http://rankexploits.com/musing.....-rejected/

Basically and analysis using GISTemp and UAH to show the IPCC’s forecast for temp increase to be falsified

Blogs are conversational media involving the author and mostly regular readers. They provide comments blogs. For that reason, the posts may not be “complete” in the sense that stand alone journal articles may be, and some choices discussed in past blog posts may not be discussed in the current one.

Oddly, book chapters do the same things. They often assume the reader has read previous chapters to understand the current one, and don’t necessarily tell “link back” to the precise page where a concept was first introduced.

In my opinion, the conversational style of blogs has little to do with the rigor of analysis. That’s basically what I mean when I respond with “this is a blog” to those sorts of complaints.

Not to pick nits, but you did say, “And yes, as for your complaints about lack of rigor: I describe this as a blog. It is not a journal article. Blogs are conversational media. If you don’t like that or think yours is something else, fine. This seems to suggest you had used the observation “this is a blog” in response to more than just linking issue. That being said, it also suggests only lacking the rigor of a journal article and says nothing about lacking rigor, which is what Robert is inferring.

General: I’ve opened up a ‘question place’, http://moregrumbinescience.blo.....ace-2.html for science questions. See also my current draft of comment policy at http://moregrumbinescience.blo.....olicy.html

Eric: I see folks on blogs refer to AGW theory, but it isn’t one I know from scientific sources. Could you, or someone else, tell me what it is that theory is supposed to be. (Preferably also what scientific source you saw it in.)

A major point, which I see I didn’t write explicitly in either of my testing ideas notes http://moregrumbinescience.blo.....eas-1.html and http://moregrumbinescience.blo.....eas-2.html is that my basic objections are physical, not statistical.

The physical issue is ‘what is climate’, which I started giving takes on in http://moregrumbinescience.blo.....imate.html and http://moregrumbinescience.blo.....imate.html. Though statistics are apparent, what’s important to me is the physical side — climate is the part of the system which is slowly varying, weather is the part that is rapidly varying. Unfortunately, we only ever directly observe weather — instantaneous measures of what’s happening.

If what you want to study is weather, you’re set, just take the observations and have at it. (Which still leaves you a very complex problem!) If you want to study climate, you have to do more work than that. So say I, Lucia disagrees. What I, and (more importantly) the people who do the observational work on climate think is that you have to disentangle the two. If you’re interested in trends in surface temperature, for instance, you need enough data so that the magnitude of your trend doesn’t depend sensitively on how long your record is (among other things). If it does, you’re looking at weather (still).

How long is ‘long enough’ … well, that’s what I was addressing in the two posts. It’s incomplete and not rigorous, but it is, at least, a start. The conventional answer in the field is ‘30 years’, which goes back to at least the early 1900s. I’ve taken a side project to see if I can find the origins of that conventional 30 year period. That it’s stood to use for a century, though, suggests a pretty good value.

But let’s make another tour of the trend issue (I was looking at average rather than trend, but several others have looked specifically to trend, including stoat, tamino, and realclimate). Yesterday the high in my area was 10 F below normal. Four days earlier it was 10 F above normal. Clearly the climate will drop below absolute zero in another few months (5 F/day, ca. 500 F to drop, 3-4 months). Except that those weren’t climate observations, they were weather observations. The former was while a front had us in the warm sector, the latter observation was while Hanna was pulling colder air down. If you want a trend, you need a long enough period that the timing of fronts and storms doesn’t change your answer. Today will probably be close to normal, dropping the trend from 5 F/day to 2 F/day. A one day change in record length can’t have that kind of effect on the trend of climate.

So now you (a figurative you) decide that you’ll avoid that sort of problem by using global averages (if the front is making me warm, it’s making someone else cool, so it averages out) and, say, average a month at a time. But look up at the plot of monthly temperature anomalies. They change by several tenths of a degree C from month to month. Your trends for ‘climate’ are then changing by plus or minus 500 C per century (century being Lucia’s preferred time unit), from month to month. This is still, clearly (to me and folks in the field) ‘weather’. And, clearly, you’d need quite a lot of months to get that sort of noise to average out, even more so when you’re arguing about signals of only about 2 C/century.

So then you try, say, about a decade, as Spencer and Christy did in their first major MSU publication. And you get burned by the fact that the record starts near a major El Nino (warm event) and ends near a major cool event (Pinatubo). Had they used the standard 30 year period, this wouldn’t have been a major problem. On the other hand, it would also mean that they’d only now be publishing their first paper. If you’re going to violate the standard, you have to be very careful about all those sorts of things that can give you false results, including: volcanoes, solar variation, ENSO, AAO, AO, MJO, QBO, PNA, AACW, PDO, … plus how all of them affect the different observing and analysis systems. (Not least, you should know all the things I left off the list — that list is far from complete, just things I happen to know of while working in areas not terribly concerned with most of them.) If you then want to compare a trend from your analysis of a short record of temperatures with a trend from a short record of model output, you also have to become expert enough about the models to know whether they can honestly be expected to model those various things, and if not, whether and how that affects their ability both to get the trend for your short period right, and how it affects their ability to get the long period changes right.

Of course that ‘have to’ is only one in the sense of ‘in order to be rigorous by my notions’. You’ll see from the testing ideas 1 note (an example that has nothing much to do with here) that this can be rather demanding. In contrast with rigor, I’ve been reminded that this is just a blog. For ‘just a blog’ purposes, the fact that my back yard was well above the models’ projections for last Wednesday, or that it was well below yesterday, or that the cooling trend was so enormous between the two, are all ’sufficient’ to lead people to disregard the models. I’ve been told this a number of times over the years (the other people using their own back yards for the ‘disproof’).

So what does all this say about whether climate models can be tested? Not much. There are a lot of aspects to climate models — global mean surface air temperature is far, far from the only variable present. There are temperatures for the ground, sea surface (vs. atmosphere), pressure levels through the troposphere and stratosphere, elevations of the pressure levels (relates to temperature in a vertically averaged sense), phenomena like ENSO that the models should be able to treat with some degree of skill, and on for quite a long list. The different variables have different observing problems associated (all observations have problems!), different variances, different expectable degrees of skill from the models, etc., etc. While global mean surface air temperature is not a good variable for making a test with only 7 years of data, there are rafts of other candidates to look at. So look at them, and verify that 7 years (or whatever) is sufficient to make a good statement about the climate with. Then have at it.

What would be a statistically significant interval, assuming this trend stays the same or only slightly warms/cools for x amount of years in regards to severely weakening or disproving AGW – my thoughts are that it already has begun weakning the argument and we might start seeing friendly neighborhood GISS henchmen lighting fires near strategic temp sites if things don’t pick up soon.

As far as the pizzas, I guess I wasn’t very good with my analogy judging from your response. In my version, when I said, “four different cooking methods” I was referring to her statistical models, which she doesn’t mix. I had thought that your earlier qualm was that in some of the tests, she’s mixing the toppings(GMST estimates/proxies). In your version of the pizza analogy, if I’m reading it right, she’s mixing the cooking methods (New York/Chicago) and I think my “toppings” transformed into your “cooking methods”. If that’s the case, the point I was trying to illustrate with my analogy (but now converted into the terms of your analogy) is that if you don’t like the New York/Chicago style mixed pizzas, that’s ok, but it doesn’t change the tests of the isolated New York style (only) pizzas.

Sorry, I promise not to use any more analogies for awhile..

Sorry, sometimes I confuse myself. Someplace along the way I got to thinking the theory says that temperatures go up if the sensitivity to GHG’s, after feedbacks, is greater than 0.

If you think it is starting get confusing, you should try being in my head.

Ok, when I re-read, “the temperature trend from the 30’s is not supposed to match the trend now,” I thought about what I would ask my son if he said the same thing (to try to stimulate his grey matter). “Really? Why not? How should it differ? Does it differ in the way expected?” I think that last question is the one I am having.

I think in your question “x” is the trend.

The original question is can we estimate a trend (x) using smaller time samples: The answer is yes. However, there are larger uncertainty intervals for smaller time samples.

The argument that we can’t use smaller time samples is founded on a number of things. One of them is the variability of “weather noise” is very large, and so our confidence intervals are so large we can’t get decent estimates. I think you are calling the “variability”, “y”. Those are similar during the two periods– and the dispute is fundamentally over y.

So, we can estimate “x”. We can distinguish “x” from some incorrect theories about “x”. The argument that we can’t do it is founded (in part) on the claim “y” is larger than it is. But, evidence suggests “y”– based on current weather– is the same value as “y” in other eras with no volcanic eruptions.

This is all getting very confusing.

N ( x , y )

My question was is our current sample consistant with past samples.
You clarified that our sample would need to be compared with other non-volcanic samples. You state that y for our sample is “roughly the same” as other non-volcanic samples. What about x? Does it matter?

The current non-volcanic period has roughly the same variability monthly temperatures relative to an trend fit as other non-volcanic periods. In contrast, volcanic periods have must larger variabilities. Also, if you calculate all 8 year trends during periods with no volcanic eruptions, those are larger than during periods with volcanic eruptions.

So, basically, volcanoes introduce “noise”. (I guess if we were going for the height analogy, a country with more immigrants fro different places — some tall, some short–might have more “height noise”.)

But, they don’t affect the long term trend much. (The do affect the short term trend a lot– that means when there are eruptions it’s harder to estimate the long term trend using short term data.

The magnitude of this variability affects of ability to estimate the mean trend.

Thanks for the patience with the dumb questions. You said our current sample of brazillian men was on the low side of samples of brazillian men entering clubs. errr… I mean you said that the current temperature trend was on the low side of other samples from non-volcanic periods.

Would adjusting for solar account for it being on the lower side of the “non-volcanic norm”?

On the dance club: Yes. If there was a dispute over sampling in a dance club, we would best resolve that by not sampling in a dance club. That’s an issue of bias, and we do need to try our best to avoid that. (If we can’t avoid it, we do get to argue about it. It’s valid to investigate possible biases, estimate their likely magnitude etc.)

Which leads me to conclude that the brazillian men who go do dance clubs are not shorter than average. (Unless of course volcanoes really do cause temeperatures to increase more than they would without them)