Scientific method and so-called skeptics

This is an addendum to the answer I have tried to give to Scott Adams‘ (the creator of Dilbert)
question on why science can’t seem to persuade climate skeptics. The basic answer is: Because they don’t want to be convinced. Maybe I should leave it at that, but I feel that maybe some people could indeed benefit by discussing the finer points of the methods of science Adams is targeting with his post. So this is for those specialists (this post not being announced anywhere but at the end of the main answer I have given).

Dear Scott Adams, I’ll dip into some of your points 1 through 14 here. I’ll go at my own pace, I am not necessarily going to jump the hoops you hold up. If you want to skip ahead: I will be dealing with

For starters this business about supplying a number, a percentage of how much of global warming was man-made. The IPCC said in 2013: “It is extremely likely that human influence has been the dominant cause of the observed warming since the mid-20th century.“ The UN-body elsewhere defines “extremely likely” as 95 to 100 percent probability. But it doesn’t give a number for “dominant”.

Let’s figure this out. Since the only other category except human influence would be non-human or natural influence we have a two-way-split. I think it would be safe then to assume that dominant means decidedly more than 50 percent. And if you look at this figure from that same IPCC report, you would think it is closer to 100 percent, wouldn’t you? Without human influence there would be little to no warming, that’s those blue bands. Add human and natural together and you get the pinkish bands around the actual observations. It’s clearer in larger regions with more data: temperatures in North America, Europe, Asia, Australia show no overlap of the colored bands anymore. And it’s most clear in global averages. And there are literally 1000s of sources in the IPCC report.

Source: IPCC, AR5, WG1, Figure SPM.6 

Maybe well over 50 to 100 percent human influence is still too vague for you. Sorry, things are not as simple as we all would like them to be. Anyways: Any single number would most certainly be evidently wrong in the precision it radiates. And attack would be sure to follow. Another of the mock debates surrounding this issue.

Next: Why do you fight being told the rate of warming was a tell-tale sign of what is happening? Rates are often as important or even more so than absolute numbers in every day life, too. Consider the task of sending a message from New York to San Francisco: The distance has stayed the same since both cities are on the map. First you could send it via stage coach, then by train, plane, now e-mail. The amount of time went from weeks to seconds and that means the speed of transmission, the rate of transport went up by orders of magnitude. Not such a difficult concept, is it? And that doesn’t mean heat and cold records aren’t relevant information, too. In fact they are, and there are way more new heat records than new cold records. By 4:1, here is a source for the US.The point is, if there was no change, then there would randomly be new cold and heat records in about the same proportion.

Arctic vs Antarctic: The ice goes away in the north and grows in the south, you say, and accuse scientist of ignoring that. Well, they aren’t, there is lots of research trying to figure out what is happening in both areas. If you think scientist are hiding that ice is growing down south then maybe you are listening to somebody who wants you to think they are hiding it. You can, for instance, look here if you interested in sea ice. For land ice, this might be a good starting point – the question, it says, in under debate.
There are several big differences between the regions that make the simple comparison you demand difficult, to put it mildly. Antarctica is a continent surrounded by ocean and fierce circumpolar currents that basically give it its own independent weather. The Arctic is an ocean surrounded by continents and tightly coupled to what happens with storms, temperatures and precipitation in many countries around it including yours and mine. The northern polar region belongs to a hemisphere that contains two thirds of earth’s land mass – including the major industrial nations with their emission histories. The southern has one third of the land including Antarctica and is otherwise 81 percent water. Why would you expect the two regions to be mirror images? It’s also not as if they were shifting ice masses between them making the changes a zero-sum-game. It is a little like adding the scores of the Chicago Bears, Cubs, Bulls, Black Hawks and Red Stars and comparing them to a sum of clubs from Los Angeles to see which is the better sports town.

And just to finish up with details, your point number 11.

  • Why aren’t insurance companies paying attention to sea level rise? They are. The US flood insurance program is recognized to be in dire need of reform because of rising seas (here, here and here, the last being a report by the National Academy of Sciences). And internationally take the example of Munich Re. It says in its annual report (page 74) that “Climate change represents one of the greatest long-term risks of change for the insurance industry”. It quotes the cost of adapting to sea level rise as 1 trillion dollars for the US alone. And university analysts have given it thumbs up on its response.
  • Why do your beaches look the same? Well, Wikipedia has you living in California, the Eastern Bay area to be exact, and I assume you might be going to Santa Cruz or thereabouts. So let’s look at a map. Here is Santa Cruz with the inundation 2010, 2060 and 2100 (Source). The water has been progressing slowly (and presumably local authorities have taken care of their beaches) but changes will accelerate.
                              Source: Monterey Bay Sea Level Assessment by Noaa

    The numbers in one projection are: seven inches rise last century, six more until 2030, another six until 2050, and two more feet on top by 2100. So you are in the middle of the two-inches-per-decade phase between 2000 und 2030. That’s not easy to notice by naked eye especially when there is maintenance going on. The water will rise, your state authorities say.
  • Why are half the top hits in web searches debunking the rise? Are you honestly asking that? It’s not about the quantity of hits but about the quality of sources. Bloggers have made it their business to attack the science and since they get shared and linked across a parallel universe the Google algorithms think the sites are trustworthy. They get ranked high like science sources who often don’t spend as much time on search engine optimization. For fun try “aliens landing earth” or “Barack Obama muslim” to see the quota of reliable information.

Now for the grand finale, your major and indeed first point: models. Maybe you have been impatiently skipping down here (or maybe you have tuned out) because the IPCC-graph I showed earlier depends on models. How else would you be able to artificially switch off human contributions? That’s nothing we can do in real life.

But why do we need models, in plural, at all? Well, that’s a tenet of science. You make multiple measurements because each one could be flawed, and then you look at what the tendency, the average, the median is, or whatever statistical analysis of them tells you. It’s easy to make fun of that, I know. In everyday life one measurement usually suffices. Has your kid gained another inch, have you lost weight, are the panels of your strip all the right size, is there enough flour in the cookie dough, how many people are at the party, is there enough pressure in your tires, how many e-mails do you get in a single day? Nobody does multiple measurements and statistical analysis for those. Well, for the e-mails you might to eliminate the effect that there could be systematically higher numbers on single days, maybe Mondays and Fridays – just guessing.

In science the questions are usually a whole lot harder, so this multiple measurement habit has proved to be a good idea. As has trusting science as a whole for society, by the way. A consequence of doing this, if you do it right, is that you get two numbers. The result you are looking for and a measure of confidence that that is a figure you can depend on. Scientists being a little nerdy often call the confidence measure “uncertainty”. And that is what it is, too. It is a measure of how uncertain we need to be – or certain we can be. In everyday language that word “uncertainty” comes across badly, just like scientists don’t know their stuff. In fact it is not a weakness but a strength of the scientific method. It has established ways of quality control that are certainly lacking from many other parts of communication – like getting directions in a strange town, remembering the weather two years ago, indicating the size of fish you caught.

When there are no measurement values to be had, because you are talking about a future that hasn’t arrived yet, you need to construct models. They simplify reality in a way that makes it computable, usually leaving out big chunks of the complicated stuff. As the saying goes, they are all wrong, but some are useful. For them also it is a good idea to have several, ideally constructed from scratch in independent, competitive research groups. If they point in different directions, all but one must be getting it wrong. If, however, they point in the same direction that can boost confidence they are getting something right.

Of course they could in principle still all be wrong. Then they all would make the same mistakes and that points to the underlying science being wrong. It’s equations about how pressure and humidity and wind and rain work together and how CO2 and other greenhouse gases influence the radiation that enters and leaves the atmosphere. Complicated stuff but pretty well tested, so the criticism against models is usually not leveled there.

This graph comes from the IPCC. It shows there’s some work to be done but not total chaos. It is a comparison of a whole lot of climate models hindcasting global temperatures over roughly 150 years and through several large volcano eruptions. Black and bold are the observations, red and bold is the average of the models, the thin lines are individual ones, the yellowish shading marks the reference period against which all temperature differences are measured.
Source: AR5, WG1, Figure 9.8, 2013

To prove that the equations and the mechanisms in the models are working together well researchers do this thing called hindcasting that irritates you so. The models start their work at some point in the distant past, fed with the initial conditions from that point in time and then run unguided through the years up to the present. The results can be compared to what actually happened and give the model makers a measurement of how good they are doing. Only if they pass that test can anyone place any confidence in them getting the future right.

„It tells you nothing of their ability to predict the future“, you say. I wonder why you would think you are in the position to judge that. Yes, hindcasting indeed tells us something about the ability to project the future, because the models prove they get the underlying science right. They calculated a future from 1850 onwards based on first principles. It’s just that for us, that future is the past.

I think it is seriously wrong to equate climate models to the financial market. You are missing something rather important. In climate science people actually know what is happening and what causes what. Obviously the minute details of dealings at the Stock Exchange are NOT known even remotely as well. If you insist on that kind of simile: The only thing that would be comparable is if you have a set of publicly available rules and based on that not one financial genius but a whole bunch of them put their predictions of the future into sealed envelopes guarded by a trusted institution only to be opened when the end year envisioned has arrived and all the data are in. And you would only count it as a success if everyone of those guys got it reasonably right.

Anyways, hindcasting of course is not the primary product of the models. It is a necessary quality check, nothing more, nothing less, and should no more be advertised than race track results for a new car model that will never go 150 mph in traffic. The results should be there for the asking, sure. But not center stage. If you see hindcasting as a major part of science communication around climate models you are looking at the communication in a different way from me. It’s safe to assume that neither of us has the right way to look so we both should be careful about what we say about them. I’ll certainly look out for exaggerated hindcasting language in the future. Maybe you can look beyond that.

Remember though I said that models were “ideally constructed from scratch in independent, competitive research groups” – well, not all of them are. There are connections between some research groups but not others. So some researchers are wondering if all the models should be treated equally, they question the model democracy, as they put it. This is brand new stuff. The suggestion is to weight the models and then to do weighted averages. In real life you would be using that technique maybe to locate a sports arena that several clubs will be using – and you want to giving the club with a 100 actives more say than the clubs with 25, 33 or 50 players. So you weight their respective answers with the number of members. In the case of the models the categories that could be used for weighting are independence – the unique efforts get more say – and skill in hindcasting. So there still is not what you wanted, the single best model, but more of a unified voice of them.

Finally to put the whole business into a framing you might understand (and I find weird examples work best to get points across): If you were to hold the secret theory that creating Dilbert was inevitable given your life’s twists and turns – how would you test that? You could recruit a number of young, very young kids and subject about half of them to the same experiences you had to see if they came up with a similar syndicated cartoon series. You would have control groups that get different experiences to test correlates: Could a woman become Scott Adams? Would there be Dilbert if there hadn’t been Little Nemo, Peanuts or Doonesbury before?

And since this in reality would be impossible or inhumane to subject people to you would think of models to give you an answer. Models, in plural, because if you only relied on one you could so easily get it wrong and never even know about it.

If you are seriously interested, it is a good idea to do this in plural. Scientists are seriously interested. And as a citizen following science news you would be better off not pushing for answers that leave out the best science has to offer. And that doesn’t just apply to models, that goes for most of your demands.

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.