March 2012 Update

I call this the March 2012 Update because the last HadCRUT3 (red line) datapoint is for March 2012. The April 2012 figure has not yet been released. 

Despite this, there is enough data to plot the prediction (blue line) up to April 2012. The prediction for April 2012 is about 0.4C and I suspect the HadCRUT3 April figure will come in at about that level too, so the recent divergence will, at least for a while, be closed.

After that the blue prediction line has an upturn in store - the increase in ENSO MEI (purple line) in the last few months is due to push the blue prediction line about 0.1C higher in coming months to about 0.5C. Whether the red HadCRUT3 line also continues rising to 0.5C levels remains to be seen.

After that the probable onset of an El Nino this year is likely to push the MEI index even higher, which would push up the blue prediction line further, probably breaching the 0.6C mark this year, maybe even 0.7C. Whether HadCRUT3 follows suit will be a big test of the hypothesis underlying this analysis.

The solar contribution (orange line) has been slightly negative in recent months, although it's contributed about 0.1C to the prediction line in the last 2 years.

December 2011 update

Diversion Continues

I think the diversion has now peaked. Apart from a few months during the Pinatubo time period the last few months show the largest difference between HadCRUT3 and the blue predicted line on the whole graph.

Why? I put it down to two factors.

1) HadCRUT3 has responded to the current La Nina more strongly than the analysis assumed it would. The current La Nina isn't as deep as the last according to MEI (purple line), but HadCRUT3 seems to have fallen about as low as it was during the last La Nina. Perhaps there's a lag effect in there? Eg the 2nd La Nina gets a boost from the first one?

2) The rise in the solar cycle (orange line) has not impacted global temperature yet. It was perhaps a bad assumption to not put a lag in there.

In either case I don't expect the blue prediction line to have monthly precision. Even 12 month precision is pushing it.

I still expect the red line and blue line will come back together shortly. There might be a longterm drift but I doubt there will be a sudden and sharp departure.

It is interesting to note though that to come back together the red line has to shoot upwards. The blue line isn't coming down much more. All the components that make up the blue line are heading upward from here. I think MEI has probably bottomed out and is about to start rising and solar is going to head higher if anything.

So really if we take the graph as a prediction it is predicting a very substantial jump in HadCRUT3 over roughly the next year. Of course it will jump as the La Nina fades, but this graph is expecting it to jump *a lot* more than typical for leaving a La Nina.

Note:
The NASA Goddard Institute for Space Studies has recently released a report for global temperature in 2011:
http://data.giss.nasa.gov/gistemp/2011/

This is of note because they discuss the impact of the solar cycle on global temperature noting:
"Because of the ocean's thermal inertia, global temperature change caused by solar variability lags solar irradiance by about 18 months. Thus the influence of the sun in 2011 continued to be a cooling effect. However, the sun's influence will change rapidly to a warming effect over the next 3-5 years."

Foster & Rahmstorf paper

It's like the analysis done here, but done properly rather than eyeballing the fits.
http://tamino.wordpress.com/2011/12/06/the-real-global-warming-signal

Here's the image of various temperature records with ENSO, solar cycle and volcanic influences removed:

September 2011 Update

Large Diversion

The blue prediction line has shot upwards to about 0.52C (note: the blue line above, due to error on my part, actually ends in October 2011 at 0.55C, but the September value is about 0.52C)

Actual hadCRUT3 temperatures seem to have topped out at 0.45C and are now falling (currently about 0.37C). So this month the blue line is about 0.15C higher than HadCRUT. That's not the largest monthly difference across the whole record but it is unusually large and given we already know next month the blue line will reach 0.55C and HadCRUT could fall further due to La Nina conditions, this divergence could grow - or more importantly, in my opinion, linger.

Of course even two months are not particularly significant as this analysis only aims for predicting longer term changes, but on the other-hand the blue line is unlikely to drop as low as 0.35C in coming months and supposedly HadCRUT3 red line will fall even further below 0.35C. It certainly will not reach the highs the blue line predicted for September and October 2011.

So what's happened? Last month the blue line was slightly below the red one. How has it diverged so much in an almost hopeless way?

Well the MEI ENSO increase back a few months ago has finally kicked in on the blue chart. Remember it's delayed about 3 months. So the blue line got pushed up in September by about 0.12C. Meanwhile HadCRUT fell about 0.04C as La Nina conditions cooled the oceans. Another contributor is the solar cycle. In the last 3 months it has contributed about 0.07C warming to the blue line. The solar cycle fit in this whole analysis is a lot more sketchy than the ENSO fit so it is possible that this solar increase is being pegged as contributing slightly too soon.

So what will happen? The blue line looks set to fall to about 0.4C into the beginning of next year, yet HadCRUT looks set to keep falling as ENSO keeps falling.

So really this is crunch time. Will something unusual happen to vindicate the assumptions behind the analysis - such as HadCRUT refusing to get any lower or bouncing back up rapidly? Or will the blue line stay far above the red line for many months, perhaps creating a very unusual or even unprecedented divergence calling the assumptions behind the analysis into question?

The ENSO relationship is unlikely to be in doubt. It's either the SSN or the green warming trend that are wrong if the blue line does diverge from the red line for a long time. It'll be interesting to see what happens. I notice across the graph there aren't really many examples of the blue line rising far above the red line (barring the Pinatubo era which I don't count).

August 2011 Update

Skipped a month, but still the update doesn't bear much discussion.

The MEI rise from 3 months ago is finally starting to bump up the blue prediction line. It will continue upwards to just above 0.5C in the next 3 months, although so far HadCRUT3 has maxed out around 0.46C. Both the blue prediction line and HadCRUT will start dropping from those levels before the end of the year as MEI has started dropping quite sharply.

June 2011 Update

HadCRUT jumps back up above prediction line


Ideally I would post updates every 6 months or every year, but it's hard to not post the update every month even though so little changes. The MEI (enso) index is up a little in June - about 0.02C worth.

Projecting forwards 3 months

If you were to zoom in on the above graph you would see the MEI purple line goes a few months beyond the blue prediction line and red hadcrut3 line. This is because the purple line is actually MEI index shifted forward (roughly) 3 months. That very recent uptick in MEI still has a few months until it will impact the blue line.

But why wait? You can see without zooming in that the green "trend" line and gray "volcanic forcing" line go forwards years. So lets project the blue line out 3 months ahead of hadcrut. The only missing data is the sunspot count, but I will just assume the sunspot count will not change much (at all) in the next 3 months.

Here in cyan is the next 3 months of the blue line assuming a constant ssn count for the next 3 months:



This means projected HadCRUT anomalies in excess of 0.45C in a few months time. Sunspot count and MEI is likely to go a bit higher yet, so the blue line will probably reach a maximum of about 0.5C. Then I expect a slide into another La Nina which means MEI will come back down again.

But there should be a long enough period of the blue line being at around the 0.45C range that the actual HadCRUT (red) anomalies should have ample time to fluctuate around that approximate range. That means, barring a fast drop in MEI in the next two months, I expect HadCRUT anomalies for September-December to average about 0.45C. That will be quite high for ENSO neutral conditions!

BTW there is a good monthly-updated expert look at future ENSO by Dr Klaus Wolter at the bottom of the page at this link under the heading Discussion and comparison of recent conditions with historic La Niña events.

The green line in the graph

What does the green line represent?

It is an approximation of the background warming trend. By background I mean the warming trend once noise such as the solar cycle and ENSO have been factored out.

Why not split the green line out into CO2, aerosols, etc?

I am not trying to attribute the recent warming, only to observe whether the background warming has continued and think about where global temperature is likely to go.

Why is the green line curved rather than linear?

Because it worked. I wanted to challenge the idea that global warming has stopped recently and demonstrating that HadCRUT3 is not only consistent with a continuing background warming trend, but even an accelerating background warming trend is a strong argument.

How did you determine the curve of the green line?

I eyeballed a curve that sorta-fitted the past data. Now the thing is I am not confident the curve is correct. In fact I expect I will have to revise it at some point. I would not be surprised at all if with more data coming in the curve turns out to be too steep or shallow, or it might turn out that in the end a linear trend is better.

Really I should have a range of possibilities for the background warming trend, but that would be too much work to update the graph each month. What I will do if the existing blue prediction line starts to diverge is continue to plot it, but add a new background warming trend line (perhaps a lighter green) and a new prediction line (a lighter blue) that fits better.

Why expect the background temperature trend to be a perfect curve or line?

I'm only using it because it works so far.