Skip to main content

Advertisement

SpringerLink
Log in

Are climate models credible worlds? Prospects and limitations of possibilistic climate prediction

  • Original paper in Philosophy of Science
  • Published:
European Journal for Philosophy of Science Aims and scope Submit manuscript

Abstract

Climate models don’t give us probabilistic forecasts. To interpret their results, alternatively, as serious possibilities seems problematic inasmuch as climate models rely on contrary-to-fact assumptions: why should we consider their implications as possible if their assumptions are known to be false? The paper explores a way to address this possibilistic challenge. It introduces the concepts of a perfect and of an imperfect credible world, and discusses whether climate models can be interpreted as imperfect credible worlds. That would allow one to use models for possibilistic prediction and salvage widespread scientific practice.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. See, e.g., Betz (2007), Stainforth et al. (2007), Parker (2010a, 2010b), Frigg et al. (2013), Frigg et al. (2014); compare Katzav et al. (2012) for a recent review. More specifically, this literature primarily questions precise probabilistic climate forecasts and does not necessarily preclude that we get reliable imprecise probabilistic predictions. My view is that as soon as we try to set up imprecise probabilistic climate forecasts, systematic variation of priors will blow up the probability ranges to the unit interval and effectively result in pure possibilistic prediction. So, for example, Meinshausen et al. (2009) determined the imprecise probability of reaching the 2-degree-target given certain CO2 emission scenarios. But the authors only considered published probability density functions of climate sensitivity as priors; in a personal communication, they conceded that a more systematic variation of priors would yield uninformative imprecise probabilities. Of course, these claims about imprecise climate prediction are in need of further justification, which however goes beyond the scope of this paper.

  2. Grüne-Yanoff (2014), however, clarifies that serious possibility requires consistency with the entire background knowledge, see also below.

  3. Note that Sugden does not explicitly answer this question in the affirmative.

  4. Nor does the claim that there exist no fossil fuels on this planet whatsoever.

  5. On false assumptions in economic models and the implications for Sugden’s account see, e.g., Rodrguez and Zamora Bonilla (2009, 115).

  6. Levi (1980, 2–5), Lorenzen (1987, 106ff.) and Grüne-Yanoff (2014, 853) make use of the same epistemic notion of possibility, which should not be confused with the metaphysical concept of real possibility, discussed by, e.g., Hartshorne (1963), Gibbs (1970) and Deutsch (1990).

  7. It might be objected that, according to this explication of serious possibility, nearly any statement about a system is seriously possible if our knowledge about the system is very limited. First of all, I would tend to accept this consequence, which I don’t consider as a counter-intuitive proliferation of relevant possibilities. More importantly, though, I’m worried that climate models don’t even verify serious possibilities in this weak sense. This worry is only multiplied if consistency with background knowledge is merely a necessary, but not a sufficient condition for being a serious possibility, as the objection has it.

  8. Maybe the explication could be modified so as to cover counterfactuals (such as “a stock crash was a serious possibility two years ago, although it actually didn’t happen”) by introducing a time-index, i.e. a statement P is seriously possible at time t if and only if P is consistent with the entire body of background knowledge K at t.

  9. So there are different kinds of serious possibilities. But which of these do ultimately count for policy making? Which of these should fuel our practical reasoning? I struggled with this problem in Betz (2009). Today I’d say that these questions are ill-posed, since they presume a false alternative. We don’t have to choose: All types of possibilistic hypotheses (verified, falsified, merely articulated ones) are decision relevant and should inform policy making. Scientists may employ these types of possibility statements to communicate their limited knowledge about future climate change in a differentiated way (Betz 2010).

  10. Parker (2010a, 267) observes: “Least controversial is the following: Insofar as each model or model version (+initial conditions) in an ensemble is plausibly adequate for the predictive task of interest, then the simulations produced indicate a set of predictive outcomes that are plausible, given current knowledge.” The 1994 IPCC Guidelines note: “[Only] general circulation models, possibly in conjunction with nested regional models […], have the potential to provide consistent and physically consistent estimates of regional climate change, which are required in impact analysis.” (Carter et al. 1994, 23) Being careful not to interpret ensemble results probabilistically, the IPCC SRREN assumes nonetheless that integrated assessment models provide relevant possibilities (Fischedick et al. 2011). Decision-theoretic discussions of climate model uncertainties tend to take for granted that (climate) model simulations identify policy-relevant possibilities (e.g., Lempert (2002, 2003); McInernery et al. (2012); Kunreuther et al. (2013)). Only Stainforth and Smith (2012) indicate that maybe not all projections generated with climate models are plausible and seriously possible, saying: “It is important to distinguish between questions for which current models are useful as prediction engines and those for which the models merely probe possibilities. The role of science is to reflect on the plausibility and relevance of such possibilities.”

  11. This is the point I already raised above, when discussing Sugden.

  12. My impression is that more and more climatologists start to question the policy-relevance of climate modelling, see, e.g., a recent interview with D. Stainforth (Steele 2013).

  13. I use “realistic” to indicate that the model is interpreted in view of domain 𝕯. So, a realistic description of a climate model is not just the presentation of its formal, mathematical structure, but includes also an interpretation of the mathematical symbols in terms of climatological variables.

  14. Note that it may turn out that climate models are imperfect credible worlds with regard to one class of implications (e.g. temperature projections) but not with regard to another class (e.g. extreme events). The broader the class of T-implications, the stronger the claim that a model represents an imperfect credible world w.r.t. T-implications.

  15. The controversy about idealisation in scientific reasoning is with us since the very beginnings of modern science. And at first glance, the task we face here may appear similar to the challenge of justifying unrealistic assumptions in models or explanations as it is extensively discussed in the literature; cf. Weisberg (2007) for a general review and Winsberg (2006) in particular for a discussion pertinent to computer simulations. Yet idealisations have typically been assessed against representational and veritistic ideals: Does an argument explain despite unrealistic premisses? Does a model truthfully represent despite idealisations? Does a method yield reliable forecasts despite contrary-to-fact assumptions? As explained in Section 5, we have departed from this representational understanding of models. If climate models aspire to be imperfect credible worlds rather than representations of the climate system, contrary-to-fact assumptions pose novel problems and have to be judged against different standards. This said, however, there will inevitably be similarities between the treatment of idealisations in a representational and a credible-worlds framework. One such correspondence seems particularly obvious. The argument from causal irrelevance and the possibilistic induction, discussed below, rely on causal premisses that closely resemble assumptions which underly so-called minimalist idealisations (Weisberg 2007, 642–5). Claims about causal irrelevance and core causal processes figure prominently in both accounts. The specific argumentative function of these claims, however, varies. (See also footnote 22.) Moreover, we will see that the argument from possibilistic induction assumes a clear-cut structural-similarity hypothesis as premiss. Representational and credible-worlds arguments are hence interwoven and the traditional problem of idealisation indirectly pertains to the justification of imperfect-credible-world hypotheses.

  16. A further problem is raised by the fact that the above conclusions don’t state that the entire climate model C (plus some of its contrary-to-fact assumptions) is an imperfect credible world. The conclusions refer to different parts of C. So they don’t rule out that Ψ1 plus C 2 (and hence Ψ1 plus C) gives rise to T-implications which are inconsistent with K.

  17. Accordingly, statement Dfw has the general form: Given the record Rt of endogenous climate variables of AOGCM, the fresh water flux at time t equals F(Rt).

  18. Let’s agree on the following notation: “PP”, “GHG”, “LUC”, etc. refer to different domains or aspects of the climate system, namely to regional precipitation patterns, the causal processes that determine greenhouse gas forcing as a function of past temperature, land use change, etc. Whereas “Dpp”, “Dghg”, “Dluc”, etc. denote specific statements about these domains (tokens), we refer to a class of statements about those domains by “PP-implications”, “PP-statements”, etc. (types).

  19. At most, the following steps provide a rough outline of the argumentation. Steps 1 to 3 are reconstructed in more detail in the Appendix. The semi-formal analysis reveals that the possibilistic argumentation is even more complex, demanding and problematic than the discussion in the main text suggests.

  20. That’s because Dfw and Dghg are stated as conditional auxiliaries with antecedent conditions that screen off other causal impacts.

  21. In terms of the detailed reconstruction in the appendix: Premiss (1) in argument E appears to entail that all PP-implications of ESM are consistent with K and that ESM is hence an imperfect credible world w.r.t. PP-implications.

  22. These tasks may profit from further exploiting the similarities between the argument schemes proposed in this paper and the literature on so-called minimalist idealisation (see Weisberg 2007), in particular the work by Michael Strevens (e.g. Strevens 2008).

References

  • Betz, G. (2007). Probabilities in climate policy advice: a critical comment. Climatic Change, 85(1–2), 1–9.

    Article  Google Scholar 

  • Betz, G. (2009). What range of future scenarios should climate policy be based on? Modal falsificationism and its limitations. Philosophia Naturalis, 46(1), 133–158.

    Article  Google Scholar 

  • Betz, G. (2010). What’s the worst case? The methodology of possibilistic prediction. Analyse und Kritik, 32(1), 87–106.

    Google Scholar 

  • Carter, T., Perry, M., Harasawa, H., & Nishioka, S. (1994). IPCC technical guidelines for assessing climate change impacts and adaptations. University College London (London, UK), National Center for Environmental Studies (Tsukuba, Japan).

  • de Donato Rodríguez, X., & Zamora Bonilla J.P. (2009). Credibility, idealisation, and model building: an inferential approach. Erkenntnis, 70(1), 101–118.

    Article  Google Scholar 

  • Deutsch, H. (1990). Real possibility. Noûs, 24(5), 751–755.

    Article  Google Scholar 

  • Fischedick, M., Schaeffer, R., Adedoyin, A., Akai, M., Bruckner, T., Clarke, L., Krey, V., et al. (2011). Mitigation potential and costs. In O. Edenhofer, R. Pichs-Madruga, Y. Sokona, K. Seyboth, P. Matschoss, S. Kadner, T. Zwickel et al. (Eds.), IPCC special report on renewable energy sources and climate change mitigation. New York: Cambridge University Press.

  • Flato, G., Marotzke, J., Abiodun, B., Braconnot, P., Chou, S. C., Collins, W., Cox, P., et al. (2013). Evaluation of climate models. In T.F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, & P.M. Midgley (Eds.), The physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change (pp. 741–866). Cambridge: Cambridge University Press.

  • Frigg, R., Bradley, S., Machete, R.L., & Smith, L.A. (2013). Probabilistic forecasting: why model imperfection is a poison pill. In H. Andersen, D. Dieks, W.J. Gonzalez, T. Uebel, & G. Wheeler (Eds.), New challenges to philosophy of science. The philosophy of science in a European perspective (pp. 479-491). Dordrecht: Springer.

  • Frigg, R., Smith, L.A., & Stainforth, D.A. (2014). The myopia of imperfect climate models: the case of UKCP09. Philosophy of Science, 80(5), 886–897.

    Article  Google Scholar 

  • Gibbs, B. (1970). Real possibility. American Philosophical Quarterly, 7(4), 340–348.

    Google Scholar 

  • Giere, R. N. (1988). Explaining science: a cognitive approach. Chicago: University of Chicago Press.

    Book  Google Scholar 

  • Grüne-Yanoff, T. (2009). Learning from minimal economic models. Erkenntnis, 70(1), 81– 99.

    Article  Google Scholar 

  • Grüne-Yanoff, T. (2014). Appraising models nonrepresentationally. Philosophy of Science, 80(5), 850–861.

    Article  Google Scholar 

  • Hansson, S.O. (1997). The limits of precaution. Foundations of Science, 2(2), 293–306.

    Article  Google Scholar 

  • Hansson, S.O. (2011). Coping with the unpredictable effects of future technologies. Philosophy & Technology, 24(2), 137–149.

    Article  Google Scholar 

  • Hartshorne, C. (1963). Real possibility. The Journal of Philosophy, 60(21), 593–605.

    Article  Google Scholar 

  • Katzav, J. (2014). The epistemology of climate models and some of its implications for climate science and the philosophy of science. Studies in History and Philosophy of Science Part B: Studies in History and Philosophy of Modern Physics, 46(2), 228–238.

    Article  Google Scholar 

  • Katzav, J., Dijkstra, H.A., & de Laat, A.T.J. (2012). Assessing climate model projections: State of the art and philosophical reflections. Studies in History and Philosophy of Science Part B: Studies in History and Philosophy of Modern Physics, 43(4), 258–276.

  • Kunreuther, H., Heal, G., Allen, M., Edenhofer, O., Field, C.B., & Yohe, G. (2013). Risk management and climate change. Nature Climate Change, 3(5), 447–450.

    Article  Google Scholar 

  • Lempert, R.J., Popper, S.W., & Bankes, S.C. (2002). Confronting surprise . Social Science Computer Review, 20(4), 420–440.

    Article  Google Scholar 

  • Lempert, R.J., Popper, S.W., & Bankes, S.C. (2003). Shaping the next one hundred years: new methods for quantitative, long-term policy analysis. Santa Monica: RAND.

    Google Scholar 

  • Levi, I. (1980). The enterprise of knowledge. An essay on knowledge, credal probability and chance. Cambridge, Mass.: MIT Press.

    Google Scholar 

  • Lloyd, E.A. (2010). Confirmation and robustness of climate models. Philosophy of Science, 77(5), 971–984.

    Article  Google Scholar 

  • Lorenzen, P. (1987). Lehrbuch der konstruktiven Wissenschaftstheorie. Mannheim, Wien, Zrich: BI-Wissenschaftsverlag.

  • McInernery, D., Lempert, R., & Keller, K. (2012). What are robust strategies in the face of uncertain climate threshold responses? Climatic Change, 112(3–4), 547–568.

    Article  Google Scholar 

  • Meinshausen, M., Meinshausen, N., Hare, W., Raper, S. C. B., Frieler, K., Knutti, R., Frame, D. J., & Allen, M. R. (2009). Greenhouse-gas emission targets for limiting global warming to 2 degrees C. Nature, 458(7242), 1158–1162.

    Article  Google Scholar 

  • Meinshausen, M., Raper, S.C.B., & Wigley, T.M.L. (2011). Emulating coupled atmosphere-ocean and carbon cycle models with a simpler model, MAGICC6-Part 1: model description and calibration. Atmospheric Chemistry and Physics, 11(4), 1417–1456.

    Article  Google Scholar 

  • Parker, W.S. (2009). Confirmation and adequacy-for-purpose in climate modelling. Proceedings of the Aristotelian Society, 83(1), 233–249.

  • Parker, W.S. (2010a). Predicting weather and climate: uncertainty, ensembles and probability. Studies in History and Philosophy of Modern Physics, 41(3), 263–272.

    Article  Google Scholar 

  • Parker, W.S. (2010b). Whose probabilities? Predicting climate change with ensembles of models. Philosophy of Science, 77(5), 985–997.

    Article  Google Scholar 

  • Schmidt, G., & Sherwood, S. (2015). A practical philosophy of complex climate modelling. European Journal for Philosophy of Science. Early on-line, doi: 10.1007/s13194-014-0102-9 .

  • Shue, H. (2010). Deadly delays, saving opportunities: creating a more dangerous world? In S.M. Gardiner, D. Jamieson, H. Shue (Eds.), Climate ethics: essential readings (pp. 146–162). New York: Oxford University Press.

  • Stainforth, D.A., Allen, M.R., Tredger, E.R., & Smith, L.A. (2007). Confidence, uncertainty and decision-support relevance in climate predictions. Philosophical Transactions of the Royal Society A-Mathematical Physical and Engineering Sciences, 365(1857), 2145–2161.

    Article  Google Scholar 

  • Stainforth, D.A., & Smith, L.A. (2012). Clarify the limits of climate models. Nature, 489(7415), 208.

    Article  Google Scholar 

  • Steele, K. (2013). Interview with David Stainforth. The Reasoner, 7(2), 15–18.

    Google Scholar 

  • Strevens, M. (2008). Depth: an account of scientific explanation. Cambridge, Mass.: Harvard University Press.

    Google Scholar 

  • Sugden, R. (2009). Credible worlds, capacities and mechanisms. Erkenntnis, 70(1), 3–27.

    Article  Google Scholar 

  • Weisberg, M. (2007). Three kinds of idealization. The Journal of Philosophy, 104(12), 639–659.

    Google Scholar 

  • Winsberg, E. (2006). Models of success vs. the success of models: reliability without truth. Synthese, 152(1), 1–19.

    Article  Google Scholar 

Download references

Acknowledgments

I’d like to thank two anonymous reviewers of EJPS and especially the guest editors Wendy Parker and Joel Katzav for their detailed and extremely helpful feedback.

Author information

Authors and Affiliations

  1. Karlsruhe Institute of Technology, Institute of Philosophy, Kaiserstr. 12, D-76135, Karlsruhe, Germany

    Gregor Betz

Authors
  1. Gregor Betz
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Gregor Betz.

Appendix

Appendix

This appendix reconstructs steps 1-3 of the serial argumentation, outlined in Section 7, in a semi-formal way. Arguments A, B and C below explicate step 1; argument D spells out step 3; and arguments E and F develop step 2. The arguments’ individual inferences are formally or materially valid.

1.1 Argument A

  1. (1)

    AOGCMcore, i.e. what AOGCM says about the PP-relevant causal dynamics, is a perfect credible world. [AOGCMcore&K is consistent.]

  2. (2)

    The auxiliaries (Dfw & Dghg & Dsol) constitute a perfect credible world. [(Dfw&Dghg&Dsol)&K is consistent.]

  3. (3)

    The auxiliaries (Dfw & Dghg & Dsol) and AOGCMcore are marginally consistent relative to K.

  4. (4)

    Thus: There exists a perfect credible world C ⋆ according to which AOGCMcore (what AOGCM says about the PP-relevant causal dynamics) and the auxiliaries (Dfw & Dghg & Dsol) are true. [AOGCMcore&(Dfw&Dghg&Dsol)&K is consistent.]

  5. (5)

    Every PP-implication of AOGCM and (Dfw & Dghg & Dsol) is a PP-implication of AOGCMcore and (Dfw & Dghg & Dsol). [PP-implications of AOGCM and its auxiliaries are fully determined by the model’s PP-relevant causal processes and auxiliaries.]

  6. (6)

    Thus: All PP-implications of model AOGCM and auxiliaries (Dfw & Dghg & Dsol) are PP-implications of a perfect credible world C ⋆.

1.2 Argument B (pro A.1)

  1. (1)

    AOGCM did get the past PP-relevant causal dynamics right.

  2. (2)

    The (future) PP-relevant causal dynamics may resemble the past PP-relevant causal dynamics.

  3. (3)

    If the T-relevant causal dynamics may resemble the T P -relevant causal dynamics and model M did get the latter right, then M core , i.e. what M says about the T-relevant causal dynamics, is a perfect credible world.

  4. (4)

    Thus: AOGCMcore, i.e. what AOGCM says about the PP-relevant causal dynamics, is a perfect credible world.

1.3 Argument C (pro A.2)

  1. (1)

    Dfw is a perfect credible world.

  2. (2)

    Dghg is a perfect credible world.

  3. (3)

    Dsol is a perfect credible world.

  4. (4)

    Any two subsets of (Dfw, Dghg, Dsol) are marginally consistent relative to K.

  5. (5)

    Thus: The auxiliaries (Dfw & Dghg & Dsol) constitute a perfect credible world. [(Dfw&Dghg&Dsol)&K is consistent.]

Comment: Steps 4-6 of the serial argumentation, not reconstructed here, support this argument. In particular, premiss C.1 is supported by a possibilistic induction, applied to ISM (step 6); premiss C.2 is supported by an argument from emulation, applied to CCM (step 5); premiss C.3 is supported by an argument from emulation, applied to the extrapolation of past solar activity (step 4).

1.4 Argument D (pro C.4)

  1. (1)

    FW-processes, GHG-processes and SOL-events, which are described by the auxiliaries Dfw, Dghg and Dsol, are pairwise causally unrelated; i.e., they have no common causes and don’t cause each other.

  2. (2)

    FW-processes, GHG-processes and SOL-events represent spatio-temporally distinct physical events and processes.

  3. (3)

    Background knowledge K implies nothing about events and processes jointly caused by FW-processes, GHG-processes and SOL-events.

  4. (4)

    If statements DX,DY and DZ describe spatio-temporally distinct physical events and processes X,Y and Z which are pairwise causally unrelated, and if background knowledge K has no implications about events and processes jointly caused by X,Y and Z, then any joint implications of DX,DY and DZ are consistent with K.

  5. (5)

    If any joint implications of statements DX,DY and DZ are consistent with K, then (any two subsets of) DX,DY and DZ are marginally consistent relative to K.

  6. (6)

    Thus: Any two subsets of (Dfw, Dghg, Dsol) are marginally consistent relative to K.

Comment: Note that this particular version of the argument from causal irrelevance seizes upon our ignorance of causal processes, cf. D.3, and his hence specifically apt to warrant possibilistic conclusions.

1.5 Argument E (pro A.3)

  1. (1)

    The coupled earth system model ESM plus its auxiliaries emulates a more complex coupled model plus auxiliaries with respect to all statements that are jointly implied by AOGCMcore and (Dfw & Dghg & Dsol), but not by any of the two alone; where the more complex coupled model plus its auxiliaries is jointly consistent with the background knowledge K.

  2. (2)

    If model M 1 X-emulates model M 2 , then all X-implications of M 1 are X-implications of M 2 . [Partial explication of the concept of emulation.]

  3. (3)

    Thus: The auxiliaries (Dfw & Dghg & Dsol) and AOGCMcore are marginally consistent relative to K.

1.6 Argument F (pro E.1)

  1. (1)

    We nearly know nothing about the interactions between different components of the earth system, except that they are complex.

  2. (2)

    If we nearly know nothing about the interactions between different components of the earth system, except that they are complex, then there exists a (arbitrarily) complex coupled model ESM⋆ which (i) describes events and processes that depend on those interactions, (ii) is consistent with K (i.e., does not rely on contrary-to-fact assumptions), and (iii) possesses (arbitrarily many) poorly constrained parameters.

  3. (3)

    If a complex model, consistent with K, has (arbitrarily many) poorly constrained parameters, then it can be fitted to yield any output values whatsoever; specifically: for any events and processes X that depend on interactions between different components of the earth system, there are auxiliaries (e.g. parameters) A⋆ consistent with the complex model ESM⋆ and K, such that ESM⋆&A⋆ yield the events X as output.

  4. (4)

    Thus: There exist a (arbitrarily) complex model ESM⋆ (plus auxiliaries), consistent with K, which agrees with ESM (plus auxiliaries) in terms of events and processes that depend on interactions between different components of the earth system. [In other words, ESM⋆ can be fitted consistently to ESM plus its auxiliaries.]

  5. (5)

    If a model M with auxiliaries A yields the same T-output as a model M ⋆ with auxiliaries A ⋆, then M&A T-emulates the complex model M ⋆ with auxiliaries A ⋆.

  6. (6)

    A statement p depends on the interactions between different components of the earth system iff p is jointly implied by AOGCMcore and its auxiliaries (Dfw & Dghg & Dsol), but not by any of the two alone.

  7. (7)

    Thus: The coupled earth system model ESM plus its auxiliaries emulates a more complex coupled model plus auxiliaries with respect to all statements that are jointly implied by AOGCMcore and (Dfw & Dghg & Dsol), but not by any of the two alone; where the more complex coupled model plus its auxiliaries is jointly consistent with the background knowledge K.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Betz, G. Are climate models credible worlds? Prospects and limitations of possibilistic climate prediction. Euro Jnl Phil Sci 5, 191–215 (2015). https://doi.org/10.1007/s13194-015-0108-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s13194-015-0108-y

Keywords

Search

Navigation