Uncertainties in Global Warming Projection

According to the IPCC Fourth Assessment Report (AR4) [ IPCC, 2007 ], the global mean surface temperature in 2100 is projected to be 1.1–6.4°C higher than the 1980–1999 mean state, with the best estimation of increase at 1.8–4.0°C. It should be noted that this is the projected range of increase rather than the so-called uncertainties in global warming projection. Specifically, 1.1°C and 1.8°C are the lower limit and the best estimation for the B1 emissions scenario (the low emissions scenario), respectively; while 4.0°C and 6.4°C are the corresponding upper limit and the best estimation for the A1F1 emissions scenario (the high emissions scenario). Therefore, the range of the warming values from IPCC AR4 just reflects the model’s response to various designed emissions scenarios. The projected value for each scenario is the average of multi-model ensemble. A total of 23 models from the climate research centers around the globe are included in the ensemble. The number of participated models is different for some time intervals and scenarios. For most of intervals, the results are available only from 16–17 models. The projected values for warming in IPCC AR4 are based on the multi-model mean, which reflects the best estimation. Calculated standard deviation shows the spread among the models.

Therefore, the wide range of global warming projection largely represents the differences among the emissions scenarios rather than the real uncertainty in climate models. As mentioned in IPCC AR4, besides projecting greenhouse gases (GHGs) emissions, for a given GHGs emissions there are still errors leading to uncertainty in climate projection. These errors are manifested by the concentration of GHGs in the atmosphere, its radiative forcing and the consequent climate changes. In IPCC AR4, the uncertainty in climate projection is defined as a measure that indicates our lack of understanding of future climate conditions [ IPCC , 2007 ]. Generally, the uncertainties come from lack of information or disagreement of knowledge. For example, they can be derived from data quality control, misunderstanding of concepts, and even inaccurate estimation of future human activity. The uncertainty can be quantified, for example, through a tunable parameter in various models, or qualitatively described, through a subjective discussion made by a group of professionals. Charney et al. [1979] in his pioneering work first indicated that global mean temperature could increase (3±1.5)°C with doubling CO₂ concentration, in which the ± 1.5°C refers to the uncertainties that people used to call expert estimation [ IPCC, 1990 ; IPCC, 1996  ;  IPCC, 2001 ] due to lack of accurate calculation. Charney’s conclusion was an early but significantly accurate estimate, which was adopted by IPCC FAR (First Assessment Report), SAR (Second Assessment Report), and TAR (Third Assessment Report). Results from most of current scientists are similar and within the range of Charney’s estimation. It is certainly necessary to use abundant observations and models to accurately calculate rather than guesstimate the uncertainties in projection of future climate.

According to Chapter 10 of IPCC AR4 [ IPCC, 2007 ], the uncertainties in climate projection came from two key sources: climatic internal variability and bias in reproducing physical and chemical processes in climate system models. The climatic internal variabilities can be partly reproduced by using ensemble methods, such as the multi-initial-condition ensemble. However, the internal variabilities cannot be explained by the observed climatic variabilities, the latter are actually the responses to natural and anthropogenic forcing. In fact, the internal variability should be defined as the responses of climate system to a condition with fixed external forcing. The bias in reproducing the physical and chemical processes in models came from three key treatments [ Forest et al ., 2002  ;  Knutti et al ., 2002 ]: equilibrium climate sensitivity [ Kiehl, 2007  ;  Knutti and Hegerl, 2008 ], ocean heat uptake [ Forest et al ., 2008  ;  Boe et al ., 2009 ], and anthropogenic aerosols [ Kiehl, 2007  ;  Wigley and Raper, 2001 ]. IPCC AR4 [ IPCC, 2007 ] proposed multi-model ensemble as the solution to quantify and reduce any single model bias. There were 29 models around the world participating the model projections in IPCC AR4. Each model should generate 1–3 members, and there were more than twenty runs in total. They were considered as a typical multi-model multi-member ensemble experiment as initiated by CMIP-3. IPCC AR4 defined the range of −40% to 60% of the global mean temperature in 2100 as the likely (> 66% probability) spread. It should be noted that CMIP-3 design reflected most, but not all, of the models’ uncertainties. Rowlands et al. [2012] further proposed perturbed physics ensemble (PPE), a better approach to address the models’ uncertainty in reproducing physical processes.

Actually, the PPE approach had already been proposed by Wigley and Raper [2001] before IPCC AR4. Furthermore, another useful ensemble method was also mentioned in AR4 [ IPCC, 2007 ]: ensemble of simple models [ Wigley and Raper, 2001 ], EMICs [ Knutti et al., 2002 ], and simplified AOGCMs [ Murphy et al ., 2004  ;  Stainforth et al ., 2005 ]. Later studies showed that this was an effective method for determining the uncertainties of real models [ Jackson et al ., 2008  ;  Collins et al ., 2011 ]. Certainly, the PPE method also helps in refining the second kind of uncertainties.

Rowlands et al. [2012] performed comprehensive simulation of global mean temperature for the period of 1920–2080 using HadCM3L. Three components, i.e., atmosphere, ocean, and sulfate aerosol, were included in the model. The emissions scenario of A1B was applied as the GHGs forcing for the years of 2000–2080. In addition, a variety of volcanic forcing, sulfate aerosol, and solar activity derived from the past years were also included in the model. Thus, not only the anthropogenic forcing but also the natural forcing was considered for 2000–2080. Specifically, they designed 153 cases of perturbation of the model. Each case was configured with 10 flux-correction schemes resulted in 1,530 experiments in total. Further, each experiment was forced by various GHGs, volcanic eruptions, solar activity, and sulfate emissions. At the end, there were 9,745 runs in total. This probably has been the biggest ensemble experiment so far to perform climate projection. Figure 1 shows the results of global mean temperature anomaly for the period of 1980–2080 relative to 1961–1990 mean state. It can be seen that the observed temperatures of 1978–2010 were successfully reproduced by the model.

Figure 1. Simulations of global mean temperature. Blue color indicates spread of all runs. The thick blue lines show the likely range (> 66% probability). Black line shows the observations. Red bars stand for the IPCC AR4 expert estimation of likely range around 2050 and 2080. All temperatures are relative to 1961–1990 mean [ Rowlands et al., 2012 ]

Figure 2 shows surface temperature hindcasts of 2001–2010 and the forecasts for 2041–2060. The vertical dashed lines indicate the likely range or 66% in statistical significance as defined in IPCC AR4. For the 2001–2010 hindcast results, the likely warming amplitude is ranging from 0.3°C to 0.9°C with a mean value of 0.6°C, which is slightly higher than the observed warming trend at 0.5°C probably due to discounting the effects of the reduced stratospheric water vapors [ Solomon et al., 2010 ] and the weakened solar activity [ Lockwood, 2010 ] in the model. For the 2041–2060 forecasts, the projected likely warming is 1.4–3.0°C, as shown in Figure 2 b. Both the top and bottom limits of this range are greater than the IPCC AR4 estimations, probably because of the fact that climatic sensitivity was enlarged, external natural forcing was included, and oceanic thermal uptake was accurately determined by bounded observations. Although the model estimated uncertainty can still be refined, this work signifies that for the first time the community was able to really calculate the climatic sensitivity in 2050 by using a great number of ensemble members, and therefore we have gone beyond the subjective expert estimation era.

Figure 2. Goodness-of-fit to temperature changes as a function of global mean warming. Colored triangles represent members of the HadCM3L perturbed-physics ensemble with colors denoting climate sensitivity (CS); D denotes model configurations differing in natural forcing and anthropogenic sulfate emissions; black crosses indicate realizations of model error and temperature changes from simulations of internal variability with the horizontal line denoting the 66% of error distribution; vertical dotted lines show the range of the HadCM3L ensemble within 66% of confidence interval; grey triangles show the cases with the adjustments outside ± 5 W m–2 . Black vertical bar and grey band in (a) show observations and likely range; horizontal bar in (b) gives the expert IPCC AR4 likely range; black filled circles show CMIP-3 simulations; black open circles show QUMP-HadCM3 simulations; arrowed larger triangles refer to model with low-response ensemble A and high-response ensemble B [ Rowlands et al., 2012 ]

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 41005035) and the Strategic Priority Research Program of Chinese Academy of Sciences (No. XDA05080801 and XDA05090104).

References

1. Boe et al., 2009 J. Boe, A. Hall, X. Qu; Deep ocean heat uptake as a major source of spread in transient climate change simulations; Geophys. Res. Lett., 36 (2009) L22701
2. Charney et al., 1979 J.G. Charney, A. Arkawa, D.J. Baker, et al.; Carbon dioxide and climate: A scientific assessment, National Academy of Sciences Press (1979), p. 33
3. Collins et al., 2011 M. Collins, B.B.B. Booth, B. Bhaskaran, et al.; Climate model errors, feedbacks and forcing: A comparison of perturbed physics and multi-model ensembles; Climate Dyn., 36 (2011), pp. 1737–1766
4. Forest et al., 2002 C.E. Forest, P.H. Stone, A.P. Sokolov, et al.; Quantifying uncertainties in climate system properties with the use of recent climate observations; Science, 295 (2002), pp. 113–116
5. Forest et al., 2008 C.E. Forest, P.H. Stone, A.P. Sokolov; Constraining climate model parameters from observed 20th century changes; Tellus, A60 (2008), pp. 911–920
6. IPCC, 1990 IPCC; J.T. Houghton (Ed.), et al. , Climate Change 1990: The IPCC Scientific Assessment, Cambridge University Press (1990), p. 365
7. IPCC, 1996 IPCC; J.T. Houghton (Ed.), et al. , Climate Change 1995: The Science of Climate Change. Contribution of Working Group I to the Second Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press (1996), p. 572
8. IPCC, 2001 IPCC; J.T. Houghton (Ed.), et al. , Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press (2001), p. 881
9. IPCC, 2007 IPCC; S.D. Solomon (Ed.), et al. , Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press (2007), p. 996
10. Jackson et al., 2008 C.S. Jackson, M.K. Sen, G. Huerta, et al.; Error reduction and convergence in climate prediction; J. Climate, 21 (2008), pp. 6698–6708
11. Kiehl, 2007 J.T. Kiehl; Twentieth century climate model response and climate sensitivity; Geophys. Res. Lett., 34 (2007) L22710
12. Knutti and Hegerl, 2008 R. Knutti, G.C. Hegerl; The equilibrium sensitivity of the Earth’s temperature to radiation changes; Nature Geoscience, 1 (2008), pp. 735–743
13. Knutti et al., 2002 R. Knutti, T.F. Stocker, J. Fortunat, et al.; Constraints on radiative forcing and future climate change from observations and climate model ensembles; Nature, 416 (2002), pp. 719–723
14. Lockwood, 2010 M. Lockwood; Solar change and climate: An update in the light of the current exceptional solar minimum; Proc. R. Soc. A., 466 (2010), pp. 303–329
15. Murphy et al., 2004 J.M. Murphy, D.M. Sexton, D.N. Barnett, et al.; Quantification of modeling uncertainties in a large ensemble of climate change simulations; Nature, 430 (2004), pp. 768–772
16. Rowlands et al., 2012 D.J. Rowlands, D.J. Frame, D. Ackerley, et al.; Broad range of 2050 warming from an observationally constrained large climate model ensemble; Nature Geoscience, 5 (2012), pp. 256–260
17. Solomon et al., 2010 S. Solomon, K.H. Rosenlof, R.W. Portmann, et al.; Contributions of stratospheric water vapor to decadal change in the rate of global warming; Science, 327 (2010), pp. 1219–1223
18. Stainforth et al., 2005 D.A. Stainforth, T. Aina, C. Christensen, et al.; Uncertainty in predictions of the climate response to rising levels of greenhouse gases; Nature, 433 (2005), pp. 403–406
19. Wigley and Raper, 2001 T.M.L. Wigley, S.C.B. Raper; Interpretation of high projections for global mean warming; Science, 293 (2001), pp. 451–454