Peter Guttorp (@pgseattle) is a Professor at the Norwegian Computing Center in Oslo, Norway, and Professor Emeritus at the University of Washington, Seattle. He is also a vice-president of the International Statistical Institute . His research focuses on stochastic models of scientific data and their statistical analysis. He has worked in seismology, hydrology, climatology, hematology and biology.
+ Full Transcript
Rosemary Pennington : Planning for uncertainty is no easy task, but Climate Change is forcing communities around the globe to plan for uncertain futures. While we know the planet is warming and that could affect everything from farming to forestry to flooding, what we don't know is just how big the impact will be, or how dire the circumstances. That hasn't stopped cities from trying to figure out how to prevent a climate-caused catastrophe. Today on Stats+Stories we're talking rising seas, and the difficulty of protecting coastal communities among other things. I'm Rosemary Pennington. Stats+Stories is a partnership of Miami University's Departments of Statistics and Media Journalism and Film, and the American Statistical Association. I'm joined in the studio by regular panelists Department of Statistics Chair John Bailer and Media, Journalism and Film Department Chair Richard Campbell. Our guest today is Peter Guttorp. Professor Emeritus of Statistics at the University of Washington, Professor at the Norwegian Computing Center, and Co-Director of The Research Network on Statistical Methods for Atmospheric and Ocean Sciences, or STATMOS. Guttorp is also one of the authors of an article in Significance Magazine, which explores how the Norwegian city of Bergen is working to create a plan to save its historic harbor from possible future flooding. Thank you so much for being here today Peter.
Peter Guttorp : Glad to be here.
Pennington : In your article you say that most climate change models don't predict a rise in sea-level, and that seems like such a pressing concern for so many communities including Bergen, could you explain what makes predicting a rise in sea-level so difficult?
Guttorp : Well, first, what I actually said was that most climate models don't compute that. Or that's what I meant anyway. So, climate models deal with the entire climate system, and including oceans and atmosphere and other things. And, the actual sea-level is not important for the physics of that. But of course, it's very important for people who are trying to keep buildings from flooding and things like that. So, you have to measure sea-level globally, you have to measure seal level locally, and those two are not the same. It's not an easy problem, but no climate problems are really easy.
John Bailer : I thought it was interesting, in your paper, Peter, when you talked about the idea of these are projections and not predictions. Could you talk a little bit more about the idea of projection versus prediction distinction?
Guttorp : Yeah. If you were trying to predict what's going to happen you'll have to have a model for how governments are going to deal with emissions, for how economies is going to deal with growth of electric cars and things like that. We don't have models at that level. And so, instead of trying to predict the future, we outline some scenarios that describe, for example, the degree of emissions of greenhouse gasses. And so, the current IPCC report uses four such scenarios. One that's business as usual, nothing much changes. A couple that are a little lower. And one that essentially says well, "what would happen if we stopped emitting greenhouse gasses around the year 2000. Now all these projections, so runs of the computer models using these scenarios of what's going to happen in the future, are run typically 3000 years back and then 100 years forward.
Bailer : Okay, to calibrate the models?
Guttorp : Yes, 3000 years back to calibrate the models and make the ocean and the atmosphere work together.
Bailer : There's an interesting use of the word "model," so there's models for prediction and models describing this physical system. Can you just say a little bit more about the types of models that are used?
Guttorp : Basically, the models are technically solutions to partial differential equations describing the behavior of the atmosphere, the behavior of the oceans, the interactions between the two, and then the further interactions with ice, and land use and things like that. So, these are mathematical models that can't be solved analytically, they have to be solved numerically, and therefore you have to approximate things, you solve on a large grid squares, and then you need to think about how does that relate? What actually happens at a point like the harbor of Bergen, where people are concerned, and the truth is that the global models don't tell you much about that. So, to go from global models to local models you have to somehow downscale it. You have to make it smaller. And the downscaling is… there's several different ways of doing it. What we're doing in our paper is called a Statistical Downscaling. We write down a statistical model, for the relationship between global sea-level and local sea-level and try to use that to predict what's going to happen in the harbor of Bergen.
Richard Campbell : Very good. Peter, as one of the two journalists in the room, I'm interested in the notion of uncertainty. This is something that when journalists report the work of statisticians or scientists, they don't usually talk about uncertainty. And this is certainly, when you try to explain this, for instance to policy makers, the people that are going to make decisions about what to do in this particular case in Bergen. How do you explain uncertainty to them? And to an audience that's not statisticians?
Guttorp : Well, there's a couple of different things one can talk about. I used to teach a course, a freshman seminar at the university called "Uncertainty is Knowledge". The idea there was really to explain, "well, how can you tell, for example, what the Global Mean Temperature is?" You can't measure it, we don't have any tools for measuring Global Mean Temperature. We can measure temperature in place, we can sometimes do satellite measurements that are well related to temperatures, but there's going to be some uncertainty because we don't measure it everywhere. And that uncertainty has to be taken into account. Our paper about the sea-level in Bergen, shows that if you don't take it into account, if you just ignore the uncertainty, there are several different pieces to the puzzle. And each piece has uncertainty associated with it. If you just ignore it and just take an average measurement for each of the pools, you're going to be way below what the cost of what damages in the harbor is going to be. And when you take uncertainty into account, you see that this could less easily be ten times more expensive than you would think, if you just gave one number.
Campbell : Very good.
Guttorp : So, it's important to give a number and some sense of how uncertain that number is. And the way that we do it is the same way that IBCC does it, we give what's called a 90% confidence interval.
Pennington : You're listening to Stats+Stories. Today we're talking about planning for an uncertain climate with co-director of STATMOS, Peter Guttorp. Now, Peter, when I was doing some background research for you, I saw that you at one point had studied Journalism before moving into studying Math and Statistics, and I think Musicology was also listed there. Did studying journalism impact the way that you present scientific material? And why did you switch gears?
Guttorp : Well, that's several questions.
Pennington : Sorry.
Guttorp : The first one is yes, it certainly has influenced how I… it has above all influenced the importance I put on communicating science to non-scientists. So, this kind of thing is something I find very important. The reason I started doing statistics in the U.S…. I got a scholarship to go to Berkeley, and the idea was to get a Master's Degree in Statistics and become a science journalist. Well, then they talked me into staying to get a Ph.D., and then I got a job and a family and the journalism went on the back burner. But it's always been there as one of the things I find very important in what I learned in journalism school is thinking about who you're writing for. And who you are as a writer. And being very explicit about that whole situation.
Richard Campbell : I also wanted to compliment you on, until I got to the actual statistics and the tables and the numbers where I'd become lost… getting to that is you're writing is really good you really drew me in, you know how to tell a story. So, that little bit of journalism experience paid off!
Guttorp : It certainly did. And thank you for that compliment. I'm sorry got lost, the idea was to not get lost.
Bailer : So, Journalism's loss is Statistic's gain here. You know, one thing in looking at this… you've talked about all of these- there's lots of moving pieces in the story that you're working on here. The idea of uncertainty that what's going to happen with sea-level changes, the uncertainty that's associated with the cost, the distribution of impact of cost. And you talk a little bit about the idea of extreme flooding and how it might affect infrastructure like bridges and buildings and storm sewers and other structure. You also mention that Bergen is a historic city, and the idea of… I was thinking how do you value impacts and changes to the loss of historic sites?
Guttorp : Good question. Obviously. What can you say about that? Well, what has happened in Bergen is, well, it used to be a Hanseatic city, so it was part of the Northern European trading network in the middle ages. Which was where a lot of the commercial activity in northern Europe happened. And so, they had an office where the Hanseatic Organization, and they had a bunch of buildings. Now, these were wooden buildings, and they burned down. In fact, they burned down at least twice. And so, the buildings that are currently there are built like the traditional buildings from the Middle Ages, but they are relatively new. And so, the damage that happens to them is not as serious as it would have been if they were actual medieval buildings. But it's still important, and they are a World Heritage Site and the Norwegian Government is required to keep things from happening. Well, they do flood quite frequently. So that's why the Bergen authorities are thinking about how they are going to keep this from happening so often, especially if sea-level goes up.
Campbell : You mentioned before you talked about one of your goals is communicating Science to non-Scientists, that's something that you're interested in. So, we live in a time where were… one our critiques in journalism is making sure that students don't get into this notion of 'false balance". You know, where you have those who believe in climate change and those that don't, and those are two equally divided segments. So, we have to work hard in journalism classes to make sure that students understand that all issues aren't divided into two equally equal sides. So, do you have any suggestions for how you confront this notion of, well, there's people that have belief systems that they don't seem to buy in to climate change, even though the science and the evidence is there for it… How do you wrestle with that? In communicating to non-scientists.
Guttorp : Well, it's first of all, difficult to deal with people who will not accept science as a process.
Campbell : Yes, you got that right.
Guttorp : It's very hard. What you can do as a journalist is try to push the person you're talking to to describe which pieces of the whole system we actually know. For example, we know that increased greenhouse gasses increase temperature. That was calculation by a Swedish chemist in the late 1800s. You know, I mean, this is old physics. What we don't know is how fast is the ice on Greenland going to melt. And you know, we have models for each of these things. The models for the ice melt in Iceland are not as precise as the models for the temperature increase when we increase greenhouse gasses. Neither of the two are completely precise. I mean we can't say that if we double CO2, what is the temperature going to be? We can give a range… but we can't give a number.
Pennington : You're listening to Stats+Stories. Out guest today is Peter Guttorp, who is the author of an article for Significance Magazine, about how Bergen, Norway is planning to combat rising seas. You mentioned a little earlier, Peter, how important you think it is to communicate science for a non-science audience, and Richard has been asking about advice for Journalists, but I wonder what advice you would have for scientists who are doing work like yours, or work that is also dealing with these big complex issues, about how they can communicate clearly what's at stake in the research that they're doing.
Guttorp : Well, it's something you have to practice. It's not something that you can do just out of nothing. I've argued for a long time that we ought to have courses in statistics for journalists, and we ought to have courses in communications for statisticians
Campbell : We agree.
Guttorp : There are lots of those for general scientists, and the difficulty there is that statisticians deal with uncertainty. And the explanation of uncertainty, as we've been talking about before, to non-statisticians it's not easy. It's a complicated concept and one has to somehow come up with better ways of talking about it. And that's why I think we need specialized courses. I've applied for grants to do these kinds of things and the National Science Foundation didn't find that terribly important, and so they didn't fund the grants. I find it terribly important.
Bailer : I was intrigued at the course that you taught: "Uncertainty is Knowledge." When I think about uncertainty and variability in systems, I think that more knowledge of a system will help reduce uncertainty but it will just better characterize variability. That's yet another distinction that sometimes seems lost.
Guttorp : Right. Some people divide uncertainty into epistemic uncertainty, uncertainty that has to do with our understanding, and aleatoric uncertainty which is due to the fact that we don't measure exactly, and things like that.
Bailer : Indeed. I want to revisit the paper, and in there, there were three options that the council, the politicians can take. And you've done some work to try to help them, so just for folks that are listening, can you do a quick recap of the options that are available and how the modeling effort you did helped and informed what might be done?
Guttorp : Well, basically there was a very expensive option, which was to build some kind of barrier. Bergen is at the end of a fjord, as most Norwegian cities are, and so this would put the barrier outside the fjord. It would probably affect the ecology of the fjord fairly substantially, and as I said, it would be very expensive. But it would keep the entire fjord system from getting substantial sea-level rise. The other alternatives had to do with building smaller barriers, and one could choose to protect different areas. We calculated for a system where they had two barriers inside the fjord, so closer to the harbor. And our calculations, which took into account the cost of flooding, took into account the likely sea-level rise and how much costs are going to increase with increased sea-level rise. All three pieces of those are uncertain. Our calculations showed that there was no point in doing the outside, expensive barrier. From a cost-benefit point of view, it was simply not worth it. It was too expensive. It was an order of magnitude it was more expensive than the cost would be, even in the worst scenario and at the highest boundary of the confidence band. But the other options we could say when they should do them. We could actually say "when would be the most pieceable way of building these things?", and it wasn't right away. It was doing it a little bit later, around 2040 or 2050 depending on which type of pessimism you want to deal with. The interesting thing about this is that the Norwegian authorities have told the planners that they are to use calculations based on the upper 90% competence band for the highest scenario. So, that would never happen in the U.S.
Bailer : They're being conservative and protective in their decision making.
Guttorp : They're being conservative and protective, and they don't want things to be bad.
Bailer : Have you consulted at all with them?
Guttorp : We talked to them, yeah.
Bailer : And have they been pretty receptive to some of the modeling and the process that you've described for deciding amongst the alternatives?
Guttorp : Yeah…
Guttorp : …who are in charge of doing that themselves, and our answers and their answers for the sea-level are pretty similar.
Campbell : Your recommendation has been the more modest… building the barriers inside the fjord, right?
Guttorp : Right.
Bailer : So, what would you recommend to someone, a student who is interested in working on these types of problems? If they were coming from the stat's side what might they do, or if they were coming from the journalism side, what might they do to equip themselves to work on these problems or report on these problems?
Guttorp : Easiest part is the first one, to say what statisticians need to do, they need to know something about time series, they need to know something about spatial statistics, they need to understand regression, and they need to know the difference between pointwise and simultaneous confidence intervals. Which I'm not here to explain that.
Bailer : Richard will summarize it.
Pennington : So, read the article…
Guttorp : As to journalists, find somebody who works on it and work with them. There's a wonderful article in the New York Times over the weekend on the effect of sea-level rise in Louisiana. And there was no mention of uncertainty. And so, that one focused on the effects and how the town is trying to protect itself and the difficulties it has getting funding and things like that. Because New Orleans gets much more funding than they do. But, you know, it's what's going to happen twenty or thirty years down the line that's going to matter to these people, more than what's happening right now. And the uncertainty in that needs to be taken into account.
Pennington : Well, Peter, that's all the time we have for today's conversation. Thank you so much for being here.
Guttorp : Thanks for having me.
Pennington : Stats+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 or iTunes. If you'd like to share your thoughts on the program send your email to firstname.lastname@example.org , and be sure to listen for future editions of Stats & Stories, where we discuss the statistics behind the stories, and the stories behind the statistics.