Guest Post by Willis Eschenbach
After many false starts, thanks to Steven Mosher and Derecho64 I was able to access the forcings used by the CCSM3 climate model. This is an important model because its successor, the CESM3 model, is going to be used in the laughably named “CIM-EARTH Project.” Anyhow, just as new telescopes have “first light” when they are first used, so here I’ll provide the first light from the CCSM3 ozone forcings. These are the forcings used by the CCSM3 model in their hindcast of the 20th Century (called the “20C3M” simulations in the trade). How well did they do with the hindcast? Not all that well … but that’s a future story. This story is about ozone concentrations. Figure 1 shows the concentration at the highest-altitude of the 18 atmospheric levels, concentrations that were used as one of the forcings for the 20C3M climate model runs.
Figure 1. Ozone concentration at about 36 km altitude (23 mi), used as input to the CCSM3 20th century (20C3M) simulations.
There are so many things wrong with using that “data” as an input to a climate model that I scarcely know where to start.
First, the provenance. Is this historical data, some kind of record of observations? Nope. Turns out that this is the output of a separate ozone model. So instead of being observations, it’s like a Hollywood movie that’s “based on a true story”, yeah, right … and even then only for part of the time.
Second, what’s up with the strange sub-annual ups and downs (darker sections) in the annual cycle? They start out in the upper part of the annual swing, and then they change to the lower part after about 1970. Nor is this the only altitude level with this kind of oddity. There are 18 levels, and most of them show this strangeness in different forms. Figure 2 shows their claimed ozone concentrations from about half that altitude:
Figure 2. Ozone concentration at about 19 km altitude (12 mi), used as input to the CCSM3 20th century simulations.
Again you can see the sub-annual cycles, but this time only post-1970. Before that, it goes up and down in a regular annual variation, as we would expect. After that, we see the strange mid-year variation. Most other altitude levels show similar oddities. Again, it appears that the modelers are not applying the famous “eyeball test”.
Third, how on earth can they justify using this kind of manufactured, obviously and ridiculously incorrect “data” as input to a climate model? If you are trying to hindcast the 20th century, using that kind of hockeystick nonsense as input to your climate model is not scientific in any sense, and at least gives the appearance that you are cooking the books to get a desired outcome.
Anyhow, that’s not why I wanted to access the forcings. I wanted to compare them to the output of the model, to see if (like the GISS model) it is functionally equivalent to a trivially simple single-line equation. I’m working on that, these things take time. I just posted this up because it was so bizarre and … well … so hockeystick-like.
More results as they occur,
w.
(Willis)“Third, if we don’t have information, we don’t have it. We can’t just take our best guess and use it as though it were an observation. Or I guess we can, but the output isn’t science without error estimates … and without observations that’s tough.”
(Steve) “Actually we do this all the time.”
(Willis) “We do? We make up fake historical “data” to feed a model, and on the basis of the results of using that fake data we claim that our model is accurate enough to forecast a century out into the future? We do that “all the time”?”
(Me) Yes, Willis, they do that all the time, but that is not the only thing ‘we’ do. If there is data that appears viable, but does not support the desired theory, that data is ignored or rewritten, such as in the infamous Hockey Stick. The long-accepted, natural climate fluctuations of the last 2,000 years were simply dismissed by one, poorly done ‘study’. All evidence to the contrary was abandoned and ignored. On the other hand, if there is actual data that we know for a fact is inaccurate, but supports the desired theory, it is accepted without question. This was the case with the studies that linked AGW to increasing hurricanes, in particular, and severe weather in general. We know for a fact that historical hurricanes were under sampled, both in number and intensity, as were floods, tornadoes, severe thunderstorms, lightning strikes, hail, snowstorms and even earthquakes. In AGW logic, none of these things actually existed before people began recording the events. It is the ‘Schrodenger’s Cat’ hypothesis of the human impact on climate.
So here is how data is handled in almost all climate science that points to a significant human influence on global climate:
1. If it supports a significant human influence on global climate, but is obviously flawed, use it and defend it as is.
2. If it does not support a significant human influence on global climate, but is robust, ignore it or discredit it with less robust data.
3. If it does not exist…make it up in such a way to support a significant human influence on global climate.
It has been like this now for 20 years at least. There is no climate ‘crisis’ science that would get a passing grade at a middle school science fair! NONE! All such studies ‘cook’ the data!
‘Because to me, it looks like the output of a bozo-simple AND INCORRECT computer model. Unless you think the historical record actually does look like Fig. 1.
A “yes” or “no” to my question will be sufficient, although of course an explanation is always welcome. Once I know which model you’re talking about, this might all be clearer.”
I’m not convinced of the following.
1. that you’ve downloaded the data properly.
2. that you’ve plotted it properly.
3. that the model actually uses the data.
Once I looked into all those matters and the actual provenance and the model
and read the papers, then I’d hazard an opinion on the data. I’d also contact
the PI.
Since I’ve seen people be wrong about the simplest things, I wouldt say something looked wrong. It would make me curious, not incredulous.
There are two forms of skepticism.
1. X is wrong
2. I dont know if X is wrong.
I always adopt #2. folks should know that by now
Wilde.
“A thermometer doesnt measure temperature, it models it. In other cases, like ozone, the models are complex.”
I think there is a level of complexity at which simple measuring segues into modelling and so the higher the level of complexity the less reliable the outcome.
#######
philosophically, there is no sharp line. Pragmatically we do act differently toward certain types of “observations” But that pragmatic difference is not really under pinned by any epistemic feature. Now of course, your assumption that
‘the higher the level of complexity” then the “less reliable the result” IS a testable hypothesis. What we would find is that this theory about complexity and reliability is most likely confirmed by our experience. But it’s not a necessary truth. It’s a good rule of thumb. What you see here is that the methods and assumptions that we actually USE in science are shot through with theory, rules of thumb, pragmatic choices.
the point of this is to be very cautious when saying ‘that’s not science, or science works only this way. The idea that there is a model over here and data over there is an ideal. An ideal that upon close philosophical inspection is not real.
None of this means that the model which produced the ozone is correct or incorrect.
But it does bear investigation. There is also the ironic issue that all satillite data relies on the very physics at the heart of the AGW debate. So that if you want to reject AGW physics in TOTAL, you’d best ask yourself why you rely on satillite data.. or your GPS for that matter. Some fellows around here, for example, questioned general relativity.
BUT they use a GPS. that device uses GR to perform its function. It relies on GR being true. Part of what I’m doing is pointing out to folks is this. Some of the data they rely on, relies on the physics of AGW. and you dont even know that. So you use something that you dont believe in. weird.
This is the first AGW discussion I’ve ever seen that devolved to The Allegory of the Cave.
” There is also the ironic issue that all satellite data relies on the very physics at the heart of the AGW debate. So that if you want to reject AGW physics in TOTAL, you’d best ask yourself why you rely on satillite data.. or your GPS for that matter.”
That’s a neat sidestep, Steve, but I’m not sure it is good enough.
The thing is that predictability out in the real world is key. If one actually measures a temperature then on the basis of well established general physical principles there are measurable and predictable outcomes in any well defined situation.
If one merely assumes a temperature (or an ozone quantity) on the basis of a model then the same does not follow and the test of the validity of the model is a comparison with separate real world measurements of any type that are verifiable well enough to constitute a meaningful comparator.
As far as I can see ALL climate science is in the latter category and the ONLY available comparator is what we can measure separately in the global environment so the further one goes towards a modelled scenario the more likely that it will diverge from any identifiable independent reality.
So then we come to the difference between our respective approaches.
I think that the reliance by you and many others on modelling and assumptions rather than measurements and logical conceptual constructs based on first principles is silly whereas you think that my reliance on measurements and logical constructs based on first principles rather than modelling is silly.
In reality there is merit in both approaches but each of us needs to accept the limitations of our differing methods. Both are faithful to the scientific method within reason but my approach is age old whereas the modelling approach is very recent.
There is an overlap in that in a sense my logical constructs are also primitive models but I think they are far more flexible by reason of their simplicity and closeness to established physical laws. The more complex a model becomes the more it adds contentious assumptions to those established physical laws and the more it will diverge from reality unless it is constantly challenged and adjusted to reduce that divergence.
There is currently a remarkable tendency on the part of AGW proponents to resist the implications of developing divergences.
Anyway my point here is to assert that my approach is as valid as anyone else’s and perhaps more so (certainly more in line with historical scientific endeavour over the ages before sophisticated models) because there are numerous bits of data that could come in and many climate events that could occur to invalidate, or require adjustment of, the propositions that I have put forward here and elsewhere.
Just don’t expect me to make any concessions to mere models or so called reconstructions. Climate science has been so influenced by wishful thinking for over 25 years now that I think EVERY reconstruction has been skewed by the biases of those who created them. I keep coming across reconstructions each from reputable sources that say the opposite of each other. A particular example arose in the BCP thread where it became clear that one reconstruction suggests El Nino getting stronger for the last 500 years and another suggests El Nino getting weaker for the last 500 years.
So all in all I would like to establish cordial relations with you but would appreciate some expression of regret for the thoughtless insults you directed at me on the BCP thread.
Thanks, Willis, your pricking the bubbles of scientific pomposity and unexamined self-righteousness that arise from the swamp that science has become always makes me smile in delight.
On the other hand, Mosh’s wild justifications for making stuff up are not funny at all.
steven mosher says:
May 10, 2011 at 4:38 pm
There is also the ironic issue that all satillite data relies on the very physics at the heart of the AGW debate. So that if you want to reject AGW physics in TOTAL, you’d best ask yourself why you rely on satillite data.. or your GPS for that matter.
That statement by you is quite non-sequitur.
It is possible to start theories/hypotheses from the same principles of physics, but introduce an error or fake data in one of them, come up with two separate, unrelated results. Then, if I say one of the results was wrong, you can’t claim that both results have to be wrong.
If the same principles of physics were used in developing temperature model based on what a satellite measures and AGW mumbo-jumbo, you cannot say i accept both or neither. We can use measurements using other techniques and validate satellite temps.
When we take the same approach to all the AGW predictions, They fail consistently. You cannot equate the science used in temperature models generated from sat measurements and AGW models
steven mosher says:
May 10, 2011 at 2:49 pm
Steve, I’m not reject the garbage shown in Figure 1 because it is the result of a model. I’m rejecting it because it is the result of a garbage model. How do I know? Because nature doesn’t look like Figure 1.
Sure, there’s lots of models out there. And often we have to use models to figure out what our observations mean.
But (as far as I can tell) the section of “data” in Figure 1 has NO UNDERLYING OBSERVATIONS for the period from 1870 to 1970.
Now, as much as you would like to gloss over that fact by claiming it’s OK because we use models all the time, you seem oblivious to the difference between:
1. A model that is used to ANALYZE observations, and
2. A model that is used to MANUFACTURE observations.
Yes, as you have repeated ad nauseum here and elsewhere, both are models … but so what? What does that have to do with anything, and in particular, how does that justify using a model to manufacture “observations”, as opposed to using a model to analyze real, actual, authentic observations.
Are you really not getting this distinction? Because it sure seems you haven’t noticed it. So I will make it brutally simple.
I’m not against models. I’m against “models all the way down”, models which (like the first century of the “data” shown in figure one) are completely data-free.
w.
steven mosher says:
May 10, 2011 at 3:30 pm
Oh, please, read what I say before you start to reply, then you won’t look so unprepared. Seems like you missed the part about the claims of accuracy to forecast a century into the future. Sure, you did what you said … but did the company build the airplane on the basis of that model?
What you have described is in fact common, when you don’t have real data particularly at the start of a project you might use “pseudo” data for testing or initial analyses.
But using made-up, imaginary data for the actual design of an actual plane? You have noticed that Boeing still employs these folks with the lovely title of “test pilot”? They are there because, unlike climatologists, airplane designers know that computers aren’t good enough to tell us if a plane will fly tomorrow … and they make no claims at all about a century from now.
So your example that purports to show ‘it’s done all the time’ merely reveals that you didn’t read what I wrote. Sure, manufacturing “testing data” and using it in a model is common as you say.
But then claiming that your model has century long forecasting powers using that manufactured “testing data”? Not common anywhere but in climate science, despite your fantasies.
w.
steven mosher says:
May 10, 2011 at 4:08 pm
And why should we care what you are “convinced of”? If you think it’s wrong, show it’s wrong or shut up about it.
As to whether I downloaded and plotted the data correctly, while mistakes are always possible, this is not my first rodeo. What I did was, I downloaded the data. I plotted each of the 18 levels of ozone data. I averaged the ones above and below 200 mb pressure. I compared them to their graph of the same data. I couldn’t see any difference between my average, and their data. So I concluded my downloading and analysis were correct.
And until and unless you’ve done the same, mosh … why are you opening your mouth? It doesn’t help your case to share your fantasies with us, go out and get some data if you think I’m wrong.
w.
PS – I don’t know if the model “actually uses the data”, Mosh. Perhaps foolishly, I believe their claim that they actually used the ozone data that they say they used.
And since the graph I just cited above is entitled “20C3M_ozone.gif”, I’m going to assume that yes, those are the forcings used in the 20C3m simulations.
And since my analysis of the data I downloaded matches that graphic exactly, I’m gonna say that your claim that we don’t know “that the model actually uses the data” is just your paranoid fantasy.
Facts, Steven. Your post is full of soaring flights of hyperbole, but it is totally fact-free. If you think I’m using the wrong data or have analyzed it improperly … shouldn’t you determine if your claims are true before uncapping your electronic pen?
Phew, that’s a relief.
Mosh getting a hammering after giving me a bit of anxiety on another thread just because of his ‘big’ name.
Naturally everyone puts a lot of themselves into their opinions so the big hitters are just as vulnerable as the rest of us.
So, whoever you are, don’t give in to intellectual bullying.
Quick question, Willis – have you looked at the other levels in that data? Did you properly area-weight the values, or just sum them? Have you thought about looking at the geographic distributions of both the stratospheric and trop ozone?
A vertical integral over the relevant sigma ranges might be interesting.
EllisM says:
May 12, 2011 at 7:26 pm
Good questions all.
1. I (apparently unlike the authors) looked at each and every one of the ozone levels in the data.
2. I area-weighted them, as did the authors, which is why we got the same answers.
3. While looking at the geographical distributions would be interesting, I haven’t done it.
4. Not sure what you mean by a “vertical integral over the relevant sigma ranges”.
Thanks,
w.
Are you sure you did area-weighting? I grabbed the file too and couldn’t replicate your plot with it. A simple sum of the values over all the grid boxes for the ozone data, then dividing by the total number of boxes did replicate your plot. Not that it really changes anything, but it’s more correct.
EllisM says:
May 15, 2011 at 4:51 pm (Edit)
Not sure what to say. I generated a cosine weighting curve and applied it to each 2-D weighting map (latitude/longitude gridcells). I’ll check it, I couldn’t swear it was applied to the Figure 1 above.
But in any case it makes little difference, as the ugly non-physical shape shown in the figures is present either way, the cosine weighting doesn’t change that at all..
w.