Robust twenty-first-century projections of ElNiño and related precipitation variability

Journal name:
Nature
Volume:
502,
Pages:
541–545
Date published:
DOI:
doi:10.1038/nature12580
Received
Accepted
Published online

The ElNiño–Southern Oscillation (ENSO) drives substantial variability in rainfall1, 2, 3, severe weather4, 5, agricultural production3, 6, ecosystems7 and disease8 in many parts of the world. Given that further human-forced changes in the Earth’s climate system seem inevitable9, 10, the possibility exists that the character of ENSO and its impacts might change over the coming century. Although this issue has been investigated many times during the past 20years, there is very little consensus on future changes in ENSO, apart from an expectation that ENSO will continue to be a dominant source of year-to-year variability9, 11, 12. Here we show that there are in fact robust projected changes in the spatial patterns of year-to-year ENSO-driven variability in both surface temperature and precipitation. These changes are evident in the two most recent generations of climate models13, 14, using four different scenarios for CO2 and other radiatively active gases14, 15, 16, 17. By the mid- to late twenty-first century, the projections include an intensification of both El-Niño-driven drying in the western Pacific Ocean and rainfall increases in the central and eastern equatorial Pacific. Experiments with an Atmospheric General Circulation Model reveal that robust projected changes in precipitation anomalies during ElNiño years are primarily determined by a nonlinear response to surface global warming. Uncertain projected changes in the amplitude of ENSO-driven surface temperature variability have only a secondary role. Projected changes in key characteristics of ENSO are consequently much clearer than previously realized.

At a glance

Figures

  1. Multi-model average (MMA) of the projected change in the structure of the standardized first EOF of interannual (high-pass-filtered, /`year-to-year/') variability for the four twenty-first-century scenarios.
    Figure 1: Multi-model average (MMA) of the projected change in the structure of the standardized first EOF of interannual (high-pass-filtered, ‘year-to-year’) variability for the four twenty-first-century scenarios.

    a, c, e, g, Surface temperature (ST); b, d, f, h, precipitation. The pattern for each model was standardized by the spatial standard deviation of EOF1 over the domain 0–360°E, 30°S to 30°N. The CMIP5 models were forced using RCP8.5 (a, b), RCP4.5 (c, d) and 1% CO2 (e, f). The CMIP3 models were forced using SRES A2 (g, h). Stippling indicates that more than 70% of models agree on the sign of change. Red shades indicate an increase in EOF1 (ST) and a decrease in EOF1 (precipitation).

  2. MMA of the difference between twentieth-century and twenty-first-century filtered ST and precipitation anomalies in El[thinsp]Nino years.
    Figure 2: MMA of the difference between twentieth-century and twenty-first-century filtered ST and precipitation anomalies in ElNiño years.

    a, c, e, g, Surface temperature (ST); b, d, f, h, precipitation. a, b, RCP8.5. c, d, RCP4.5. e, f, 1% CO2. g, h, The CMIP3 models were forced using SRES A2. The corresponding averages for ElNiño–LaNiña years are very similar although larger in magnitude (not shown). The contour lines show the MMA of the twentieth-century anomalies during ElNiño years. Stippling indicates agreement in more than 70% of models on the sign of change. Red shades indicate warming in ST, drying in precipitation.

  3. Precipitation along the Equator over the Pacific in the AGCM.
    Figure 3: Precipitation along the Equator over the Pacific in the AGCM.

    Experiments 1–5 as described in Extended Data Table 2. Values displayed in a and b are differences between precipitation with α1 (that is, with αSSTA_EN applied) and the corresponding experiment with α = 0. ‘1EN’, for example, corresponds to the case α = 1. ‘20C’ values are differences relative to precipitation in the experiment with α = 0 under twentieth-century conditions. ‘21C’ values are differences relative to precipitation in the experiment with α = 0 under twenty-first-century conditions (that is, with SSTA_GW applied). a, RCP8.5; b, A2. Values for the twentieth century (solid lines) and the twenty-first century with (dashed lines) and without (dotted lines) structural changes in the ElNiño SST anomaly (ΔSSTA_EN) are shown in a and b. The key in b also applies to a. Values displayed in c and d show the impact of global warming (solid lines) and structural changes in the ElNiño SST anomaly (dotted lines) on precipitation for α = 1, 2, 3 and 4: c, RCP8.5; d, A2. ΔP21C is the impact of global warming (that is, 21C20C) on the El Niño precipitation anomaly, and PdSSTα is the impact of structural change in the El Niño SST anomaly on the El Niño precipitation anomaly. The key in d also applies to c. See Methods for further details on the AGCM experiments conducted.

  4. Diagram illustrating main findings.
    Figure 4: Diagram illustrating main findings.

    It is often assumed that projected changes in ENSO amplitude are critically important in projections of all other ENSO impacts. We show instead that uncertainty in ENSO-driven SST variability does not necessarily exclude robust changes in other forms of ENSO-driven variability. This is apparent because robust projected changes in ENSO-driven rainfall variability in the equatorial Pacific are shown to occur despite uncertain changes in ENSO-driven ST variability. The rainfall response is caused by a nonlinear response to robust changes in background SST due to global warming (1), unchanged twentieth-century ENSO-driven SST variability (2), and uncertain changes in ENSO-driven SST variability (3). Components (1) and (2) dominate, resulting in more certainty in the rainfall projections than in the SST projections. The greater uncertainty in projected changes in ENSO-driven SST variability (3) results from uncertainty in projected change in the amplitude of SST variability (3a), and not from robust changes in the standardized pattern of ENSO-driven SST variability (3b). Grey shading is used to identify components making up the precipitation response that have uncertain projections.

  5. Leading patterns (standardized, first EOFs) of interannual variability in surface temperature (ST) and precipitation.
    Extended Data Fig. 1: Leading patterns (standardized, first EOFs) of interannual variability in surface temperature (ST) and precipitation.

    Observations29, 30 (top: a, ST; b, precipitation), CMIP5 models (c, ST; d, precipitation) and CMIP3 models (e, ST; f, precipitation). The model patterns are MMAs of the first EOF of each individual model. All data were spectrally filtered to remove variability with periods greater than 13years before EOF analysis and are standardized as described in Methods.

  6. Scatter plot showing the amplitude of EOF1 time series in the twenty-first century (y[thinsp]axis) and the twentieth century (x[thinsp]axis) in individual models under the four scenarios.
    Extended Data Fig. 2: Scatter plot showing the amplitude of EOF1 time series in the twenty-first century (yaxis) and the twentieth century (xaxis) in individual models under the four scenarios.

    Dots above the line (indicated by the dashed line) are projected increases; dots below the line y = x are projected decreases. The amplitude is measured here using the standard deviation. The time series for each model was re-scaled as described in the text. The vertical (dotted) line gives an observational estimate. The figures in the bottom right of the plot give the percentage of models above the diagonal; that is, with a projected increase.

  7. The SST anomalies (SSTAs) used in the AGCM experiments.
    Extended Data Fig. 3: The SST anomalies (SSTAs) used in the AGCM experiments.

    a, SST_EN; b, SSTA_GW (RCP8.5); c, SSTA_GW (A2); d, ΔSSTA_EN (RCP8.5); e, ΔSSTA_EN (A2). SSTA_EN is the observed SST anomaly averaged over all ElNiño events between 1978 and 2009, SSTA_GW is the change in background SST projected for the late twenty-first century, and ΔSSTA_EN is the projected multi-model average change in the filtered ElNiño SST pattern obtained from climate models.

  8. Multi-model average (MMA) of projected change in mean SST under the four scenarios.
    Extended Data Fig. 4: Multi-model average (MMA) of projected change in mean SST under the four scenarios.

    Note the high degree of similarity in the spatial structures between the scenarios and the enhanced equatorial warming towards the east.

  9. Impact of bias correction (or /`shifting/') on the MMA of [Dgr]EOF1 of surface temperature and precipitation for the four different twenty-first-century scenarios.
    Extended Data Fig. 5: Impact of bias correction (or ‘shifting’) on the MMA of ΔEOF1 of surface temperature and precipitation for the four different twenty-first-century scenarios.

    Each model’s first EOF was shifted east to maximize the spatial correlation coefficient with the observed first EOF. On average, roughly 20° (CMIP3 models) and 14° (CMIP5 models) of shifting was required. Some models did not require adjustment; others required eastward shifts up to 22.5°. The MMA of both unadjusted EOFs (‘raw’) and bias-corrected EOFs (‘shifted’) are shown. The dateline is indicated by the red dashed vertical line in each panel for reference only. ST: ad, raw; eh, shifted. Precipitation: hl, raw; mp, shifted. Stippling is applied if more than 70% of models agree on the sign of change. See Methods for further details.

  10. Precipitation along the Equator over the Pacific in AGCM experiments 6 and 7.
    Extended Data Fig. 6: Precipitation along the Equator over the Pacific in AGCM experiments 6 and 7.

    See Methods and Extended Data Table 2 for more details on these experiments. Values in the top two panels represent differences between precipitation with α1 and the corresponding experiment with α = 0. The twentieth-century figures, for example, are differences relative to precipitation in the experiment with α = 0 under twentieth-century conditions. The twenty-first-century figures are differences relative to precipitation in the experiment with α = 0 and SSTA_GW, and either twentieth-century CO2 concentrations (dotted lines) or late twenty-first-century CO2 concentrations (dashed lines) from experimental sets 6 and 7. a, RCP8.5; b, A2. The lower two panels depict the impact of warming only (21C20C, solid lines) and the impact of increasing CO2 concentrations only (21C SST&CO221C, dotted lines). c, RCP8.5; d, A2. The results show that the SST changes are primarily responsible for the precipitation response (a, b), and that the impact of CO2 change—over and above the impact it has as a result of the changes in SST it causes—is small compared with the nonlinear response (c, d).

  11. Contribution of PTH, PMCD, PCOV and E to the response in the AGCM.
    Extended Data Fig. 7: Contribution of PTH, PMCD, PCOV and E to the response in the AGCM.

    a, RCP8.5; b, A2. α = 1 and 4 only. Symbols are defined in Methods.

  12. Ability of CMIP5 models to simulate the spatial structure of observed EOF2 of ST, and projected changes in the relative frequency of central Pacific and eastern Pacific El[thinsp]Ninos.
    Extended Data Fig. 8: Ability of CMIP5 models to simulate the spatial structure of observed EOF2 of ST, and projected changes in the relative frequency of central Pacific and eastern Pacific ElNiños.

    a, Observed EOF2 of ST. b, Spatial correlation coefficient between observed EOF2 and both EOF2 and EOF3 in each model. c, Change in the relative frequency of central Pacific and eastern Pacific ElNiños under the RCP8.5 scenario.

Tables

  1. Climate models analysed
    Extended Data Table 1: Climate models analysed
  2. Description of the AGCM experiments conducted
    Extended Data Table 2: Description of the AGCM experiments conducted

References

  1. Ropelewski, C. F. & Halpert, M. S. Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Clim. 2, 268284 (2007)
  2. Australian Bureau of Meteorology and CSIRO. Climate Change in the Pacific: Scientific Assessment and New Research (eds Hennessy, K. Power, S. & Cambers, G.). (2011)
  3. Power, S., Casey, T., Folland, C., Colman, A. & Mehta, V. Interdecadal modulation of the impact of ENSO on Australia. Clim. Dyn. 15, 234319 (1999)
  4. Donelly, J. P. & Woodruff, J. D. Intense hurricane activity over the past 5,000 years controlled by ElNiño and the west African monsoon. Nature 447, 465468 (2007)
  5. Callaghan, J. & Power, S. B. Variability and decline in the number of severe tropical cyclones making land-fall over eastern Australia since the late nineteenth century. Clim. Dyn. 37, 647662 (2011)
  6. Hammer, G. L. et al. Advances in the application of climate prediction in agriculture. Agric. Syst. 70, 515553 (2001)
  7. McPhaden, M. J., Zebiak, S. E. & Glantz, M. H. ENSO as an integrating concept in earth science. Science 314, 17401745 (2006)
  8. Sari Kovats, R. S. et al. ElNiño and health. Nature 361, 14811489 (2003)
  9. Meehl, G. A. et al. in Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the 4th Assessment Report of the Intergovernmental Panel on Climate Change (eds Solomon, S. et al.) 747– 845 (Cambridge Univ. Press, 2007)
  10. Peters, G. P. et al. The challenge to keep global warming below 2°C. Nature Clim. Change 3, 46 (2013)
  11. Collins, M. et al. The impact of global warming on the tropical Pacific Ocean and ElNiño. Nature Geosci. 3, 391397 (2010)
  12. Vecchi, G. A. & Wittenberg, A. T. ElNiño and our future climate: where do we stand? Wiley Interdisc. Rev. Clim. Change 1, 260270 (2010)
  13. Meehl, G. et al. The WCRP CMIP3 multimodel dataset: a new era in climate change research. Bull. Am. Meteorol. Soc. 88, 13831394 (2007)
  14. Taylor, K. E., Stouffer, R. J. & Meehl, G. A. An overview of the CMIP5 and the experimental design. Bull. Am. Meteorol. Soc. 93, 485498 (2012)
  15. Moss, R. H. et al. The next generation of scenarios for climate change research and assessment. Nature 463, 747756 (2010)
  16. van Vuuren et al. The representative concentration pathways: an overview. Clim. Change 109, 531 (2011)
  17. Nakicenovic, N. et al. Special Report on Emissions Scenarios: A Special Report of Working Group III of the Intergovernmental Panel on Climate Change (Cambridge Univ. Press, 2000)
  18. Neelin, J. D. et al. ENSO theory. J. Geophys. Res. 103, 1426114290 (1998)
  19. Coelho, C. A. S. & Goddard, L. ElNiño-induced tropical droughts in climate change projections. J. Clim. 22, 64566476 (2009)
  20. Kug, J. S., An, S. I., Ham, Y. G. & Kang, I. S. Changes in ElNiño and LaNiña teleconnections over North Pacific–America in the global warming simulations. Theor. Appl. Climatol. 100, 275282 (2010)
  21. Meehl, G. A. & Teng, H. Multi-model changes in ElNiño teleconnections over North America in a future warmer climate. Clim. Dyn. 29, 779790 (2007)
  22. Lorenz, E. N. Empirical Orthogonal Functions and Statistical Weather Prediction (Massachusetts Inst. Technology Sci. Rep. no. 1, 1956)
  23. Xie, S.-P. et al. Global warming pattern formation: SST and rainfall. J. Clim. 23, 966986 (2010)
  24. Yeh, S.-W. et al. ElNiño in a changing climate. Nature 461, 511514 (2009)
  25. Guilyardi, E. et al. Understanding ElNiño in ocean–atmosphere general circulation models: progress and challenges. Bull. Am. Meteorol. Soc. 90, 325340 (2009)
  26. Power, S. B., Delage, F., Colman, R. & Moise, A. Consensus on twenty-first-century rainfall projections in climate models more widespread than previously thought. J. Clim. 25, 37923809 (2012)
  27. Chung, C. T. Y. et al. Nonlinear precipitation response to ElNiño and global warming in the Indo-Pacific. Clim. Dyn.. http://dx.doi.org/10.1007/s00382-013-1892-8 (2013)
  28. Ashok, K., Behera, S. K., Rao, S. A., Weng, H. & Yamagata, T. ElNiño Modoki and its possible teleconnection. J. Geophys. Res. 112, C11007 (2007)
  29. Xie, P. & Arkin, P. A. Global precipitation: a 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Am. Meteorol. Soc. 78, 25392558 (1997)
  30. Rayner, N. A. et al. Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res. 108, 4407 (2003)

Download references

Author information

Affiliations

  1. Centre for Australian Weather and Climate Research, Bureau of Meteorology, Docklands, Melbourne, Victoria, 3008, Australia

    • Scott Power,
    • François Delage,
    • Christine Chung,
    • Greg Kociuba &
    • Kevin Keay

Contributions

S.P. drafted the paper, devised the hypotheses and was primarily responsible for designing the methods used. F.D., C.C., G.K. and K.K. contributed to method design, implemented the very complex code required, performed extensive data analyses, contributed to the writing of this paper, produced the plots and assisted in their interpretation. C.C. also conducted the AGCM experiments.

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Author details

Extended data figures and tables

Extended Data Figures

  1. Extended Data Figure 1: Leading patterns (standardized, first EOFs) of interannual variability in surface temperature (ST) and precipitation. (482 KB)

    Observations29, 30 (top: a, ST; b, precipitation), CMIP5 models (c, ST; d, precipitation) and CMIP3 models (e, ST; f, precipitation). The model patterns are MMAs of the first EOF of each individual model. All data were spectrally filtered to remove variability with periods greater than 13years before EOF analysis and are standardized as described in Methods.

  2. Extended Data Figure 2: Scatter plot showing the amplitude of EOF1 time series in the twenty-first century (yaxis) and the twentieth century (xaxis) in individual models under the four scenarios. (134 KB)

    Dots above the line (indicated by the dashed line) are projected increases; dots below the line y = x are projected decreases. The amplitude is measured here using the standard deviation. The time series for each model was re-scaled as described in the text. The vertical (dotted) line gives an observational estimate. The figures in the bottom right of the plot give the percentage of models above the diagonal; that is, with a projected increase.

  3. Extended Data Figure 3: The SST anomalies (SSTAs) used in the AGCM experiments. (624 KB)

    a, SST_EN; b, SSTA_GW (RCP8.5); c, SSTA_GW (A2); d, ΔSSTA_EN (RCP8.5); e, ΔSSTA_EN (A2). SSTA_EN is the observed SST anomaly averaged over all ElNiño events between 1978 and 2009, SSTA_GW is the change in background SST projected for the late twenty-first century, and ΔSSTA_EN is the projected multi-model average change in the filtered ElNiño SST pattern obtained from climate models.

  4. Extended Data Figure 4: Multi-model average (MMA) of projected change in mean SST under the four scenarios. (1,062 KB)

    Note the high degree of similarity in the spatial structures between the scenarios and the enhanced equatorial warming towards the east.

  5. Extended Data Figure 5: Impact of bias correction (or ‘shifting’) on the MMA of ΔEOF1 of surface temperature and precipitation for the four different twenty-first-century scenarios. (488 KB)

    Each model’s first EOF was shifted east to maximize the spatial correlation coefficient with the observed first EOF. On average, roughly 20° (CMIP3 models) and 14° (CMIP5 models) of shifting was required. Some models did not require adjustment; others required eastward shifts up to 22.5°. The MMA of both unadjusted EOFs (‘raw’) and bias-corrected EOFs (‘shifted’) are shown. The dateline is indicated by the red dashed vertical line in each panel for reference only. ST: ad, raw; eh, shifted. Precipitation: hl, raw; mp, shifted. Stippling is applied if more than 70% of models agree on the sign of change. See Methods for further details.

  6. Extended Data Figure 6: Precipitation along the Equator over the Pacific in AGCM experiments 6 and 7. (460 KB)

    See Methods and Extended Data Table 2 for more details on these experiments. Values in the top two panels represent differences between precipitation with α1 and the corresponding experiment with α = 0. The twentieth-century figures, for example, are differences relative to precipitation in the experiment with α = 0 under twentieth-century conditions. The twenty-first-century figures are differences relative to precipitation in the experiment with α = 0 and SSTA_GW, and either twentieth-century CO2 concentrations (dotted lines) or late twenty-first-century CO2 concentrations (dashed lines) from experimental sets 6 and 7. a, RCP8.5; b, A2. The lower two panels depict the impact of warming only (21C20C, solid lines) and the impact of increasing CO2 concentrations only (21C SST&CO221C, dotted lines). c, RCP8.5; d, A2. The results show that the SST changes are primarily responsible for the precipitation response (a, b), and that the impact of CO2 change—over and above the impact it has as a result of the changes in SST it causes—is small compared with the nonlinear response (c, d).

  7. Extended Data Figure 7: Contribution of PTH, PMCD, PCOV and E to the response in the AGCM. (182 KB)

    a, RCP8.5; b, A2. α = 1 and 4 only. Symbols are defined in Methods.

  8. Extended Data Figure 8: Ability of CMIP5 models to simulate the spatial structure of observed EOF2 of ST, and projected changes in the relative frequency of central Pacific and eastern Pacific ElNiños. (389 KB)

    a, Observed EOF2 of ST. b, Spatial correlation coefficient between observed EOF2 and both EOF2 and EOF3 in each model. c, Change in the relative frequency of central Pacific and eastern Pacific ElNiños under the RCP8.5 scenario.

Extended Data Tables

  1. Extended Data Table 1: Climate models analysed (284 KB)
  2. Extended Data Table 2: Description of the AGCM experiments conducted (75 KB)

Additional data