E.T. Jaynes
“Probability Theory: the logic of science”‘

http://rankexploits.com/musings/2008/falsifying-is-hard-to-do-β-error-and-climate-change/

John M Reynolds

But as the models are of course simply models of a very complex climate systems, they simply cannot be perfect and it must be possible to test them against actual changes to climate to see how well their understandings of the linkages between various aspects of climate correspond to actual changes in claimte conditions.

I would presume that this process of testing and improving the models goes on all the time, so I am curious how you conclude that “issues of falsification are simply ignored or avoided.”

It also seems to me that your conclusion – that the “”consistent with” game being played with climate models by activist scientists … is every bit as misleading as the worst arguments offered by climate skeptics and a distraction from the challenge of effective policy making on climate change” – is stronger than can be justified, or at least that you have not shown us the basis on which you ground such a conclusion.

Regards,

Tom

Thanks for all your comments. You write, “Mark: please let me know if I’ve misrepresented or misunderstood your argument.”

I think you’ve misunderstood my argument a bit. I think that’s mostly because I never got around to responding directly to the most important point of Roger’s post. Roger asked, “Are climate models falsifiable?”

And his answer was, “I am not sure.”

I don’t really agree that that’s the most important question. A more important question is, “Are the projections in the IPCC AR4 falsifiable?”

And an even more important question–for a guy like Roger, and for the public in general–is, “Are the projections in the IPCC AR4 useful for making good policies?”

I think the answer to the last question is that the projections in the IPCC AR4 clearly are not useful for making policies that have not already been decided. That’s because the projections don’t have estimated probabilities of occurrence for different scenarios. As I’ve pointed out, the IPCC projections without probabilities estimated essentially provides no estimate of the most likely warming in the 21st century, absent intervention by governments. It could be nearly 100 percent certain of more than 6 degrees Celsius warming, or nearly 100 percent certain of warming less than 1 degree Celsius.

But to go back to the less-important question of whether the IPCC AR4 projections are falsifiable…I maintain they aren’t. At least not in any practical sense. Roger has called the IPCC scenarios “conditional forecasts.” But conditional forecasts are only falsifiable if the condition(s) actually come to pass. For example, I can forecast that, if I win $10+ million in the lottery on a Tuesday, in a month starting with an “M,” the excitement will cause me to have a massive heart attack. Obviously, the conditions on which that forecast is made are almost certainly never to occur (especially since I’ve never played the lottery).

The IPCC scenarios are an analogous situation, although obviously less dramatic. Roger has compared warming since 1990 with warming under the A1F1 scenario. However, he apparently based this exclusively on CO2 emissions. The actual climate forcing of any scenario is based on emissions/atmospheric concentrations of ALL forcing agents, both positive (CO2, CH4, black carbon, N2O, etc.) and negative (SO2, organic carbon, and others). This give defenders of the models and scenarios an “out” if the CO2 emissions match the scenario, but the temperature increases do not. They can simply say, “Oh, well, if the projections for CO2 AND all the other forcings (CH4, black carbon, SO2, organic carbon, etc.) had been accurate, the actual temperatures would have been exactly as projected.”

There’s a way this could be addressed, if the IPCC was really interested in doing science. (I think it’s clear they are not.) The solution would be to isolate each of the individual forcings, and to provide estimates of the most probable (i.e., 50 percent probability the value will be lower, 50 percent probability the value will be higher) values for each of the forcings (as well as estimates of the spread). In other words, develop probability density functions for each of the forcings separately. That way, there would be a single scenario that had the most probable forcing for each of the variables. And then update the probability density functions during each six-year analysis interval. In essence, this would do away with the “scenario” situation entirely…because the scenario situation simply isn’t scientific.

Tom Wigley and Sarah Raper had a paper in Science in July 2001 that at least improved on the Third Assessment Report (and in fact was ***better*** than the Fourth Assessment Report…a clear sign of pathological science that something produced 6 years earlier was even better than the AR4). In the Wigley and Raper paper, it was assumed that all the scenarios were equally likely, and a most probable warming from 1990 to 2100 of 3.06 deg Celsius was calculated.

Mark

By the way, I checked the IPCC AR4 for regions I think I know something about: essential areement and the IPCC express quite a bit of uncertainty regarding their predictions. Modest of them.

I happen to know that a recent paper presented at the latest AGU desagrees with the IPCC AR4 regarding the Amazon basin. Show that the science progresses (at leat changes) faster than one can hope to summerize.

I also note that IPCC AR4 does a climate model evaluation in Chapter 8.

From IPCC FAQ 2.1: “Human activities contribute to climate change by causing
changes in Earth’s atmosphere in the amounts of greenhouse gas-
es, aerosols (small particles), and cloudiness. The largest known
contribution comes from the burning of fossil fuels, which releases
carbon dioxide gas to the atmosphere. Greenhouse gases and aero-
sols affect climate by altering incoming solar radiation and out-
going infrared (thermal) radiation that are part of Earth’s energy
balance. Changing the atmospheric abundance or properties of
hese gases and particles can lead to a warming or cooling of the
climate system. Since the start of the industrial era (about 1750),
he overall effect of human activities on climate has been a warm-
ng influence. The human impact on climate during this era greatly
exceeds that due to known changes in natural processes, such as
solar changes and volcanic eruptions.”

So you doubt it then you admit there are differences? You seem confused. Here is the reference: (IPCC, 2007, p. 675, based on Santer et al, 2003. See also IPCC, 2007, Appendix 9C). Those charts show the warming signature of the atmosphere for 5 scenarios from 90N to 90S from ground level to about 30 km up.

John M Reynolds

Thanks for your detailed comments…especially yours Mark. Let me try to answer your last question. I’ll first respond generally why I think falsifiability cannot and should not be applied to all types of science. In a second post (if this one gets too long), I’ll give my very non-expert opinion on the IPCC.

So I take issue, as does David, with the idea that falsifiability can always be used to demarcate science from non-science. Falsifiability works very well when, e.g., you’re trying to determine if a given mixture is an acid or a base. If you hypothesize acid but the indicator says base, then you’ve falsified your hypothesis very easily.

But in systems science like climate science or evolution, the theories are based on cumulative evidence from many different fields. In evolution, facts from zoology, genetics, paleontology, radiometric dating, etc. add together to give a coherent picture. Because no single piece of evidence proves evolution, no single piece of evidence can disprove it. Evolution therefore does not falsify very easily. I think the same is true for climate science.

I think David hit the nail on the head when he said that science in practice does not always abide by a strict Popperian definition of falsifiability. Apparently even in general relativity, falsifiability does not necessarily apply.

In short, I (politely:)) disagree with your characterization that being unfalsifiable means something is not “scientifically valid.” Falsifiability can be a useful metric in some, but not all, areas of science.

I wrote you have an “overly simplified view of science,” because it seems you think that ALL of science can be easily described by a single idea. I think it’s really important to realize that science is not a single, monolithic concept. It changes from field to field. For certain questions, e.g., acid or base, the scientific method works as described in 8th grade textbooks. In other fields, the rules aren’t that straight forward. As a graduate student who studies space physics, I would argue that you cannot really experiment with the geophysical environment. When geophysicists use the word experiment, we mean something entirely different than someone in, e.g., particle physics.

If you accept that space physicists are scientists, you have to accept that being a scientist does not always require following the scientific method because we don’t really do controlled experiments. Just as all of science does not necessarily require the scientific method, all of science does not necessarily have to be falsifiable.

Mark: please let me know if I’ve misrepresented or misunderstood your argument.

You wrote, “While David’s critique appears to be valid (i.e. Mark’s view of science is overly simplified),…”

In what way do you think my “view of science is overly simplified”?

I’ve written that the IPCC projections are not falsifiable, and therefore aren’t scientifically valid.

Is that what you think is “too simplified”? Could you elaborate?

Thanks,
Mark