Opening Remarks
Lee C. Bollinger, President, Columbia University

President Lee C. Bollinger:

John Mutter: I'd like to thank all the morning speakers, and particularly Dr. Yee-Cheong for his excellent presentations. All of these were more than we hoped for, greater than what we expected, and I'm delighted with the talks so far. I'd like to have us take a short break, I know that's difficult, there's more than a thousand people in the auditorium, and I know that puts demands on the flushing systems and the coffee, but if at all possible if you could come back within ten minutes everything you need is very nearby. Thank you very much, see you soon.

Please take your seats. We're a little bit behind. We'd like to start the next session. Okay, close enough. Professor Kelley will again lead the remainder of this session.

Darcy Kelley: Okay. You do realize that my only qualification for this is herding unruly students in, even more unruly faculty. I call it herding cats.

So our next speaker has an interesting background, and I think it points out to us the kind of strong intellectual power that you need to handle complicated, multidimensional issues. And of course what I'm talking about is the weather. So Tim Palmer is the head of Probability and Seasonal Forecasting Division at the European Center for Medium Range Weather Forecast, based in the United Kingdom. But his background is actually in general relativity theory. He was trained as a physicist, he calls himself the “academic grandson of Paul Dirac, and these computational and theoretical strengths have been brought to bear on what is arguably going forward going to be something extremely important to us. So we realize it here with Katrina, but if you think about it worldwide the entire global weather system changing as it is in front of us is going to have an enormous impact on our ability to go for sustainable development.

Dr. Palmer received his Ph.D. from Oxford University, and his title is “Climate Prediction on Timescales of Seasons to Decades: A Requirement for Sustainable Development in the 21st Century.”

Tim Palmer: Thank you very much.

So I want to talk a bit about climate prediction, not only about what's going to happen sort of decades from now with climate change, but also make a link to shorter time scale climate forecasting on sort of a season or two ahead, because I want to claim that in the future as climate becomes more and more extreme our ability to make shorter range predictions will become an important aspect of sustainable development in the 21st Century.

Actually Jeff talked about global warming in the news. And I checked into the hotel last night and switched on the news, and literally within five seconds there was an article on this Time magazine special issue on global warming. And just the other week, about two weeks ago, I was on the train up to London and picked up one of the free newspapers on the train, and here were the headlines, “CO2 levels hit thirty million year high.” This was the report of the latest 380 parts per million levels of carbon dioxide, and this was a quote actually from the UK government chief scientist, Sir David King. So it's very much in everyone's consciousness, and yet on the other hand I think some of my fellow passengers on the train might have been a little bit bemused by this headline, because in Europe this winter we've had actually a pretty cold winter, not a winter you might associate with global warming. And I believe even in New York here you had one of the record breaking snowstorms of the last few years. So what's going on here? I mean this is a question people ask, is global warming somehow on hold now?

Well, here's a little animation of the atmosphere as seen from space. And it's showing two things. First of all it's showing how incredibly complex and complicated, even chaotic, our atmosphere is. If you can run it again, I don't know if that's possible, but it also shows the Earth seen from the infrared pass of the spectrum, and what we're actually seeing here is the main greenhouse blanket, which keeps the Earth warm. This is a natural blanket, and it's due to in fact water vapor, which is the key greenhouse gas. And the whole science, or one of the key parts of the science, behind global warming is how increases in carbon dioxide, which is a secondary greenhouse gas, will interact with water vapor, which is the primary greenhouse gas in the atmosphere, to produce a kind of potentially accelerated or enhanced greenhouse warming. So you see to understand this problem we've got a very complex system, and to actually try and solve the problem we use, if you can animate this, it shows the Earth now seen not from space but from the inside of a computer, and we're trying to simulate the complexity of that signal that's received from space using scientific laws, basically laws that were discovered by people like Isaac Newton many hundreds of years ago, and by some of the great thermodynamic scientists in the 19th Century, so laws of physics that were discovered well before global warming was ever considered to be a problem, so these are kind of very dispassionate and disinterested, if you like, models of global warming. They're just responding to the laws of physics.

Now some people have said, and I don't know if you've read Michael Crichton's interesting book called “States of Fear.” He even suggests because of this kind of complexity and chaotic nature of the atmosphere that we actually can't really predict global warming. So I thought I'd just spend a few minutes on a little tutorial example here which might be useful in sort of conceptualizing the problem.

This is an executive decision maker. This is actually a good example of a chaotic system actually, I guess there are a few executives in the audience so it may resonate. Because it's chaotic you can't actually predict the particular path of the pendulum as it goes from these different magnets. But actually one thing you can say, and if you could just animate that again, one thing you can say is that there's a kind of probability, an equal probability, that the pendulum will reside at any particular time at any of the four magnets. There's a 25% chance of finding the pendulum on any magnet. And if you think of these magnets as representing weather states, so warm winters, cold winters, dry winters, wet winters, this could be a representation of the 20th Century climate. Now what are we doing with climate change? So the next slide shows us a little wedge, and this wedge is going to be slipped under the executive decision maker. and the system goes off again, and again it's still chaotic, you still can't predict what's going to happen. But if you look at it closely you'll see that bob is visiting this warm winter, let's call it, state much more frequently than the other states. But its chaotic, and occasionally it might even slip over here and produce the 2006 cold winter. But overwhelmingly the statistics are now biased towards this magnet, which is my picture of a warm winter.

So the right way to think actually about the global warming problem in the presence of this complex, chaotic system that is our climate is that by increasing levels of carbon dioxide we're influencing the probability of occurrence. This is a key word in climate science, probability of occurrence, of drought, flood, heat or cold. And this is what we're trying to do, to estimate how CO2, increasing CO2, is changing this probability of occurrence. And the technique we use is a technique called ensemble climate forecasting. We run our models many times, so we have a climate model, we run it many times with different initial conditions and different model approximations. And we have a most likely prediction, we say what's most likely to happen, and estimates of uncertainty and estimates of probability of what's going to happen.

And here are some examples. So this is based on an ensemble of the world's global climate models, made for the IPCC fourth assessment. So years that have a one in twenty percent chance of occurring, so they're so warm that in the 20th Century they would only occur once every twenty years, that's the probability of occurrence in the 20th Century, that according to these forecasts much of the world, wherever the area where it's shaded red, that type of - well summer mean temperature will be occurring almost every year. So we're going out from a 5% probability in the 20th Century up to something like 90 or 100% probability across much of the globe in the - this is now towards the end of this century. We're looking basically at a time when CO2 has doubled over pre-industrial revolution values, so around 560 parts per million, estimated to be sometime towards the end of this century. I'll just focus on one thing here, which is the Arctic, because it's estimated on this basis that at least in the summer, in June, July, August, the Arctic is going to be completely ice free towards the end of the century, and there are obviously implications for the survivability of things like polar bears. But if you don't care about polar bears particularly you might want to worry about things like the Greenland ice cap, because if that melts then we're looking at seven meters, twenty-three feet of sea level rise. And there's disagreement about how fast the Greenland ice cap may indeed melt, and it's to do with uncertainty about the actual, if you like, dynamics of ice sheets and how they may disintegrate under prolonged warming.

Now warming is one aspect of climate change. We also need to address things like droughts and floods, as Jeff mentioned in his talk. So here's a probabilistic estimate of climate change from this probabilistic perspective in terms of how one in twenty year dry summers, I'm talking boreal summer here, June, July, August, one in twenty year occurrences based on 20th Century climate, how frequently those types of events will occur again towards the end of this century with doubling of carbon dioxide. And, for example, you see there are regions like over Europe here, and over central and southern America where a one in twenty year event has gone up to, this is 35% probability, so about one every three years.

Now if we focus on Europe those of us who come from Europe will know that just a couple of years ago we had an example of one of these very dry and warm summers, tens of thousands of mortalities literally arising from it, but also another statistic from that summer is that the vegetation cover over Europe was actually acting as a carbon source for the atmosphere, not a carbon sink, because photosynthesis had completely shut down. Now that may or may not be a concern globally, but what is probably more of a concern is that sort of thing happening in this part of the world, where again you see over Amazonia and parts of central America a significant increase in the probability of these extreme dry seasonal conditions. And again some of you may remember in the year 1997-98 when we had a big El Niño event there was considerable dryness over the Amazon and forest fires and such like.

So this raises again the question of the sustainability of the Amazon rainforest, and again the Amazon is a major carbon sink. So we have these bioprocesses which are feeding back onto the climate system and possibly accelerating it yet further.

Drought is just one aspect. We can also talk about flood. And here's a corresponding map which shows estimates of the probability of occurrence of a one in twenty year wet June, July, August. And the area I want you to focus on here is the Asia monsoon, because this is showing up as an area of increased probability of flood. Now places like Bangladesh are already, as you know, prone to severe flooding during the monsoon season, and this is going to further exacerbate the problem. This is, by the way, completely separately to the sea level rise issue, which is another factor which will obviously hit low lying countries like Bangladesh. So we have to now start questioning whether how sustainable these countries will be in the future with climate change. And then there are whole issues of what to do with a hundred odd million displaced people.

So the talk so far is to say that this ensemble climate forecasting is a way to provide quantitative risk assessments in a way which is completely consistent with the notion of the climate as a chaotic and complex physical non-linear dynamical system. And climate forecasts are clearly very important to guide strategy and policy on reducing greenhouse gases, so we can advise governments about risks of certain types of climate change, and also to plan for infrastructure to adapt to I think what has to be now seen as a certain amount of inevitable climate change. As I say, we're already 381 parts per million and increasing it 2 parts per million per year.

But I want to just focus the last bit of my talk on the use of climate models also for anticipating specific occurrences of climate extremes, not now decades ahead but maybe months or seasons ahead. So if we're saying that in the future climate extremes are becoming more prevalent, either drought or flood, can we use these climate models as a tool to warn society of specific occurrences of such climate extremes in order that actions can be taken?

And I want to just review a couple of examples of how this is already occurring. One issue which we've already heard about is the malaria problem. And actually in a study which a group of us in Europe combined forces with scientists at the Earth Institute, in fact at the International Research Institute for Climate and Society which is part of the Earth Institute, we looked at a particular type of malaria called epidemic malaria, which tends to occur in certain semi-arid countries of the world. Our particular study was focused on Botswana which is an example of semi-arid country which has epidemic malaria, so it occurs not year in, year out, but specific irregularly, certain years there will be a massive outbreak, and Botswana has good epidemiological records of malaria so we can study our ability to make useful predictions based on past records. This refers to a paper which actually appeared in Nature just a few weeks ago. By looking at climate models just on these short seasonal time scales, because it's understood now that the outbreak of epidemic malaria is strongly linked to climate, and in Botswana the epidemics tend to occur when the rainy season is well above average. And with these climate models we can predict ahead of the occurrence of the rainy season whether that rainy season will be below or above average, and by linking the climate model to malaria models we can thereby make predictions, probabilistic predictions, of risk of epidemic malaria.

So again, I just want to make this as an issue that will become increasingly important in the future as these seasonal extremes of climate become more prevalent. This is a slide from the paper which I think I'll pass over, but it shows the ability of the system to predict years which either have high or low malaria incidence in this probabilistic framework.

Another application in the health area is actually in meningitis, which in sub-Saharan Africa again has a link with climate, because years with high meningitis tend to be dry, dusty years. And again the ability of climate models to predict the sort of dry, dusty conditions is something that they're able to do.

And I mentioned about Bangladesh and flooding. And again we have a project jointly now with the Georgia Institute of Technology down in Atlanta, again with scientists in my group in the UK, looking at seasonal time scale prediction of the river basin drainage of the Brahmaputra and the Ganges in Bangladesh. And these forecasts are given to the Bangladesh MetService [?] in order to make them prepared ahead of the monsoon season for potential flooding problems.

And the final application is in this area of food, and again this is a study of ground nut or peanut crops in Gujarat in India, where again we found seasonal climate forecasts have an ability to couple to crop yield models, to predict whether the crop yields are going to be above or below average.

My just very last scientific point is to say that this is a sort of latest generation of climate model, where we've got down a one kilometer resolution, so we're starting to really resolve the cloud. This is actually of a developing hurricane simulation. The problem is to run this one day, even on the top supercomputer in the world, takes about one day. So if you want to wait a hundred years for the forecast it's not going to be much use. The point being that we're in a sort of a slightly unsatisfactory situation I think as climate scientists, in that we know actually the laws of physics very, very well, but we're constrained by computational technology to produce the most accurate forecasts we can possibly produce. And certainly a number of us, and I'm one of them, are trying to advocate for much more powerful supercomputers that can be dedicated to the climate problem. And after all, the whole issue about the sustainability of New Orleans in the future is actually a complicated one, and it's a combination actually of two factors. One is whether hurricanes are going to become more intense as global warming increases and the amount of water vapor in the atmosphere increases, allowing hurricanes to spin up. But counter to that, or on the other side of the equation, if you like, is what's happening to the global circulation of the climate system. And there are actually indications that if the climate goes into what's called a more permanent El Niño state, the circulation patterns over the Caribbean and up into the Gulf of Mexico may be such as to suppress hurricane activity. So if we ask the question is New Orleans sustainable we have to kind of balance these two factors. And at the moment with climate models with rather limited resolution, this is actually a very difficult question to answer.

So, I'll just come to my conclusions now. I've tried to indicate that the way to think about this problem in the sort of context of a complex and chaotic system which climate is, is to think about how increased levels of carbon dioxide are changing the probability of occurrence, for example, of extremes of weather. And the way we can make such probabilistic estimates are through these ensemble forecast techniques where we don't just have one climate model, we maybe have many different climate models which have different approximations, the approximations we have to make to the laws of physics to fit them on a computer are not in a sense precisely determined, so we have different possible approximations. And by running multiple forecasts we can get this kind of probabilistic estimate of climate change. And these are the things which decision makers and policymakers can make decisions about.

And I also wanted to stress that these very same techniques and these very same models can also be used in the shorter time scale forecasts. So the seasonal forecast problem is a little bit like just forecasting the movement of that pendulum bob from one magnet sort of to the next, so from one season to the next season. And this is going to be an increasingly valuable tool to help society and, I would say especially in developing countries, adapt to the changing probabilities of these climate extremes.

And then my plea as a scientist is that I believe we can improve the accuracy of these forecasts significantly with better computational resources, and as a community we're certainly trying to lobby for this in the future.

So thank you very much indeed.