This Subcommittee asks important questions: What areimplications of climate change for national security, economicdevelopment and public health?
The answers to the important questions about the implications ofclimate change are driven by a series of computer models andmathematical simulations. First, one estimates changes in climate.Then changes these are input into a series of subsidiary models toestimate their impact. Finally, one compares putative costs of theclimate change compared to the costs of mitigation by reduction orstabilization of the concentration of atmospheric carbondioxide.
We often hear that "the science is settled" on global warming.This is hardly the case. While almost all scientists agree thatglobal surface temperature is warmer than it was a century ago,there is considerable debate about the ultimate magnitude ofwarming, as evinced by the broad range of future mean surfacetemperature given by the United Nations' Intergovernmental Panel onClimate Change.
The primary drivers of the impact models are therefore themodels for climate change itself. I must report that our models arein the process of failing. When I say that, I mean the ensemble of21 models used in the midrange projection for climate change by theIPCC. I am an active participant on this Panel, providing extensivereviews and comment on several iterations of their scientificsummaries, as well as invited text for their Second Assessment.
If it is demonstrable that these models have failed, then thereis no real scientific basis for any estimates of the costs ofinaction. I will now perform that demonstration.
Remember this: a climate model is really nothing more than ascientific hypothesis. If a hypothesis is consistent withobservations, then it is standard scientific practice to say thatsuch a hypothesis can continue to be entertained. In this case,that hypothesis can then serve as a basis for other subsidiarymodels or, in reality, subsidiary hypotheses.
If the hypothesis is not consistent with observations, it mustbe rejected. That does not mean that human-induced climate changemay or may not be real, but it does mean that (in this case) themagnitude of prospective change has-with high probability-beenoverestimated. That means that all subsidiary hypotheses oneconomic costs, strategic implications, or effects on health aresimilarly overestimated.
Figure 1 shows the various model projections for the IPCC "A1B"emissions scenario for the period 2000-2020. This is the "midrange"estimate. Actual emissions rates that are above these values willproduce higher projected rates of warming, and vice-versa for lowerones. The actual accumulation of carbon dioxide in the atmosphere,in parts per million, has been very close to the A1B estimates, soit serves as a very useful point of analysis.
Figure 1. Climate model projections (colored lines) andclimate model ensemble mean (black circles) of global averagesurface temperature anomalies, 2000-2020, under the IPCC A1Bemissions scenario.
Figure 2 shows the observed surface temperatures from theUniversity of East Anglia record since the second warming of the20th century commenced in 1977. This history, designated HadCru3,and its predecessor versions, are the most cited histories by theIPCC. For designation, I refer to this as the IPCC surface datahereafter in this testimony.
Figure 2. Annual global average temperature anomalies,1977-2008, from the HadCru3 temperature history.
Several things should be apparent.
First, the ensemble behavior of the A1B models is largely linearin this time frame. In other words, the tendency of both theindividual models (colored lines) and the average of the models isa constant rate of warming. Indeed, the observed warming in theHadCru3 record, back to 1977 (when the second warming of the 20thcentury commences) is also constant. This is true despite a lack ofoverall trend since 1998, but it is noteworthy that 1998 was anobvious high point in the observed record because of a strong ElNino and an active sun, in addition to the warming pressure fromincreasing carbon dioxide.
We now examine the distribution of warming trends within the 21A1B models for various time periods. We use the set of modelsavailable at http://climexp.knmi.nl/, a standardreference. The models begin in 2001 and end in 2020. Note that themodeled warming rates in the first half of this period, which weare nearly through (by 2010), are the same as they are in thesecond half. In other words, the modeled rate of warming isconstant.
We first analyze various modeled trends beginning with afive-year window and then on up to 15 years, using the 2001-2020reference period. We ran successive monthly iterations of eachmodel. Consequently the sample size is very, very large. Theresults are shown in Figure 3.
Figure 3. Climate model 95% confidence range of projectedsurface temperature trends of varying lengths (gray area) and thecurrent observed values for these trends (through December 2008)(black line).
We then calculated the percentile ranges of temperature changefor the model ensembles at the .025 level on both the "warm" and"cold" sides of the model distributions. This is analogous to the95% confidence bounds for the model ensemble. Generally speaking,hypotheses are either rejected or continued to be entertained atthe .95 level, so our test of the models is consistent with normalscientific practice.
Also in Figure 3 are the observed temperature trends for periodsfrom five to fifteen years from the IPCC history, ending inDecember, 2008. It is very clear that temperatures are running atthe lower limit for the .95 confidence level. In other words, theensemble of the AIB models is failing.
While much ado has been made about the lack of warming from 1998through now, the analysis is clearly quite stable across othertrend periods. However, the longer that the current regimepersists, the worse the models fail. Figure 4 assumes that 2009mean surface temperatures are the same as 2008, which is a veryreasonable assumption at this time. We are currently in the coldphase of El Nino, called La Nina, which decreases the likelihoodthat this will be a very warm year.
Figure 4. Climate model 95% confidence range of projectedsurface temperature trends of varying lengths (gray area) and theexpected values for these trends assuming the temperature in thecoming year is similar to the temperature in 2008 (blackline).
In Figure 5, we run the analysis for the last 20 years ofobserved IPCC temperatures (1989-2008), rather than the last 15.There is a clear warming trend in this period, but, again, it is solow as to fall again along the .95 level. The ensemble modelfailure is not a product of the selection of recent years; ratherit is a systematic failure of the models as a whole to accommodatetemperatures in recent decades.
Figure 5. Climate model 95% confidence range of projectedsurface temperature trends of varying lengths (gray area) and thecurrent observed values for these trends (through December 2008)(thick black line) and when the observations are adjusted toaccount for the impact of Mt. Pinatubo (dotted blackline).
The failure becomes even more obvious when the effect of the1991 eruption of Mt. Pinatubo is removed. This results in a moreappropriate comparison of the model ensemble with observationsbecause the models themselves contain no volcanoes. Being near thebeginning of the 20-year analysis period, Pinatubo introduced atemporary cooling early in the study, which results in more"apparent" warming than was observed. As a consequence of thisadjustment, the observed temperature trends fall away from the .95level for trends of 15 to 20 years in length.
"The Science is Settled"?
One implicit assumption in calculating the "costs of inaction"is that we know with reasonable confidence indeed what climaticchanges will ensue as atmospheric carbon dioxide concentrationsincrease. With regard to climate, we often assume a commonWashington mantra: with regard to global warming, "the science issettled".
This demonstration shows how far from the truth thisoft-repeated sentence actually is. One can say this. "The scienceis settled" inasmuch as surface temperatures have increased fromthe late 1970s. That this is shown in the surface record has notbeen in dispute, so claiming some finality for such a truism ishardly noteworthy. What is true, however, is that the rates ofwarming, on multiple time scales, have now invalidated the midrangesuite of IPCC climate models. No, the science is not settled. Infact, judging from these results, it's time for climate scientiststo get back to work and generate models which will be able toestimate the recent past and present within their normal confidenceranges.
Until that is done, all we know is this: calculations of thecosts of inaction, based upon models that are clearlyoverestimating warming to the point that they can no longer berelied upon, are likely to be similarly overestimated. In thateventuality, the costs of drastic action can easily outweigh thecosts of a more measured response, consistent with what is beingobserved, rather than what is being erroneously modeled.
Brohan P, et al., (2006). Uncertainty estimates in regional andglobal observed temperature changes: a new dataset from 1850.Journal of Geophysical Research, 111, D12106.
IPCC (2007), Climate Change 2007: The Physical ScienceBasis. Contribution of Working Group I to the FourthAssessment Report of the Intergovernmental Panel on Climate Change.In: Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB,Tignor M, Miller HL (eds) Cambridge University Press, Cambridge,United Kingdom and New York, NY, 996 pp.
IPCC (1996), Climate Change 1995: The Science of ClimateChange. Contribution of Working Group I to the SecondAssessment Report of the Intergovernmental Panel on Climate Change.In: Houghton JT, Meira Filho LG, Callander BA, Harris N, KattenbergA, Maskell K (eds). Cambridge University Press, Cambridge, UK.,572pp.
ClimateExplorer, The Royal Netherlands Meteorological Institute, asdescribed in: Oldenborgh, G.J. van, et al, 2009. Western Europe iswarming much faster than expected. Climate of the Past, 5, 1-12.