[btilly]

First of all a note. I've made a substantial fraction of my living over the last decade from my understanding of statistics. When I look at a projected statistical range of 1.0-3.0 that later got refined to 0.9-2.0 I see that as a fit. Sure, the bottom end of the new range moved out of the bottom end of the old range. But if you watch an A/B test run, you'll see that this is entirely expected. But the median prediction of the new range - the most likely outcome - is 1.55 which is (assuming that the original range was a 95% confidence interval) is inside of 1 standard deviation of the prediction.

Secondly when you say that the warming is insufficient to create the run-away warming predicted in climate models you're in disagreement with the vast majority of people who have actually tried to run the numbers. Having just seen you draw an incorrect "not a fit to the statistics" from something that I know very well looks exactly in line with what I'd expect a fit to look like, I'm going to trust that scientists understand their own numbers better than you understand them.

Thirdly your claim that the new historical record has 10-20 years of no statistically significant warming is just plain false. The article this discussion started about finds that if you just use data from the last decade and project that forward you get an average projection of increasing 1.55C in a period that previous models had said would increase 1-3C. That doesn't look to me like you're not warming.

And finally I'm glad that scientists don't let their models sit still for decades. It is a fact that the models have huge error bars. I want them to improve the models, to bring them down. And the fact that the new models are in good statistical agreement with the old is confirmation that the old models were reasonable (if less accurate than desired). Until we see a statistical lack of fit between old and new data (which has yet to happen) - there is no statistical reason to doubt the science.

In the meantime I'm concerned that the 10 years with the least arctic ice in the summer all happened in the last 10 years. You may dismiss that data point. But in all of the discussions about newly available oil drilling locations and transport routes, it is worth noting that it is a very visible sign of a major global phenomena.

BTW if you want to dig farther, I recommend http://onlinelibrary.wiley.com/doi/10.1029/2011GL048794/abst... for a detailed energy budget of where heat appears to be going right now.

[glenra]

> The article this discussion started about finds that if you just use data from the last decade and project that forward you get an average projection of increasing 1.55C in a period that previous models had said would increase 1-3C

No, for some definitions of "the last decade" the trend line is cooling and for others it really is flat.

If you define "the last decade" as the ten year period ending this month, the temperature trend looks like this (slight cooling):

http://www.woodfortrees.org/plot/wti/last:120/plot/wti/last:...

(The "woodfortrees index" shown is built from an average of several standard temperature series - if you like HADCRU or GISS or some other specific one you can select it from the popup menu and hit the "plot" button to see that instead.)

The 15-year trend is rising, but just barely so: http://www.woodfortrees.org/plot/wti/last:180/plot/wti/last:...

The 20-year trend is still positive, but the recent flattening is real and it has already been flat enough long enough that it's starting to pose a serious problem for the model predictions, hence articles like this one.

> But the median prediction of the new range - the most likely outcome - is 1.55 which is (assuming that the original range was a 95% confidence interval) is inside of 1 standard deviation of the prediction.

Er, no. You're assuming the probability distribution is a normal distribution with the median in the middle - it isn't. IIRC, some of the newer attribution-based papers that have been forcing them to shift the window to the left have a positive skew - the median peak is way on the left side and then there's a "long tail" on the right. So depending on which papers they use it's actually possible the new median could be outside the 95% confidence interval of the old range.

[btilly]

Interestingly when I went looking I came up with http://www.guardian.co.uk/commentisfree/2008/jun/07/climatec... which is a prediction from 5 years ago citing two studies saying that there would be several cooling years ahead.

But more fundamentally, the natural variation over a decade is +- 0.2 C. If the warming trend for a century is 2 C, then the warming trend per decade is also 0.2 C. Thus a flat decade is readily explainable by normal variation. If the warming trend is instead 1.5 C per century, then a 15 year stall becomes even more reasonable than a decade had been under a faster rise.

The golden question then becomes how this study found a 0.15 C rise in temperature from data over the last decade. I have not read the study. But I know that correctly modeling these things is far more complex than just looking at a long-term average. And given that I know that detailed attempts to produce actual heat budgets for where energy is going have consistently found that the planet has been steadily absorbing large amounts of heat in recent years, I'm inclined to believe their figures over an argument from a temporary stall in global atmospheric temperature.

[glenra]

> The golden question then becomes how this study found a 0.15 C rise in temperature from data over the last decade.

In my skimming of the original article I didn't see where that claim was presented. But if I had to guess, there are at least three ways somebody could accidentally reach that wrong conclusion.

(1) Using something like the woodfortrees link, specify a "From date" year but don't specify an "end date", so you plot, say, from 2002 to "today". The leftmost point on your plot is in January; the rightmost point is in the middle of summer because that's when it is now. Boom, you've got some instant extra warming, assuming you picked a temperature trend influenced by surface thermometer readings.

(2) Instead of looking at actual temperatures, look at a heavily smoothed moving average of temperatures. Or average every decade into a single point and then compare those points. This gives you a plausible excuse to ignore much of the most recent data and the most recent trend ("Tamino" aka Grant Foster often pulls this trick on the readers of his blog Open Mind.)

(3) Instead of looking at the most recent data, google up an old study that ended in, say, 2000 and interpret all talk about "the last decade" as referring to the last ten years shown in that study. :-)

UPDATE: I just thought of another:

(4) Timing. The annual temperature trend is pretty noisy, so over any given ten-year period it might increase or decrease. If you're a "warmist", your favored information sources are going to publish new studies and trumpet their findings whenever the most recent decade now seems to show a big jump compared to the trend it seemed to show in prior years, with headlines like "it's worse than we thought!". So any time this issue comes up, those are the studies/stats that you'll remember and cite.

(the original Mac/PC debates had exactly this dynamic - both Mac fanboys and PC fanboys were generally convinced that their platform was better in all the ways that mattered; their certainty was almost entirely a matter of salience bias and the timing of news release events.)

[gwright]

> Secondly when you say that the warming is insufficient to create the run-away warming predicted in climate models you're in disagreement with the vast majority of people who have actually tried to run the numbers.

I think we are talking about two different things here. The ranges you mentioned (1.0 - 3.0) and (0.9 - 2.0) are predictions for the net change in temperature over 100 years when all mechanisms are factored in.

When I said the 'warming is insufficient' I'm talking only about one factor, the greenhouse effect caused by atmospheric CO2. This one factor has been the focus of most of the policy discussions related to global warming since it is the one most closely associated with human activities.

The CO2 greenhouse effect is a single factor that contributes to the ranges you described and by itself is insufficient to create the catastrophic warming represented by the upper end of those ranges or even the mild warming represented by the lower end of those ranges. All the models presume some sort of positive feedback that adds to the warming created by the CO2.

I'm not going to dispute your statement that the new error bars overlap with the old ones but the catostrophic scenarios are obviously associated more with one particular end of the error bars so that the shift in the error bars, while not invalidating the model in their entirety, suggest that the most severe scenarios are less and less likely and it is exactly those scenarios and the policy recommendations associated with them that has been most in dispute.