BAMS-88-9-1383.

    • Search Google Scholar
    • Export Citation
  • Miao, C., Q. Duan, L. Yang, and A. G. L. Borthwick, 2012: On the applicability of temperature and precipitation data from CMIP3 for China. PLoS ONE, 7, e44659, doi:10.1371/journal.pone.0044659.

    • Search Google Scholar
    • Export Citation
  • Qian, W., and A. Qin, 2006: Spatial-temporal characteristics of temperature variation in China. Meteor. Atmos. Phys., 93, 116, doi:10.1007/s00703-005-0163-6.

    • Search Google Scholar
    • Export Citation
  • Ren, G., Y. Ding, Z. Zhao, J. Zheng, T. Wu, G. Tang, and Y. Xu, 2012: Recent progress in studies of climate change in China. Adv. Atmos. Sci., 29, 958977, doi:10.1007/s00376-012-1200-2.

    • Search Google Scholar
    • Export Citation
  • Riahi, K., and Coauthors, 2011: RCP 8.5—A scenario of comparatively high greenhouse gas emissions. Climatic Change, 109, 3357, doi:10.1007/s10584-011-0149-y.

    • Search Google Scholar
    • Export Citation
  • Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K. Averyt, M. M. B. Tignor, and H. L. Miller Jr., Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.

  • Su, F., X. Duan, D. Chen, Z. Hao, and L. Cuo, 2013: Evaluation of the global climate models in the CMIP5 over the Tibetan Plateau. J. Climate, 26, 31873208, doi:10.1175/JCLI-D-12-00321.1.

    • Search Google Scholar
    • Export Citation
  • Tang, G., and G. Ren, 2005: Reanalysis of surface air temperature change of the last 100 years over China (in Chinese). Climatic Environ. Res., 10, 791798.

    • Search Google Scholar
    • Export Citation
  • Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485498, doi:10.1175/BAMS-D-11-00094.1.

    • Search Google Scholar
    • Export Citation
  • Thomson, A., and Coauthors, 2011: RCP4.5: A pathway for stabilization of radiative forcing by 2100. Climatic Change, 109, 7794, doi:10.1007/s10584-011-0151-4.

    • Search Google Scholar
    • Export Citation
  • van Vuuren, D. P., and Coauthors, 2008: Temperature increase of twenty-first century mitigation scenarios. Proc. Natl. Acad. Sci. USA, 105, 15 25815 262, doi:10.1073/pnas.0711129105.

    • Search Google Scholar
    • Export Citation
  • van Vuuren, D. P., and Coauthors, 2011: The representative concentration pathways: An overview. Climatic Change, 109, 531, doi:10.1007/s10584-011-0148-z.

    • Search Google Scholar
    • Export Citation
  • Wang, G., S. Sun, and R. Mei, 2011: Vegetation dynamics contributes to the multi-decadal variability of precipitation in the Amazon region. Geophys. Res. Lett., 38, L19703, doi:10.1029/2011GL049017.

    • Search Google Scholar
    • Export Citation
  • Wang, L., and W. Chen, 2014: A CMIP5 multimodel projection of future temperature, precipitation, and climatological drought in China. Int. J. Climatol., doi:10.1002/joc.3822, in press.

    • Search Google Scholar
    • Export Citation
  • Wang, S., and D. Gong, 2000: Enhancement of the warming trend in China. Geophys. Res. Lett., 27, 25812584, doi:10.1029/1999GL010825.

  • Wang, S., J. Ye, D. Gong, J. Zhu, and T. Yao, 1998: Construction of mean annual temperature series for the last one hundred years in China (in Chinese). Quart. J. Appl. Meteor., 9, 392401.

    • Search Google Scholar
    • Export Citation
  • Wang, S., D. Gong, and J. Zhu, 2001: Twentieth-century climatic warming in China in the context of the Holocene. Holocene, 11, 313321, doi:10.1191/095968301673172698.

    • Search Google Scholar
    • Export Citation
  • Wang, X., F. Chen, E. Hasi, and J. Li, 2008: Desertification in China: An assessment. Earth Sci. Rev., 88, 188206, doi:10.1016/j.earscirev.2008.02.001.

    • Search Google Scholar
    • Export Citation
  • Wild, M., 2009: How well do IPCC-AR4/CMIP3 climate models simulate global dimming/brightening and twentieth-century daytime and nighttime warming? J. Geophys. Res., 114, D00D11, doi:10.1029/2008JD011372.

    • Search Google Scholar
    • Export Citation
  • Willmott, C. J., and S. M. Robeson, 1995: Climatologically aided interpolation (CAI) of terrestrial air temperature. Int. J. Climatol., 15, 221229, doi:10.1002/joc.3370150207.

    • Search Google Scholar
    • Export Citation
  • Xu, Y., X. Gao, Y. Shen, C. Xu, Y. Shi, and F. Giorgi, 2009: A daily temperature dataset over China and its application in validating a RCM simulation. Adv. Atmos. Sci., 26, 763772, doi:10.1007/s00376-009-9029-z.

    • Search Google Scholar
    • Export Citation
  • Yao, T., J. Pu, A. Lu, Y. Wang, and W. Yu, 2007: Recent glacial retreat and its impact on hydrological processes on the Tibetan Plateau, China, and surrounding regions. Arct. Antarct. Alp. Res., 39, 642650, doi:10.1657/1523-0430(07-510)[YAO]2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Yu, R., B. Wang, and T. Zhou, 2004: Climate effects of the deep continental stratus clouds generated by the Tibetan Plateau. J. Climate, 17, 27022713, doi:10.1175/1520-0442(2004)017<2702:CEOTDC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zhao, Z., S. Wang, Y. Xu, G. Ren, Y. Luo, and X. Gao, 2005: Attribution of the 20th century climate warming in China (in Chinese). Climatic Environ. Res., 10, 808817.

    • Search Google Scholar
    • Export Citation
  • Zhou, T., and Z. Li, 2002: Simulation of the East Asian summer monsoon using a variable resolution atmospheric GCM. Climate Dyn., 19, 167180, doi:10.1007/s00382-001-0214-8.

    • Search Google Scholar
    • Export Citation
  • Zhou, T., and R. Yu, 2006: Twentieth-century surface air temperature over China and the globe simulated by coupled climate models. J. Climate, 19, 58435858, doi:10.1175/JCLI3952.1.

    • Search Google Scholar
    • Export Citation
Save