I firmly believe CO2 will tend to cause global warming and the effect is sufficiently large to worry about. That said, I’m also interested in reading the questions and arguments presented by skeptics. Many of these questions can be found in the unthreaded posts at Climate Audit.
Not being an expert in all things, I sometimes read recurring questions that I cannot answer. It may well be that experts can, or have, answered the questions, but the fact that I cannot do so prods me to seek out the answers were they exist.
One big question asked by those who actually ask questions is:
Have any GMCs been demonstrated to predict Annual Mean Global Temperature (AMGT)?
This is a rather narrow question and should be tractable. Logically, to answer this question, one must:
- find older papers that predicted future values for Annual Mean Global Temperature (AMGT.)
- find data collected after the predictions were made and finally,
- compare the predictions to the data, and using all information available, decide whether the agreement is “good” or “bad”.
So…. has the comparison been made?
Not being a climate modeler, today, Dec 10, 2008, my answer is “I, Lucia, don’t know.”
I did a bit of hunting, and found two papers well worth reading. I’m going to discuss them a bit over the course of the next few days.
The papers I will be discussing are are Hansen et. al 1988 and Hansen et al. 2006.
Based on questions in my mind, which revolve solely around empirical testing of GCM predictions for Annual Mean Global Temperature, these features struck me:
- Hansen et. al 1988 describes the results of 3 transient simulations.
- Hansen et. al 1988 describes the results of a 100 year control run holding CO2 at 1958 levels.
- Hansen et al. 2006 compares the predictions in Hasen et. al 1988 to AMGT data collected from 1988 to 2006.
So, it appears that this pair of papers provides the basis for determining whether the predictions of a particular GCM agreed with data collected after predictions were made.
I will be posting reflections describing what I can conclude based on these two papers. Believe it or not, I am planning 3 papers posts, one for each bullet above.
As I am not an expert, my musing may be idiotic, but after all, this is a blog. Should anyone who knows more arrive, I will be quiz them for more information that might help me in my quest to give more complete answers to questions regarding the empirical testing of GCM predictions.
Interesting paper, although if I could choose which of my predictions to test myself, I reckon I’d do pretty well also 🙂
Ignoring the detail of the paper (hey that would take time) my observations, apart from the ex post prediction selection, would be:
1. Fig 2 is the one everyone talks about, although I suspect there is more in the rest of the paper, but that is what I’m going to talk about now…
2. In Fig 2, draw a line from 1987/8 station data (just below .5 deg) out to the right at the edge of the graph.
3. Consider the volume between this horizontal line and the top prediction (scenario A). We’ll call this “the Zone”. If the (highly autocorrelated) temperature ended up anywhere in the zone, Hansen would claim a success, irrespective of which scenario was most closely followed (because he can always argue no scenarios are a perfect emissions/event match).
4. Had the temperature curve gone above the Zone, we would have heard claims of “OMG! It’s worse than I thought!!!!11!”
5. Had the temperature curve gone below the Zone, his prediction would have been falsified (and that study would never have been published).
So my claim is that with respect to figure 2, the chances of Hansen being right or wrong were about 50/50; if temps went up at all, he could claim a victory, if they went up a lot he would have gone into hyperbole central. If they went down, he would have lost. Basically, he tossed a coin (actually he probably tossed several coins as he could pick and choose which prediction he went with) and called heads.
What is nice now though, is that he has chosen that prediction as his standard, so we can watch for the next 5-10 years for a genuine “out-of-sample” test.
Interestingly, if you plot 2006 and 2007 GISS+dSST they both click beneath scenario C. Only a couple of years of course, not exactly significant. I look forward to his 2015 paper where he documents how things went from there… I suspect it is “heads I win, tails you lose” though – if his prediction diverges from reality they will just claim to have “moved on” with better models today (and therefore more trustworthy! Of course!)
Spence UK. Have you looked at figure 1? 🙂
I can see you’re trying to persuade me to read the papers here!
Start at the beginning…
(I assume you refer to fig 1 of the 1988 paper. Fig 1 of the 2006 paper seemed to include little relating to GCMs)
My observation above really relates to the statistical aspects of the prediction, rather than the detailed technical aspects on how the GCMs perform. I suspect any analysis I attempt carry out in that regard would be considerably worse than yours, judging by your contributions on CA. I will continue to read with interest.
Yep. I mean figure 1 of the 1988 paper. I’ve posted my discussion of that now.
What would happen if CO2 stayed at 1958 levels?
I suspect everyone’s analysis will be different.
However, though Hansen leaves out information from figure 1 (1988) when discussing relative agreement or disagreement bewteen data and the prediction on figure 2 in 2006, I think it’s important to remember figure 1 exists.
So… by what criteria does figure 1 disagree with the data? And also, what would we have expected to happen if CO2 were at 1958 levels (as in figure 1?)
Hint: Figure 1 (1988) says temperatures would have been up quite a bit even if we hadn’t added CO2.
OK I started to read a few bits of the paper. Couching fig 1 in terms of natural, “unforced variability” is interesting. As you note, the natural variability is such that the warming of 88-05 could be entirely natural. This doesn’t allow us to reject the models of course; that would be denying the consequent (hey, then we’d be real denialists), but it shows there is little support for the models from this paper.
Interesting to look into the 1988 paper and stumble across terms unlikely to be seen in a contemporary climate science paper. A nod by Hansen to the issue of large scale unpredictability (even citing Ed Lorenz). Quite a significant admission; shame they didn’t explore that aspect a bit further, before the political “don’t discuss the uncertainties” edict was imposed.
Yep. I’ll be writing more and talking about figure 3 (1988) and figure 2 (2006) and some numbers discussed in the narrative.
Before I write up the last bit, I need to get the temperature number for 2006. 2007 isn’t over, but we should have those numbers soon.