• Ajayamohan, R. S., S. A. Rao, J.-J. Luo, and T. Yamagata, 2009: Influence of Indian Ocean Dipole on boreal summer intraseasonal oscillations in a coupled general circulation model. J. Geophys. Res., 114 , D06119. doi:10.1029/2008JD011096.

    • Search Google Scholar
    • Export Citation
  • Alexander, M. A., M. S. Timlin, and J. D. Scott, 2001: Winter-to-winter recurrence of sea surface temperature, salinity and mixed layer depth anomalies. Prog. Oceanogr., 49 , 4161.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., J.-J. Luo, S. Masson, P. Delecluse, S. Gualdi, A. Navarra, and T. Yamagata, 2005: Paramount impact of the Indian Ocean dipole on the East African short rain: A CGCM study. J. Climate, 18 , 45144530.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., J.-J. Luo, S. Masson, S. A. Rao, H. Sakuma, and T. Yamagata, 2006: A CGCM study on the interaction between IOD and ENSO. J. Climate, 19 , 16881705.

    • Search Google Scholar
    • Export Citation
  • Bekryaev, R. V., I. V. Polyakov, and V. A. Alexeev, 2010: Role of polar amplification in long-term surface air temperature variations and modern Arctic warming. J. Climate, 23 , 38883906.

    • Search Google Scholar
    • Export Citation
  • Boer, G. J., 2000: A study of atmosphere–ocean predictability on long time scales. Climate Dyn., 16 , 469477.

  • Cai, M., C.-S. Shin, H. M. Van den Dool, W. Wang, S. Saha, and A. Kumar, 2009: The role of long-term trends in seasonal predictions: Implication of global warming in the NCEP CFS. Wea. Forecasting, 24 , 965973.

    • Search Google Scholar
    • Export Citation
  • Cherchi, A., S. Gualdi, S. Behera, J.-J. Luo, S. Masson, T. Yamagata, and A. Navarra, 2007: The influence of tropical Indian Ocean SST on the Indian summer monsoon. J. Climate, 20 , 30833105.

    • Search Google Scholar
    • Export Citation
  • Chowdary, J. S., S.-P. Xie, J.-J. Luo, J. Hafner, S. Behera, Y. Masumoto, and T. Yamagata, 2011: Predictability of Northwest Pacific climate during summer and the role of the tropical Indian Ocean. Climate Dyn., 36 , 607621. doi:10.1007/s00382-009-0686-5.

    • Search Google Scholar
    • Export Citation
  • Church, J. A., N. J. White, and J. M. Arblaster, 2005: Significant decadal-scale impact of volcanic eruptions on sea level and ocean heat content. Nature, 438 , 7477.

    • Search Google Scholar
    • Export Citation
  • Collins, M., 2002: Climate predictability on interannual to decadal time scales: The initial value problem. Climate Dyn., 19 , 671692.

    • Search Google Scholar
    • Export Citation
  • Collins, M., and M. R. Allen, 2002: Assessing the relative roles of initial and boundary conditions in interannual to decadal climate predictability. J. Climate, 15 , 31043109.

    • Search Google Scholar
    • Export Citation
  • Compo, G. P., and P. D. Sardeshmukh, 2009: Oceanic influences on recent continental warming. Climate Dyn., 32 , 333342.

  • Cox, P., and D. Stephenson, 2007: A changing climate for prediction. Science, 317 , 207208.

  • Doblas-Reyes, F. J., R. Hagedorn, T. N. Palmer, and J.-J. Morcrette, 2006: Impact of increasing greenhouse gas concentrations in seasonal ensemble forecasts. Geophys. Res. Lett., 33 , L07708. doi:10.1029/2005GL025061.

    • Search Google Scholar
    • Export Citation
  • Dommenget, D., 2009: The ocean’s role in continental climate variability and change. J. Climate, 22 , 49394952.

  • Dong, B., J. M. Gregory, and R. T. Sutton, 2009: Understanding land–sea warming contrast in response to increasing greenhouse gases. Part I: Transient adjustment. J. Climate, 22 , 30793097.

    • Search Google Scholar
    • Export Citation
  • Fyfe, J. C., N. P. Gillett, and D. W. J. Thompson, 2010: Comparing variability and trends in observed and modelled global-mean surface temperature. Geophys. Res. Lett., 37 , L16802. doi:10.1029/2010GL044255.

    • Search Google Scholar
    • Export Citation
  • Griffies, S. M., and K. Bryan, 1997: Predictability of North Atlantic multidecadal climate variability. Science, 275 , 181184.

  • Gualdi, S., A. Navarra, E. Guilyardi, and P. Delecluse, 2003: Assessment of the tropical Indo-Pacific climate in the SINTEX CGCM. Ann. Geophys., 46 , 126.

    • Search Google Scholar
    • Export Citation
  • Guilyardi, E., P. Delecluse, S. Gualdi, and A. Navarra, 2003: Mechanisms for ENSO phase change in a coupled GCM. J. Climate, 16 , 11411158.

    • Search Google Scholar
    • Export Citation
  • Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19 , 56865699.

  • Hermanson, L., and R. T. Sutton, 2009: Case studies in interannual to decadal climate predictability. Climate Dyn., 35 , 11691189. doi:10.1007/s00382-009-0672-y.

    • Search Google Scholar
    • Export Citation
  • Hong, C.-C., T. Li, and J.-J. Luo, 2008: Asymmetry of the Indian Ocean dipole. Part II: Model diagnosis. J. Climate, 21 , 48494858.

  • Izumo, T., C. Montegut, J.-J. Luo, S. Behera, S. Masson, and T. Yamagata, 2008: The role of the western Arabian Sea upwelling in Indian monsoon rainfall variability. J. Climate, 21 , 56035623.

    • Search Google Scholar
    • Export Citation
  • Jin, E. K., and Coauthors, 2008: Current status of ENSO prediction skill in coupled ocean–atmosphere models. Climate Dyn., 31 , 647664.

    • Search Google Scholar
    • Export Citation
  • Joshi, M. M., J. M. Gregory, M. J. Webb, D. M. H. Sexton, and T. C. Johns, 2008: Mechanisms for the land/sea warming contrast exhibited by simulations of climate change. Climate Dyn., 30 , 455465.

    • Search Google Scholar
    • Export Citation
  • Karnauskas, K. B., R. Seager, A. Kaplan, Y. Kushnir, and M. A. Cane, 2009: Observed strengthening of the zonal sea surface temperature gradient across the equatorial Pacific Ocean. J. Climate, 22 , 43164321.

    • Search Google Scholar
    • Export Citation
  • Keenlyside, N. S., M. Latif, J. Jungclaus, L. Kornblueh, and E. Roeckner, 2008: Advancing decadal-scale climate prediction in the North Atlantic sector. Nature, 453 , 8488.

    • Search Google Scholar
    • Export Citation
  • Kug, J.-S., T. Li, S.-I. An, I.-S. Kang, J.-J. Luo, S. Masson, and T. Yamagata, 2006: Role of the ENSO–Indian Ocean coupling on ENSO variability in a coupled GCM. Geophys. Res. Lett., 33 , L09710. doi:10.1029/2005GL024916.

    • Search Google Scholar
    • Export Citation
  • Lambert, F. H., and J. C. H. Chiang, 2007: Control of land–ocean temperature contrast by ocean heat uptake. Geophys. Res. Lett., 34 , L13704. doi:10.1029/2007GL029755.

    • Search Google Scholar
    • Export Citation
  • Lambert, F. H., and M. J. Webb, 2008: Dependency of global mean precipitation on surface temperature. Geophys. Res. Lett., 35 , L16706. doi:10.1029/2008GL034838.

    • Search Google Scholar
    • Export Citation
  • Lee, J.-Y., and Coauthors, 2010: How is the performance on mean climate related to seasonal prediction skill in coupled climate models? Climate Dyn., 35 , 267283.

    • Search Google Scholar
    • Export Citation
  • Liepert, B. G., and M. Previdi, 2009: Do models and observations disagree on the rainfall response to global warming? J. Climate, 22 , 31563166.

    • Search Google Scholar
    • Export Citation
  • Liniger, M. A., H. Mathis, C. Appenzeller, and F. J. Doblas-Reyes, 2007: Realistic greenhouse gas forcing and seasonal forecasts. Geophys. Res. Lett., 34 , L04705. doi:10.1029/2006GL028335.

    • Search Google Scholar
    • Export Citation
  • Lorenz, E. N., 1975: The physical bases of climate and climate modelling. Climate Predictability, GARP Publication Series, Vol. 16, WMO, 132–136.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., and T. Yamagata, 2001: Long-term El Niño–Southern Oscillation (ENSO)-like variation with special emphasis on the South Pacific. J. Geophys. Res., 106 , 2221122227.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, P. Delecluse, S. Gualdi, A. Navarra, and T. Yamagata, 2003: South Pacific origin of the decadal ENSO-like variation as simulated by a coupled GCM. Geophys. Res. Lett., 30 , 2250. doi:10.1029/2003GL018649.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, E. Roeckner, G. Madec, and T. Yamagata, 2005a: Reducing climatology bias in an ocean–atmosphere CGCM with improved coupling physics. J. Climate, 18 , 23442360.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, S. Shingu, and T. Yamagata, 2005b: Seasonal climate predictability in a coupled OAGCM using a different approach for ensemble forecasts. J. Climate, 18 , 44744497.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, and T. Yamagata, 2007: Experimental forecasts of the Indian Ocean dipole using a coupled OAGCM. J. Climate, 20 , 21782190.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Behera, Y. Masumoto, H. Sakuma, and T. Yamagata, 2008a: Successful prediction of the consecutive IOD in 2006 and 2007. Geophys. Res. Lett., 35 , L14S02. doi:10.1029/2007GL032793.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, and T. Yamagata, 2008b: Extended ENSO predictions using a fully coupled ocean–atmosphere model. J. Climate, 21 , 8493.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., R. Zhang, S. Behera, Y. Masumoto, F.-F. Jin, R. Lukas, and T. Yamagata, 2010: Interaction between El Niño and extreme Indian Ocean dipole. J. Climate, 23 , 726742.

    • Search Google Scholar
    • Export Citation
  • Madec, G., P. Delecluse, M. Imbard, and C. Levy, 1998: OPA 8.1 ocean general circulation model reference manual. LODYC/IPSL Tech. Rep. Note 11, 91 pp.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., and Coauthors, 2001: North Atlantic climate variability: Phenomena, impacts and mechanisms. Int. J. Climatol., 21 , 18631898.

    • Search Google Scholar
    • Export Citation
  • Masson, S., and Coauthors, 2005: Impact of barrier layer on winter-spring variability of the southeastern Arabian Sea. Geophys. Res. Lett., 32 , L07703. doi:10.1029/2004GL021980.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., J. M. Arblaster, K. Matthes, F. Sassi, and H. van Loon, 2009: Amplifying the Pacific climate system response to a small 11 year solar cycle forcing. Science, 325 , 11141118.

    • Search Google Scholar
    • Export Citation
  • Navarra, A., and Coauthors, 2008: Atmospheric horizontal resolution affects tropical climate variability in coupled models. J. Climate, 21 , 730750.

    • Search Google Scholar
    • Export Citation
  • Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85 , 369432.

  • Palmer, T. N., and Coauthors, 2004: Development of a European multimodel ensemble system for seasonal-to-interannual prediction (DEMETER). Bull. Amer. Meteor. Soc., 85 , 853872.

    • Search Google Scholar
    • Export Citation
  • Rao, S. A., S. Masson, J.-J. Luo, S. K. Behera, and T. Yamagata, 2007: Termination of Indian Ocean dipole events in a coupled general circulation model. J. Climate, 20 , 30183035.

    • Search Google Scholar
    • Export Citation
  • Rao, S. A., J.-J. Luo, S. K. Behera, and T. Yamagata, 2009: Generation and termination of Indian Ocean dipole events in 2003, 2006 and 2007. Climate Dyn., 33 , 751767.

    • Search Google Scholar
    • Export Citation
  • Rashid, H., S. B. Power, and J. R. Knight, 2010: Impact of multidecadal fluctuations in the Atlantic thermohaline circulation on Indo-Pacific climate variability in a coupled GCM. J. Climate, 23 , 40384044.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Roeckner, E., and Coauthors, 1996: The atmospheric general circulation model ECHAM-4: Model description and simulation of present-day climate. Max-Planck-Institut für Meteorologie Rep. 218, 90 pp.

    • Search Google Scholar
    • Export Citation
  • Smith, D. M., S. Cusack, A. W. Colman, C. K. Folland, G. R. Harris, and J. M. Murphy, 2007: Improved surface temperature prediction for the coming decade from a global climate model. Science, 317 , 796799.

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

    • Search Google Scholar
    • Export Citation
  • Sutton, R., B. Dong, and J. M. Gregory, 2007: Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. Geophys. Res. Lett., 34 , L02701. doi:10.1029/2006GL028164.

    • Search Google Scholar
    • Export Citation
  • Tozuka, T., J.-J. Luo, S. Masson, S. Behera, and T. Yamagata, 2005: Annual ENSO simulated in a coupled ocean–atmosphere model. Dyn. Atmos. Oceans, 39 , 4160.

    • Search Google Scholar
    • Export Citation
  • Tozuka, T., J.-J. Luo, S. Masson, and T. Yamagata, 2007: Decadal modulations of the Indian Ocean dipole in the SINTEX-F1 coupled GCM. J. Climate, 20 , 28812894.

    • Search Google Scholar
    • Export Citation
  • Tozuka, T., J.-J. Luo, S. Masson, and T. Yamagata, 2008: Tropical Indian Ocean variability revealed by self-organizing maps. Climate Dyn., 31 , 333343.

    • Search Google Scholar
    • Export Citation
  • Troccoli, A., and T. N. Palmer, 2007: Ensemble decadal predictions from analysed initial conditions. Philos. Trans. Roy. Soc., A365 , 21792191.

    • Search Google Scholar
    • Export Citation
  • Valcke, S., L. Terray, and A. Piacentini, 2000: The OASIS coupler user guide version 2.4. CERFACE Tech. Rep. TR/CGMC/00-10, 85 pp.

  • Wang, B., and Coauthors, 2008: How accurately do coupled climate models predict the Asian-Australia monsoon interannual variability? Climate Dyn., 31 , 605619.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and Coauthors, 2009: Advance and prospectus of seasonal prediction: Assessment of the APCC/CliPAS 14-model ensemble retrospective seasonal prediction (1980–2004). Climate Dyn., 33 , 93117.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., L. Ricciardulli, K. Hilburn, and C. Mears, 2007: How much more rain will global warming bring? Science, 317 , 233235.

  • Yamagata, T., S. Behera, J.-J. Luo, S. Masson, M. Jury, and S. A. Rao, 2004: Coupled ocean–atmosphere variability in the tropical Indian Ocean. Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 189–212.

    • Search Google Scholar
    • Export Citation
  • Zhang, R., T. L. Delworth, and I. M. Held, 2007: Can the Atlantic Ocean drive the observed multidecadal variability in Northern Hemisphere mean temperature? Geophys. Res. Lett., 34 , L02709. doi:10.1029/2006GL028683.

    • Search Google Scholar
    • Export Citation
  • Zhang, X., F. W. Zwiers, G. C. Hegerl, F. H. Lambert, N. P. Gillett, S. Solomon, P. A. Stott, and T. Nozawa, 2007: Detection of human influence on twentieth-century precipitation trends. Nature, 448 , 461465.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–92. J. Climate, 10 , 10041020.

  • View in gallery
    Fig. 1.

    (a) Surface 2-m air temperature (T2m) anomaly (°C, relative to the 1983–2006 climatology) averaged over the globe (60°S–75°N, black line), land (red line), and ocean (blue line) based on the NCEP atmospheric reanalysis (available online at http://www.esrl.noaa.gov/psd/). All time series were smoothed with a 5-month running mean to filter out high-frequency variability. (b) T2m difference between 1997–2008 and 1983–94.

  • View in gallery
    Fig. 2.

    As in Fig. 1 but for the model 9-member ensemble mean initial conditions (ICs) and 6-month lead predictions.

  • View in gallery
    Fig. 3.

    (a) As in Fig. 1a but for the detrended global mean T2m variations based on the NCEP reanalysis (black line), model ICs (gray line), and predictions at 6-, 12-, and 24-month lead (colored lines). (b) Linear trends of the global mean T2m anomalies from 1982 to 2008 based on the NCEP data (black square), model 9-member mean ICs (gray square, blue circles for individual member), and predictions at lead times of 1–24 months (medium blue line, dashed-curves for each member). (c),(d) As in (b) but for model prediction skill and RMSEs. Red (blue) solid lines denote the practical (potential) predictability for the nondetrended (line with closed circle mark) and detrended global mean T2m anomalies (line without closed circle mark). Red (blue) short-dashed lines show the predictability if a perfect warming trend of the NCEP observations (model ensemble mean ICs) was predicted at all lead times. Black (gray) zonal solid line in (d) indicates one standard deviation of the nondetrended observations (model ICs).

  • View in gallery
    Fig. 4.

    As in Fig. 3 but for global mean SST anomalies. The SST observations are based on monthly NCEP analysis (Reynolds et al. 2002).

  • View in gallery
    Fig. 5.

    As in Fig. 3 but for global mean T2m anomalies over land.

  • View in gallery
    Fig. 6.

    As in Fig. 3 but for T2m anomalies averaged in the extratropical Northern Hemisphere (20°–75°N).

  • View in gallery
    Fig. 7.

    As in Fig. 3 but for T2m anomalies averaged over the tropics (30°S–30°N).

  • View in gallery
    Fig. 8.

    The T2m anomaly correlations (colored scale) between the nondetrended NCEP reanalysis and model 9-member mean predictions at different lead times for the period 1982–2008. Shown also are the skill improvements (contour interval: ±0.1, ±0.2, ±0.3, …; thick solid lines indicate a 0.2 contour) by assuming a perfect warming trend of the NCEP global mean T2m anomalies in the model predictions. The skill is measured based on 4° by 4° grid cells.

  • View in gallery
    Fig. 9.

    As in Fig. 3 but for global mean precipitation anomalies. The observational data is based on monthly GPCP analysis (available online at http://precip.gsfc.nasa.gov/). The model ensemble mean produces a much smaller variability (see the right y axis scale in Fig. 9a). Linear trends based on Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) (available online at http://www.esrl.noaa.gov/psd/) and NCEP CPC convection-allowing models (CAMS)–outgoing longwave radiation precipitation index (OPI) (available online at http://iridl.ldeo.columbia.edu/) datasets are also shown in (b).

  • View in gallery
    Fig. 10.

    Radiative imbalance at the top of the atmosphere (black solid line, downward positive), surface heat flux over the ocean (black short-dashed line, left scale), and over land (gray long–short-dashed line, right scale) averaged during 1982–2008 over the globe (90°S–90°N) based on model ensemble mean ICs and predictions. We note that results averaged at 60°S–75°N are similar.

  • View in gallery
    Fig. 11.

    As in Figs. 3a and 3b but for global mean temperature anomalies averaged from the sea surface to 500-m depth (T500). The observations are based on monthly NCEP ocean reanalysis (available online at http://www.cpc.ncep.noaa.gov/products/GODAS/).

  • View in gallery
    Fig. 12.

    (a) Global (60°S–75°N, right scale) and tropical (30°S–30°N, left scale) mean monthly T500 based on the SINTEX-F 100-yr free model simulations with fixed GHG concentrations (blue and black lines), model spinup from 1971 to 1981 (green and gray lines), and model ICs during 1982–2008 generated by the coupled SST-nudging approach (purple and orange lines). Sea surface net heat fluxes averaged over the two domains and individual time lengths are also shown. (b) As in (a) but for the surface heat flux damping term during the model spinup and SST-nudging periods. Restoring of sea ice to observed monthly climatology leads to some local heat damping in high latitudes (green line). (c) Global (blue lines) and tropical (black lines) mean temperature difference between 1997–2008 and 1983–94 based on the model ICs (solid lines) and NCEP ocean reanalysis (short dashed lines).

  • View in gallery
    Fig. 13.

    As in Fig. 5c but for model prediction skills based on the JRA-25 (black lines). The skills calculated based on the NCEP reanalysis (red lines) are reproduced from Fig. 5c for comparison.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 437 173 7
PDF Downloads 196 56 4

Impact of Global Ocean Surface Warming on Seasonal-to-Interannual Climate Prediction

Jing-Jia LuoResearch Institute for Global Change, JAMSTEC, Yokohama, Japan

Search for other papers by Jing-Jia Luo in
Current site
Google Scholar
PubMed
Close
,
Swadhin K. BeheraResearch Institute for Global Change, JAMSTEC, Yokohama, Japan

Search for other papers by Swadhin K. Behera in
Current site
Google Scholar
PubMed
Close
,
Yukio MasumotoDepartment of Earth and Planetary Science, The University of Tokyo, Tokyo, and Research Institute for Global Change, JAMSTEC, Yokohama, Japan

Search for other papers by Yukio Masumoto in
Current site
Google Scholar
PubMed
Close
, and
Toshio YamagataDepartment of Earth and Planetary Science, The University of Tokyo, Tokyo, and Research Institute for Global Change, JAMSTEC, Yokohama, Japan

Search for other papers by Toshio Yamagata in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Surface air temperature (SAT) over the globe, particularly the Northern Hemisphere continents, has rapidly risen over the last 2–3 decades, leading to an abrupt shift toward a warmer climate state after 1997/98. Whether the terrestrial warming might be caused by local response to increasing greenhouse gas (GHG) concentrations or by sea surface temperature (SST) rise is recently in dispute. The SST warming itself may be driven by both the increasing GHG forcing and slowly varying natural processes. Besides, whether the recent global warming might affect seasonal-to-interannual climate predictability is an important issue to be explored. Based on the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) climate prediction system in which only observed SSTs are assimilated for coupled model initialization, the present study shows that the historical SST rise plays a key role in driving the intensified terrestrial warming over the globe. The SST warming trend, while negligible for short lead predictions, has substantial impact on the climate predictability at long lead times (>1 yr) particularly in the extratropics. The tropical climate predictability, however, is little influenced by global warming. Given a perfect warming trend and/or a perfect model, global SAT and precipitation could be predicted beyond two years in advance with an anomaly correlation skill above ∼0.6.

Without assimilating ocean subsurface observations, model initial conditions show a strong spurious cooling drift of subsurface temperature; this is caused by large negative surface heat flux damping arising from the SST-nudging initialization. The spurious subsurface cooling drift acts to weaken the initial SST warming trend during model forecasts, leading to even negative trends of global SAT and precipitation at long lead times and hence deteriorating the global climate predictability. Concerning the important influence of the subsurface temperature on the global SAT trend, future efforts are required to develop a good scheme for assimilating subsurface information particularly in the extratropical oceans.

Corresponding author address: Jing-Jia Luo, Research Institute for Global Change, JAMSTEC, 3173-25 Showa-machi, Kanazawa-ku, Yokohama, Kanagawa 236-0001, Japan. Email: luo@jamstec.go.jp

Abstract

Surface air temperature (SAT) over the globe, particularly the Northern Hemisphere continents, has rapidly risen over the last 2–3 decades, leading to an abrupt shift toward a warmer climate state after 1997/98. Whether the terrestrial warming might be caused by local response to increasing greenhouse gas (GHG) concentrations or by sea surface temperature (SST) rise is recently in dispute. The SST warming itself may be driven by both the increasing GHG forcing and slowly varying natural processes. Besides, whether the recent global warming might affect seasonal-to-interannual climate predictability is an important issue to be explored. Based on the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) climate prediction system in which only observed SSTs are assimilated for coupled model initialization, the present study shows that the historical SST rise plays a key role in driving the intensified terrestrial warming over the globe. The SST warming trend, while negligible for short lead predictions, has substantial impact on the climate predictability at long lead times (>1 yr) particularly in the extratropics. The tropical climate predictability, however, is little influenced by global warming. Given a perfect warming trend and/or a perfect model, global SAT and precipitation could be predicted beyond two years in advance with an anomaly correlation skill above ∼0.6.

Without assimilating ocean subsurface observations, model initial conditions show a strong spurious cooling drift of subsurface temperature; this is caused by large negative surface heat flux damping arising from the SST-nudging initialization. The spurious subsurface cooling drift acts to weaken the initial SST warming trend during model forecasts, leading to even negative trends of global SAT and precipitation at long lead times and hence deteriorating the global climate predictability. Concerning the important influence of the subsurface temperature on the global SAT trend, future efforts are required to develop a good scheme for assimilating subsurface information particularly in the extratropical oceans.

Corresponding author address: Jing-Jia Luo, Research Institute for Global Change, JAMSTEC, 3173-25 Showa-machi, Kanazawa-ku, Yokohama, Kanagawa 236-0001, Japan. Email: luo@jamstec.go.jp

1. Introduction

Predictability of the earth’s climate can be categorized into two types: one rooted in the initial value and the other in the boundary forcing (Lorenz 1975). A well-known example for predictability of the first type is weather forecasting. Given a perfect model or ignoring model errors, predictability of weather is essentially determined by the accuracy of atmospheric initial conditions (ICs). Projection of climate change in the coming 30 years to one century, an example of the second type of predictability, is to assess how the earth’s climate would change under the radiative forcing of increasing greenhouse gases (GHGs) and other anthropogenic agents (Solomon et al. 2007). For decadal climate predictions, both initial conditions (mainly ocean) and external radiative forcing due to anthropogenic and natural agents appear to be important, as suggested by recent studies (e.g., Smith et al. 2007; Troccoli and Palmer 2007; Keenlyside et al. 2008). On seasonal-to-interannual time scales, climate predictability primarily originates from the ocean–atmosphere coupled variability in the tropics, where the ocean provides key memory and lower-boundary forcing of the atmosphere. The tropical climate signals were found to be predictable on the order of a few seasons to 1–2 yr based on current state-of-the-art fully coupled models (e.g., Palmer et al. 2004; Luo et al. 2007, 2008b; Jin et al. 2008). It is also believed that a modest predictability of global climate anomalies can be gained owing to significant impacts by the predictable tropical signals, particularly El Niño–Southern Oscillation (ENSO) events.

Predictability of climate variations and climate changes has usually been treated separately. A majority of studies on seasonal predictions including real-time experimental forecasts have neglected the potential influence of changing anthropogenic radiative forcing, whereas much evidence shows that increasing GHG concentrations may have already exerted an important role in global warming, particularly over the recent decades (Solomon et al. 2007). The changing climate background can affect daily-to-interannual climate variations and the occurrence frequency of climate extreme events. Indeed, nowadays debates often arise among climate experts when seeking possible causes for the occurrence of influential climate anomalies and extreme events. Therefore, it becomes important and of practical use to examine the role of historical GHG forcing in global climate predictability on seasonal-to-interannual time scales (e.g., Doblas-Reyes et al. 2006; Liniger et al. 2007).

Two general approaches are often adopted to assess climate predictability. One is for potential predictability assessment by assuming both model and initial conditions are perfect (e.g., Griffies and Bryan 1997; Boer 2000; Collins 2002; Collins and Allen 2002; Hermanson and Sutton 2009). The other is for practical predictability assessment by measuring model ability in forecasting the observed climate from as realistic as possible initial conditions. A number of perfect model experiments with historical anthropogenic radiative forcing suggested that GHG forcing can be potentially important for global climate predictability at lead times O(10 yr) (e.g., Collins and Allen 2002; Troccoli and Palmer 2007). On lead times <10 yr, anthropogenic radiative forcing was usually believed to be unimportant (Cox and Stephenson 2007). However, a recent study for practical predictability assessment with and without historical GHGs on the basis of the European Centre for Medium-Range Weather Forecasts (ECMWF) prediction system suggested that increasing GHGs may also be important to seasonal prediction of summer global-mean surface air temperature (SAT) even at three months lead (Doblas-Reyes et al. 2006). The anthropogenic radiative forcing is found to be helpful to improve the short lead prediction of the warming trends of seasonal mean global SAT (Liniger et al. 2007). By assuming a perfect warming trend of global SAT, Cai et al. (2009) also found a small but robust increase of skill in predicting annual mean global SAT at lead times longer than 4–5 months on the basis of the National Centers for Environmental Prediction (NCEP) 9-month prediction system.

Figure 1a shows a clear warming trend of global SAT over the recent two to three decades on which large interannual variations are superimposed, particularly for the terrestrial SAT. Impacts of some major ENSO events, such as the warming related to 1982–83, 1987–88, 1997–98, 2002–03, 2006–07 El Niños and the cooling related to 1984–86, 1988–89, 1995–96, 1999–2000, 2005–06, and 2007–08 La Niñas, are clearly seen. The global SAT is relatively lower before 1997 and jumps to a warmer state following the strong 1997–98 El Niño episode. The linear trends of global, terrestrial, and ocean mean SAT from 1982 to 2008 are 0.14°, 0.21°, and 0.10°C decade−1, respectively. We note that the warming contrast between land and ocean is a well-known fingerprint of global warming. The exacerbated terrestrial warming, particularly in mid-high-latitude Northern Hemisphere (NH), compared to the ocean surface warming (Fig. 1b) has been attributed to a more rapid response of the land surface to anthropogenic radiative forcing owing to its small heat capacity. However, recent findings have suggested that this hypothesis may be wrong (e.g., Sutton et al. 2007; Lambert and Chiang 2007). Based on atmospheric model simulations with historical sea surface temperature (SST) forcing only, Compo and Sardeshmukh (2009) have found that most of the land warming in recent decades is caused by SST rise rather than by its local response to increasing GHG forcing. We note that the SST warming itself may be driven by both increasing GHG forcing and slowly varying natural processes (Solomon et al. 2007). The SST change was found to play a dominant role in determining the land–ocean warming contrast probably via complex hydrodynamic–radiative teleconnections (Joshi et al. 2008; Compo and Sardeshmukh 2009; Dong et al. 2009). A similar surface temperature anomaly contrast between land and ocean, as well as their close phase relationship, is also seen on interannual time scales (cf. the red and blue lines in Fig. 1a); this implies a similar role of SST forcing in generating natural terrestrial SAT variations (e.g., Dommenget 2009).

These previous studies have suggested that the historical SST rise may have exerted an important influence on global warming. Given the relatively large heat capacity of the sea surface mixed layer and that SST can be influenced by internal ocean circulations, which have much longer memory, it is possible that the global warming trend and its impacts on the earth’s climate can be predicted (at least partly) up to certain lead times, even if all anthropogenic radiative forcing is fixed at current levels during model forecasts. This provides hope for enhanced prediction of climate anomalies (interannual variations + warming trend) under the increasing GHG forcing. Indeed, both ECMWF and NCEP climate prediction systems with fixed GHG concentrations showed some modest skill in predicting the recent seasonal or annual mean global SAT warming trends at short–mid lead times (i.e., Doblas-Reyes et al. 2006; Liniger et al. 2007; Cai et al. 2009). Thus, it is important to study the predictability of the long-term global SST trend and its influence on global climate and seasonal-to-interannual predictability, as a first step toward better understanding and predicting the global warming and its impacts.

For this study, we have adopted the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) ensemble climate prediction system, which is built from a fully coupled ocean–atmosphere general circulation model (GCM). In contrast to the ECMWF and NCEP prediction systems in which forecasts are initialized from realistic atmosphere and ocean conditions by assimilating all available observations, ocean and atmosphere initial conditions in the JAMSTEC prediction system are generated by simply assimilating observed SSTs into the coupled model (Luo et al. 2005b, 2008b). Despite this simplicity, the JAMSTEC prediction system has turned out to be superior to many other existing systems for ENSO prediction (Jin et al. 2008). This system also provides a unique tool for studying the role of SST rise in causing the intensified land warming over recent decades and for assessing the potential role of the SST warming trend in seasonal-to-interannual climate predictability. The JAMSTEC prediction system and retrospective forecast experiments are described in section 2. Section 3 assesses the influence of SST trend on global climate and its predictability up to 2-yr lead. Discussed also are possible reasons for limited predictability of the SST warming trend at mid–long lead times. The summary and discussions are given in section 4.

2. The JAMSTEC climate prediction system and hindcast experiment

The JAMSTEC climate prediction system was built on the basis of the Scale Interaction Experiment-Frontier (SINTEX-F) fully coupled global ocean–atmosphere GCM (Luo et al. 2003, 2005a; Masson et al. 2005). The coupled model was developed at JAMSTEC under the European Union–Japan collaboration (Gualdi et al. 2003; Guilyardi et al. 2003). The atmospheric component (ECHAM4.6) of the SINTEX-F model has a resolution of 1.1° (T106) with 19 vertical levels (Roeckner et al. 1996). Its oceanic component (OPA8) has a relatively coarse resolution of a 2° Mercator horizontal mesh but with an equatorial intensification up to 0.5° in the meridional direction (Madec et al. 1998). It has 31 levels in the vertical of which 20 lie in the top 500 m with a high resolution of 10 m from the sea surface to 110-m depth. Heat, water, and momentum fluxes across the air–sea interface are exchanged every two hours without any corrections using a standardized coupler (Valcke et al. 2000). No sea ice model is incorporated into the current system; sea ice cover is relaxed toward observed monthly climatology in the ocean GCM. For model details and performance, readers are referred to Luo et al. (2005a,b). The SINTEX-F model has been applied to various climate studies and proved to have good performance in simulating the tropical climate (e.g., Yamagata et al. 2004; Luo et al. 2005a; Masson et al. 2005; Behera et al. 2005, 2006; Tozuka et al. 2005, 2007, 2008; Kug et al. 2006; Rao et al. 2007, 2009; Cherchi et al. 2007; Navarra et al. 2008; Izumo et al. 2008; Hong et al. 2008; Ajayamohan et al. 2009).

To generate realistic initial conditions for coupled model forecasts by developing complex schemes to assimilate available ocean and atmosphere observations is important work, but it requires much man and computer power. Instead, we have adopted a simple but effective initialization approach as an attempt to produce realistic and well-balanced ocean–atmosphere initial conditions by assimilating only observed SSTs in a coupled way (e.g., Luo et al. 2005b). After a 11-yr spinup, model SSTs since 1 January 1982 are strongly nudged toward daily observations, which are interpolated from weekly NCEP analysis (Reynolds et al. 2002), by applying three large feedback values (−2400, −1200, and −800 W m−2 K−1) to the surface heat flux (Luo et al. 2007, 2008b). They correspond to 1-, 2-, and 3-day restoring time for temperature in a 50-m surface mixed layer, respectively. Interannual variations of the equatorial Pacific thermocline over the past 2–3 decades are well captured by using the above coupled SST-nudging initialization scheme (see Luo et al. 2005b, 2010). Concerning large uncertainties in surface wind stress estimations, model-coupling physics is further perturbed in three different ways by considering or neglecting ocean surface current contributions (Luo et al. 2005a). Therefore, our ensemble prediction system attempts to measure uncertainties of both initial conditions and model errors for forecasts. Based on this semimultimodel ensemble prediction system, we have performed 9-member retrospective forecasts for 24 target months from the first day of each month from February 1982 to March 2009. Concentrations of GHGs (CO2, CH4, N2O, CFC11, CFC12, HCFC113, CFC114, CFC115, HCFC22, and CCL4) in the atmospheric model during the model initialization and forecasts are fixed to their values in 1990 (i.e., 353 ppmv, 1.72 ppmv, 310 ppbv, 280 pptv, 484 pptv, 60 pptv, 15 pptv, 5 pptv, 122 pptv and 146 pptv, respectively); this is a common way adopted in most current state-of-the-art climate prediction systems. With the fixed GHGs and other external radiative forcing, seasonal-to-interannual predictability of both the natural variations and long-term trends of global climate essentially arises from the SST forcing.

When calculating model forecast anomalies, we have removed model climate drifts at each lead time in a posteriori manner (see Luo et al. 2005b). The JAMSTEC climate prediction system has shown good skill in predicting ENSO up to 1–2 yr ahead (Luo et al. 2005b, 2008b, 2010; Jin et al. 2008) and the Indian Ocean dipole (IOD) and Asian monsoon up to a few seasons in advance (e.g., Luo et al. 2007, 2008a; Wang et al. 2008, 2009; Lee et al. 2010; Chowdary et al. 2011). Since 2005, the JAMSTEC prediction system has been used for real-time forecast experiments and demonstrated excellent performance, including the successful forecasts of the 2006 IOD event up to 1 yr ahead and of the long-lasting La Niña episode during mid-2007 to early 2009 up to 2 yr in advance (information available online at http://www.jamstec.go.jp/frcgc/research/d1/iod/index.html).

3. Assessing the influence of SST trend on global warming and climate predictions

a. Role of SST rise in global warming

With assimilating merely historical SST observations, the SINTEX-F coupled model reproduces realistic interannual variations and long-term trend of the global SAT during 1982–2008 (cf. Figs. 1a and 2a, solid curves). The ENSO-related interannual signals and colder (warmer) states before (after) the 1997/98 climate shift are correctly captured. The model nine-member ensemble mean response to observed SST forcing also shows enhanced warming over land, particularly in mid-high latitudes of the NH, compared to the ocean surface warming (Fig. 2b). The global mean terrestrial warming during 1997–2008 (relative to 1983–94) reaches 0.45°C, which is almost twice as large as that of the ocean surface warming (0.24°C). This land–ocean warming contrast is close to the NCEP (0.36° versus 0.20°C, recall Fig. 1b) and NCEP2 reanalysis (0.39° versus 0.19°C, not shown).1 The land/ocean warming ratio of nearly 2 is consistent with recent studies and is found in coupled model projections of future climate change under various warming scenarios (e.g., Sutton et al. 2007; Lambert and Chiang 2007). This suggests the importance of SST rise in causing the intensified land warming. At 6-months lead, the model predicts most of the interannual signals of the global SAT and the 1997/98 climate shift despite that the predicted warming trend during 1982–2008 is weak compared to the observations as well as model initial conditions (Fig. 2a, dashed curves).2 Correspondingly, the recent warming over the global land and ocean surface is realistically predicted despite the weak strength particularly in mid-high latitudes of the NH continent (Fig. 2c). The predicted land mean warming during 1997–2008 is 0.20°C, again about twice as large as that over the ocean surface (0.09°C). Significant cooling biases in the northern North Atlantic, the Southern Ocean from 40°S to 60°S, and the equatorial eastern Pacific are found in the model prediction. We speculate that the much reduced warming (even slight cooling) in Europe as shown in the 6-month lead predictions might be caused by the spurious cooling signal in the North Atlantic [e.g., Marshall et al. (2001) for a review].

It is interesting to note that observed SSTs in the eastern tropical Pacific have cooled slightly in recent decades (recall Figs. 1b and 2b). This appears to be against the El Niño–like or whole tropical Pacific warming response projected by climate models under global warming scenarios (Solomon et al. 2007). Whether the La Niña–like trend in recent decades might be caused by natural processes or by oceanic response to global warming (e.g., Karnauskas et al. 2009) is an interesting issue to be explored. Despite this uncertainty, Cai et al. (2009) assumed the spatial pattern of SAT trends (similar to what shown in Fig. 1b) to be a global warming signal and assessed its potential impact on global SAT predictability up to a 9-month lead based on the NCEP prediction system. Considering large uncertainties in local climate response to GHG forcing but the relatively robust response of the global mean temperature (Solomon et al. 2007), here we examine the potential influence of the trend of global mean surface temperature on climate predictability at lead times of up to 2 yr. Note that we do not perform an additional set of ensemble forecasts with time-evolving GHG concentrations; this requires too much computer power. Instead, we simply analyze the influence of limited predictability of the global mean warming trend on seasonal-to-interannual climate predictability. This is similar to what was done by Cai et al. (2009).

b. Predictability of warming trends and global SAT anomalies

Figure 3a shows observed and model predicted global mean 2-m SAT from 1982 to 2008 with linear trends removed. The ENSO-related interannual fluctuations can be seen more clearly from the detrended time series (compared to Fig. 1a). Forced by the historical SSTs only, the model reproduces almost all observed interannual signals (gray line in Fig. 3a) with a correlation of ∼0.8 (Fig. 3c, red solid line without circles) as well as the linear warming trend (Fig. 3b, gray square mark). The interannual variations of the global mean SAT can be realistically predicted up to 6–12-months lead with an anomaly correlation coefficient (ACC) score above 0.5 (Figs. 3a and 3c). At 24-month lead, the model basically predicts low-frequency signals including the impact of two long-lasting La Niña events from late 1983 to early 1986 and from late 1998 to early 2001 as well as the decadal signal since 2002 despite some phase delay (blue curve in Fig. 3a). These results are similar to the model ENSO predictions at lead times of up to 2 yr (i.e., Luo et al. 2008b).

As found in the ECMWF and NCEP predictions systems with fixed GHG forcing (i.e., Doblas-Reyes et al. 2006; Liniger et al. 2007; Cai et al. 2009), the warming trend of the global mean SAT decreases with forecast lead time (Fig. 3b, medium blue lines). Our model predicts a warming trend up to 10–11 months ahead; beyond this lead time, the predicted trend becomes surprisingly negative. The cooling trend appears in all 9-member predictions beyond 16–18-months lead (dashed lines in Fig. 3b), which cannot be explained by the fixed GHG forcing. This indicates that other processes may operate to deteriorate the trend in our prediction system. The positive trends predicted at short–mid lead times help to slightly improve the prediction skill, whereas the spurious cooling trends at lead times beyond 10 months largely deteriorate the skill (cf. the solid red lines with and without closed circle mark in Fig. 3c). This suggests the importance of correctly predicting the long-term trend. If superimposing the observed trend (black square in Fig. 3b) upon the detrended forecasts (i.e., assuming a perfect trend) at each lead time and recalculating the correlation skill, the model 9-member ensemble mean forecasts show high skill—greater than 0.66 at leads of up to 2 yr with root-mean-square errors (RMSEs) being much smaller than one standard deviation of the observations (short-dashed red lines in Figs. 3c and 3d). The rather slow decline of the skill within the 2-yr forecast period (from 0.87 to 0.66) suggests that the global mean SAT anomaly would be predictable at lead times far beyond 2 yr if the warming trend could be predicted correctly. Assuming a perfect model gives similar results (blue lines in Figs. 3c and 3d). If one replaces the predicted trends with the trend of model initial conditions produced by the observed SST forcing (i.e., the gray square mark in Fig. 3b), the skill would be highly enhanced, retaining above 0.76 at leads of up to 2 yr with much reduced RMSEs.

The rapid decrease of the predicted global mean SAT trends from realistic warming at short lead times to spurious cooling at long lead times is closely linked with the similar decrease of SST trends in model predictions (Fig. 4b, possible reasons for the latter are discussed in section 3d). Besides, the interannual fluctuations of global mean 2-m SAT and SST in observations, model initial conditions and mid–long lead predictions bear a remarkable resemblance (cf. Figs. 3a and 4a). These clearly indicate the importance of SST forcing in determining both interannual variations and the long-term trend of the global mean SAT. Compared to the prediction skill of the interannual fluctuations, the spurious cooling trends of the global mean SST at long lead times again largely degrade the predictability (Figs. 4c and 4d). Similarly, correctly predicting the real global SST warming trend would significantly enhance its practical and potential predictability, attaining skill higher than 0.6–0.7 at leads of up to 2 yr with much reduced RMSEs.

As discussed above, the SSTs play a key role in generating the interannual variability and long-term warming trend of global terrestrial SAT based on the observations and model initial conditions. This is reconfirmed by the model 9-member ensemble mean predictions at mid–long lead times (cf. the colored curves in Figs. 5a and 4a). Corresponding to the rapid decrease of the SST trend with forecast lead time, the predicted warming trend of the global mean SAT over land also decays quickly and becomes negative beyond 14-months lead (Fig. 5b). Prediction skill of the interannual fluctuations of the observed global mean SAT over land appears to be rather limited; nevertheless, they would be potentially predictable up to about 1-yr lead with a perfect model (cf. the red and blue solid lines without circles in Fig. 5c). Again, assuming a perfect warming trend for the forecasts would largely increase the practical and potential predictability, attaining high skill above 0.6–0.7 up to a 2-yr lead with much smaller RMSEs (Figs. 5c and 5d, red and blue short-dashed lines).

It is interesting to note that a positive trend of the global terrestrial SAT is predicted up to 13-month lead, which is longer compared to the SST trend predictions (up to only 8–9 months ahead, cf. Figures 4b and 5b). As a result, the land/ocean warming trend ratio increases from ∼2 at 1-month lead to ∼4 at 7-month lead; this might be related to a delay of response of the terrestrial SAT to SST forcing as well as the memory of land surface itself in sustaining the local warming signals. Beyond the lead time of 15–16 months, the land/ocean cooling trend ratio is almost constant (slightly smaller than one); this is in contrast to the large land/ocean cooling ratio of interannual variations (recall Figs. 1a and 2a, cf. Figs. 4a and 5a) and the study of Joshi et al. (2008). By assuming that global uniform SST cooling may generate an equal temperature cooling in the lower troposphere over the globe, Joshi et al. (2008) argued that the land–ocean cooling contrast can be simply explained by the difference between the dry and moist lapse rate of air temperature over land and ocean. We note that the underlying reasons for the weak land/ocean cooling trend ratio in our model predictions need to be explored further.

NH SAT has experienced the largest rise over the past decades. From 1982 to 2008, the 2-m SAT averaged at 20°–75°N has warmed at a rate of ∼0.31°C decade−1 (Fig. 6b), about twice as large as that of the global mean SAT warming intensity. This large warming rate is reproduced well by the model with historical SST forcing only. Distinct from the low-frequency variations of the global mean SAT, the NH mean SAT shows predominant annual fluctuations superimposed on interannual–decadal variations (cf. Figs. 6a and 3a). This is related to similar annual fluctuations of the NH SST (not shown), presumably associated with the “reemergence process” as a result of the seasonal cycle of ocean surface mixed layer depth (e.g., Alexander et al. 2001). We note that the predominant annual fluctuations are also seen in the NH mean SAT over land (not shown). The annual fluctuations are realistically captured by the model initial conditions (gray line in Fig. 6a) but poorly predictable even if a perfect model is given (Fig. 6c, solid blue line without circles). This is consistent with a rather limited predictability of the NH SST variations (not shown) and the common notion that extratropical climate is largely influenced by unpredictable synoptic weather events. The rapid decrease of the NH SAT warming trend with forecast lead time (Fig. 6b) is again closely linked with a similar decrease of the NH SST trend. The weak warming trends predicted at short-mid lead times help to attain skillful scores (>0.5) up to 9-month lead; however, skill beyond this lead time declines quickly due to the impact of spurious cooling trends (Fig. 6c, compare the red/blue solid lines with circles and those without circles). Assuming a perfect warming trend would again largely enhance the practical and potential predictability, attaining high skill above 0.7–0.8 at leads of up to 2 yr with much reduced RMSEs (Figs. 6c and 6d). We note that results for the NH mean SST and terrestrial SAT predictions are similar.

Given that the SST warming trend is poorly predicted at mid-long lead times and that the SST trend has large impact on the global climate predictability, it sounds surprising that the JAMSTEC prediction system is able to predict ENSO and its global climate impact out to 1–2 yr ahead (i.e., Luo et al. 2005b, 2008b). To understand this, we examine model predictability of the tropical (30°S–30°N) mean SAT (Fig. 7). Interannual fluctuations of the tropical SAT are predominantly forced by SST variations. The close resemblance between the tropical and global mean SAT variations in the observations, model initial conditions, and predictions clearly suggests the importance of the tropical SSTs in driving global climate (cf. Figs. 7a and 3a). The warming trend in the tropics, partly due to the increasing GHG forcing (Liniger et al. 2007), is about two-thirds of the global warming (Fig. 7b). The tropical SAT warming trend again decreases quickly during the 2-yr forecasts, following a similar decrease in predicted tropical SST trends. However, the long-term trend has little-to-small impact on the tropical climate predictability out to 2 yr ahead (Figs. 7c and 7d). The spurious cooling trends at long lead times and assuming a prefect warming trend do not affect the predictability much. In other words, global warming has little influence on tropical climate prediction. This is presumably attributed to the strong intrinsic ocean–atmosphere interactions in the tropics. The dominant influence of the interannual variability on the tropical climate predictability can be seen by its much shorter persistence compared to that of the global mean SST (not shown), despite that these two indices have comparable warming trends from 1982 to 2008.

Figure 8 shows the spatial map of global SAT prediction skill and potential contributions of a perfect global warming trend to the predictability. At short-mid lead times, the model predicts a warming trend of the global mean SAT (recall Fig. 3) and shows good skill in predicting the SAT anomalies over most parts of the globe (Figs. 8a and 8b). Medium-to-high predictability is basically found over the oceans, whereas predictability of terrestrial SAT anomalies is generally low-to-modest except for the tropics and some coastal areas. The lowest skill appears in the northern North Atlantic; skill there rapidly goes down to negative values even at 3-months lead (Fig. 8a). The negative skill in the North Atlantic becomes much worse and expands across the whole northern basin with increasing lead times; this appears to reduce the predictability over a broad area of the Eurasian continent at mid–long lead times (Figs. 8b–d). As expected from the strong tropical air–sea interactions, the highest predictability of the SAT anomalies at 1–2-yr lead is mostly confined in the tropics (Figs. 8c and 8d). Modest predictability at long lead times can also be found in some extratropical regions such as the North and South Pacific in association with ENSO teleconnections (see also Luo et al. 2008b).

Assuming a perfect warming trend of the global mean SAT, the predictability of the SAT anomalies over major areas of the globe (except the Arctic Ocean) does not change much at 3–6-months lead (Figs. 8a and 8b, see also Fig. 3c). The skill gain over the Arctic Ocean reaches as high as 0.7 at a 2-yr lead (Fig. 8d). The large improvement over that region is related to the strong warming in past decades associated with a rapid decrease of sea ice cover there (recall Fig. 1b); whereas sea ice cover in the model is restored toward observed monthly mean climatology. At 12–24-months lead, considerable improvements are found not only in the NH continents but also in the tropical Indian Ocean–western Pacific and the tropical North Atlantic over where the SAT shows warming trends in recent decades (Figs. 8c and 8d, recall Fig. 1b). Over regions where the SAT has cooling trends in the tropical eastern Pacific and Southern Oceans, the predictability would be reduced by assuming a homogeneous warming trend over the globe. It appears to be surprising that the skill enhancements over the midlatitute Eurasian continent are not large despite the strong observed warming there. One possible reason for this is that the global mean warming trend assumed as a perfect one is not sufficient to compensate for the strong local warming. Another possible reason is due to the strong cold SST bias in the northern North Atlantic in model predictions (recall Fig. 2c). Starting from an initial warming trend in that region, predicted SSTs there rapidly cool and become colder by more than 3°C in 1997–2008 compared to those in 1983–94 at mid-long lead times. This leads to a considerable cooling trend of SAT over the Eurasian continent at long lead times (not shown) and may severely limit the predictability of the SAT anomalies there. Further efforts are required to improve the prediction of the North Atlantic SST, which is an important driver for the NH and even global climate variations and changes (e.g., Marshall et al. 2001).

c. Predictability of global mean precipitation

Interannual fluctuations of global mean precipitation mainly arise from the tropics, dominated by tropical climate events such as the ENSO and IOD. Despite the large diversity and uncertainty that exists in precipitation observations and simulations, the increase (decrease) of global precipitation during 1982–83, 1987–88, 1997–98, 2002–03, and 2006–07 El Niño (1984–86, 1988–89, 1999–2001, and 2007–09 La Niña) years can be identified from the Global Precipitation Climatology Project (GPCP) observations, model initial conditions, and predictions (Fig. 9a). Besides, the decadal fluctuation since 2002 appears to be reasonably well captured and predicted up to 2-yr lead. Comparing with one observational sample, the model 9-member ensemble mean tends to reproduce robust but weak signals. Predictability of the GPCP global mean precipitation is rather limited even with prescribed historical SSTs; modest skill (∼0.4) is found up to 6-month lead (Fig. 9c, red solid lines). Assuming a perfect model would enhance the predictability of the precipitation, attaining useful skill above 0.5 at leads of up to 8 months (Fig. 9c, blue solid lines).

The hydrological cycle response to global warming is of much concern recently (e.g., X. Zhang et al. 2007; Wentz et al. 2007), though the sensitivity of the global mean precipitation to the surface temperature change is in dispute. Various climate model results showed that global mean precipitation would increase by only 1%–3% for 1-K surface warming (e.g., Held and Soden 2006; Lambert and Webb 2008); this is much smaller than that (∼7% K−1) given by the Clausius–Clapeyron equation for the saturation vapor pressure.3 Various observational precipitation datasets for the past 2–3 decades do not show consistent trends (Fig. 9b), in accordance with the conclusion of Solomon et al. (2007). Forced by the historical SSTs with a linear warming trend of ∼0.13°C decade−1 from 1982 to 2008, our model simulation produces an increase of the global mean precipitation at a rate of ∼4.8 mm yr−1 decade−1, equivalent to a hydrological sensitivity of ∼3.4% K−1 (or ∼2.6% K−1 on the basis of the global mean 2-m SAT). This is comparable to the sensitivity estimated under global warming scenarios (i.e., Held and Soden 2006). Consistently, following the rapid decrease of the global SST warming trend during the 2-yr forecasts, the rate of precipitation increase also decays quickly and becomes negative beyond 10–11-months lead (Fig. 9b). The hydrological sensitivity, however, remains almost constant with forecast lead time; this is true even beyond 16-month lead when the spurious SST cooling trends lead to negative trends of the precipitation (not shown). Given a perfect model and a perfect long-term trend, it is encouraging that potential predictability of the global mean precipitation would be high (above ∼0.6) with small RMSEs at leads of up to 2 yr (Figs. 9c and 9d, blue short-dashed lines). This suggests an important influence of the long-term trend on the global mean precipitation predictability.

d. Why the global SST warming trend rapidly decays during model forecasts?

The results shown above clearly suggest the importance of SST rise to the global terrestrial warming and seasonal-to-interannual climate predictability, particularly at mid-long (>6–12 months) lead times. However, it is rather peculiar that the SST warming trend weakens rapidly during model forecasts and even becomes negative at long lead times. As discussed in many global warming studies, global SSTs may have risen in response to the increasing anthropogenic radiative forcing (Solomon et al. 2007). Figure 10 shows the global-mean heat flux at the top of the atmosphere and ocean–land surface in the model initial conditions and predictions averaged during 1982–2008. The model has a positive radiative imbalance at the top of the atmosphere (2.0–2.4 W m−2) during the forecasts; this leads to considerable heating of the ocean–land surface. The heat gain over the ocean surface decreases with forecast lead time but is always larger than the initial gain (short-dashed line in Fig. 10). The decreasing heat gain over the ocean surface may tend to weaken the SST warming trend during the forecasts. However, this cannot explain why the SST trend becomes negative at long lead times.

According to a one-dimensional thermodynamical equation, the ocean surface mixed layer temperature is influenced not only by the surface heat flux but also by the vertical advection/entrainment/diffusion of cold water from the subsurface. The global mean temperature in the top 100 m shows similar variability and predictability to those of the SST, and assuming a perfect warming trend would again produce high potential predictability up to a 2-yr lead (not shown). This suggests that the surface mixed layer temperature can be realistically reproduced by assimilating historical SSTs only. However, model global mean temperature in the top 500 m (T500) displays a strong cooling trend (with an ensemble mean value of −0.19°C decade−1) in the initial conditions, in contrast to a weak warming trend (∼0.01°C decade−1) of the NCEP ocean reanalysis (Fig. 11b). The initial cooling trend maintains and becomes slightly worse during the 2-yr forecasts. This suggests that the strong spurious cooling drift in the subsurface (below about 100-m depth) in the model initial conditions is the key for the rapid decrease of the SST warming trend during the forecasts. Despite a high potential predictability of the T500, the model fails to capture the dominant decadal variations as revealed in the NCEP data (Fig. 11a). The results suggest that, unlike the equatorial Pacific where realistic surface wind stress forcing may be sufficient to generate realistic thermocline variations (i.e., Luo et al. 2005b), assimilating ocean subsurface information appears to be indispensable for obtaining realistic initial conditions in the extratropical oceans.

As described in Luo et al. (2005b), to generate initial conditions for the forecasts, the SINTEX-F model was first spun up from 1971 to 1981 in a decoupled way using observed monthly SSTs. The global mean T500 during the spinup period closely follows the free coupled model integration, showing a quick rise in the first 10 yr due to large heat gain over the ocean surface (Fig. 12a, green and blue lines). The upper-ocean temperature will continue to rise slowly in response to a positive radiative imbalance at the top of the atmosphere. Starting from Levitus annual mean climatology without motion, the tropical mean T500 decreases in the first 10–15 yr despite a large local surface heat gain and then increases following the global mean T500 (black and gray lines in Fig. 12a). However, sharp cooling drifts of both the global and tropical mean T500 appear immediately when the coupled SST-nudging initialization starts from January 1982 (purple and orange lines in Fig. 12a). The large cooling drifts persist for about 10 yr and then appear to be saturated with small cooling further after the early 1990s.

The cooling drift of global mean T500 from 1982 to 2008 is driven by negative surface heat flux (−4.989 W m−2); this is induced by a large negative heat damping associated with the SST-nudging process (Fig. 12b). The strong SST nudging is required to produce realistic SSTs but unfortunately induces large spurious negative surface heat flux particularly during the first several years; this causes the strong cooling drifts of global T500 in the model initial conditions and predictions. The sudden cooling in the upper ocean is somehow similar to what is induced by intensive volcanic eruptions (e.g., Church et al. 2005). We note that the large subsurface cooling in the tropics is not directly related to the SST-nudging process (orange line in Fig. 12b) but appears to be caused by the intrusion of cooled waters from the extratropical oceans. Forced by the large negative surface heat flux damping, warming trends in the upper ocean weaken rapidly from the sea surface to 90-m depth (Fig. 12c, solid lines); the subsurface temperature in a whole column below that depth is strongly cooled with maximum values at 364-m depth. This is in contrast to the NCEP reanalysis, which shows significant warming during 1997–2008 (relative to 1983–94) from the surface down to the depth of about 2200 m with two peaks at the surface and around 1400-m depth (Fig. 12c, dashed lines).

4. Summary and discussions

The global mean SAT has continuously risen in recent decades with major warming mainly over the land, particularly the NH mid–high-latitude continents. However, whether the global terrestrial warming may be caused by the local response to increasing GHG forcing or by remote SST influence (as we have already known from the impact of El Niño events) is controversial. Limited seasonal prediction experiments initialized from realistic ocean and atmosphere conditions showed the importance of increasing GHG concentrations in predicting the global warming trend (i.e., Doblas-Reyes et al. 2006; Liniger et al. 2007). In this study, based on the JAMSTEC climate prediction system, which assimilates only observed SSTs, we have examined impacts of the SST forcing on global warming and seasonal-to-interannual climate predictability. Our results suggest that historical SSTs play a major role in driving the global warming over land and the land–ocean warming contrast; this is consistent with a few recent studies (i.e., Compo and Sardeshmukh 2009).

We note that the SST warming itself may be driven by both increasing GHG forcing and slowly varying natural processes (Solomon et al. 2007). Recent studies based on observations and climate model simulations with historical GHG forcing have suggested a similar warming trend of global mean SAT during 1950–2000 by removing and retaining natural atmospheric variability, ENSO, and volcanic signals (e.g., Fyfe et al. 2010). However, differentiating between anthropogenic changes and intrinsic decadal–multidecadal climate variations remains as a long-standing challenge; they may also interact in a complex way. Known long-term natural fluctuations that have significant impacts on global climate include the multidecadal variability of Atlantic meridional thermohaline circulation (e.g., R. Zhang et al. 2007; Rashid et al. 2010), Pacific decadal–interdecadal variations (e.g., Zhang et al. 1997; Luo and Yamagata 2001; Luo et al. 2003), strong multidecadal fluctuations in the Arctic area (e.g., Bekryaev et al. 2010), and the 11-yr solar cycle forcing (e.g., Meehl et al. 2009). Quantifying the natural and anthropogenic contributions to the SST warming trend over the recent 2–3 decades (if possible) could improve our understanding of the potential predictability of the historical SST rise and of the impacts of natural/anthropogenic trends on seasonal-to-interannual climate prediction. This is, however, beyond the scope of the present study.

Our results suggest that the global warming trend over the past 2–3 decades, predominantly induced by the SST rise, has a substantial impact on global climate predictability, particularly at long lead times (>12 months). At short lead times, however, its influence is negligible. By assuming a perfect warming trend and/or a perfect model, SAT and precipitation averaged over the globe or NH would be potentially predictable at lead times of up to 2 yr. It is worth examining the differences in results caused by different observations. We do not find much difference using the NCEP2 reanalysis. However, the Japanese 25-yr reanalysis (JRA-25) (Onogi et al. 2007) produced a large warming rate (0.345°C decade−1) of the global mean terrestrial SAT during 1982–2008, which is about 60% higher than that of the NCEP reanalysis (0.213°C decade−1), despite the fact that the interannual variations produced by both reanalyses are similar. Consequently, the potential predictability of global SAT over land based on the JRA-25 would be about 0.1–0.2 higher than that based on the NCEP reanalysis (Fig. 13, dashed lines). Similar differences can also be seen from the NH SAT and NH terrestrial SAT potential predictability (not shown). The warming rates of the two variables based on the JRA-25 are 0.375° and 0.481°C decade−1 respectively, higher than the 0.309° and 0.321°C decade−1 based on NCEP data.

It is unfortunate that the SST warming trend (and hence upward trends of the global climate) in the present model rapidly decays during the forecasts and even becomes negative at long lead times. This is caused by a strong spurious cooling drift of the ocean subsurface temperature in the model initial conditions owing to large negative surface heat flux damping when the model SSTs are strongly restored toward the observed values. The surface heat flux from the atmosphere is actually positive, associated with a positive radiative imbalance at the top of the atmosphere due to the model bias. The rapid decrease of the SST warming trend during the forecasts deteriorates the global climate predictability. However, this is not against the good performance of the JAMSTEC prediction system in predicting the tropical climate up to 1–2 yr ahead. Despite the influence of global warming, the tropical climate is dominated by strong local ocean–atmosphere interactions and, hence, is highly predictable on seasonal-to-interannual time scales.

Future efforts are required to obtain subsurface initial conditions that are as realistic as possible, particularly in the extratropical ocean. This can be achieved by assimilating available subsurface information from in situ and satellite observations. A simple way is to improve the coupled SST-nudging initialization process with a better-tuned model and a longer spinup period prior to the hindcasts. We note that assimilating observed SST anomalies rather than SSTs themselves might also help to reduce the climate drifts (e.g., Keenlyside et al. 2008). Regarding the importance of SSTs in driving the global warming and the crucial influence on SSTs of the subsurface temperature, which has a much longer memory, global warming/cooling trends could be potentially predictable (at least partly) on seasonal-to-interannual time scales if model forecasts are initialized from realistic ocean conditions. This provides an explanation for the limited predictability of the long-term global warming trend by the ECMWF and NCEP prediction systems with fixed GHG concentrations. The intrinsic predictability of global warming, which arises from the long memory of ocean warming, provides hope for the enhanced prediction of climate anomalies (interannual variations + warming trend) under increasing GHG forcing.

Acknowledgments

NCEP reanalysis and CMAP precipitation data are provided by the NOAA/OAR/ESRL PSD. All model forecasts were carried out on the JAMSTEC Earth Simulator. We thank two anonymous reviewers for their valuable comments that helped to improve the manuscript.

REFERENCES

  • Ajayamohan, R. S., S. A. Rao, J.-J. Luo, and T. Yamagata, 2009: Influence of Indian Ocean Dipole on boreal summer intraseasonal oscillations in a coupled general circulation model. J. Geophys. Res., 114 , D06119. doi:10.1029/2008JD011096.

    • Search Google Scholar
    • Export Citation
  • Alexander, M. A., M. S. Timlin, and J. D. Scott, 2001: Winter-to-winter recurrence of sea surface temperature, salinity and mixed layer depth anomalies. Prog. Oceanogr., 49 , 4161.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., J.-J. Luo, S. Masson, P. Delecluse, S. Gualdi, A. Navarra, and T. Yamagata, 2005: Paramount impact of the Indian Ocean dipole on the East African short rain: A CGCM study. J. Climate, 18 , 45144530.

    • Search Google Scholar
    • Export Citation
  • Behera, S. K., J.-J. Luo, S. Masson, S. A. Rao, H. Sakuma, and T. Yamagata, 2006: A CGCM study on the interaction between IOD and ENSO. J. Climate, 19 , 16881705.

    • Search Google Scholar
    • Export Citation
  • Bekryaev, R. V., I. V. Polyakov, and V. A. Alexeev, 2010: Role of polar amplification in long-term surface air temperature variations and modern Arctic warming. J. Climate, 23 , 38883906.

    • Search Google Scholar
    • Export Citation
  • Boer, G. J., 2000: A study of atmosphere–ocean predictability on long time scales. Climate Dyn., 16 , 469477.

  • Cai, M., C.-S. Shin, H. M. Van den Dool, W. Wang, S. Saha, and A. Kumar, 2009: The role of long-term trends in seasonal predictions: Implication of global warming in the NCEP CFS. Wea. Forecasting, 24 , 965973.

    • Search Google Scholar
    • Export Citation
  • Cherchi, A., S. Gualdi, S. Behera, J.-J. Luo, S. Masson, T. Yamagata, and A. Navarra, 2007: The influence of tropical Indian Ocean SST on the Indian summer monsoon. J. Climate, 20 , 30833105.

    • Search Google Scholar
    • Export Citation
  • Chowdary, J. S., S.-P. Xie, J.-J. Luo, J. Hafner, S. Behera, Y. Masumoto, and T. Yamagata, 2011: Predictability of Northwest Pacific climate during summer and the role of the tropical Indian Ocean. Climate Dyn., 36 , 607621. doi:10.1007/s00382-009-0686-5.

    • Search Google Scholar
    • Export Citation
  • Church, J. A., N. J. White, and J. M. Arblaster, 2005: Significant decadal-scale impact of volcanic eruptions on sea level and ocean heat content. Nature, 438 , 7477.

    • Search Google Scholar
    • Export Citation
  • Collins, M., 2002: Climate predictability on interannual to decadal time scales: The initial value problem. Climate Dyn., 19 , 671692.

    • Search Google Scholar
    • Export Citation
  • Collins, M., and M. R. Allen, 2002: Assessing the relative roles of initial and boundary conditions in interannual to decadal climate predictability. J. Climate, 15 , 31043109.

    • Search Google Scholar
    • Export Citation
  • Compo, G. P., and P. D. Sardeshmukh, 2009: Oceanic influences on recent continental warming. Climate Dyn., 32 , 333342.

  • Cox, P., and D. Stephenson, 2007: A changing climate for prediction. Science, 317 , 207208.

  • Doblas-Reyes, F. J., R. Hagedorn, T. N. Palmer, and J.-J. Morcrette, 2006: Impact of increasing greenhouse gas concentrations in seasonal ensemble forecasts. Geophys. Res. Lett., 33 , L07708. doi:10.1029/2005GL025061.

    • Search Google Scholar
    • Export Citation
  • Dommenget, D., 2009: The ocean’s role in continental climate variability and change. J. Climate, 22 , 49394952.

  • Dong, B., J. M. Gregory, and R. T. Sutton, 2009: Understanding land–sea warming contrast in response to increasing greenhouse gases. Part I: Transient adjustment. J. Climate, 22 , 30793097.

    • Search Google Scholar
    • Export Citation
  • Fyfe, J. C., N. P. Gillett, and D. W. J. Thompson, 2010: Comparing variability and trends in observed and modelled global-mean surface temperature. Geophys. Res. Lett., 37 , L16802. doi:10.1029/2010GL044255.

    • Search Google Scholar
    • Export Citation
  • Griffies, S. M., and K. Bryan, 1997: Predictability of North Atlantic multidecadal climate variability. Science, 275 , 181184.

  • Gualdi, S., A. Navarra, E. Guilyardi, and P. Delecluse, 2003: Assessment of the tropical Indo-Pacific climate in the SINTEX CGCM. Ann. Geophys., 46 , 126.

    • Search Google Scholar
    • Export Citation
  • Guilyardi, E., P. Delecluse, S. Gualdi, and A. Navarra, 2003: Mechanisms for ENSO phase change in a coupled GCM. J. Climate, 16 , 11411158.

    • Search Google Scholar
    • Export Citation
  • Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19 , 56865699.

  • Hermanson, L., and R. T. Sutton, 2009: Case studies in interannual to decadal climate predictability. Climate Dyn., 35 , 11691189. doi:10.1007/s00382-009-0672-y.

    • Search Google Scholar
    • Export Citation
  • Hong, C.-C., T. Li, and J.-J. Luo, 2008: Asymmetry of the Indian Ocean dipole. Part II: Model diagnosis. J. Climate, 21 , 48494858.

  • Izumo, T., C. Montegut, J.-J. Luo, S. Behera, S. Masson, and T. Yamagata, 2008: The role of the western Arabian Sea upwelling in Indian monsoon rainfall variability. J. Climate, 21 , 56035623.

    • Search Google Scholar
    • Export Citation
  • Jin, E. K., and Coauthors, 2008: Current status of ENSO prediction skill in coupled ocean–atmosphere models. Climate Dyn., 31 , 647664.

    • Search Google Scholar
    • Export Citation
  • Joshi, M. M., J. M. Gregory, M. J. Webb, D. M. H. Sexton, and T. C. Johns, 2008: Mechanisms for the land/sea warming contrast exhibited by simulations of climate change. Climate Dyn., 30 , 455465.

    • Search Google Scholar
    • Export Citation
  • Karnauskas, K. B., R. Seager, A. Kaplan, Y. Kushnir, and M. A. Cane, 2009: Observed strengthening of the zonal sea surface temperature gradient across the equatorial Pacific Ocean. J. Climate, 22 , 43164321.

    • Search Google Scholar
    • Export Citation
  • Keenlyside, N. S., M. Latif, J. Jungclaus, L. Kornblueh, and E. Roeckner, 2008: Advancing decadal-scale climate prediction in the North Atlantic sector. Nature, 453 , 8488.

    • Search Google Scholar
    • Export Citation
  • Kug, J.-S., T. Li, S.-I. An, I.-S. Kang, J.-J. Luo, S. Masson, and T. Yamagata, 2006: Role of the ENSO–Indian Ocean coupling on ENSO variability in a coupled GCM. Geophys. Res. Lett., 33 , L09710. doi:10.1029/2005GL024916.

    • Search Google Scholar
    • Export Citation
  • Lambert, F. H., and J. C. H. Chiang, 2007: Control of land–ocean temperature contrast by ocean heat uptake. Geophys. Res. Lett., 34 , L13704. doi:10.1029/2007GL029755.

    • Search Google Scholar
    • Export Citation
  • Lambert, F. H., and M. J. Webb, 2008: Dependency of global mean precipitation on surface temperature. Geophys. Res. Lett., 35 , L16706. doi:10.1029/2008GL034838.

    • Search Google Scholar
    • Export Citation
  • Lee, J.-Y., and Coauthors, 2010: How is the performance on mean climate related to seasonal prediction skill in coupled climate models? Climate Dyn., 35 , 267283.

    • Search Google Scholar
    • Export Citation
  • Liepert, B. G., and M. Previdi, 2009: Do models and observations disagree on the rainfall response to global warming? J. Climate, 22 , 31563166.

    • Search Google Scholar
    • Export Citation
  • Liniger, M. A., H. Mathis, C. Appenzeller, and F. J. Doblas-Reyes, 2007: Realistic greenhouse gas forcing and seasonal forecasts. Geophys. Res. Lett., 34 , L04705. doi:10.1029/2006GL028335.

    • Search Google Scholar
    • Export Citation
  • Lorenz, E. N., 1975: The physical bases of climate and climate modelling. Climate Predictability, GARP Publication Series, Vol. 16, WMO, 132–136.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., and T. Yamagata, 2001: Long-term El Niño–Southern Oscillation (ENSO)-like variation with special emphasis on the South Pacific. J. Geophys. Res., 106 , 2221122227.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, P. Delecluse, S. Gualdi, A. Navarra, and T. Yamagata, 2003: South Pacific origin of the decadal ENSO-like variation as simulated by a coupled GCM. Geophys. Res. Lett., 30 , 2250. doi:10.1029/2003GL018649.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, E. Roeckner, G. Madec, and T. Yamagata, 2005a: Reducing climatology bias in an ocean–atmosphere CGCM with improved coupling physics. J. Climate, 18 , 23442360.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, S. Shingu, and T. Yamagata, 2005b: Seasonal climate predictability in a coupled OAGCM using a different approach for ensemble forecasts. J. Climate, 18 , 44744497.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, and T. Yamagata, 2007: Experimental forecasts of the Indian Ocean dipole using a coupled OAGCM. J. Climate, 20 , 21782190.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Behera, Y. Masumoto, H. Sakuma, and T. Yamagata, 2008a: Successful prediction of the consecutive IOD in 2006 and 2007. Geophys. Res. Lett., 35 , L14S02. doi:10.1029/2007GL032793.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., S. Masson, S. Behera, and T. Yamagata, 2008b: Extended ENSO predictions using a fully coupled ocean–atmosphere model. J. Climate, 21 , 8493.

    • Search Google Scholar
    • Export Citation
  • Luo, J.-J., R. Zhang, S. Behera, Y. Masumoto, F.-F. Jin, R. Lukas, and T. Yamagata, 2010: Interaction between El Niño and extreme Indian Ocean dipole. J. Climate, 23 , 726742.

    • Search Google Scholar
    • Export Citation
  • Madec, G., P. Delecluse, M. Imbard, and C. Levy, 1998: OPA 8.1 ocean general circulation model reference manual. LODYC/IPSL Tech. Rep. Note 11, 91 pp.

    • Search Google Scholar
    • Export Citation
  • Marshall, J., and Coauthors, 2001: North Atlantic climate variability: Phenomena, impacts and mechanisms. Int. J. Climatol., 21 , 18631898.

    • Search Google Scholar
    • Export Citation
  • Masson, S., and Coauthors, 2005: Impact of barrier layer on winter-spring variability of the southeastern Arabian Sea. Geophys. Res. Lett., 32 , L07703. doi:10.1029/2004GL021980.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., J. M. Arblaster, K. Matthes, F. Sassi, and H. van Loon, 2009: Amplifying the Pacific climate system response to a small 11 year solar cycle forcing. Science, 325 , 11141118.

    • Search Google Scholar
    • Export Citation
  • Navarra, A., and Coauthors, 2008: Atmospheric horizontal resolution affects tropical climate variability in coupled models. J. Climate, 21 , 730750.

    • Search Google Scholar
    • Export Citation
  • Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85 , 369432.

  • Palmer, T. N., and Coauthors, 2004: Development of a European multimodel ensemble system for seasonal-to-interannual prediction (DEMETER). Bull. Amer. Meteor. Soc., 85 , 853872.

    • Search Google Scholar
    • Export Citation
  • Rao, S. A., S. Masson, J.-J. Luo, S. K. Behera, and T. Yamagata, 2007: Termination of Indian Ocean dipole events in a coupled general circulation model. J. Climate, 20 , 30183035.

    • Search Google Scholar
    • Export Citation
  • Rao, S. A., J.-J. Luo, S. K. Behera, and T. Yamagata, 2009: Generation and termination of Indian Ocean dipole events in 2003, 2006 and 2007. Climate Dyn., 33 , 751767.

    • Search Google Scholar
    • Export Citation
  • Rashid, H., S. B. Power, and J. R. Knight, 2010: Impact of multidecadal fluctuations in the Atlantic thermohaline circulation on Indo-Pacific climate variability in a coupled GCM. J. Climate, 23 , 40384044.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Roeckner, E., and Coauthors, 1996: The atmospheric general circulation model ECHAM-4: Model description and simulation of present-day climate. Max-Planck-Institut für Meteorologie Rep. 218, 90 pp.

    • Search Google Scholar
    • Export Citation
  • Smith, D. M., S. Cusack, A. W. Colman, C. K. Folland, G. R. Harris, and J. M. Murphy, 2007: Improved surface temperature prediction for the coming decade from a global climate model. Science, 317 , 796799.

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

    • Search Google Scholar
    • Export Citation
  • Sutton, R., B. Dong, and J. M. Gregory, 2007: Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. Geophys. Res. Lett., 34 , L02701. doi:10.1029/2006GL028164.

    • Search Google Scholar
    • Export Citation
  • Tozuka, T., J.-J. Luo, S. Masson, S. Behera, and T. Yamagata, 2005: Annual ENSO simulated in a coupled ocean–atmosphere model. Dyn. Atmos. Oceans, 39 , 4160.

    • Search Google Scholar
    • Export Citation
  • Tozuka, T., J.-J. Luo, S. Masson, and T. Yamagata, 2007: Decadal modulations of the Indian Ocean dipole in the SINTEX-F1 coupled GCM. J. Climate, 20 , 28812894.

    • Search Google Scholar
    • Export Citation
  • Tozuka, T., J.-J. Luo, S. Masson, and T. Yamagata, 2008: Tropical Indian Ocean variability revealed by self-organizing maps. Climate Dyn., 31 , 333343.

    • Search Google Scholar
    • Export Citation
  • Troccoli, A., and T. N. Palmer, 2007: Ensemble decadal predictions from analysed initial conditions. Philos. Trans. Roy. Soc., A365 , 21792191.

    • Search Google Scholar
    • Export Citation
  • Valcke, S., L. Terray, and A. Piacentini, 2000: The OASIS coupler user guide version 2.4. CERFACE Tech. Rep. TR/CGMC/00-10, 85 pp.

  • Wang, B., and Coauthors, 2008: How accurately do coupled climate models predict the Asian-Australia monsoon interannual variability? Climate Dyn., 31 , 605619.

    • Search Google Scholar
    • Export Citation
  • Wang, B., and Coauthors, 2009: Advance and prospectus of seasonal prediction: Assessment of the APCC/CliPAS 14-model ensemble retrospective seasonal prediction (1980–2004). Climate Dyn., 33 , 93117.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., L. Ricciardulli, K. Hilburn, and C. Mears, 2007: How much more rain will global warming bring? Science, 317 , 233235.

  • Yamagata, T., S. Behera, J.-J. Luo, S. Masson, M. Jury, and S. A. Rao, 2004: Coupled ocean–atmosphere variability in the tropical Indian Ocean. Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 189–212.

    • Search Google Scholar
    • Export Citation
  • Zhang, R., T. L. Delworth, and I. M. Held, 2007: Can the Atlantic Ocean drive the observed multidecadal variability in Northern Hemisphere mean temperature? Geophys. Res. Lett., 34 , L02709. doi:10.1029/2006GL028683.

    • Search Google Scholar
    • Export Citation
  • Zhang, X., F. W. Zwiers, G. C. Hegerl, F. H. Lambert, N. P. Gillett, S. Solomon, P. A. Stott, and T. Nozawa, 2007: Detection of human influence on twentieth-century precipitation trends. Nature, 448 , 461465.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–92. J. Climate, 10 , 10041020.

Fig. 1.
Fig. 1.

(a) Surface 2-m air temperature (T2m) anomaly (°C, relative to the 1983–2006 climatology) averaged over the globe (60°S–75°N, black line), land (red line), and ocean (blue line) based on the NCEP atmospheric reanalysis (available online at http://www.esrl.noaa.gov/psd/). All time series were smoothed with a 5-month running mean to filter out high-frequency variability. (b) T2m difference between 1997–2008 and 1983–94.

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 2.
Fig. 2.

As in Fig. 1 but for the model 9-member ensemble mean initial conditions (ICs) and 6-month lead predictions.

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 3.
Fig. 3.

(a) As in Fig. 1a but for the detrended global mean T2m variations based on the NCEP reanalysis (black line), model ICs (gray line), and predictions at 6-, 12-, and 24-month lead (colored lines). (b) Linear trends of the global mean T2m anomalies from 1982 to 2008 based on the NCEP data (black square), model 9-member mean ICs (gray square, blue circles for individual member), and predictions at lead times of 1–24 months (medium blue line, dashed-curves for each member). (c),(d) As in (b) but for model prediction skill and RMSEs. Red (blue) solid lines denote the practical (potential) predictability for the nondetrended (line with closed circle mark) and detrended global mean T2m anomalies (line without closed circle mark). Red (blue) short-dashed lines show the predictability if a perfect warming trend of the NCEP observations (model ensemble mean ICs) was predicted at all lead times. Black (gray) zonal solid line in (d) indicates one standard deviation of the nondetrended observations (model ICs).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 4.
Fig. 4.

As in Fig. 3 but for global mean SST anomalies. The SST observations are based on monthly NCEP analysis (Reynolds et al. 2002).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 5.
Fig. 5.

As in Fig. 3 but for global mean T2m anomalies over land.

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 6.
Fig. 6.

As in Fig. 3 but for T2m anomalies averaged in the extratropical Northern Hemisphere (20°–75°N).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 7.
Fig. 7.

As in Fig. 3 but for T2m anomalies averaged over the tropics (30°S–30°N).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 8.
Fig. 8.

The T2m anomaly correlations (colored scale) between the nondetrended NCEP reanalysis and model 9-member mean predictions at different lead times for the period 1982–2008. Shown also are the skill improvements (contour interval: ±0.1, ±0.2, ±0.3, …; thick solid lines indicate a 0.2 contour) by assuming a perfect warming trend of the NCEP global mean T2m anomalies in the model predictions. The skill is measured based on 4° by 4° grid cells.

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 9.
Fig. 9.

As in Fig. 3 but for global mean precipitation anomalies. The observational data is based on monthly GPCP analysis (available online at http://precip.gsfc.nasa.gov/). The model ensemble mean produces a much smaller variability (see the right y axis scale in Fig. 9a). Linear trends based on Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) (available online at http://www.esrl.noaa.gov/psd/) and NCEP CPC convection-allowing models (CAMS)–outgoing longwave radiation precipitation index (OPI) (available online at http://iridl.ldeo.columbia.edu/) datasets are also shown in (b).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 10.
Fig. 10.

Radiative imbalance at the top of the atmosphere (black solid line, downward positive), surface heat flux over the ocean (black short-dashed line, left scale), and over land (gray long–short-dashed line, right scale) averaged during 1982–2008 over the globe (90°S–90°N) based on model ensemble mean ICs and predictions. We note that results averaged at 60°S–75°N are similar.

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 11.
Fig. 11.

As in Figs. 3a and 3b but for global mean temperature anomalies averaged from the sea surface to 500-m depth (T500). The observations are based on monthly NCEP ocean reanalysis (available online at http://www.cpc.ncep.noaa.gov/products/GODAS/).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 12.
Fig. 12.

(a) Global (60°S–75°N, right scale) and tropical (30°S–30°N, left scale) mean monthly T500 based on the SINTEX-F 100-yr free model simulations with fixed GHG concentrations (blue and black lines), model spinup from 1971 to 1981 (green and gray lines), and model ICs during 1982–2008 generated by the coupled SST-nudging approach (purple and orange lines). Sea surface net heat fluxes averaged over the two domains and individual time lengths are also shown. (b) As in (a) but for the surface heat flux damping term during the model spinup and SST-nudging periods. Restoring of sea ice to observed monthly climatology leads to some local heat damping in high latitudes (green line). (c) Global (blue lines) and tropical (black lines) mean temperature difference between 1997–2008 and 1983–94 based on the model ICs (solid lines) and NCEP ocean reanalysis (short dashed lines).

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

Fig. 13.
Fig. 13.

As in Fig. 5c but for model prediction skills based on the JRA-25 (black lines). The skills calculated based on the NCEP reanalysis (red lines) are reproduced from Fig. 5c for comparison.

Citation: Journal of Climate 24, 6; 10.1175/2010JCLI3645.1

1

We note that the Japanese 25-year Reanalysis (JRA-25) with a high resolution (T106) produces a much bigger land–ocean warming ratio (0.53°C versus 0.09°C, not shown). Underlying reasons for this are unclear at the current stage. However, using the JRA-25 for global SAT predictability assessment will not change the conclusions of the present study (see also the discussions given in section 4).

2

The sharp cooling signal in 1992 associated with the Mount Pinatubo eruption (June 1991), however, is not predicted owing to a fixed volcanic forcing in the atmosphere model (see also Fig. 3a).

3

Liepert and Previdi (2009) found high interdecadal variability of the global hydrological sensitivity in climate model simulations and argued that this may influence the sensitivity estimation over a 20-yr period.

Save