Dan Cooley is a Professor or Statistics at Colorado State University and is a past member and chair of the ASA’s Advisory Committee on Climate Change Policy. Dan’s research is primarily focuses on developing statistical methods for the study of extreme values and is largely motivated by problems in atmospheric science.
Dr. Michael F. Wehner is a senior staff scientist in the Computational Research Division at the Lawrence Berkeley National Laboratory. Dr. Wehner’s current research concerns the behavior of extreme weather events in a changing climate, especially heat waves, intense precipitation, drought and tropical
+ Full Transcript
Rosemary Pennington: Stories about historic rain and flooding in the US Midwest, catastrophic wildfires in Australia, and massively destructive hurricanes in the Caribbean can make it seems as though the weather is out of control. It can also spawn arguments about how much of this is part of the normal ebb and flow of the earth’s life cycle, and how much of it is actually the result of a changing climate. Today’s Stats and Stories explores what it means to say we’re observing extreme environmental events. I’m Rosemary Pennington. Stats and Stories is a production of Miami University’s Departments of Statistics and Media, Journalism and Film, as well as the American Statistical Association. Joining me in the studio are regular panelists John Bailer, Chair of Miami University’s Statistics Department and Richard Campbell, former Chair of Media, Journalism and Film. We have two guests joining us today. The first is Colorado state university’s statistics professor Dan Cooley, Dan, thank you so much for being here today.
Dan Cooley: Thanks for having me.
Pennington: Our second guest is Michael Wehner, Senior Staff Scientist in the computational research division at the Lawrence Berkley National Laboratory. Thank you so much for joining us as well today Michael.
Michael Wehner: Thank you, it’s a pleasure.
Pennington: Late last year the two of you co-authored an editorial arguing that climate science needs more professional statisticians. Why did the two of you feel compelled to write that?
Cooley: The article came out of some discussions we had at the JSM in Vancouver in 2018 where there were several sessions on statistics and climate, and one of the recurring themes in the sessions and in discussions I’ve had with Michael was the challenge we’ve had in trying to get statistical expertise involved in climate science, and integrating the latest and greatest methods to answer climate questions. So, Michael and I sat down at that JSM and started to hash out this article, which took us some time to put together, but I think that was the start of it anyway.
Wehner: I’d like to say that I am not a statistician. I’m not trained formally that way. But climate is the statistics of weather. And weather is the day to day noise we live in, and climate is kind of the overall description of that weather, whether it be the average or the tales of distributions or however you want to phrase it. And I’ve found in the last five-seven years maybe, that my interactions with professional statisticians, trained statisticians, has been extremely fruitful. And part of it is keeping me out of trouble by not misusing statistics. You know the old saying about statisticians and liars. And one doesn’t want to be the liar. But beyond that we’ve been able to apply much more contemporary techniques, statistical techniques to the climate science, in particular in my work with extreme temperature and extreme precipitation that the climate science community didn’t know about, and can’t really know about because they’re not really privy to that literature. And so, the collaboration between statisticians and probably any physical scientist can be extremely rewarding. It is, however, not easy because we speak a different language. English, but…
John Bailer: I liked how the article you were writing tied to the analogy of the bio stat collaboration and methodological work in biomedicine and public health. I thought that was a nice model, can you talk a little more about that?
Cooley: I think that that discipline- the interaction between statistics and biomedical research is just taken for granted now. And certainly, biomedical research would not be where it is today without the statistical contributions to show what are meaningful advances and what aren’t. And that assumed or default collaboration in that discipline is really, I think not necessarily what you find in climate science, and probably in a lot of other sciences as well, but as Michael said if we describe climate as the distribution of weather, then it’s inherently statistical. And so, statisticians are needed at the beginning when designing the study or the science plan, and throughout the study, and in drawing statistical conclusions, and then making important inference or meaningful statements about the science at the end of the study. So, I think that way of thinking is something we’d like to se more of in climate science.
Richard Campbell: So, I’m kind of interested in one of the articles I read about extreme value analysis. Can you talk a little bit about that and extreme weather in general? Because I think this is the area that you’re really interested in.
Cooley: I think the area for both of us, and I’ll describe it from a statistical side and try to do it briefly, and then Michael can chime in from the climate science side. But there’s- extreme value theory is nicely based on very formal probabilistic results that tell us how to describe the tale of the distribution, and that’s what we’re interested in when we try to describe extreme events. Classically, people have tried to describe things like the so called 100-year flood or that quantile which the annual maximum will exceed with probability 1/100th in a given year, and people have had less than 100 years to make those estimates. And so, extrapolation has been something that’s been needed to do and to solve practical problems, and extrapolation is something that statisticians are naturally wary of because we are told early on in our statistics classes not to extrapolate. But there’s nice theory that tells us the right way to do that. That the tale of the distribution should be converging to a specific type of distribution. So, the practice of extreme value theory is to start with those probabilistic foundations, and then fit models described by that, and we do it in an unusual way where we throw away the bulk of the data. The idea being that- at least puristically- that the bulk of the data, the usual sorts of day to day weather phenomena if we’re studying weather, tell us very little about the tale and in fact if we fit a model to the entire distribution, there’s so much data there in the middle that we can miss the tale of the distribution dramatically. So, in practice the classical approach is to keep a very small subset of data in the tale and fit one of these models from the theory that allows us to say that intelligent things about 100 years flood, even if we’ve got only 40 years of data. And of course, not only when I say intelligent things it’s not only that point estimate but also the uncertainty associated with that. And there are large uncertainties when one is extrapolating, and I think that’s part of the story that classical extreme value theory can do very well.
Wehner: So, I stumbled across extreme value theory about 20 years ago. I started asking some statisticians about it and there weren’t many who knew too much that I knew, but I stumbled across some books and I tried to understand that as best I could and all of the sudden- out of the blue as far as I was concerned- there were some papers who were at that time in Canada, and the way they framed the question of changes in extreme precipitation and changes in extreme temperature through extreme value theory methods struck a chord with me, and I realized right away that this was the way we needed to go because the rather elegant statistical theory would allow us to do those kinds of extrapolations that we have to do because of the limited record of the observations. We can run models for a lot longer and make up for some of that, but the observational record is by its nature, limited. And the reason I thought this was so important is because the way that people are adversely affected by the weather is usually when there is some extreme storm where a big heat wave, where people get sick or die, or it could be a big rainstorm where they’re flooded out, or a drought or what have you. And as climate change continues to increase, you know things are getting warmer all the time it seems, part of the manifestation of climate change is through changes in extreme weather, and given that this is such an impactful thing- extreme value theory gave us a rigorous way to quantify. You know bioscience is a much larger industry than climate science, which is largely funded academically at national laboratories and universities rather than large corporate interests. But a lot of the problems are very similar and as a matter of fact one of the more recent developments in climate science is called extreme event attribution, and in extreme event attribution what we try to do is look at individual weather events, like say Hurricane Harvey, and ascertain what effect the climate change that’s already happened has impacted that individual storm. And this was a rather remarkable development that came about after the 2003 European heat wave that has 70,000 excess deaths attributed to it. Where people were asking how much worse is this because of climate change? And this has been extended now to a whole class of extreme weather, of heat waves being the first, flooding, and now hurricanes being the state of the art. And these techniques we really borrowed from epidemiology, and so in some ways it’s not new in terms of the statistics and the math, but the application is certainly new, and I believe, very important.
Bailer: Oh, very cool.
Pennington: You’re listening to Stats and Stories and today we’re talking statistics, weather and climate with Dan Cooley of Colorado State University and Michael Wehner of the Lawrence-Berkley National Laboratory.
Bailer: So, clearly you both have written and spoken about the impact of human’s contribution to climate change; that’s been a big part of it. When was the first evidence that there might be human impact on climate change?
Wehner: Well that’s actually very old. This is not rocket science. It’s steam engine science. [Laughter] The recognition that greenhouse gases, carbon dioxide and other radiated active in the atmosphere affected the climate goes back into the 19th century. But it was Svante Arrhenius in 1904 who first phrased the problem at hand, which was if you increase the amount of carbon dioxide how much will the climate change? And what he asked is- the question was if you doubled the amount of carbon dioxide how much more would it get? And he came up with the answer five degrees centigrade for the global temperature. And in the most recent IPCC, the Intergovernmental Panel for Climate Change report, we now call that quantity the equilibrium climate sensitivity, and the IPCC report gave a very likely range of two degrees to five degrees centigrade. So, Arrhenius is at the upper end of that report, although this most recent report is almost certainly going to increase that upper range, and so this is not new stuff. Campbell: My question has to do with- a lot of the work that you guys do gets translated to the general public through journalists, do you feel a particular responsibility to communicate with policy makers, the public? What do we need to know that you’re doing that’s not getting to the public? And not getting to it in ways that we need to know about it?
Cooley: Michael is probably more on the front line of this that I, so it will be interesting to hear his take on it. As an academic statistician I feel a little bit removed from the public interaction with- our currency in academia is to produce papers and often those are in statistics journals and so we’re trying to push methodology, but we’re also trying to push methodology that will innovate methodology that will be useful for people who are answering climate questions. And certainly, I have written papers which try to answer those climate questions too, but still it seems maybe a little bit removed. One of the most direct interactions I’ve had with the public or with policy makers was to- I attended Climate Science Day back in 2016 as a member of the ASA’s committee on climate change policy, and that is an opportunity where scientists from all branches of science who are studying climate change go and visit people on Capitol Hill, and really tell our story. And at that time the story, essentially, was we’re doing very good science, we’re here as a resource to your policy makers, and please reach out to us to get the best information you can from science. And that was a really eye opening experience for me; I’m certain that I learned more by my visit to Capitol Hill than they learned from me, but it was- I think that message of the science we’re doing is very good science, and the scientific discipline in this nation works, and that the answers that we’re coming up with are extremely well founded and justified, and we’re describing the system as best we can, and we’d love to serve as a resource to the policy makers when it comes time to ask questions of the science to influence policy.
Wehner: You know, at the end of the day I’m just a geek with access to very good computing and a very good national laboratory, [Laughter] and I am not supposed to be on TV.
Campbell: But I’ve seen you on TV.
Bailer: We all have. You did a great job.
Wehner: And let me tell you there is nothing quite so terrifying, [Laughter] and as a result of that experience, which I never expected would happen, I advise young scientists all the time take the media training. The physical union offers- more the American meteorological society and probably the American Statistical Association, but in any event these media training courses I think are – I wish I’d had them. I think they could have helped, because our message is important and it is policy relevant, relevant to individual’s lives, and it is particularly relevant to our children and our grandchildren. And talking to the media is difficult. I learned early on that I tend to be too jargony and explaining what would appear on the surface to be complex topics in plain English is actually quite a challenge.
Bailer: The aspects of uncertainty and communicating that seems to be a particular challenge. You report- you know, you’ve written more than one half of global mean temperatures since 1951 can be attributed to human impact, but actually you argue beyond that that it supports that all change could be due to human impact in this. And you know there’s arguments that are based on models and confidence bounds, and lower bounds exceeding certain levels- there’s a lot of nuance to this story, and is this nuance part of the place where people will base their objection to some of the assertion that are made to human contribution to climate change?
Wehner: That particular statement, I was part of writing that for the US National Climate Assessment and climate scientists have generally tended- and maybe could be faulted for being rather conservative, and so we would talk about the lower bound on complex intervals. At least half the warming is attributed to statements like that, when in fact-and this is what we tried to do in the fourth national assessment – our best estimate is that all of the climate change that’s been observed is due to humans. And in fact, our very likely bound is that 90-120% is due to humans. And you might even ask well, why it would be more than 100%? And that’s because there could be more natural phenomena that would be trying to cool us. And so, it is a subtle point. We do have what we call the calibrated language, which is inherently- well it’s sort of two parts. There’s the likelihood language and the confidence language. And the likelihood language is when we say something, and Dan might want to chime in on that, but we say a statement is very likely, but that some amount of anything is very likely, we’re saying that there’s a 90% chance of that statement being true. And then we have the confidence language where we would say – which is much more of a subjective statement that we have high confidence that this is very likely or something of that nature. And so, we do have a language that is supposed to translate from the technical art of it to something that’s plain English. I do think this is something that could be improved, however.
Campbell: So when I’ve taught journalism classes in the past we talk about this notion of false balance that journalists often say you’re supposed to go out and get two sides to every story, and climate change is one of those ones where there may not be two sides. We have dramatic climate change, and we ask them often to get a number of sources, find the best evidence, you know, you’ve got to demonstrate in your story that there are documents and they have evidence that shows this; I’ve often wondered and sometimes I’ve heard people say well if you do this right you’re going to go out and interview 98 climate scientists who know that there’s been climate change and it’s attributable to humans, and you’ll find 2% that don’t believe that. I want to know who those guys are. Who are those 2% of the climate scientists that don’t believe it? And are there? Why are we having this debate? I guess I always just wonder why this got so politicized.
Wehner: That’s a really good question. And the answer is pretty simple. I mean, you go to some talk and you hear about the evidence for gravity waves and people sit there and they listen about this very dramatic and very bizarre phenomena that is quite now well accepted, but you talk about climate change and you talk about the weather; everybody thinks they know about the weather. You don’t experience gravity waves, but you experience weather, and so there’s a preconception that you’re just going to have to have because you’re human- but beyond that there are- climate change doesn’t come without significant costs. Whether you do something or don’t do something, it’s going to cost a lot of money, it’s going to cost a lot of anxiety in people’s lives, in ways that other kinds of more esoteric branches of science don’t.
Cooley: Just getting back to your comment, Richard, about the journalistic balance, I think that if the lay person is watching a news program and they have two scientists arguing different sides of climate change, the lay person is going to assume that represents a representative example of climate scientists, and we know that that’s not true, right? So, I think the misperception can arise from that. Just reflecting back – we were talking about uncertainty before and Michael talked about the language that IPCC uses of when they attribute something as very likely it’s 95%, correct? Very likely corresponds to 95% Michael, is that right?
Wehner: I’ll have to look it up.
Bailer: That can be a homework assignment for listeners.
Cooley: Michael talked about the calibrated language, and right in the IPCC they will denote very likely indicates such and such percent, and likely indicates such and such percent. It’s a way of trying to speak about uncertainty in language that captures it for policy makers, and for people that the IPCC report is supposed to reach. And this is not an idea that’s new to anybody on the statistics side of things. Talking about uncertainty is very difficult. We are in our community we’re in the midst of talking about it again nearly 100 years later with all the discussion about significance and p-values and how to [inaudible]. I don’t think there’s anything wrong with a p-value, so long as the person who hears the p-value reported understands what it is. But I think the misuse of p-values has, at least partially, arises from people misunderstanding the idea of significance and so uncertainty is incredibly difficult to talk about. And people are – as Michael was saying, one of the challenges- the reason there’s opposition to climate change is doing something about climate change means making changes in the way we live our day to day lives, and the way that society is structured, and so before people are willing to do that or are eager to do that, they love to talk about things with certainty, and in science we’re trained not to do that. We’re trained to convey what we’re very certain about and what we’re not, and I think the public can take those qualifiers of uncertainty and run with them, and then say then we shouldn’t act. And I think it’s very clear that the cost of not acting, even when, well- let me put it this way, we act with uncertainty in almost every decision we make in our day to day lives, and the conclusions that are being made about our changing climate are pretty darn certain. So, we can act, even given the levels of uncertainty that do exist, and I think that they imply a direction and it’s something that policy makers and lay people can understand and act.
Bailer: So, I’m going to close with a real quick question, or at least I’m trying to close with a quick question, but Rosemary may hit me anyway. If you could- so I’m going to ask you two things. One is if you could change one thing to try to mitigate the adverse climate change what would it be? That’s the first part and what should a student do if they want to join you in working on this problem? Pennington: We said one question John.
Bailer: Well, you didn’t say it couldn’t be a compound question, come on.
Cooley: How about I take the second one and Michael takes the first?
Bailer: Perfect.
Cooley: As far as a student is- if the student is interested in statistics, I think this paper that we- this opinion piece that we wrote about the need for climostatisticians like [inaudible] statisticians, and I think there are lots of entry points for students to do meaningful statistics that can then be applied to answer very important questions in climate science, and hopefully there will be an increasing role for well-trained statisticians to really be part of scientific teams studying the big questions. So that’s sort of a long-winded answer to study hard and do good things.
Wehner: What was the first part of the question?
Bailer: The first part was if you could do one thing to impact this adverse trend, what would it be?
Wehner: Well, the problem is that there isn’t one thing. I think it’s important to recognize scientists like Dan and I do not have the responsibility to make decisions about what we should do as a society. That’s not the role of scientists. The role of scientists is to inform, and let the public make the decision that it wants based on the best information through its elected officials, and if there is only one thing, it’s to make sure that the information that the public gets is of the highest quality, is as honest as can be, and that may be where the biggest problem is, is that there’s a lot of misinformation out there from special interest groups that feel, for one reason or another, that the most widely accepted and most credible explanations of the observed climate change are somehow not trustworthy. That kind of thing is the most dangerous thing we face right now.
Pennington: Well, on that note, Dan and Michael thank you so much for joining us for this conversation today.
Cooley: Thanks for having us.
Pennington: Stats and Stories is a partnership between Miami University’s Departments of Statistics and Media, Journalism and Film, and the American Statistical Association. You can follow us on Twitter, Apple podcasts or other places you can find podcasts. If you’d like to share your thoughts on the program you can email us at statsandstories@miamioh.edu or check us out at statsandstories.net, and be sure to listen for future editions of Stats and Stories, where we discuss the statistics behind the stories and the statistics behind the stories.