Arguments
Lessons from Past Climate Predictions: Syun-Ichi Akasofu
Posted on 7 July 2011 by dana1981
In 2009, Syun-Ichi Akasofu (geophysicist and director of the International Arctic Research Center at the University of Alaska-Fairbanks) released a paper which argued that the recent global warming is due to two factors: natural recovery from the Little Ice Age (LIA), and "the multi-decadal oscillation". The paper was trimmed down and published in 2010 by a new journal called Natural Science, which is published by Scientific Research Publishing, which has been called "what may be the world’s strangest collection of academic journals." Most of the following discussion focuses on the unpublished version of the paper.
Akasofu argues that the current warming trend of approximately 0.5°C per century actually began in the early 1800s, and that:
"This trend (0.5°C/100 years) should be subtracted from the temperature data during the last 100 years when estimating the manmade contribution to the present global warming trend. As a result, there is a possibility that only a small fraction of the present warming trend is attributable to the greenhouse effect resulting from human activities."
He also argues that the "multi-decadal oscillation" can explain some of the warming over the past 35 years:
"This particular natural change had a positive rate of change of about 0.15°C/10 years from about 1975 (positive from 1910 to 1940, negative from 1940 to 1975), and is thought by the IPCC to be a sure sign of the greenhouse effect of CO2. However, the positive trend from 1975 has stopped after 2000. One possibility of the halting is that after reaching a peak in 2000, the multi-decadal oscillation has begun to overwhelm the linear increase, causing the IPCC prediction to fail as early as the first decade of the 21st century."
Akasofu presents his argument graphically (Figure 1). Basically he argues that there is a linear warming trend caused by "recovery from the LIA", with natural oscillations superimposed upon it.
Figure 1: Akasofu's interpretation of the global warming trend and causes since 1880
In order for this to be a physically sound argument, Akasofu must explain the physical mechanism behind the "LIA recovery", and why this 0.5°C global warming trend continues to persist. What is the underlying cause? Surely a geophysicist will examine this question.
Akasofu's Physical and Logical Failure
Unfortunately, nowhere in the 55 pages of his unpublished paper does Akasofu examine the physical cause of his purported 0.5°C per century warming trend since 1825. Most of the paper is spent looking at various regional temperature proxies, as well as ice data, to show that the purported warming trend exists.
"Obviously, the LIA was caused by a natural change or changes. Further, the fact that an almost linear change of the temperature rise had been progressing until 2000 suggests that the linear change is a natural change, because the rapid increase of CO2 began only after 1946 (Figure 2c). The linear change began from 1800~1850, at least one hundred years before the rapid increase of CO2 in the atmosphere."
In the published version, Akasofu devotes a section to a discussion about galactic cosmic rays, but does not attempt to quantify their effect. In fact he begins the section by stating:
"It is not the purpose of this section to discuss any major causes of climate change."
Well that's a relief, I was afraid he was going to sneak some science in there! Instead, Akasofu appears to assume that the planet will naturally revert back to its previous state after a significant climate change as in the LIA. However, Rea et al. (2010) found that the climate does not behave in this manner:
"we present a new way of looking at long memory in these reconstructions and proxies, which gives support to them being described by the non-stationary models. The implications for climatic change are that the temperature time series are not mean reverting. There is no evidence to support the idea that the observed rise in global temperatures are a natural fluctuation which will reverse in the near future."
Not only does Akasofu fail to examine the physical causes of the warming since 1825, but he also fails to consider the possibility that a number of different factors are at play. For example, increased solar activity and low volcanic activity (and even human greenhouse gas emissions) contributed to the early 20th Century warming, but solar and volcanic activity have not contributed significantly to the warming since the mid-20th Century. It is a logical failure to assume that a warming over nearly two centuries must have the same physical cause throughout the 200 years, and this argument is contradicted by the observational data (i.e. increasing solar activity in the early 1900s, but no increase since mid-century).
Unsupported Linear Warming Assumption
A further failure of Akasofu's analysis is that while the linear warming trend over the past two centuries is approximately 0.5°C per century, nearly all of that warming has occurred over the past 100 years, as Akasofu's own Figure 7a shows (Figure 2).
Figure 2: Various temperature reconstructions in Akasofu Figure 7a, showing that most of the temperature increase over the past 200 years has occurred since 1900.
Akasofu also fails to justify his assumption of a linear warming trend over the past two centuries. A slight warming in the 1800s, followed by faster warming in the early 1900s, followed by even faster warming over the past few decades - sounds rather like an exponential trend, doesn't it? If you're going to fit a certain trend to the data, you first need a physical justification - what's the cause? Akasofu does not have this justification, and without a physical reason the choice of trend is essentially arbitrary.
Indeed, in a previous post, Riccardo fitted various trends to the global temperature data (Figure 3), and found that whether the residuals form a clear cycle depends on the choice of trend. In fact, the residual is more cyclical for an n=2 exponentlal fit than a linear fit (Figure 4). Therefore, Akasofu's argument for fitting the data with a trend plus multi-decadal oscillations makes more sense for an exponential trend.
Figure 3: HadCRUT3 monthly data (grey) and the fits for n=1 (red), 2 (green) and 4 (blue).
Figure 4: Residuals calculated with the Figure 3 trend curves shown with n=1 (red), 2 (green) and 4 (blue).
Thus we find that Akasofu's entire premise is faulty on many different levels: physical, logical, and statistical.
Natural Variability
Akasofu does finally address the issue of physical causality later in his paper, when he states that the multi-decadal oscillations such as the Pacific Decadal Oscillation (PDO) can explain the acceleration in global warming from approximately 1977 to 2000. While it's true that a positive PDO can explain some of the warming of surface temperatures, PDO and natural variability cannot explain either the ocean warming at the same time, or the long-term warming trend, whose cause Akasofu has still failed to address.
Akasofu's Prediction
Considering that Akasofu's paper almost entirely neglects physics and fails to address the causes of the observed warming trends, one might expect very little accuracy in his predictions of future temperature changes. However, although he does not explain its cause, Akasofu does assume that the 0.5°C per century warming trend discussed in his paper will continue into the future. So if the anthropogenic + natural warming trend is limited to 0.05°C per decade, Akasofu will have stumbled onto a correct prediction (Figure 5).
Figure 5: Akasofu vs. IPCC global temperature predictions
In order to evaluate the accuracy of Akasofu's prediction so far, we digitized Figure 5 and compared it to the Wood for Trees Index, which is a composite of the global temperature datasets compiled by NASA GISS, HadCRU, UAH, and RSS (Figure 6).
Figure 6: Akasofu's prediction from 2000 to 2015 (blue) vs. the Wood for Trees Index five year running average (red).
Akasofu's Results
As you can see, Akasofu predicted a very slight cooling (approximately 0.02°C) between 2000 and 2011, whereas the Wood for Trees Index has warmed approximately 0.1°C over that period. So his prediction has not been terribly inaccurate yet - there hasn't really been sufficient time to evaluate its accuracy.
However, given the expected atmospheric CO2 increase over the next century, in order for Akasofu's predicted 0.05°C per decade warming trend to hold true, climate sensitivity would have to be in the range of 0.5 to 1.5°C for doubled CO2, depending on how rapidly CO2 continues to increase. Coincidentally, this is approximately the same low climate sensitivty range that his fellow "skeptics" like Spencer, Lindzen, and Christy argue is accurate.
However, it's also important to note that as in Don Easterbrook's temperature predictions, Akasofu has completely ignored the warming effects of increasing atmospheric CO2 in his predictions, assuming that whatever caused the pre-industrial warming is also causing the current warming. Thus Akasofu is really arguing that the climate sensitivity to CO2 is effectively zero, and that the observed and projected warming is due to some other 'natural' effect which he has not identified, but describes as "LIA recovery". This is not a physical argument; Akasofu implies that extra heating from CO2 (which we've directly measured with satellites) isn't causing any warming - the heat somehow magically disappears. So even if his fellow "skeptics" are somehow correct about low climate sensitivity and Akasofu's prediction turns out to be accurate, it will simply be due to sheer luck.
Any temperature prediction like Easterbrook's and Akasofu's which totally ignores the warming effects of CO2 is fundamentally physically incorrect. Akasofu assumed a linear trend of unknown cause, an unknown periodic variability, and assumed that these two unknown phenomena will continue in the future, while disregarding what we know about the physics of the climate system. Unfortunately there's no knowledge to be gained from Akasofu's paper, except how not to predict future climate change.
NOTE: As you can see at the top right corner of this post, John Cook has created a snazzy new button for the Lessons from Past Climate Predictions series. We have also added some older posts which analyzed climate predictions, which you can see if you click on the button.
I notice that nobody has commented yet on my main point that it is perfectly reasonable for somebody to point out an apparent natural pattern, or cycle, without knowing or hypothesising a cause.
Were the early solar astronomers "solarstrologers" when they spotted the sun spot cycle?
If you ignore this point again I will assume you concede it.
Thanks Tom and MA for pointing out that the trend is roughly between the IPCC and Akasofu. Very diplomatic of it! My only remaining point in Akasofu's defence is that at least he predicted a falling rate of change in the short term whereas the IPCCs rate of change rises almost montonically (with wide error/noise bands) and exponentially.
In the long term I suspect the IPCC is going to be right that the change will be exponential (if maybe not as rapid) but what the Akafosu study suggests is that we may see fluctuations of a much larger extent, more regular and of longer duration than is output from the current crop of current climate models. However I think his saying that the fluctuations will be a regular sine wave is leaving himself a hostage to fortune almost as much as making precise predictions of future global temperatures.
It is important that papers like the one from Akasofu get published. There is an "apparent" cycle. Whether it is real or an artifact of noise needs to be studied and debated. I think his talents were very well employed!
I did it again! I must put Akasofu into my spell checker. Also in my second last paragraph the phrase in brackets should ideally have read "(if maybe not as rapid as they predict before the end of this century)". Yes, I do know what "exponentially" means :-)
Mathew L wrote:
You assume wrong. I suspect that there was so little intellectual meat on this particular bone that none of the regular commenters could be bothered responding.
Maybe you could flesh out your sunspots analogy with some detail. How many cycles were observed before the hypothesis of periodicity was raised? Were there alternative explanations available?
There is a vast literature on time series analysis, and there are statistical techniques for deciding whether an apparent pattern is sufficiently different from random noise to represent a worthy object of further study. As far as I can tell, Akafosu has not even paid lip-service to those tried and tested techniques. Merely pointing out a wobble in the noise that looks a bit like a couple of sine waves is not science.
Thanks for biting Leto! I thought that challenge was sharp enough to elicit a snap-back ;-)
Well sunspots had been observed long enough for the periodicity to be blindingly obvious before 1800. The actual reason was only discovered around 30 years ago and even today is not fully understood. I am not sure if there were alternative explanations put forward, but they would have been nonsense, of course, particularly as they did not even know what sunspots actually were for most of the time they were observed. Certainly it was possible to predict an approximate periodicity long before the cause was known.
The early stages of any scientific endeavour are about observation. Sometimes it is possible to make predictions based on observation without making a hypothesis. Validation by prediction is a vaild way to determine if a pattern is real before you even attempt to explain it.
Hypothesis is the second stage. It is not a "scientific" requirement to append an explanation to every set of observations. Surely it is better to say "no idea why this is happening" than to offer a pointless and arbitary explanation just because you feel that you won't get published unless you do.
Re your last paragraph, "pointing out the wobble" IS science, early stage and incomplete of course, but NOT pointing out the wobble, or dismissing it out of prejudice might mean missing something important.
Of course there are statistical methods to decide whether something is signal or noise. And it is open to anybody to do that analysis and refute Akasofu. That is what science is all about.
I would have a go, if I knew how. Perhaps you would care to?
Just a few points to throw into the topic.
To detect a cyclical pattern one needs more than 2 cycles. Otherwise all you can say for certain is that there were 2 events. In fact one can probably do a statistical confidence test for the existence of a cyclical pattern, just as one does when looking at other data and the level of confdence in what you find will be related to the length of the data set.
And if there is a cycle, it should have some sort of reasonably steady periodicity.
Next, when looking at the Surface Air Temperature Record (SAT) it is so easy to forget that it is the composite of changes from all over the globe. Does each point on the graph (or in a cycle) correspond to the same degree of change in all locations around the world? Or is it a mix of some change in one region and quite diffferent patterns of change in other regions?
GISS do a much better job than the other SAT teams at providing tools to allow people to explore the temperature record. They have one graph that doesn't get nearly as much attention as it deserves. This is a plot of temperature anomaly (as colours) plotted against latitude and time. So how temps have varied in each latitude band over time. It is sort of a 1/2 way house between the global SAT graphs we are all used to, and the maps of Temp anomaly that give a snapshot of the world at one instant.
Unfortunately with whatever troubles GISS have had with their server, this graph isn't available at present. Keep an eye out for it's return, it is worth it.
What it shows is that the 1930's/40's warming was partly ocean warming. However some of this was likely due to biases in the measurement of SST's from ships because of differences in how it was done and the impact that WWII had on the mix of nationalities taking those measurements - the latest Hadley SST data set attempts to correct for these biases.
The remainder of the warming was substantially Arctic warming, not global. Consider the following. We have had high warming in the Arctic recently; Sea ice is vanishing, and indications are that recent extreme weather events in the NH may be related to these changes through their impact on the Polar Jet Stream. During the 1930's there seems to have been Arctic warming. Sea ice areas were lower than in later decades (although nothing like what is happening now). And weather extremes may have been involved - the droughts that contributed to the dust-bowl years in the US for example. Perhaps a similar pattern to what we are seeing now, although now is more extreme.
So if there is a cycle, the data suggests it is more likely to be a cycle in the Arctic, rather than a cycle for the whole planet.
There is an important caveat to this. Station coverage. Until the 1920's there was no coverage in the high Arctic. Then, only slowly, did a few stations appear; this was driven by the USSR engaging with it's vast Siberian wilderness, adding stations across Siberia, and then WWII and the addition of stations across the Canadian Arctic & Greenland. The point is that at exactly the time that the physical observations suggest a warming was happening in the Arctic, the station coverage up there was in flux. I'm not talking about going from 50 stations to 100, more like 2 to 10. So biases due to fluctuating station coverage cannot be ruled out.
One possible cycle that has been suggested at times is a fluctuation between the Arctic & Antarctic, possibly with a 60 year cycle. However we hit another problem trying to look for this in the SAT data; decent coverage in the Antarctic lagged behind the Arctic. We didn't really get reasonable station coverage in the Antarctic until the 1950's, particularly the International Geophysical Year in 1957. So the evidence we might look for to compare the north and south polar regions isn't available.
If we then wish to look back before the 1920's, for a possible earlier cycle according to Akasofu, station coverage gets worse at higher latitudes - in fact in many regions. If there is a polar cycle, and we have no data on the polar regions from the 19th century, how can we tell whether there was or not.
My take home from that is that if one wishes to look for patterns in the SAT data going back before the 1950's, one needs to take these changes in station coverage very seriously in evaluating the past record. At a minimum, to do an apples vs apples comparison, one should only use data from regions where there is adequate station coverage over the entire period one is studying. Akasofu (and others who engage in similar analyses) aren't doing this! And that is ultimately mathturbation. Clever numbers games that shed very little light. Which is a pity because there may well be real insights to be gained from an appropriately careful study of the data.
That said, could there be a cycle that fluctuates between Northern and Southern hemispheres? Yes, quite plausible. Even the most casual observation of fluid mechanics shows that in seemingly chaotic fluid flows there are often 'beat' phenomena; quasi oscillations that have a rough periodicity - vortex spins off to the left, vortex spins off to the right, vortex spins off to the left ....
So too that is possible here. But the key point about why it is possible is that it would happen most easily due to a 'beat' in ocean currents; that is where most of the heat is, so small changes in the flow in these currents could have much bigger impacts on SATs.
The worlds oceans have 2 major axes. North/South along the Atlantic, and East/West across the Atlantic/Indian/South Pacific. The North Pacific is more contained because the Bering straight is tiny. So any 'beat' that influences currents in the Atlantic could easily have large regional climate impacts between the hemispheres.
Really the North Atlantic basin is about the worst region in the world for us to look for evidence of global changes due to how easily local climate can vary because of the shape of the ocean basins, even if globally it doesn't vary much. But the North Atlantic Basin looms large in our (western) imagination and psyche's - Vikings in Greenland anyone?
But what we have now, that we don't have for earlier periods, is data on the oceans as well. Not enough yet to see cycles, we need another century of that. But enough data to know that heat accumulation in the oceans - where 90% of the extra heat is going - is so large that there is no terrestrial heat source large enough to have supplied it.
Akasofu's cycles might exist, there just isn't enough data to say that with any confidence. But the underlying trend is clear, and definitely extra-terrestrial in origin. And when Akasofu (or others) make claims for these cycles, if one doesn't see an attempt to drill down into the data to remove time based biases, then I would call their analysis suspect at best.
Something has screwed with the Earth's energy balance with space in the last 1/2 century. No other conclusion is possible from the data! That ain't cycles.
MatthewL wrote "Of course there are statistical methods to decide whether something is signal or noise. And it is open to anybody to do that analysis and refute Akasofu. That is what science is all about."
Statistical methods are not straightforward to use in this case, as the forcings are not approximately constant on a centennial scale. Akasofu's theory seems to be based on inappropriate use of overly simplistic statistics, whilts ignoring the physics (for instance changes in solar forcing, aerosols etc., never mind GHGs). I am a statistician, and I would find a phsyical model (i.e. a GCM) far more convincing than a statistical argument based purely on correlations, without even adequate statistical hypothesis testing (against a null hypothesis that is more than just a straw man).
Secondly, promulgating a theory based on statistical models, with no quantifiable physical basis, and putting the onus on others to refute it is about as far from "what science is all about" as you can get. In reality, if you want to promulgate a theory, the onus is on you to demonstrate that your theory provides a more plausible explanation of the observations, or better still that the observations are inconsistent with previous theories. Akasofu has not done this, which is presumably why his paper was published in an obscure journal. What science is all about is actually self-skepticism.
Matthew L @51:
1) Akasofu does not predict a neat sine wave. In fact, he never, in all three of his papers on the subject I have read goes so far as to make a numerical prediction at all. Instead he is vague about the nature of the cycle, and appears to believe it varies arbitrarilly in cycle length and magnitude. Without specification, so vague a model is consistent with any short term cycle. If, however, you try to treat him as actually makiing falsifiable predictions, and model his theory, the resultant prediction is fairing poorly.
2) Although the observed trend in GMST since 2000 is being very democratic, lying only slightly closer to the IPCC precition of 2006 than Aklasofu's predicition of 2009, it is possible to notice the short term effects that lie outside either prediction. If you do, and remove those effects, the resulting trend (0.16 C per decade) is very close to the IPCC prediction. In other words, even though Akasofu had three more years data to work with when he made his prediction, he is still doing far worse than the IPCC.
3) The IPCC does not predict a monotonic increase in temperature. The IPCC shows a prediction for the multimodel mean, but notes that short term fluctuations are large relative to annual increase in the mean so that short term negative trends (in some cases as long as 20 years, form the models) are to be expected. The IPCC does not explicitly predict those short term trends because they are essentially unpredictable. It only predicts that the short term fluctuations will exist; which is why it states it projections in terms of the change in mean temperature between distinct 20 year intervals.
This may seem like nit-picking to you, but ignoring these facts seems to me like fudging the issue so that Akasofu's unjustified and very poor prediction can seem to be fairing relatively well compared to that of the IPCC. It is not. The best that can be said for it is that he has made it so vague that it is not yet falsified.
Thanks Glenn, excellent post. Have you (or others) posted on this in detail elsewhere? I would be very interested in reading more.
Have satellites added much to our understanding of these global temperature fluctuations? Are they yet showing up the kind of north-south "beat" you describe?
I think one can be too hard on Akasofu. It was just one paper and it was an interesting thing to do, even if it had the taint of "denial" about it. His problem was saying that an interesting point on climate fluctuations and cycles somehow puts AGW in doubt, which of course it does not. It got so much attention because it was seized on so vigorously by the WUWT crowd.
It will be interesting to see how the attribution for the flatter temperature trend to volcanoes, solar and ENSO develops. It may be that ENSO, PDO and the Atlantic Multidecadal Oscillation are somehow expressions of that N/S "beat". Now we have Argo and satellites covering the southern oceans hopefully that will become apparent.
MatthewL, the problem is that Akasofu's paper is not a scientifically interesting thing to do. All he has really done is to point out that temperatures have increased over the last hundred years or so with some wiggles on the top, which is something that is immediately apparent just by looking at the graph. What Akasofus has not done is provide plausible reasoning to support that the observed linear increase actually is caused by a recovery from the little ice age, rather than a change in solar and GHG forcing. Neither does he provide a plausible cause for the 35 year oscillation, that explains why is is not caused by known changes in the focings.
Thanks Tom #57, points taken. I have learnt a lot from this dialogue. Certainly more than I have from the Daily Mail!
Dikran #56. Did Akasofu go as far as promulgating a theory? I thought he was just pointing out an apparent cycle (observation) and projecting it forward? (Albeit in rather fudged and vague way as Tom points out).
At Uni I was always told that science went Observation > Hypothesis > Testing > Theory. I don't think Akasofu got anywhere much beyond a rather incomplete hypothesis.
MatthewL - yes, he posited that the linear trend is due to a recovery from the little ice age and hence there is a possibility that the warming due to GHGs had been over-estimated. He provides nothing to substantiate that.
He does point out that there is an apparent cycle; however, as I pointed out, this is immediately obvious to anyone who has looked at the graph, but there is plenty of evidence from looking at changes in the forcings that the cycle is probably spurious and the result of a number of separate changes in focings that appear to give a cycle.
Yes, "Observation > Hypothesis > Testing > Theory" is a common and reasonable model of the scientific process, however Akasafu has only done the Observation and hypothesis bit, but without testing and finding plausible physics, it doesn't justify publication, and is bad science when it is prematurely published.
Dikran #59
Well I found it interesting! Probably says more about me than the paper, probably because it chimed with my own statistical dabblings (mathturbation?).
I agree that simply saying it is recovery from the LIA without saying anything about the mechanism is not useful. I just ignored that bit and focused on the cycle on a trend bit. As I said earlier, what he does not point out is the accelleration in the rate of change, which is also obvious from the linear correlation charts.
I would dispute that a major ~60 year fluctuation can be dismissed as a "wiggle". I think it bears more serious investigation than that. Hopefully Glenn will be able to point me at some more informative research.
Dikran #61. Does every scientific paper have to go all the way from Observation to Theory before it bears publication? Sounds like a recipe for some pretty thin scientific journals to me.
MatthewL I didn't say that the 60 year fluctuation can be dismissed as a "wiggle", as it happens it is pretty well explained by known changes in forcings, as descirbed in the last IPCC WG1 report (you need both natural and anthropogenic forcings to explain the fluctuations). It has already been investigated, and while there is uncertainties involved, there is no good reason to invoke some unknown mechanism to explain them.
That doesn't mean that mainstream science does not have an open mind on the issue (ask e.g. Ray Pierrehumbert for example), it is just that the evidence for an oscillations is currently merely statistical, which is not very compelling.
MatthewL - it should go at least as far as testing, and should at least show some familiarity with the existing science that has a direct bearing on the question. In that case, Akasafu's paper fails on both counts, as far as I can see.
One of the problems with current scientific publishing is that far too much low quality science is published because of the "publish or perish" culture that pervades science. The vast majority of papers never achieve more than a handful of citations, often for good reasons. Thinner journals might not be a bad idea.
Two quick points. The first is an erratum. My graph of the "Akasofu prediction" @41 displaces the my reconstruction of Akasofu's prediction upwards by 0.04 C. This has no effect on the trends, but improves the fit with temperaure anomalies in the late twentieth century. For the record, the fit is established by ensuring the linear trend of the Akasofu reconstruction has the same mean as the observed data.
Second, coming from a background in philosophy and logic, I consider the usage of "hypothesis" and "theory" in science thoroughly inconsistent and confused. This is, firstly, because in practise scientists and historians of science do not reserve the term "theory" only for well established hypotheses, and in some cases continue to refer to hypotheses as hypotheses even once they are well established. What variouse theories are called is more a matter of historical accident than of consistent usage.
Further, it is because in logic (and in my usage) a "theory" is a set of propositions closed under logical implication. That is, the theory includes all propositions that are logical implications of any subset of the propositions in the theory. On that basis, all hypotheses are theories, assuming they are not logically inconsistent. Some hypotheses are well established, and would be better called "well established hypotheses" or "well confirmed hypotheses" (but not "well verified hypotheses") than ellevating "theory" to a usage that is inconstent with the use in philosophy and in common usage. Just my two cents worth.
Matthew L @54, in fact the sunspot cycle was not hypothesized until 1843, and that on the basis of observations of around one and a half solar cycles. Heinrich Schwabe reported that:
"From my earlier observations, which I have reported every year in this journal, it appears that there is a certain periodicity in the appearance of sunspots and this theory seems more and more probable from the results of this year."
There is a marked contrast between Schwabe's tentative "seems more and more probable" and Akasofu's rather definite "We learn that the recovery from the LIA has proceeded continuously, roughly in a linear manner, from 1800-1850 to the present".
Schwabe's conjecture did not attract widespread interest until 1850, when the number of observations had been extended to three solar cycles. Early interest lead Rudolf Wolf to not only continue the observations himself, but to reconstruct the cycle from historical observations back to 1745, a task completed in 1868. That means that before the hypothesis of sunspot cycles was accepted, there existed known observations of a total of around eleven cycles.
So using your chosen example, yes, you can establish the existence of a regular pattern in science without having a physical explanation; but no, you can't do it on just one or two cycles. At the very least, Akasofu should have compared his theory with known variations in past temperatures:
Unfortunately, if he did so, it would be evident that neither the straight line "recovery" from 1800-1850 to the present, nor his regular fluctuations are supported by evidence of past temperatures. In fact, the retrodiction of the linear trend appears to fail at any date prior to 1880; and while there are small fluctuations in the temperature record, which might be regular (and might just as easilly be pure chance), they are nowhere near the magnitude required by Akasofu's theory. (Note, the reconstruction used was chosen as most acceptable to "skeptics", and is biased towards the North Atlantic, and hence is likely to overstate rather than understate fluctuations in global temperature.)
Following on from Tom's comment, another difference between Schwabe and Akasofu is that Schwabe did not ignore generally accepted physics (such as observed changes in solar, GHGs and aerosol forcing) that already adequately explain the observations. For Akasofu's hypothesis to be correct, we would need to be wrong on a fair amount of existing physics (which would then make a lot of paleoclimate very hard to explain).
Thanks Tom for taking the time with those comments. The first one went over my head somewhat, but the second with the history on the sunspot cycle research and backwards projection of Akasofu's chart was very informative.
This example has also clarified in my mind that climate prediction isn't amenable to simple linear models - there are just way too many factors involved.