I recently gave this talk at Columbia’s Nevis Labs, in their evening public lecture series. The video is here – you can watch it all or in bite-size chunks. It’s about the relationship between extreme weather events and climate change: what we know and don’t know, what evidence we have so far, projections for the future, etc.
I have a post on the Oroville dam failure and what it does or doesn’t have to do with climate change here.
In this article, Andrew Freedman of Mashable argues that the Saffir-Simpson scale should be changed because it doesn’t measure rainfall, and yet rainfall causes disasters in many hurricanes – including Matthew in North Carolina just now. The Governor of North Carolina has been saying the same thing. The National Weather Service disagrees, and so do I.
In fact the Saffir-Simpson scale used to be based on not just wind, but also pressure and storm surge. It was simplified to just wind a while back because the different variables often wouldn’t match for a single storm – e.g., Sandy would have been a cat 1 due to wind but a cat 3 due to surge. There is no way to make a single number capture all the different hazards without being confusing. You wouldn’t know just from the number what the reason for the number was, and would still have to read the fine print in the forecast to find that out.
In addition, heavy floods can happen from storms that are not tropical cyclones at all. E.g., Louisiana a couple months ago. How would a change to the Saffir-Simpson scale help with those?
The rational approach is to warn on the hazard. That is, de-emphasize the Saffir-Simpson category, and issue warnings about each specific hazard separately. This is in fact the direction the National Weather Service (including the National Hurricane Center) has been moving. For example, they have recently introduced Storm Surge Watches and Warnings.
The NWS system can surely still be improved – apparently, the NWS forecasts of heavy rain didn’t reach everyone in North Carolina, even though they did exist and were quite accurate. So it would be reasonable, for example, for NHC to issue a formal Heavy Rainfall Watch or Warning (or whatever one wants to call them) when it is warranted in association with a tropical cyclone.
But multiple hazards from a tropical cyclone will remain multiple hazards. Combining them into one number wouldn’t eliminate the need for people in harm’s way to understand that. It would just disguise the problem, and disguising problems isn’t usually the best way to solve them.
My previous post was about the notion that models don’t constitute “evidence” for propositions in climate science. In this post I write about the criticism, also expressed to me in comments after my New York Times op-ed, that climate models (or the tools and methods of climate science more broadly) have been disproved by observations, so we should ignore them.
The people who wrote this to me didn’t spell out what disagreement they were talking about. But I assume that these comments were stimulated by the way I framed the op-ed piece, in that I wrote at the start about the major hurricane “drought” in the US over the last decade. The commenters seemed to be implying that since climate scientists had predicted stronger hurricanes in the past, the fact that the last ten years of Atlantic hurricane seasons have been relatively quiet disproves climate models, or climate science generally.
I set up the piece that way to be provocative, and to make a connection between current events in the US and the broader questions addressed in our Science paper that motivated the Times piece. But the point of the whole op-ed piece, as well as the Science paper, was that our recent history is, in fact, consistent with our current projections of long-term trends towards stronger hurricanes worldwide. My most vehement critics didn’t seem to seriously consider the arguments I made, as I didn’t get any responses which refuted them in any substantive way. But let’s elaborate a bit here on the two basic issues which our Science paper raises, one at a time, and then come back to the issue of whether the models, or the predictions made by climate scientists overall, are any good or not.
The first issue, very well known by the scientific community, is that there is large natural variability in hurricane activity. Hurricane seasons in different parts of the world are doing different things at different times, and they all fluctuate due to natural causes. So if you want to see climate change trends, you have to wait a while for them to become clear. Think about investing in the stock market for your retirement. Let’s say you start investing in your twenties and plan to retire at 65. Is the market going up or down over your working lifetime? You can’t tell by what happens from one day, one year, or even one decade to the next. That long term is what climate change predictions are about.
And you have to wait longer to see trends in hurricanes than you do to see trends in other variables. Take temperature for example. There are many more measurements of temperature than there are of hurricanes, since there aren’t many hurricanes while temperature can be measured anywhere. So statistical analyses of temperature data are much easier than those of hurricane data, because the sample is larger. Looking for temperature trends is like looking for stock market trends in a broad index like the S&P 500, made up of many stocks, while looking for hurricane trends is more like looking at an average of just a few stocks. The average of just the few will be more subject to the vagaries of what’s going on with those few companies day-to-day, and will have much more volatility, while the S&P 500 will show broader market trends more clearly. Looking at the hurricanes in just a single basin, like the Atlantic, reduces the sample size still further and makes the data even less representative of global trends. Continuing our stock market analogy, looking at the Atlantic only is like looking at just a single stock. It may be important, if you’re interested in that particular one, but it doesn’t necessarily tell you much about the bigger picture of the whole market.
Now, there are several credible studies by very established researchers that do, in fact, find that increasing trends in hurricane activity are already evident in recent historical data, either in individual basins or globally. But there is some disagreement between studies on the of statistical significance of these trends. So one argument is that the long-term trends are there, but that we need to wait longer to see them rise further above the noise so that everyone agrees they are there and significant.
Regardless, though, the notion that the hurricane drought in the Atlantic has somehow disproved the consensus projections of climate science is wrong, because the drought is still a relatively short-term fluctuation in a single basin, while the projections are for long-term global trends.
The second issue raised in our Science paper (now available free, see bottom of this post) is that perhaps there shouldn’t yet have been substantial long-term trends in hurricane intensity – whether we would be able detect them above the natural variability or not – because until the last couple of decades, aerosol cooling effects on hurricanes have been counteracting the effects of greenhouse gas warming. Though not a new idea either, this is one that’s perhaps a bit less understood than the one about natural variability. It has been raised before mostly with regard to the Atlantic, while to my knowledge our paper may be the first to show evidence from the IPCC climate models that aerosol effects have been important for hurricanes globally. It’s a subtle argument, because aerosol cooling has clearly been less than greenhouse warming – if not, the planet wouldn’t have gotten warmer over the last century. But it appears that aerosol effects on hurricane intensity are disproportionately stronger than their effects on the climate overall, because they reduce incoming sunlight while greenhouse gases act through longwave (infrared) terrestrial radiation, and changes in sunlight have a bigger influence on tropical cyclone intensity than changes in infrared radiation do.
To the extent the aerosol cooling estimates in the climate models are accurate, potential intensity theory (discussed in the previous post) implies that hurricane intensities shouldn’t have started to increase at all until the 1980s or 1990s, even though by that point the planet had warmed quite a bit. This would further delay time when the increases would be detectable above the noise of natural variability. The aerosol cooling in the models may be wrong to some extent, but is almost certainly in the ballpark enough so that the aerosols have compensated for the greenhouse gas effects on hurricane intensity to a significant degree, even if the exact degree is different than what’s shown in our paper.
So, to summarize: the projection that a world warmed by greenhouse gases – absent other compensating factors – will have stronger hurricanes still holds, but a) large natural variability means that it may take a while to see the trends, and the Atlantic “drought” is just part of that variability, and b) aerosol cooling can compensate the greenhouse effect on hurricane intensity, and has done so for much of the historical record until recently, slowing down the intensity increases. The role of natural variability in obscuring long-term trends has been understood (at least broadly) for a long time. The role of aerosols has emerged over the last decade and is still not fully appreciated.
Back to the broader issue of whether climate models have been disproved, though: the questions I’ve just written about are all about how climate change affects hurricanes, and not about the basic fact of human-induced climate change itself. Because hurricanes are so rare, and their natural variability so large, they are among the least clear indicators of climate change. There are plenty of much clearer ones, from increasing surface temperature (the global warming itself), to melting of ice on land and sea, to the long-term cooling of the stratosphere, increasing intensity of heavy rain events, etc.
When we consider all these climate change predictions as a whole, there is now a good track record. Climate scientists have been making predictions for several decades. The evidence shows that not only have these predictions come true, but they have tended to be conservative, understating the rapidity, extent and impacts of human-induced climate change. There are still many uncertainties, but those are reasons to be more concerned, not less, since they mean the reality could just turn out worse than we expect just as easily as better, and that’s what seems to have been happening so far.
PS: If you want to read our Science paper, it is now available for free – legally – via links on my academic publications page. (I can’t put the link directly to the paper here, as Science gives us this free link on the condition that it only can be used from a single referring page.) It’s currently at the top of the list of papers on that page, but in any case look for Sobel, Camargo, Hall, Lee, Tippett, and Wing, 2016: Human influence on tropical cyclone intensity. Science, 353, 242-246, DOI: 10.1126/science.aaf6574.
Since my op-ed on hurricanes was published in the New York Times, I have gotten a number of emails and comments on this blog about it. Some of these can be characterized as climate-denying hate mail. This is inevitable any time one publishes virtually anything about climate in a prominent forum. It isn’t practical to respond to every one individually, and in some cases I don’t think it would be productive to do so. But I do spend some time thinking about what people write, if it seems that they’ve put any thought into it – even if that thought is expressed in a hostile way – and want to respond in a general way to some of it.
I should explain first that you won’t see the comments I am talking about here. This is a moderated site, moderated by me, and I don’t approve climate denialist comments. You can tell me I’m wrong about something and that’s fine – I’ll still approve it, if it’s written in a way that suggests you’re open to evidence and argument. But I don’t approve comments that parrot the standard tropes you hear on Fox News and other standard anti-science sources in order to justify complete rejection mainstream climate science. Typical indicators of such comments are conspiracy theories, mentions of Al Gore*, calling me or other climate scientists “liars” etc. There are plenty of places to vent this kind of stuff online; I’m not going to host it here for my little blog’s twelve readers.
That said, today I want to respond to one criticism that was common to a number of the responses I got, though. This is that my argument was based to a large extent on models and theory. Several critics wrote that this is not real “evidence”.
(One can also argue that the models are bad, having been disproved by observations, and so should be disregarded. In a few days, I’ll put up a second part to this post in which I’ll respond to that. Today I’m just writing about the general question of whether models and theory can constitute evidence for a scientific claim at all.)
When we are talking about the future, models and theory are essential, because there can be no prediction of the future without a model or theory of some kind. If you think that the only valid evidence in a scientific argument is an observation of something that has already happened, then you can’t talk about the future at all. It hasn’t happened yet, so there are no observations of it.
Science is about taking observations and integrating them within the context of ideas about how the world works – theory and models – using them to test those ideas, and then, to the extent the ideas hold up, using the ideas to make predictions of things that haven’t been observed yet. If you want to disqualify predictions based on models and theory from science, then you also have to argue that science can’t make any predictions about the future of anything. If you really believe that, then say that. But don’t say there’s something wrong with climate science in particular because it uses models.
(Of course, there are some predictions that are so trivially easy that they seem not to require a model. For example, I predict the sun will rise tomorrow. We could argue about whether I am still using an implicit model of some kind to make that prediction, but it doesn’t matter here because climate change is not a case of this kind. Let’s restrict ourselves to nontrivial predictions, where we can’t rely on very simple patterns from past experience to tell us what will happen.)
That said, there are models and there are models. I argued in the op-ed that “physics” agreed with “models” and that that showed consistency between two lines of evidence. One commenter argued that the models are also based on physics, so these are not independent. Depending on how you interpret the words “physics” and “models”, this can be read as a fair criticism – assuming you read only the op-ed, which had to skip over a lot of the argument in our paper in Science on which it was based. I used the word “physics” to mean the same thing as what I am calling “theory” here.
(I’m sorry that the Science paper itself is behind a paywall; I am told by the journal that I will soon get a link allowing me to legally provide the full paper for free from my academic publications web page, but I don’t have that yet.)
The Science paper focuses on a quantity called potential intensity. This derives from a theory in the tradition of physics. (This does not mean it is speculative, or just a hypothesis unsupported by evidence. It means it is a set of logically connected statements that explains and synthesizes many observations. Newton’s theory of gravity is a theory too.) Potential intensity theory contains an explicit set of statements about what a hurricane is and how it works. It then takes the equations of classical mechanics and thermodynamics and makes a set of approximations and simplifications based on those statements. (For example, the hurricane is assumed to be circularly symmetric, so the equations simplify from three dimensions to two, radius and height.) This leads to a prediction of how strong a hurricane can get within a given local climate. Most hurricanes don’t reach their potential intensities, but how strong a real hurricane gets, on average, is still related to the potential intensity. So if potential intensity increases, we predict that real hurricanes will get stronger. Knowing whether potential intensity will increase or not requires us to make a detailed prediction of how climate will change – not just the surface temperature, but the temperature and humidity throughout the atmosphere as well. We can do this with climate models.
The distinction between a “theory” like potential intensity theory and a “model” like the ones used to make weather forecasts or climate change projections is one of complexity. The models we use for real weather and climate prediction are very big, complex computer codes that solve the equations of physics as applied to the atmosphere. The solutions can’t be written down in simple form, but can only be represented approximately by huge numbers of numbers, which we typically visualize in maps and charts. The models make some approximations too, but not nearly as many or as severe as a theory like potential intensity does. They retain as much of the atmosphere’s full complexity as they can (quite a lot, on today’s powerful computers) and the results of their simulations can’t generally be predicted ahead of time.
The climate models used to make climate change projections generally don’t simulate hurricanes well, because they don’t have enough resolution – just like a digital camera image without enough pixels to make out something small. They do simulate the climate well, though. (Not perfectly, of course, but remarkably well; but defending that statement is not today’s topic.) We take those models’ predictions of the local climate all over the world in the future, and calculate the potential intensity from it. We compare that to the potential intensity that the same models produce in the present climate, and use the difference to predict how hurricanes’ intensities should change. That is one piece of evidence.
Then there are some climate models that have high enough resolution to simulate hurricanes well. With those models, we don’t need to use potential intensity theory. We can just run them in a warmer climate and see how their hurricanes change compared to a cooler climate. That is another piece of evidence. It’s different than the one in the previous paragraph, because it involves direct simulation of hurricanes and does not require us to accept potential intensity theory as valid. The two kinds of models are similar in how they predict the climate itself, but completely different in how they represent hurricanes. And my argument in the op-ed is not about climate change itself – I take for granted that we know how the climate as changing, at least in broad outline – but rather how that climate change affects hurricanes. So for this purpose, it’s two pieces of evidence, both of which indicate that warming due to greenhouse gases, acting alone – that is, without too much aerosol cooling to counteract it – should cause hurricanes to strengthen.
Yes, both pieces of evidence are from models. But again, if you want to predict the future there is nothing else.
*Al Gore’s movie, An Inconvenient Truth, was reasonably accurate, as I recall, given the climate science of its time. And while I haven’t watched it since then, I’d say it’s probably still mostly right, at least in broad outline. I do recall, however, that the material about hurricanes was overstated even given the science of the time, and the science on that topic has advanced a lot further since then. But in any case, Al Gore is a politician, not a scientist, and his movie was not a statement by the scientific community. When people say climate science is wrong because they don’t like Al Gore or his movie, it is an indicator – usually accompanied by others – that they are forming their opinions based on political affiliation rather than consideration of evidence or argument.
Solomon Hsiang and I have a new paper out in Scientific Reports.A blog post about it by Stacy Morford is on the Lamont site. The paper points out the ramifications of a basic fact about the structure of earth’s temperature for potential climate change-induced migrations of species, including humans. In one sentence: tropical temperature gradients are small so that it takes a long migration to cool off if you start near the equator.
This new paper is a little different than my usual, in that it’s about impacts of climate change rather than about the physics of weather and climate itself. It does connect the physics to the impacts, in that the uniformity of tropical temperatures is a consequence of basic geophysical dynamics and a fact that I have made extensive use of in my normal (purely meteorological) research. It’s a theoretical study and not to be taken as a prediction of any specific migration, but we think the main point is nonetheless relevant to more practical and realistic discussions on this topic.
The lead author, Sol Hsiang, got his Ph.D. at Columbia some years ago in our sustainable development program, and has gone on to become a star in the rapidly growing field of climate impacts. He deserves most credit for this paper; the only reason I could participate in a study like this is because he had the idea and kindly asked me to be involved.