Yesterday I tried to find out from Richard Betts of the UK Met Office at what point if I was confident of my forecast that I should make it known if I was predicting cooling. He didn’t reply but after thinking about it, I’ve decided that I would criticise an academic for keeping quiet. So whilst I am very hesitant to do so, I’m going to issue my first and very provisional estimate of global temperature out to 2030.
But first an apology is due to Girma Orssengo. When I first saw suggestions of a 60-70 year cycle in the global temperature in WattsUpWithThat, I was extremely dismissive adding comments suggesting that sceptics were falling for the same mistake as alarmists and mistaking natural variation for patterns.
After further of the variation of climate trends with time, I’m having to rethink that criticism. For around a week since I wrote that article I’ve been pouring over the graphs and trying to understand how the information should best be interpreted. Whilst I am still very uncertain of what is best, each time I look at it I’m coming to the same conclusion and that is that I cannot simply dismiss the 60 year cycle which Girma Orssengo identified. But neither can I say for certain it is real or has predictive power.
However, it is also clear from my analysis that this cycle is not present earlier in the Central England Temperature record (above). Therefore it is not a fixed cycle that is always present and which is waht a quick reading of Girma’s article might imply. Instead it appears to be akin to a “resonance” in the climate.
A”resonance” responding to other variations, would tend to have a cycle which fluctuates. So we would expect to see patterns of a similar length occurring for a few cycles and then perhaps disappearing, only to reoccur later on. The Atlantic Multi-decadal Oscillation is a very good candidate for this “resonance”. As Roger Tallbloke as suggested, the two may well be linked. However the record is too short to discern much.
I had decided that I was probably over-interpreting this apparent cycle, and without any long term climate cycle I was doubtful and thought the correlation of global temperature with the Atlantic Multi-decadal Oscillation may be either a coincidence or that it may not be cyclic.
Then Tony at StevenGoddard showed an intriguing graph of Californian droughts shown below. Here I can see six distinct troughs from 1600 to 2000. 400/6 is 67 years which fits nicely with Girma’s work.
Now that I’ve seen this cycle express itself elsewhere in the climate I am on balance going to say that it is more likely than not that something like a 60-70 year “resonance” exists in the climate.
With low sunspots at the moment no discussion of what may happen would be complete without mentioning sunspots. To a large extent, these should just be part of the natural variation I have already included. And just as 1/f natural variation is easily misinterpreted, so many people are both blinded by the sunspots into believing they have a massive effect. Others are similarly blinded by their closed minds and/or politics into believing nothing exists.
My view is very much shown above right. Sunspots do closely match temperature over much of the available record, but they clearly don’t account for everything we see as the post 1980 mismatch in the graph above shows. I think it fairly safe to say, is that when we have lower solar activity as seen by sun-spots, then we are more likely to see cooling than warming.
However, that does not help me much because not only do I not know if sunspots might cause more cooling than expected, they might also rise and if low sunspots cause cooling, then rising sunspots might cause warming.
So, in effect I’m going to add a hefty 1.5σ (0.3C) additional error to my analysis in the hope it will cover what I don’t know about the effects of sunspots on climate.
There is a huge amount of thinking behind this prediction which I’m not even going to try to reproduce largely because some of it is very subjective. I’m having to make judgements using techniques which are not only new to me, but apparently new to everyone else. The biggest concern I have is concerning whether I can reasonably say that the 60-70 year period discussed by Girma is a real cycle or just a random manifestation of climate noise. However, even if it is entirely random, it will still have certain properties and those strongly suggest cooling.
Possible human caused warming does affect my calculation, but it is only a small factor both increasing the uncertainty attributing recent temperature rises to the 60-70 year cycle and also itself adding a potential increase to temperature. In making this calculate I’ve made no additional assumption except (like Girma) that past trends will continue.
Girma’s cycle suggests a cooling of temperature of about 0.4C from the maximum around 2000 with a minimum around 2030. My own analysis works in a different way and is based on there being no cycle, but suggests that the rapid warming peaked in 1990 and that we are likely to see no net change in temperature from 1990-2030. That means a prediction of around 0.15C (HADCRUT4) which is 0.31C colder than the 0.46C HADCRUT4 average from 2000-2009 (0.34 colder than the 2005-2014 period).
I don’t like being so close to Girma, I feel I should be predicting less cooling than he does. But this is my best prediction of what would happen if all we had was natural variation and any underlying trend (whether human or not). My inclination is to reduce the scale of cooling. But to offset this is that the current low level of sunspots makes it more likely we will see a drop in temperature. So I’m going to stick at 0.34 whilst restating that this is yet another reason to have big error bars.
Errors are a tricky thing here. I probably should add variances but I know there’s a lot I don’t know. So to keep them big I’m going to just add them together and see whether it looks sensible.
There are three main errors I know. The first is simply how uncertain I am of the reading from the graph which is ±0.2C. The second is an estimation of what I don’t know about sunspots and whether or not climate has a 60-70 year cycle. I guess that to be ±0.3C. The third is easy to calculate which is the actual predicted variation in that trend.
From this my best estimate is that we will see a 0.35 drop in temperature from the 2000s till 2030 (-0.05C 1990-2030). This gives the likely change in the next 15 years as:
Predicted change = -0.35 ± 0.7C (using 1σ trend)
Predicted change = -0.35 ± 1.1C (using 3σ trend)
These are not strictly “3σ” but instead estimates using a 3σ trend error. The ratio is around 2:3 so I’m going to call these “moderate” confidence for which 65% seems appropriate and “high” confidence by which I mean 95%.
This compares with an IPCC prediction of 1.4 to 5.8C from 1990 to 2000. So, the final initial table comparing the two forecasts is follows:
|This analysis||-0.75C||-0.05C||0.65C||Moderate (65%)|
First a month ago I would have said it was insane to try to predict natural variation. But assuming I am still sane, I don’t see that a slight cooling from 1990 is in anyway extreme. Cooling or warming of 1C over the next 30 years would be by far the largest cooling seen. Perhaps the confidence should be “very high” (99%). However I can say that if I’m alive at that point, I will have to admit I’m likely wrong if it has warmed by more than 0.6C
And all I can add is that I cannot do better than that at the moment.