U.S. House of Representatives: Committee on Energy and Commerce, RepublicansPrepared Witness Testimony: Michaels, Patrick J.

Subcommittee on Oversight and Investigations
July 25, 2002
09:30 AM
2322 Rayburn House Office Building

Dr. Patrick J. Michaels
Professor and Virginia State Climatologist Department of Environmental Sciences
University of Virginia
PO Box 400123
Charlottesville, VA, 22904

. The two climate models that serve as the basis for the U.S. National Climate Change Assessment were not representative of the consensus of climate models, but rather represent extreme predictions for temperature and precipitation changes over the United States.

. The two climate models that serve as the basis for this Assessment performed worse than a table of random numbers when asked to simulate U.S. temperature changes as the atmosphere has changed. Under the ethics of science, they should have then been abandoned or modified, rather than used as input to a document with substantial policy implications.

. The current U.S. National Climate Change Assessment should be redacted from the public record.

. Another Assessment should be undertaken, this time with a much more diverse synthesis team selected by a more diverse political process.

. Professional interpreters of climate information, who will be called upon to explain or defend any future Assessment, such as the State Climatologists, should provide strong input to any new report.

. Any new Assessment must be based only upon hypotheses that can be verified by observed data.

TESTIMONY OF PATRICK J. MICHAELS TO THE SUBCOMMITTEE ON OVERSIGHT AND INVESTIGATIONS OF THE COMMITTEE ON ENERGY AND COMMERCE, U.S. HOUSE OF REPRESENTATIVES, JULY 25, 2002

This testimony makes no official representation for the University of Virginia or the Commonwealth of Virginia, and is tendered under the traditional protections of academic freedom.

Effects have causes. Confronting our society today is a potentially serious effect, climate change, caused by human influence on our global atmosphere.

The quantitative tools of mathematics and science are what we use to inform rational analysis of cause and effect. Science, in particular, obeys a rigid standard: that the tools we use must be realistic and must conform to observed reality. If they do not, we modify or abandon them in search of other analytical methods. Whenever the federal government releases a comprehensive science report, the public naturally assumes that it has passed these tests. The documents we will discuss today failed those tests. This failure was ignored in the public review process.

There is no doubt that the issue of climate change rightly provokes private citizens and our government to ask what its potential effects might be on the United States. That was the purpose of the recent report Climate Change Impacts on the United States: The Potential Consequences of Climate Variability and Change. This document is often called the "U.S. National Assessment" (USNA) of climate change. This report forms much of the basis for Chapter 6 of the U.S. Climate Action Report-2002, a chapter on "Impacts and Adaptation" to climate change.

The USNA began with a communication from President Clinton's National Science and Technology Council (NSTC), which was established in 1993. According to the USNA, "This cabinet-level council is the principal means for the President to coordinate science, space and technology policies across the Federal Government." "Membership consists of the Vice President [Al Gore], the Assistant to the President for Science and Technology, Cabinet Secretaries and Agency heads." The Council is clearly a political body ("coordinating.policies") rather than a scientific one.

This NSTC was, in turn, composed of several committees, including the Committee on Environment and Natural Resources, chaired in 1998 by two political appointees, D. James Baker and Rosina Bierbaum. Baker developed a further subcommittee of his committee, the Subcommittee on Global Change Research, to "provide for the development.of a comprehensive and integrated.program which will assist the Nation and the world to understand, assess, predict [emphasis added], and respond to human-induced and natural processes of global change." Ultimately, this resulted in the selection of the National Assessment Synthesis Team (NAST).

NAST was confronted with a daunting task, detailed in the schematic below. The chain of cause and effect begins with industrial activity and the combustion of compounds that alter the atmosphere's radiative balance. These are then distributed through the atmosphere. These affect the climate of the United States. Then, those changes in climate are input to a subsidiary series of computer models for forest growth, agriculture, etc.

An understanding of the effects of climate change on the United States requires that there be no substantially weak links in this catena. As an example of a relatively strong link, I would estimate that we understand about 70 percent of the changes in atmospheric carbon dioxide that result from human activity. The reason this number is not 100 percent largely stems from the fact that the current concentration of carbon dioxide seems low, given the amount emitted and assumptions about how it distributes through the atmosphere and the biosphere, and how it eventually returns to the soil and the ocean bottom.

There are two main ways to assess the most important of these linkages, which is between "Atmospheric Changes" and "Climate Changes in the United States." One involves the use of computer simulations, known as General Circulation Models (GCMs) to estimate how climate changes as a result of atmospheric alterations. An alternative method for assessment is described on page 10 of this Testimony.

There are literally dozens of GCMs currently available, and the USNA considered a subgroup of these models. Eventually, they selected two, the Canadian Climate Centre model, acronymed CGCM1, and another from the United Kingdom Meteorological Office, known as HadCM2[1]. The prime outputs of these models that are important for the assessment of climate change are temperature and precipitation.

In using GCMs to project future climate at regional scales, the USNA clearly placed itself squarely against the consensus of world climate science. In 2001, the United Nations' Intergovernmental Panel on Climate Change (IPCC) compendium on climate change, the Third Assessment Report, states:

"Despite recent improvements and developments.a coherent picture of regional climate change.cannot yet be drawn. More co-ordinated efforts are thus necessary to improve the integrated hierarchy of models.and apply these methods to climate change research in a comprehensive strategy."

In other words, even three years after the Assessment team began its report relying on GCMs, the consensus of world climate science was that they were inappropriate for regional estimates, such as those required for the United States.

As shown in the IPCC's Third Assessment Report of climate change, the average behavior of GCMs is to produce a linear (constant) rate of warming over the projectable future. In other words, once warming begins from human influence, it takes place at a constant, rather than an exponentially increasing rate.

However, the CGCM1 is an outlier among the consensus of models, producing a warming that increases as a substantial exponent. This behavior can be seen in Figure 1a, taken directly from the USNA, in which the CGCM1 clearly projects more warming than the others illustrated in the USNA.

The USNA also illustrates a similarly disturbing behavior for precipitation. Figure 1b, again taken directly from the USNA, shows that the other model employed, HadCM2, predicts larger precipitation changes than the others that are illustrated in the USNA.

A close inspection of Figure 1a reveals that CGCM1 predicts that the temperatures in the United States at the end of the 20^th century should be about 2.7°F warmer than they were at the beginning, but the observed warming during this time, according to the most recent analysis from the National Climatic Data Center, is 0.9°F. CGCM1 is making a 300 percent error in its estimation of U.S. temperature changes in the last 100 years.

My colleague Thomas Karl, Director of the National Climatic Data Center and co-chair of the USNA synthesis, explained that the reason CGCM1 was chosen was because it was one of only two models (the other was HadCM2) that produced daily temperature output, and that this was required to drive some of the subsidiary models, such as those for forest impacts.

Michael MacCracken, Executive Director of the National Assessment Coordination Office, told me otherwise. He said that the two models were selected because they gave extreme results, and that this was a useful exercise. How the explanations of the co-chair and the Executive Director could be so different is still troubling to me.

The Failure of the Models

GCMs are nothing more than hypotheses about the behavior of the atmosphere. The basic rule of science is that hypotheses do not graduate into facts unless they can be tested and validated against real data.

As part of my review of the USNA in August 2000, I performed such a test. The results were very disappointing. Both CGCM1 and HadCM2 were incapable of simulating the evolution of ten-year averaged temperature changes (1991-2000, 1990-1999, 1989-1998, etc.back to 1900-1909) over the United States better than a table of random numbers. In fact, the spurious 300 percent warming error in CGCM1 actually made it worse than random numbers, a dubious scientific achievement, to say the least.

The Synthesis Team was required to respond to such criticism. Publicly, they deflected this comment by stating that both U.S. temperatures and model temperatures rose in the 20^th century, so use of the models was appropriate!

This was a wildly unscientific response in the face of a clear, quantitative analysis. The real reason for the models' failure can be found in the USNA itself (Figure 11 in Chapter 1 of the USNA Foundation document). It is reproduced here as our Figure 2. The discrepancies occur because:

1) U.S. temperatures rose rapidly, approximately 1.2°F, from about 1910 to 1930. The GCMs, which base their predictions largely on changes in atmospheric carbon dioxide, miss this warming, as by far the largest amounts of emissions were after 1930.

2) U.S. temperatures fell, about 1.0°F, from 1930 to 1975. This is the period in which the GCMs begin to ramp up their U.S. warming, and

3) U.S. temperatures rose again about 1.0°F from 1975 to 2000, recovering their decline between 1930 and 1975.

It is eminently clear that much of the warming in the U.S. record took place before most of the greenhouse gas changes, and that nearly one-half of the "greenhouse era," the 20^th century, was accompanied by falling temperatures over the U.S. These models were simply too immature to reproduce this behavior because of their crude inputs.

Despite their remarkably unprofessional public dismissal of a rigorous test of the USNA's core models, the Synthesis Team indeed was gravely concerned about the criticism. So much so, in fact, that they replicated my test, not just at 10 year-intervals, but at scales ranging from 1 to 25 years.

At the larger time scales, they found the models applicable to global temperatures. But over the U.S., not surprisingly, they found exactly what I had. The models were worse than random numbers.

It is difficult for me to invoke any explanation other than political pressure that would be so compelling as to allow the USNA to continue largely unaltered in this environment. And so the USNA was rushed to publication, ten days before Election Day, 2000.

Given the failure of the models when directly applied to U.S. temperatures, there were other methods available to the USNA team. One would involve scaling various global GCMs to observed temperature changes, and then scaling the prospective global warming to U.S. temperatures. The first part of this exercise has been performed independently by many scientists in recent years, and published in many books and scientific journals. It yields a global warming in the next 100 years of around 2.9°F, which is at the lowest limit of the range projected by the IPCC in its Third Assessment Report.

If applied to the United States this would similarly project a much more modest warming than appears in the USNA. Perhaps that is the reason such an obviously logical methodology was not employed after the failure of the models was discovered by a reviewer and then independently replicated by the USNA itself.

Effect of the USNA

This discussion would be largely academic if the USNA were an inconsequential document. But, as noted above, it served largely as the basis for Chapter 6 of the U.S. Climate Action Report-2002. Further, it served as the basis for legislative findings for S. 556, a comprehensive proposal with extensive global warming related provisions, and it was clearly part of the findings for legislation restricting carbon dioxide emissions recently passed by the California Legislature. Hardly a week goes by without some press reference to regional alterations cited by the USNA. Would the USNA have such credibility if it were generally known that the driver models had failed?

Solving the Structural Problems with the USNA

The USNA synthesis team contains only two individuals who can logically claim, in my opinion, to be climatologists. Of the entire 14-member panel, there is not one person who has expressed considerable public skepticism about processes that were creating increasingly lurid scenarios for climate change with little basis in fact. As noted above, the administrative structure that selected the synthesis team was clearly directed by political appointees, which no doubt contributed to this imbalance.

"Finally, we come to the subject of bias in selection of USNA participants. There are plenty of knowledgeable climatologists, including or excluding this reviewer, who have scientific records that equal or exceed those of many of USNA's participants and managers. They would have picked up the model problem [that extreme versions were selected, and that they could not simulate U.S. temperatures] at an early point and would not have tried to sweep it under the rug. Where is Bob Balling? Where is Dick Lindzen? Where are [Roger] Pielke Sr., [a participant in this hearing], [Gerd] Weber or [Roy] Spencer?"

My review was tendered shortly after attending the annual meeting of the American Association of State Climatologists (AASC) in Logan, Utah, in August 2000. The AASC is the only professional organization in the U.S. devoted exclusively to climatology. Membership consists largely of senior scientists who are tasked by their states, usually through the state's major universities, to bring climate information and services to the public. Until 1972, the State Climatologists were employees of the U.S. Department of Commerce.

"Yesterday.I returned from the annual meeting of the American Association of State Climatologists (I am a past president of AASC). There were roughly 100 scientists present. I can honestly state that not one positive comment was tendered to me about the USNA, out of literally dozens made. If the report is published in anything like its current form, I predict it will provoke a public examination of how and why the federal science establishment [could have produced such a document]."

Besides being research scientists, the State Climatologists are interpretive professionals who deal with the climate-related problems of their states on a day-to-day basis. It's hard to imagine a better-suited team of professionals to provide a significant leadership role in any new Assessment.

2. Another Assessment should be undertaken, this time with a much more diverse synthesis team selected by a more diverse political process.

3. Professional interpreters of climate information, who will be called upon to explain or defend any future Assessment, such as the State Climatologists, should provide strong input to any new report.

4. Any new Assessment must be based only upon hypotheses that can be verified by observed data.

The 2000 document, Climate Change Impacts on the United States: The Potential Consequences of Climate Variability and Change, which served as the basis for an important chapter in the new Climate Action Report-2002, was based on two computer models which were extreme versions of the suite of available models. The two selected models themselves performed no better than a table of random numbers when applied to U.S. temperatures during the time when humans began to subtly change the composition of the earth's atmosphere. As a result, both reports are grounded in extremism and scientific failure. They must be removed from the public record.

This scientific debacle resulted largely from a blatant intrusion of a multifaceted political process into the selection process for those involved in producing the U.S. National Assessment. The clear lesson is that increased professional diversity, especially intermingling state-based scientists with the federal climatologists, would have likely prevented this tragedy from ever occurring.

IPCC (2001). Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Houghton JT, Ding Y, Griggs DJ, Moguer M, van der Linden PJ, Dai X, Maskell K, Johnson CA, (eds). Cambridge University Press, Cambridge, UK, 881pp.

National Assessment Synthesis Team (2001). Climate Change Impacts on the United States: The Potential Consequences of Climate Variability and Change. Report for the U.S. Global Change Research Program, Cambridge University Press, Cambridge, UK, 620pp.

U. S. Department of State. U. S. Climate Action Report-2002. Washington, D.C., May 2002.

Figure 1a. (top) Future temperature changes for the United States projected by models considered for inclusion in the U.S. National Assessment . Figure 1b. (bottom) Future precipitation changes for the United States projected by models considered for inclusion in the U.S. National Assessment. (Source: U.S. National Assessment, Foundation Document)

Figure 2. Time histories of the changes in annual temperature for the United States during the 20th century based on observations and on simulations from the Canadian and Hadley models, calculated as 10-year running means from 1900 to 2000. (Source: U.S. National Assessment, Foundation Document).

[1] In 1998, the National Research Council report Capacity of U.S. Climate Modeling to Support Climate Change Assessment Activities strongly remonstrated against the use of foreign models to assess U.S. climate. According to the NRC, ".it is inappropriate for the United States to rely heavily upon foreign centers to provide high-end modeling capabilities. There are a number of reasons for this including.[the fact that] decisions that might substantially affect the U.S. economy might be based upon considerations of simulations.produced by countries with different priorities than those of the United States."

The U.S. National Climate Change Assessment: Do the Climate Models Project a Useful Picture of Regional Climate?

The Failure of the Models

Effect of the USNA

Solving the Structural Problems with the USNA