Probabilistic Integrated Assessment of "Dangerous" Climate Change

Science

Volumn 304, Issue 5670, 2004, Pages 571-575

  Mastrandrea, Michael D ^a   Schneider, Stephen H ^a  

^a STANFORD UNIVERSITY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

DECISION MAKING; ENVIRONMENTAL IMPACT; PROBABILITY; PUBLIC POLICY; SOCIAL ASPECTS;

ANTHROPOGENIC INTERFERENCE;

CLIMATE CHANGE;

ANTHROPOGENIC EFFECT; CLIMATE CHANGE; POLICY MAKING; PROBABILITY;

ARTICLE; CLIMATE CHANGE; DECISION MAKING; DENSITY; GREENHOUSE GAS; POLICY; PRIORITY JOURNAL; TEMPERATURE;

EID: 2142644460     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.1094147     Document Type: Article

References (33)

1. United Nations Framework Convention on Climate Change (UNFCCC, 1992), available at www.unfccc.int.
2. A. Patwardhan, S. H. Schneider, S. M. Semenov, "Assessing the science to address UNFCCC Article 2" (IPCC Concept Paper, available at www.ipcc.ch/activity/cct3.pdf).
3. J. B. Smith et al., in Climate Change 2001: Impacts, Adaptation, and Vulnerability, J. J. McCarthy, O. F. Canziani, N. A. Leary, D. J. Dokken, K. S. White, Eds. (Cambridge Univ. Press, Cambridge, UK, 2001), chap. 19.
4. S. H. Schneider, Environ. Model. Assess. 2, 229 (1997).
5. M. G. Morgan. M. Henrion, Eds., Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis (Cambridge Univ. Press, New York, NY, 1990).
6. R. N. Pittock, R. N. Jones, C. D. Mitchell, Nature 413, 249 (2001).
7. S. H. Schneider, B. L. Turner, H. Morehouse Garriga, J. Risk Res. 1, 165 (1998).
8. A. S. Manne, R. G. Richels, Energy J. 16, 1 (1995).
9. J. K. Hammitt, R. J. Lempert, M. E. Schlesinger, Nature 357, 315 (1992).
10. M. O. Webster, Energy J. 23, 97 (2002).
11. R. H. Moss, S. H. Schneider, in Guidance Papers on the Cross Cutting Issues of the Third Assessment Report of the IPCC, R. Pachauri, T. Taniguchi, K. Tanaka, Eds. (World Meteorological Organization, Geneva, 2000), pp. 33-51.
12. S. H. Schneider, Nature 411, 17 (2001).
13. A. Grübler, N. Nakicenovic, Nature 412, 15 (2001).
14. S. Dessai, M. Hulme, "Does climate policy need probabilities?" (Working Paper 34, Tyndall Centre, UK, 2003).
15. We do not imply preference of the optimization framework over the methods used in other studies; in truth, we openly acknowledge the flaws many have identified in the general method (29) and in the DICE model we use (17), having ourselves contributed to this literature (16). There are other effective frameworks for analysis, such as the "robust strategies" approach (29). However, as the DICE family of models does allow a quantitative opportunity to explore methods and model uncertainties explicitly, has influenced policy decisions in the past, and continues to be used, we chose to present the influence of the most current scientific information on IAM results. The main virtue of this approach is transparency through a method to reframe the diagnosis of DAI with probabilistic IAMs.
16. M. D. Mastrandrea, S. H. Schneider, Clim. Pol. 1, 433 (2001).
17. W. D. Nordhaus, Science 258, 1315 (1992).
18. T. Roughgarden, S. H. Schneider, Energy Pol. 27, 415 (1999).
19. Materials and methods are available as supporting material on Science Online.
20. There are many ways that DAI could be interpreted from this figure, or from other sources. Each reason for concern could provide its own probability distribution for DAI; each reason for concern could be given its own weight based on some definition of its likelihood; or categories or metrics for assessing DAI other than those displayed could be used. The evaluation of DAI will likely be different depending on geographical location, socioeconomic standing, and ethical value system. For a first attempt at this definition, we used the simplest possibility-equal weight for all five categories. As seen in Figs. 3 and 4, we may already have committed to exceedance of a DAI[100%o] (1.476°C) threshold, as the probability of this threshold being crossed by 2100 is near unity. As the temperature increase exceeds the orange-to-red threshold of more categories, we believe a greater number of people will agree that dangerous change is occurring or will likely occur. Thus, we use all reasons for concem instead of just one, assuming equivalence of the danger from each category. We present a traceable account (11) of our assumptions in creating this definition (19), and we believe a similar account should be made each time such a definition is created by any analyst. We know of one other effort to create a CDF for dangerous climate change, presented by Wigley (30). Previously, Azar and Rodhe chose 2°C as their threshold for DAI (31), and O'Neill and Oppenheimer chose thresholds between 1° and 3°C for individual examples of DAI (32), without specifying ranges or percentiles in any of these cases.
21. R. H. Moss, S. H. Schneider, in Elements of Change 1996, S. J. Hassol, J. Katzenberger, Eds. (Aspen Global Change Institute, Aspen, CO, 1997), pp. 90-135.
22. U. Cubasch et al., in Climate Change 2001: The Scientific Basis, Summary for Policy Makers, J. T. Houghton et al., Eds. (Cambridge Univ. Press, Cambridge, UK, 2001).
23. N. G. Andronova, M. E. Schlesinger, J. Geophys. Res. 106, 22605 (2001).
24. C. E. Forest, P. H. Stone, A. P. Sokolov, M. R. Allen, M. D. Webster, Science 295, 113 (2001).
25. Transient temperature change in 2100 is not, in general, equilibrium change. The inertia of the climate system is such that climate change will continue long after greenhouse gas concentrations are stabilized or emissions eliminated. Some outcomes that avoid exceeding a DAI threshold until 2100 will exceed that threshold in the next century. Therefore, the time horizon of analysis will affect the potential for DAI. However, what is "dangerous" is itself a function of adaptive capacity, not a static quantity, dependent on social and economic development. So, the very threshold for any percentile X, DAI[X%o], can itself change with time and social conditions.
26. In the DICE model, carbon taxes serve as a proxy for general climate, policy controls. Thus, we do not present carbon tax data as a preferred method for mitigation or a required method to produce our results. Instead, these results should be seen as a method to provide insights into coupled model behavior, using the carbon tax in DICE as a measure of the magnitude of climate policy controls.
27. Results such as this are extremely sensitive to the discount rate. For example, the increase in the climate damage function indicated above that produces a ∼45% reduction in the probability of DAI[S0%o] with a 0% PRTP produces a reduction of only ∼10% and an order of magnitude lower "optimal" carbon tax when we used a 3% PRTP, the value employed by the original DICE model. We chose to use a 0% PRTP for Fig. 3 exactly for this reason-that using a high discount rate masks the variation in model results because of changes in parameters other than the discount rate, and observing variation in model results due to other parameters is central to our analysis.
28. 2 gases and for other anthropogenic radiative forcing agents such as aerosols would also affect the potential for DAI.
29. R. J. Lempert, M. E. Schlesinger, Clim. Change 45, 387 (2000).
30. T. Wigley, Clim. Change, in press.
31. C. Azar, H. Rodhe, Science 276, 1818 (1997).
32. B. C. O'Neill, M. Oppenheimer, Science 296, 1971 (2002).
33. We thank T. Wigley, K. Kuntz-Duriseti, J. Bushinsky, and M. Hayes for constructive comments on previous drafts. Supported by the Global Change Education Program of the Department of Energy and the interdisciplinary Graduate Program in Environment and Resources at Stanford University (M.D.M.) and by the Winslow Foundation (S.H.S.).