HadCrut is in! The January temperature anomaly was 0.370C up 0.303C in December. Clearly, January was warmer than December.
Now I could do the usual and show how the trends. However, I’ve noticed new people often pop in and suggest the models would look good if only I examined the temperatures in a different way, using different baselines etc. So, I thought I’d pour the number into a different analysis. Today, I’ll show the anomalies and the difference between the multimodel mean anomaly from a projection and the current anomaly.
Century Long Differences
As you recall, when we were discussing the differences century long trend in the difference between HadCrut and GISSTemp, BarryW asked me to subtract Hadley from GISSTemp. After I did that, I thought I’d subtract the model-mean temperature anomaly from each of the two observational data sets. Since I was focusing on long series, I decided to re-baseline everything to 1900-1999 inclusive. (I usually use the baseline from 1980-1999.)
The following figure compares the the 12 month lagging average of observations and simulations compare using the century long baseline:
Notice that using this baseline, the earth’s temperature anomaly as observed by HadCrut and GISSTemp is currently near the lower range of the model mean temperature anomalies. So, while I have generally focused on trends (which the models overproject in the recent period), if we use 1900-1999 as a baseline, the multi-model mean anomaly projection also lies above the earth’s temperature anomaly.
Notice that using this particular baseline, the 1998 El Nino event doesn’t pop-out of the range of model projections. (Bear in mind, the AR4 projections are relative to 1980-1999, so things will look different using that baseline. Still . . . )
Difference between projections and simulations
When giving figure 1 the old eyeball, it might seem that the difference between the 12 month lagging average of projections and models hit a maximum recently. To check, I subtracted:
- The 30 year average of the model mean and observations are forced to match during the baseline period from Jan 1900-Dec 1999.
- The model mean currently exceeds observed temperatures.
- Based on 12 month averages, maximum over prediction occurred in October, 2007. (That is, the period based on Oct 2007- Sept 2008). For this month, the 12 month average model anomaly overshot HadCrut by 0.3511 C.
- The maximum under-prediction occurred in Sept. 1943. This is about 26 1/4 years from the baseline period. On this month, the 12 month average model anomaly undershot HadCrut by 0.3645C
What does this mean?
Well, if the model-mean is assumed to correctly predict the anomalies relative to the 1900-1999 baseline, then earth’s weather was pretty far off track quite recently. Based on a 12 month average, the multi-model mean never exceeded the observation by this much during the 1900-1999 hind cast period.
However, the multi-model mean have been that far off on the low side. It happened way back in 1943. Rumor has it the sea surface record has data quality issue related to war activities during that period.
Recent Trends
But of course, you all do want to see the trends, right? Here you go:
(Bear in mind, this uses the 1900-1999 baseline. The agreement won’t like quite so bad if I use the 1980-1999 baseline. I’ll be showing that more often.)
As for the trends: Yes, observed trends since 2001 remain negative. Admittedly, this graph shows the least squares trends computed using annual averages computed monthly, which is not the best choice. I usually show trends based on monthly values later on. They remain negative.
Lucia,
Wouldn’t it be more accurate to say that the observed trends since 2001 remain indistinguishable from zero? I mean, I could say that the observed trends from 2000 remain positive, but they are also not significantly different from zero. I think that people tend to put a bit too much stock in “warming” vs “cooling” even though all we can confidently say is that temperatures have been roughly static for the past eight years or so.
Per the above comment,
If I calculate correctly, the HadCRUt trends are:
Jan 2000-Jan 2009 = 0.0135 +/- [edit: oops, wrong residuals! Correcting now]
Jan 2001-Jan 2009 = -0.1125 +/- [edit: oops, wrong residuals! Correcting now]
Granted, I’m probably not correcting for autocorrelation and who knows how many other things here, so excuse any statistical ignorance on my part.
They are indistinguishable from zero but, if you start from 2001, they are distinguishable from the IPCC projections. One can say either thing.
I get slightly different trends depending on whether it’s monthly data, annual average data, and the precise choice of baseline.
These are annual average, but computed monthly. It’s a bit of a weird thing to use.
I’ll post the values based on monthly tomorrow. I can do both from 2000 and 2001.
Zeke – It actually would probably be even more accurate to say that the calculated trends are still slightly negative but are statistically indistinguish from 0. But that gets to be a mouthful so you could just be like me and say that the temp trend has remained flat since 2001.
Since we are discussing accuracy of statements, it is important that the temp trends do not indicate continued warming or accelerated warming, though certain recent news releases would want you to believe that. Make no doubt, in my mind, anthropogenic effects on the climate are real, but the magnitude of those effects and what it means for our future are very much still an open question in my mind at least.
Something I’ve wondered about the graphs you’ve posted, Lucia, is that the squiggles in the model mean (and CM4) always seem to be 180 deg out of phase with the squiggles in the GMST.
.
No clue what it means (if anything). I just thought it was interesting.
In fig 1 , are we not seeing that the carefully chosen parameters that fit the data before, no longer constrain the output to follow reality?
I have said it before that the tools in the GCM s are essentially first order approximations on all fronts of very divergent nonlinear solutions of the coupled differential equations that apply to the problem. It is inevitable that when the time stepping goes beyond the forced fit to preexisting data, that the divergences from the true solutions will appear. The reason that the divergences are not all over the place but are biased for warmth must be, I guess, that the warmth bias is programmed in.
RyanO–
The squiggles being out of phase wouldn’t mean anything. The models don’t predict weather, and from the point of view of “weather” individual realizations aren’t initialized with the same weather state. So, we don’t expect el ninos, pdos etc. to happen at the same time. The squiggles themselves almost certainly mean we need to average over more runs or more models.
Lucia,
It seems that the only way to make the anamolies significant at all, extreme increases in the temp axis is required. Additionally, it seems non-credible that accuracies of 0.Xo C can be achieved when accuracy is more on the order X.0oC. I wonder if the graphs actually have any meaning. I do not write this to disparage your work in any way, but rather to call into question how this entire issue has been perceived.
Apologies for being a bit off topic and for repeating my thoughts from another thread but …
One of the traces in particular is oscillating very strongly. I think it is cnrm_cm3, but for refefernce it is has the highest temperature at the start of 2001 on the 2001-2009 plot.
It then drops about 0.5C in less than 2 yrs before heading North again. Visually it has the same oscillation (2-3yr period) in the 1900-2009 plot.
This is not Earth-like, but is of interest. It’s wanderings appear to be more harmonic than chaotic but I will have to look at it in detail to get a better feel.
Now when you see something like this you can:
a) Have a giggle,
b) Ponder what it is trying to tell us.
An instant reaction would be that it under-damped (ringing like a bell) due to insufficient conductance (too low a value for CS or possibly too little coupling to the ocean).
Anyway I think the I will but the CMIP people on my Christmas list. This is wonderful stuff and I suspect a bit cringe-making for some of the modellers.
Alex