C.1: Introduction and Conceptual Framework
In this appendix, we present a brief overview of the methodologies and methodological issues for detection and attribution of climate change. Attributing an observed change or an event partly to a causal factor (such as anthropogenic climate forcing) normally requires that the change first be detectable.1 A detectable observed change is one which is determined to be highly unlikely to occur (less than about a 10% chance) due to internal variability alone, without necessarily being ascribed to a causal factor. An attributable change refers to a change in which the relative contribution of causal factors has been evaluated along with an assignment of statistical confidence (e.g., Bindoff et al. 2013;2 Hegerl et al. 20101 ).
As outlined in Bindoff et al.,2 the conceptual framework for most detection and attribution studies consists of four elements: 1) relevant observations; 2) the estimated time history of relevant climate forcings (such as greenhouse gas concentrations or volcanic activity); 3) a modeled estimate of the impact of the climate forcings on the climate variables of interest; and 4) an estimate of the internal (unforced) variability of the climate variables of interest—that is, the changes that can occur due to natural unforced variations of the ocean, atmosphere, land, cryosphere, and other elements of the climate system in the absence of external forcings. The four elements above can be used together with a detection and attribution framework to assess possible causes of observed changes.
C.2: Fingerprint-Based Methods
A key methodological approach for detection and attribution is the regression-based “fingerprint” method (e.g., Hasselmann 1997;3 Allen and Stott 2003;4 Hegerl et al. 2007;5 Hegerl and Zwiers 2011;6 Bindoff et al. 20132 ), where observed changes are regressed onto a model-generated response pattern to a particular forcing (or set of forcings), and regression scaling factors are obtained. When a scaling factor for a forcing pattern is determined to be significantly different from zero, a detectable change has been identified. If the uncertainty bars on the scaling factor encompass unity, the observed change is consistent with the modeled response, and the observed change can be attributed, at least in part, to the associated forcing agent, according to this methodology. Zwiers et al.7 showed how detection and attribution methods could be applied to the problem of changes in daily temperature extremes at the regional scale by using a generalized extreme value (GEV) approach. In their approach, a time-evolving pattern of GEV location parameters (i.e., “fingerprint”) from models is fit to the observed extremes as a means of detecting and attributing changes in the extremes to certain forcing sets (for example, anthropogenic forcings).
A recent development in detection/attribution methodology8 uses hypothesis testing and an additive decomposition approach rather than linear regression of patterns. The new approach makes use of the magnitudes of responses from the models rather than using the model patterns and deriving the scaling factors (magnitudes of responses) from regression. The new method, in a first application, gives very similar attributable anthropogenic warming estimates to the earlier methods as reported in Bindoff et al.2 and shown in Figure 3.2. Some further methodological developments for performing optimal fingerprint detection and attribution studies are proposed in Hannart,9 who, for example, focuses on the possible use of raw data in analyses without the use of dimensional reductions, such as projecting the data onto a limited number of basis functions, such as spherical harmonics, before analysis.
C.3: Non-Fingerprint Based Methods
A simpler detection/attribution/consistency calculation, which does not involve regression and pattern scaling, compares observed and simulated time series to assess whether observations are consistent with natural variability simulations or with simulations forced by both natural and anthropogenic forcing agents.10 ,11 Cases where observations are inconsistent with model simulations using natural forcing only (a detectable change), while also being consistent with models that incorporate both anthropogenic and natural forcings, are interpreted as having an attributable anthropogenic contribution, subject to caveats regarding uncertainties in observations, climate forcings, modeled responses, and simulated internal climate variability. This simpler method is useful for assessing trends over smaller regions such as sub-regions of the United States (see the example given in Figure 6.5 for regional surface temperature trends).
Delsole et al.12 introduced a method of identifying internal (unforced) variability in climate data by decomposing variables by time scale, using a measure of their predictability. They found that while such internal variability could contribute to surface temperature trends of 30-years’ duration or less, and could be responsible for the accelerated global warming during 1977–2008 compared to earlier decades, the strong (approximately 0.8°C, or 1.4°F) warming trend seen in observations over the past century was not explainable by such internal variability. Constructed circulation analogs13 ,14 is a method used to identify the part of observed surface temperature changes that is due to atmospheric circulation changes alone.
The time scale by which climate change signals will become detectable in various regions is a question of interest in detection and attribution studies, and methods of estimating this have been developed and applied (e.g., Mahlstein et al. 2011;15 Deser et al. 2012b16 ). These studies illustrate how natural variability can obscure forced climate signals for decades, particularly for smaller (less than continental) space scales.
Other examples of detection and attribution methods include the use of multiple linear regression with energy balance models (e.g., Canty et al. 201317 ) and Granger causality tests (e.g., Stern and Kaufmann 201418 ). These are typically attempting to relate forcing time series, such as the historical record of atmospheric CO2 since 1860, to a climate response measure, such as global mean temperature or ocean heat content, but without using a full coupled climate model to explicitly estimate the response of the climate system to forcing (or the spatial pattern of the response to forcing). Granger causality, for example, explores the lead–lag relationships between different variables to infer causal relationships between them and attempts to control for any influence of a third variable that may be linked to the other two variables in question.
C.4: Multistep Attribution and Attribution without Detection
A growing number of climate change and extreme event attribution studies use a multistep attribution approach,1 based on attribution of a change in climate conditions that are closely related to the variable or event of interest. In the multistep approach, an observed change in the variable of interest is attributed to a change in climate or other environmental conditions, and then the changes in the climate or environmental conditions are separately attributed to an external forcing, such as anthropogenic emissions of greenhouse gases. As an example, some attribution statements for phenomena such as droughts or hurricane activity—where there are not necessarily detectable trends in occurrence of the phenomenon itself—are based on models and on detected changes in related variables such as surface temperature, as well as an understanding of the relevant physical processes linking surface temperatures to hurricanes or drought. For example, some studies of the recent California drought (e.g., Mao et al. 2015;19 Williams et al. 201520 ) attribute a fraction of the event to anthropogenic warming or to long-term warming based on modeling or statistical analysis, although without claiming that there was a detectable change in the drought frequency or magnitude.
The multistep approach and model simulations are both methods that, in principle, can allow for attribution of a climate change or a change in the likelihood of occurrence of an event to a causal factor without necessarily detecting a significant change in the occurrence rate of the phenomenon or event itself (though in some cases, there may also be a detectable change in the variable of interest). For example, Murakami et al.21 used model simulations to conclude that the very active hurricane season observed near Hawai‘i in 2014 was at least partially attributable to anthropogenic influence; they also show that there is no clear long-term detectable trend in historical hurricane occurrence near Hawai‘i in available observations. If an attribution statement is made where there is not a detectable change in the phenomenon itself (for example, hurricane frequency or drought frequency) then this statement is an example of attribution without detection. Such an attribution without detection can be distinguished from a conventional single-step attribution (for example, global mean surface temperature) where in the latter case there is a detectable change in the variable of interest (or the scaling factor for a forcing pattern is significantly different from zero in observations) and attribution of the changes in that variable to specific external forcing agents. Regardless of whether a single-step or multistep attribution approach is used, or whether there is a detectable change in the variable of interest, attribution statements with relatively higher levels of confidence are underpinned by a thorough understanding of the physical processes involved.
There are reasons why attribution without detection statements can be appropriate, despite the lower confidence typically associated with such statements as compared to attribution statements that are supported by detection of a change in the phenomenon itself. For example, an event of interest may be so rare that a trend analysis for similar events is not practical. Including attribution without detection events in the analysis of climate change impacts reduces the chances of a false negative, that is, incorrectly concluding that climate change had no influence on a given extreme events22 in a case where it did have an influence. However, avoiding this type of error through attribution without detection comes at the risk of increasing the rate of false positives, where one incorrectly concludes that anthropogenic climate change had a certain type of influence on an extreme event when in fact it did not have such an influence (see Box 3.1).
C.5: Extreme Event Attribution Methodologies
Since the release of the Intergovernmental Panel on Climate Change’s Fifth Assessment Report (IPCC AR5) and the Third National Climate Assessment (NCA3),23 there have been further advances in the science of detection and attribution of climate change. An emerging area in the science of detection and attribution is the attribution of extreme weather and climate events.24 ,25 ,26 According to Hulme,27 there are four general types of attribution methods that are applied in practice: physical reasoning, statistical analysis of time series, fraction of attributable risk (FAR) estimation, and the philosophical argument that there are no longer any purely natural weather events. As discussed in a recent National Academy of Sciences report,24 possible anthropogenic influence on an extreme event can be assessed using a risk-based approach, which examines whether the odds of occurrence of a type of extreme event have changed, or through an ingredients-based or conditional attribution approach.
In the risk-based approach,24 ,27 ,28 one typically uses a model to estimate the probability (p) of occurrence of a weather or climate event within two climate states: one state with anthropogenic influence (where the probability is p1) and the other state without anthropogenic influence (where the probability is p0). Then the ratio (p1/p0) describes how much more or less likely the event is in the modeled climate with anthropogenic influence compared to a modeled hypothetical climate without anthropogenic influences. Another common metric used with this approach is the fraction of attributable risk (FAR), defined as FAR = 1 – (p0/p1). Further refinements on such an approach using causal theory are discussed in Hannart et al. 29
In the conditional or ingredients-based approach,24 ,30 ,31 ,32 an investigator may look for changes in occurrence of atmospheric circulation and weather patterns relevant to the extreme event, or at the impact of certain environmental changes (for example, greater atmospheric moisture) on the character of an extreme event. Conditional or ingredients-based attribution can be applied to extreme events or to climate changes in general. An example of the ingredients-based approach and more discussion of this type of attribution method is given in Box C.2.
Hannart et al.29 have discussed how causal theory can also be applied to attribution studies in order to distinguish between necessary and sufficient causation. Hannart et al.33 further propose methodologies to use data assimilation systems, which are now used operationally to update short-term numerical weather prediction models, for detection and attribution. They envision how such systems could be used in the future to implement near-real time systematic causal attribution of weather and climate-related events.
- Allen, M. R., and P. A. Stott, 2003: Estimating signal amplitudes in optimal fingerprinting, Part I: Theory. Climate Dynamics, 21, 477–491, doi:10.1007/s00382-003-0313-9. ↩
- Anderegg, W. R. L., E. S. Callaway, M. T. Boykoff, G. Yohe, and T. y L. Root, 2014: Awareness of both type 1 and 2 errors in climate science and assessment. Bulletin of the American Meteorological Society, 95, 1445–1451, doi:10.1175/BAMS-D-13-00115.1. ↩
- Bindoff, N. L., P. A. Stott, K. M. AchutaRao, M. R. Allen, N. Gillett, D. Gutzler, K. Hansingo, G. Hegerl, Y. Hu, S. Jain, I. I. Mokhov, J. Overland, J. Perlwitz, R. Sebbari, and X. Zhang, 2013: Detection and attribution of climate change: From global to regional. T.F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P.M. Midgley, Eds., Cambridge University Press, 867–952. URL ↩
- Canty, T., N. R. Mascioli, M. D. Smarte, and R. J. Salawitch, 2013: An empirical model of global climate – Part 1: A critical evaluation of volcanic cooling. Atmospheric Chemistry and Physics, 13, 3997–4031, doi:10.5194/acp-13-3997-2013. ↩
- Christensen, J. H., K. Krishna Kumar, E. Aldrian, S.-I. An, I. F. A. Cavalcanti, M. de Castro, W. Dong, P. Goswami, A. Hall, J. K. Kanyanga, A. Kitoh, J. Kossin, N.-C. Lau, J. Renwick, D. B. Stephenson, S.-P. Xie, and T. Zhou, 2013: Climate phenomena and their relevance for future regional climate change. T.F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P.M. Midgley, Eds., Cambridge University Press, 1217–1308. URL ↩
- DelSole, T., M. K. Tippett, and J. Shukla, 2011: A significant component of unforced multidecadal variability in the recent acceleration of global warming. Journal of Climate, 24, 909–926, doi:10.1175/2010jcli3659.1. ↩
- Deser, C., L. Terray, and A. S. Phillips, 2016: Forced and internal components of winter air temperature trends over North America during the past 50 years: Mechanisms and implications. Journal of Climate, 29, 2237–2258, doi:10.1175/JCLI-D-15-0304.1. ↩
- Deser, C., R. Knutti, S. Solomon, and A. S. Phillips, 2012: Communication of the role of natural variability in future North American climate. Nature Climate Change, 2, 775–779, doi:10.1038/nclimate1562. ↩
- Easterling, D. R., K. E. Kunkel, M. F. Wehner, and L. Sun, 2016: Detection and attribution of climate extremes in the observed record. Weather and Climate Extremes, 11, 17–27, doi:10.1016/j.wace.2016.01.001. ↩
- Emanuel, K. A., 2013: Downscaling CMIP5 climate models shows increased tropical cyclone activity over the 21st century. Proceedings of the National Academy of Sciences, 110, 12219–12224, doi:10.1073/pnas.1301293110. ↩
- Hannart, A., 2016: Integrated optimal fingerprinting: Method description and illustration. Journal of Climate, 29, 1977–1998, doi:10.1175/jcli-d-14-00124.1. ↩
- Hannart, A., A. Carrassi, M. Bocquet, M. Ghil, P. Naveau, M. Pulido, J. Ruiz, and P. Tandeo, 2016: DADA: Data assimilation for the detection and attribution of weather and climate-related events. Climatic Change, 136, 155–174, doi:10.1007/s10584-016-1595-3. ↩
- Hannart, A., J. Pearl, F. E. L. Otto, P. Naveau, and M. Ghil, 2016: Causal counterfactual theory for the attribution of weather and climate-related events. Bulletin of the American Meteorological Society, 97, 99–110, doi:10.1175/bams-d-14-00034.1. ↩
- Hasselmann, K., 1997: Multi-pattern fingerprint method for detection and attribution of climate change. Climate Dynamics, 13, 601–611, doi:10.1007/s003820050185. ↩
- Hegerl, G. C., F. W. Zwiers, P. Braconnot, N. P. Gillett, Y. Luo, J. A. M. Orsini, N. Nicholls, J. E. Penner, and P. A. Stott, 2007: Understanding and attributing climate change. S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor, and H.L. Miller, Eds., Cambridge University Press, 663–745. URL ↩
- Hegerl, G. C., O. Hoegh-Guldberg, G. Casassa, M. P. Hoerling, R. S. Kovats, C. Parmesan, D. W. Pierce, and P. A. Stott, 2010: Good practice guidance paper on detection and attribution related to anthropogenic climate change. T.F. Stocker, C.B. Field, D. Qin, V. Barros, G.-K. Plattner, M. Tignor, P.M. Midgley, and K.L. Ebi, Eds., IPCC Working Group I Technical Support Unit, 1–8. URL ↩
- Hegerl, G., and F. Zwiers, 2011: Use of models in detection and attribution of climate change. Wiley Interdisciplinary Reviews: Climate Change, 2, 570–591, doi:10.1002/wcc.121. ↩
- Horton, R. M., J. S. Mankin, C. Lesk, E. Coffel, and C. Raymond, 2016: A review of recent advances in research on extreme heat events. Current Climate Change Reports, 2, 242–259, doi:10.1007/s40641-016-0042-x. ↩
- Hulme, M., 2014: Attributing weather extremes to “climate change.” Progress in Physical Geography, 38, 499–511, doi:10.1177/0309133314538644. ↩
- Knutson, T. R., F. Zeng, and A. T. Wittenberg, 2013: Multimodel assessment of regional surface temperature trends: CMIP3 and CMIP5 twentieth-century simulations. Journal of Climate, 26, 8709–8743, doi:10.1175/JCLI-D-12-00567.1. ↩
- Lackmann, G. M., 2015: Hurricane Sandy before 1900 and after 2100. Bulletin of the American Meteorological Society, 96 (12), 547–560, doi:10.1175/BAMS-D-14-00123.1. ↩
- Magnusson, L., J.-R. Bidlot, S. T. K. Lang, A. Thorpe, N. Wedi, and M. Yamaguchi, 2014: Evaluation of medium-range forecasts for Hurricane Sandy. Monthly Weather Review, 142, 1962–1981, doi:10.1175/mwr-d-13-00228.1. ↩
- Mahlstein, I., R. Knutti, S. Solomon, and R. W. Portmann, 2011: Early onset of significant local warming in low latitude countries. Environmental Research Letters, 6, 034009, doi:10.1088/1748-9326/6/3/034009. ↩
- Mao, Y., B. Nijssen, and D. P. Lettenmaier, 2015: Is climate change implicated in the 2013–2014 California drought? A hydrologic perspective. Geophysical Research Letters, 42, 2805–2813, doi:10.1002/2015GL063456. ↩
- Melillo, J. M., T. (T. C. . Richmond, and G. W. Yohe, eds., 2014: Climate Change Impacts in the United States: The Third National Climate Assessment. U.S. Global Change Research Program, 841 pp. ↩
- Murakami, H., G. A. Vecchi, T. L. Delworth, K. Paffendorf, L. Jia, R. Gudgel, and F. Zeng, 2015: Investigating the influence of anthropogenic forcing and natural variability on the 2014 Hawaiian hurricane season [in “Explaining Extreme Events of 2014 from a Climate Perspective”]. Bulletin of the American Meteorological Society, 96 (12), S115–S119, doi:10.1175/BAMS-D-15-00119.1. ↩
- NAS, 2016: Attribution of Extreme Weather Events in the Context of Climate Change. The National Academies Press, 186 pp. ↩
- Reed, A. J., M. E. Mann, K. A. Emanuel, N. Lin, B. P. Horton, A. C. Kemp, and J. P. Donnelly, 2015: Increased threat of tropical cyclones and coastal flooding to New York City during the anthropogenic era. Proceedings of the National Academy of Sciences, 112, 12610–12615, doi:10.1073/pnas.1513127112. ↩
- Ribes, A., F. W. Zwiers, J.-M. Azaïs, and P. Naveau, 2017: A new statistical approach to climate change detection and attribution. Climate Dynamics, 48, 367–386, doi:10.1007/s00382-016-3079-6. ↩
- Shepherd, T. G., 2016: A common framework for approaches to extreme event attribution. Current Climate Change Reports, 2, 28–38, doi:10.1007/s40641-016-0033-y. ↩
- Stern, D. I., and R. K. Kaufmann, 2014: Anthropogenic and natural causes of climate change. Climatic Change, 122, 257–269, doi:10.1007/s10584-013-1007-x. ↩
- Stott, P. A., D. A. Stone, and M. R. Allen, 2004: Human contribution to the European heatwave of 2003. Nature, 432, 610–614, doi:10.1038/nature03089. ↩
- Stott, P., 2016: How climate change affects extreme weather events. Science, 352, 1517–1518, doi:10.1126/science.aaf7271. ↩
- Trenberth, K. E., J. T. Fasullo, and T. G. Shepherd, 2015: Attribution of climate extreme events. Nature Climate Change, 5, 725–730, doi:10.1038/nclimate2657. ↩
- Williams, A. P., R. Seager, J. T. Abatzoglou, B. I. Cook, J. E. Smerdon, and E. R. Cook, 2015: Contribution of anthropogenic warming to California drought during 2012–2014. Geophysical Research Letters, 42, 6819–6828, doi:10.1002/2015GL064924. ↩
- Zwiers, F. W., X. B. Zhang, and Y. Feng, 2011: Anthropogenic influence on long return period daily temperature extremes at regional scales. Journal of Climate, 24, 881–892, doi:10.1175/2010jcli3908.1. ↩
- van Oldenborgh, G. J., F. J. Doblas Reyes, S. S. Drijfhout, and E. Hawkins, 2013: Reliability of regional climate model trends. Environmental Research Letters, 8, 014055, doi:10.1088/1748-9326/8/1/014055. ↩
- van den Dool, H., J. Huang, and Y. Fan, 2003: Performance and analysis of the constructed analogue method applied to U.S. soil moisture over 1981–2001. Journal of Geophysical Research, 108, 8617, doi:10.1029/2002JD003114. ↩