Statisticians Comment on Status of Climate Change Science
The following are letters from ASA members in response to the March Amstat News article “Statisticians Comment on Status of Climate Change Science,” by Richard L. Smith, L. Mark Berliner, and Peter Guttorp. The authors answered additional questions online on March 31. Click here to view a transcript of the discussion.
I read the recent article by Smith, Berliner, and Guttorp explaining why the ASA Climate Change Policy Advisory Committee recommended that Sally Morton sign the letter of consensus of climate change science.
I appreciate the hard work the committee obviously put into this and I recognize their sincerity and deep concern about the environment. However, I don’t think the article addressed the correct questions.
No one doubts that the climate is changing; it is always changing. No one doubts that there has been a rise in average temperature. No one doubts that carbon dioxide plays a role in average temperature. And no one doubts that humans release carbon dioxide into the atmosphere. Below, I list what I see as the real questions. I would be interested in knowing the committee’s view on these questions.
Question 1: The main concern is not that the temperature is increasing. The temperature of the Earth is never constant. The concern is that the rate of increase is possibly much larger than during past increases. In other words, the concern is about the derivative of the current warming period compared to the derivative of past warmings. How certain can we be that the derivative of the current warming is exceptional?
To answer this question, we need to compare the current climate to past climate. And to do this, we need to use climate reconstructions. You acknowledge that some reconstructions have been the object of great skepticism. As you noted, Ed Wegman affirmed, in his testimony before Congress, that the criticisms due to McIntyre and McKitrick are legitimate. Yet, you dismiss the criticisms by merely saying “… the research community has responded successfully to these challenges.” I don’t feel that you have really addressed this issue.
My understanding is that temperature reconstructions are based on combining inverse regressions of various surrogates (such as tree rings). The opportunities for unknown biases and uncertainties entering such a calculation seem endless. It defies statistical intuition that we could accurately reconstruct the average temperature of the Earth hundreds or thousands of years ago from inverse regressions based on nonrandomly sampled surrogates. Imagine the numerous omitted variables, not to mention lack of random sampling. And remember, we need to estimate the derivative, which is difficult even from the best data.
Question 2: Carbon dioxide is a small part of the greenhouse gas in the atmosphere, water vapor being the largest component. Humans account for only part of the carbon dioxide that is in the atmosphere. What fraction of warming is due to human-released carbon dioxide compared to the other greenhouse gases? How accurately can this number be determined? How large are the biases and variance?
Question 3: The climate is always changing. There is nothing optimal about our current climate. A warmer climate will have both positive and negative consequences. (For example, far more people die from cold than from heat.) How strong is the statistical evidence that the negative consequences outweigh the positive consequences?
Let me emphasize that I appreciate the committee’s efforts. And I am not suggesting we should be unconcerned about the environment. But, if anyone should hold climate scientists to the highest statistical standards, it should be statisticians. I’d like some assurance that the ASA is doing so.
Professor, Department of Statistics
Carnegie Mellon University
I am embarrassed by the “defense” of the anthropogenic global warming (AGW) theory published in the March 2010 issue of Amstat News. As an agnostic on the issue, I was excited about the article. I expected an objective, dispassionate analysis—and possibly conclusion—by professional statisticians trained in judging theories based on evidence supported by data. Instead, the authors either tried to cast doubt on any contradictory evidence, or, worse, determined the state of knowledge of alternative theories was not sufficient to overcome the presumption of validity of AGW.
Listen to the tortured denial of solar activity as a cause of global warming: “There is no credible physical theory that would deny the GHG [green house gas] influence.” I believe that there is also no credible physical theory that would deny the solar influence.
Or consider this gem: Solar activity theory could make future temperature projections more uncertain, but “there is already plenty of uncertainty about those projections.” So, accepting solar influence makes no sense because the projections are so uncertain already that a little more uncertainty wouldn’t be noticed?
I have a simple question: Does anyone have a model, based on GHG readings of a specific location or locations, that successfully predicts any future, observable, measurable temperatures—whether it is a monthly average of satellite readings, a set of specific ground temperature stations for a specific month, or even the thermometer on my deck—for five or 10 years in the future? Given the years of data and analysis thereof we supposedly have, that seems like a pretty low bar. Is anyone so certain of AGW that they would publish such projections?
I don’t believe the ASA should have signed onto the letter to lawmakers based on the article I read.
Terry G. Meyer
President, TECH Consulting
The article “Statisticians Comment on Status of Climate Change Science” in the March 2010 issue of Amstat News left me wondering whether the dues I pay to the American Statistical Association are paying for any part of the work by the ASA’s Climate Change Policy Advisory Committee. Please know that I do not want my dues funding this committee, and I do not want the ASA making position statements on the status of “climate change science” on its behalf. Additionally, if the ASA makes statements on behalf of its members, which includes me, about the notion of anthropogenic global warming, I will terminate my membership in the association.
I remain a skeptic of the “overwhelming” science of anthropogenic global warming because, as an industrial statistician, model bias from failing to include important factors leads to precarious conclusions, at best. It seems to me that because there is evidence that CO2 levels were twice as high as they are now millions of years ago, long before humankind walked the Earth, there are surely factors missing from current models, factors that would have caused prior warming periods and periods of elevated CO2.
Furthermore, plotting temperature data collected from small patches of the Earth’s surface over the past 140 years would have no discernible (hockey stick) trend (i.e., be part of the noise) if plotted over 400,000 years or 4,000,000 years or 4,000,0000,000 years. Don’t we teach our statistics students to plot data scaled in context to avoid making false claims?
In the March 2010 issue of Amstat News, Richard Smith, Mark Berliner, and Peter Guttorp discuss various aspects of global warming and conclude that (1) the climate is warming, (2) humans are likely responsible, and (3) mitigation measures are needed. They discuss (1) and (2) at length, but mention (3) only in the last sentence, as if it is self-evident. The whole climate debate in general, in fact, largely ignores this part of the problem. A “second front” is needed in the discussion, one that takes as a given that warming is happening and tackles questions like the following:
- What are the projected effects of global warming, and what is the strength of the evidence that they are happening and caused by global warming?
- How much harm (or benefit) will those consequences bring?
- What are the feasibility and cost of potential remedies?
As a preface to our response to letters regarding our article, the following points are germane to most of the letters. Please recall that we wrote, “We comment on some of the most common arguments that climate change is not happening, or humans are not responsible.” Our article was not intended to be a summary of the basis of any recommendation regarding ASA endorsement of any letter. Any such recommendation would involve much more than brief comments on selected points of controversy. In particular, we did not review the case for anthropogenic climate change. Such a review would involve combinations of statistical and scientific arguments leading to a weight of evidence.
Responding specifically, now, to each of the four writers, starting with Wasserman’s three questions:
Question 1: We do not agree that the main purpose of paleoclimatology is to prove the derivative during the last 25 years is greater than in the past 1,000 years (or the past 100,000) in particular, since the climate response to CO2 increase is far from instantaneous, nor do we agree that it really matters whether the 1990s were the warmest decade of the last millennium, though this statement is often made. Rather, we believe the main functions of paleoclimatology in the context of projecting future climate change are to learn about feedback mechanisms and gain some indication of the scale of natural variability that has occurred in the past and may be expected to occur in the future.
Recent work on paleoclimatic reconstruction from a statistical point of view includes the forthcoming paper by Li, Nychka, and Ammann in JASA and a forthcoming paper by Tingley and Huybers in Journal of Climate. Finally, an in-depth discussion of aspects of paleoclimatology can be found in the U.S. Congress–commissioned 2006 National Academy of Sciences report Surface Temperature Reconstructions for the Last 2,000 Years (available free as a download from National Academy Press). This report includes various studies confirming the “hockey stick” shape.
Question 2: This point relates to the general area of attribution. Climate models are run with both natural and anthropogenic forcing. These results are a major contributor to the confidence of the Intergovernmental Panel on Climate Change (IPCC) in their Fourth Assessment Report (AR4) that anthropogenic influences explain the observed temperatures.
We each have concerns regarding the quality and uncertainties associated with the inputs and forcings used. Further, interactions and climate feedbacks regarding forcings are subject to uncertainty and substantial continuing research. Indeed, characterizing the response of the hydrological cycle (water vapor, cloud properties, etc.) has been a holy grail in climate science. We note that Solomon et al. (2010) indicates that the post-1998 period of stable temperatures can be explained in terms of changing patterns of water vapor. The paper also indicates that different greenhouse gases can have different effects, so that changes in CO2 do not necessarily indicate the same changes in other greenhouse gases.
Question 3: There are uncertainties associated with climate-scale projections, with the resulting sea level rise and regional and local weather impacts and with the resulting effects on human outcomes (e.g., health) and human endeavors (e.g., agriculture). We do not believe all these uncertainties have been quantified, nor taken into account. Fortunately, researchers from many disciplines, including ours, are engaged in research in these directions. We also note that statisticians support decisionmaking in the presence of uncertainty.
Finally, we have not seen any clear evidence that “far more people die from cold than from heat.” The death rates in Europe and North America are higher during the winter months, but that does not imply that temperatures in winter are the cause of death. Generally speaking, the evidence points to long stretches of extreme temperatures as being most dangerous. For example, stretches of unusually cold weather in Spain (where unusually cold may mean just freezing) is much more harmful than stretches of unusually cold weather in Finland (where it means temperatures below -40˚ F).
In any case, this query is symptomatic of the difficulty and potential misunderstanding of “climate change” versus “global warming.” Regarding the latter, scientists are not concerned that the future climate would be identical to the current one, save a uniform, simple shift of 1˚ F or 2˚ F over this century. Rather, climate change represented in sea level rise and changes in the distribution of weather trends and extremes (perhaps both hot and cold) are the keys.
Regarding Meyer’s letter, there is credible theory as to the solar influence on climate (which is substantial from a historical perspective and explains much of the observed past climate change). However, there is no theory that both explains past climate change and current observations based on solar forcing alone.
The second-to-last paragraph betrays the writer’s misunderstanding of climate and, perhaps, statistics. “Climate” is the distribution of weather, and hence predicting the temperature at a particular location at a particular time would have similar uncertainty to predicting a single draw from a probability distribution. The writer’s challenge is a bit like asking for a prediction of the date of death due to lung cancer of a specified cigarette smoker.
Responding to Trocine, the evidence for anthropogenic contribution to global warming is not limited to CO2. Current climate models include a variety of greenhouse gases, as well as other forcings of the climate. There is interaction and feedback in the relationship between temperature and CO2. While it is correct that the last 140 years are a blip compared to previous changes, what one really needs to look at is a kind of residual plot, namely the difference between what you get without anthropogenic forcing and what you get with all observed forcings. That is the right scale of comparison for the kind of scientific judgment we are talking about here.
Finally, we fully agree that the issues Bruce raised are critical and merit substantial development and participation by statisticians. As mentioned above, we had a limited scope in this article.