Scientists use AI to predict sunspot cycles

For the first time, scientists have used artificial intelligence not only to predict sunspots but also to correct the incomplete record of past sunspot activity.

A new paper just published in Advances in Space Research by Dr Victor Velasco Herrera, a theoretical physicist at the National Autonomous University of Mexico, Dr Willie Soon, an award-winning solar astrophysicist at the Harvard-Smithsonian Center for Astrophysics, and Professor David Legates, a climatologist at the University of Delaware and former director of the U.S. Global Change Research Program, predicts that the new 11-year solar cycle that has recently begun will show near-record low sunspot activity that will last until mid-century.

Both Dr Soon (in 2004) and Dr Velasco Herrera (in 2008) had previously published papers speculating that the first half of the 21st century would be a period of unusually few sunspots, potentially slowing the rate of global warming

Sunspots matter. When there are many sunspots and the Sun is active, there is a danger that a strong solar ejection directed towards the Earth could damage or even destroy the thousands of satellites on which the world depends for everything from radio, telephone, television and internet communications to monitoring the climate and observing the farthest reaches of the universe.

Worse, a really strong solar storm could damage the largely unshielded terrestrial electricity grid. Most power lines and transformers are above ground and thus acutely vulnerable. Solar panels, too, could have their lives shortened by intense solar radiation.

The three scientists taught a machine-learning algorithm how to recognize underlying patterns and cycles in the past 320 years’ sunspot record. The algorithm then discovered a hitherto-unnoticed interaction between the 5.5-year solar half-cycles (blue) and the 120-year Gleissberg double cycles (red dotted lines) which allowed it to confirm the earlier predictions of a quiet half-century to come – predictions which are now shared by solar physicists.

That interaction between the two periodicities led the algorithm to indicate that from the 1730s to the 1760s, early in the modern sunspot record (the gray band below), sunspots appear to have been under-recorded: as the 120-year cycle approached its maximum amplitude, sunspots should have been more numerous than reported at the time.

Periods of minimum and maximum solar activity from 1700 to 2020 analyzed by machine learning.

The algorithm then predicted the sunspots from 2021 to 2100. It suggests that the current low solar activity is likely to continue until 2050:

The Sun may be quiet for half a century

Dr Velasco Herrera said: “Not everyone agrees with our expectation that solar activity will continue to be low for another three solar cycles. A paper in Solar Physics by Dr Scott McIntosh of the U.S. National Center for Atmospheric Research, says the coming solar cycle will be unusually active, with a peak sunspot number of 233, compared with our estimate of less than 100. Place your bets in the Battle of the Solar Cycles!”

Dr Soon said, “The machine-learning algorithm, with its interesting interplay between the very short 5.5-year cycle and the long 120-year cycle, confirms our results of 10-15 years ago suggesting that the next three or four solar cycles will be comparatively inactive. This is the first time that the twin problems of hindcasting incomplete past records and forecasting the future have been combined in a single analysis.”

Dr Legates said: “President Trump realized the importance of space weather, and particularly of the Sun, in influencing global climate. It was he who signed the October 2020 ProSwift Act into law to assist in studying and forecasting space weather. Given the history of previous periods of comparative solar activity, the weather may get a little cooler between now and 2050. If we are right, our electricity grids and our satellites should be safe until then.”


Interested readers may download the paper for free from the journal before June 6

https://authors.elsevier.com/a/1cwLK%7E6OilW7X

4.8 16 votes
Article Rating
105 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Jean Parisot
April 23, 2021 6:26 am

 If we are right, our electricity grids and our satellites should be safe until then.”

Wow, how refreshing to see a paper that doesn’t predict Armageddon is Nigh.

John Tillman
Reply to  Jean Parisot
April 23, 2021 6:33 am

Our electric power generation facilities however are in need of winterization.

Vuk
Reply to  John Tillman
April 23, 2021 10:28 am

A bit of ‘cyclomania’ (?!).

WBrowning
Reply to  John Tillman
April 23, 2021 1:12 pm

Our “Real” power generation facilities (Natural Gas and Nuclear) need protection from Democrat politicians more than the weather.

Reply to  WBrowning
April 24, 2021 5:25 am

WBrowning – you are correct.

Corrupt and imbecilic politicians have created the “perfect storm”, harming our vital energy grids with costly, intermittent, destructive wind and solar power generation at a time when we need more cheap, reliable, abundant electrical energy due to increasing demand and global cooling. This will end badly.

Regards, Allan

My post from 2013, warning of our current problems – to deaf ears.

https://wattsupwiththat.com/2013/10/31/blind-faith-in-climate-models/#comment-1130954
 
An Open Letter to Baroness Verma
Baroness Verma was Parliamentary Under-Secretary of State in the UK Department of Energy and Climate Change from 2012 to 2015.
 
“All of the climate models and policy-relevant pathways of future greenhouse gas and aerosol emissions considered in the Intergovernmental Panel on Climate Change’s (IPCC) recent Fifth Assessment Report show a long-term global increase in temperature during the 21st century is expected. In all cases, the warming from increasing greenhouse gases significantly exceeds any cooling from atmospheric aerosols. Other effects such as solar changes and volcanic activity are likely to have only a minor impact over this timescale”.
– Baroness Verma
…..
[excerpt]

So here is my real concern:

IF the Sun does indeed drive temperature, as I suspect, Baroness Verma, then you and your colleagues on both sides of the House may have brewed the perfect storm.
 
You are claiming that global cooling will NOT happen, AND you have crippled your energy systems with excessive reliance on ineffective grid-connected “green energy” schemes.
 
I suggest that global cooling probably WILL happen within the next decade or sooner, and Britain will get colder.
 
I also suggest that the IPCC and the Met Office have NO track record of successful prediction (or “projection”) of global temperature and thus have no scientific credibility.
 
I suggest that Winter deaths will increase in the UK as cooling progresses. 
 
I suggest that Excess Winter Mortality, the British rate of which is about double the rate in the Scandinavian countries, should provide an estimate of this unfolding tragedy.

As always in these matters, I hope to be wrong. These are not numbers, they are real people, who “loved and were loved”.

Best regards to all, Allan MacRae

starzmom
Reply to  Jean Parisot
April 23, 2021 7:11 am

Just because a coronal mass ejection is less likely in a period of low sunspots doesn’t mean it won’t happen or can’t happen. While Armageddon may not be nigh, this would seem to be a good time to harden our electrical grid and electronic systems. Of all the possible avoidable catastrophes we just can’t afford, that seems pretty high on the list.

Last edited 1 month ago by starzmom
Reply to  starzmom
April 23, 2021 9:18 am

Yes, the Carrington storm was not during solar maximum. It was caused by a multiple CME event which “cleared a path” for the largest one to travel on the sun-earth line with very little bow shock.

Sara
Reply to  Jean Parisot
April 23, 2021 2:06 pm

The algorithm then predicted the sunspots from 2021 to 2100. It suggests that the current low solar activity is likely to continue until 2050: – article

As I expected, this forecast is more likely to come within at least barking distance of the mark, and be reasonably accurate. If we’re going into a chilly pattern, I’d rather know ahead of time and be prepared for it by making minor adjustments to shopping for food and finding/supporting local growers than having to worry about droughts. We could have a drought in cold weather; winter wheat usually sprouts in January or February, and less precipitation in the winter on those plantings can definitely have an effect on prices of grains, which affect the consumer market. Same thing with corn: less rain in spring and summer means less field corn in the commodities markets, which means chicken and livestock feeds will cost more, raising the price of poultry, beef and hogs.

So we’ll see how this turns out. I guess I’ll go back to tracking the weather again. It snowed on Tuesday this week. I have photos from my front steps. One late spring in odd. Several in a row may mean a change in weather patterns. Over a longer term, it can mean an even deeper change.

Jay Willis
April 23, 2021 6:34 am

“sunspots should have been more numerous than reported at the time.”

Yeah but they weren’t. Better make adjustments to those data, rather than face the possibility that something else happened. Sometimes it’s just so difficult to predict how the past is going to change.

Rick C
Reply to  Jay Willis
April 23, 2021 9:27 am

Maybe it was just really cloudy around the 1750s? 😉

Reply to  Jay Willis
April 23, 2021 5:36 pm

Or maybe they were, but they were not observed – as evidenced by the historical records of number of observations

April 23, 2021 6:44 am

changed the world and is one of the most influential resources on global warming this is the non of sun

Redge
April 23, 2021 7:08 am

Scientists making measurable predictions. Don’t they want funding.

My only issue is the political comment at the end – the paper can stand for itself without invoking the bogeyman

czechlist
Reply to  Redge
April 23, 2021 9:13 am

“Caveat emptor: Past performance is not an indicator of future results”. Many variables. As Rummy said in short – There are known knowns, known unknowns and unknown unknowns. Modelling is based on knowns and to some degree on known unknowns. What if we had melted the Arctic based on what we thought we knew back in the 70s? I’ll trust Gaia to make her own adjustments, thank you.

Alan the Brit
Reply to  Redge
April 23, 2021 9:57 am

Was it not some Swiss born Jewish fella, called Bert Onestone or summit like that who said, “predictions are very tricky, especially about the future!…..?

Timo Soren
Reply to  Alan the Brit
April 23, 2021 11:06 am

A Danish politician was the first. It has been attributed badly to many….

JPM
Reply to  Alan the Brit
April 23, 2021 6:13 pm

No, that was Yogi Berra.

Reply to  Redge
April 23, 2021 6:15 pm

Some scientists make bold predictions about future events, but most do not. It is dangerous because you can be proved wrong, but it is also courageous, the scientific equivalent of full-contact vs touch football.

I respect scientists like Sallie Baliunas, Tim Patterson, Willie Soon and Nir Shaviv, who publish their opinions about the climate big picture, who say the Sun is the primary driver of century-scale climate, and live or die by their observations.

There is increasing evidence that the following  three statements, published in 2002, are decades ahead of the pack – both warmist and skeptic – especially since our 2002 prediction of solar-driven global cooling, starting about now, is more and more probable.

All warmists and most skeptics argue about the magnitude of climate-sensitivity-to-CO2, and whether the resulting CO2-driven global warming will be hot-and-dangerous or warm-and-beneficial. Both groups are probably wrong, because solar-driven global cooling is happening now.

Regards, Allan Macrae in Calgary

Post Script:
 
I wrote this note to Sallie Baliunas and Tim Patterson this morning:
 
Dear Sallie and Tim,
 
Our three key predictions from 2002 are looking good – the perfect trifecta.
 
I’d prefer that global cooling did not happen, but it should put an end to the fraudulent CAGW hypothesis.
 
Best regards, Allan MacRae in Calgary
 
[excerpt]
OUR PREDICTIVE CLIMATE AND ENERGY RECORD IS CORRECT-TO-DATE
 
Our 2002 predictions are among the most accurate on the planet.
 
In 2002, co-authors Dr Sallie Baliunas, Astrophysicist, Harvard-Smithsonian, Dr Tim Patterson, Paleoclimatologist, Carleton U, Ottawa and Allan MacRae, P.Eng. (now retired), McGill, Queens, U of Alberta, published:
 
1. “Climate science does not support the theory of catastrophic human-made global warming – the alleged warming crisis does not exist.”
 
2. “The ultimate agenda of pro-Kyoto advocates is to eliminate fossil fuels, but this would result in a catastrophic shortfall in global energy supply – the wasteful, inefficient energy solutions proposed by Kyoto advocates simply cannot replace fossil fuels.”
 
Allan MacRae published in the Calgary Herald on September 1, 2002, based on a conversation with Dr Tim Patterson:
 
3. “If [as we believe] solar activity is the main driver of surface temperature rather than CO2, we should begin the next cooling period by 2020 to 2030.”
 
MacRae updated his global cooling prediction in 2013:
 
3a. “I suggest global cooling starts by 2020 or sooner. Bundle up.”
 
*******************************

Reference, being updated with more cold events:
https://thsresearch.files.wordpress.com/2020/09/climate-change-covid-19-and-the-great-reset-readonly.docx

Reply to  ALLAN MACRAE
April 23, 2021 7:29 pm

Thanks, Allan. As you may know, I’ve been looking for any evidence of a sunspot-related effect on the climate for a while now without success.

I started out a true believer that variations in the sun had a large effect on the climate. But when I actually looked for evidence to back up my belief, to my great surprise, I couldn’t find any.

After repeated failures to find any such effect on any surface-level phenomena (rainfall, lake levels, tree rings, ice cores, cloudiness, river flow, sea levels, etc, etc), I followed the dictum ascribed to Paul Samuelson, viz:

“When the facts change, I change my mind. What do you do, sir?”

So my question to you is, what do you think is the solidest, most unassailable evidence that any sunspot-related change affects surface weather? I’d like to find such evidence, so a link to any such study would be greatly appreciated.

Awaiting your answer, I remain,

Your friend,

w.

Reply to  Willis Eschenbach
April 24, 2021 4:08 am

My friend Willis wrote:
“So my question to you is, what do you think is the solidest, most unassailable evidence that any sunspot-related change affects surface weather? I’d like to find such evidence, so a link to any such study would be greatly appreciated.
 
First , I have studied this subject since 1985 so have read countless relevant studies. Tomorrow I might cite a few of them, if I can find the time.
 
Tonight, after midnight my time, I will discuss the greater picture. My discipline is Engineering, and I have two degrees. Unlike many other professions, including law, accounting, tax, the social sciences, etc that try to make simple issues complicated, in order to make themselves important and justify their fees, the practice of engineering actually tackles problems that truly are complex, and our critical skill is to simplify them sufficiently such that the problems can actually be solved.
 
My path started with the Maunder and Dalton Minimums, two very cold periods which corresponded with very low sunspot numbers.
 
Then I reviewed the work of Tim Patterson, a Paleoclimatologist at Carleton University, and had discussions with him. Tim had written a number of papers and his research pointed towards the Sun’s Gleissberg Cycle as a major driver of warming and cooling. I discussed the Gleissberg with Tim, and observed that the correlation of global temperatures with the Gleissberg was good but not perfect.
 
For the record, up to at least the end of 2002 I was unaware of the work of certain others, including Theodor Landscheidt, who wrote a series of papers predicting global cooling peaking circa 2030 due to low solar activity. I had also not read Jack Eddy’s important work.
 
In 2002 , I was writing an article for the Calgary Herald, and called Tim Patterson, with whom I had co-authored a recent paper. I said to Tim “We both believe that century-scale global temperatures are solar-driven and roughly cyclical, correct?” He agreed. I said “OK Tim – when is it going to get colder again?” After a few seconds, he replied “2020 to 2030”. I asked why, and he said “Because the last time it started to get colder was 1940, the Gleissberg cycle lasts 80 to 90 years, so it’s 2020 to 2030.”I quizzed him as some length as to why it was not the PDO which reportedly had a ~60-year cycle, but he was adamant that the 80-90-year Gleissberg was the cycle he saw repeatedly in his field research.
 
I published that 2020-2030 start date for global cooling on 1Sept2002 in the Calgary Herald. Then circa 2008, I observed significant global cooling at the end of Solar Cycle 23, cold events which lasted for several years as the PDO went negative. I recalibrated the 2020-2030 start date for cooling, and given the cycle time of SC23 and SC24, revised my estimate of the start of measurable global cooling to “2020 or sooner”. That has proved to be accurate to date – some significant cold events started in 2018 and in 2019 there was the huge crop failure across the Great Plains of North America – and now there are many extreme-cold events in this Winter of 2020-2021. We’ll see how long the cold lasts – I say decades based on low solar activity, but I hope to be wrong.
 
My observation is that most analytical methods to predict global cooling based on solar activity fail because they are too simplistic – the reality is that there are several variables that impact warming and cooling – not just solar cycles, but also the major ocean cycles, particularly the ENSO and the PDO but also the AMO, which follow their own cycles and are not generally in-phase with solar cycles. A model that adequately describes the ups and downs of global temperature will have to incorporate both solar cycles and oceanic cycles.
 
Regards, Allan

Reply to  ALLAN MACRAE
April 24, 2021 12:36 pm

ALLAN MACRAE April 24, 2021 4:08 am

My friend Willis wrote:

“So my question to you is, what do you think is the solidest, most unassailable evidence that any sunspot-related change affects surface weather? I’d like to find such evidence, so a link to any such study would be greatly appreciated.”

 

First , I have studied this subject since 1985 so have read countless relevant studies. Tomorrow I might cite a few of them, if I can find the time.

Thanks, Allan. As you know, I’m not much for history such as your comment contains. I know you’re qualified and have studied this a lot, but I like facts. So I’ll wait for your links, and see if their facts stand up to examination.

Please note that I’m asking for links to studies of actual data, not the output of climate models or reanalysis models. And please, send me the strongest, best ones. I don’t have time to faff about with the typical shabby studies that pass for much of climate science these days.

Finally, as a ham radio operator (Hotel 44 Whisky Echo), I know that sunspots affect the ionosphere. I’m asking for studies that show an effect down here near the ground.

Much appreciated,

w.

posa
Reply to  Willis Eschenbach
April 24, 2021 5:52 pm

My studies have landed me in the Willis Eschenbach- Leif Svalgaard Sunspot Denialist Camp.

Even Svensmark says:

It’s useless to pretend that the problem of solar influence has been solved.

Reply to  Willis Eschenbach
April 24, 2021 9:36 pm

Hi Willis,

I cannot help you as you request. There was no one study or small group of studies that led to this prediction. Back in 2002, hardly anyone was writing about global cooling – that was sacrilege within the global warming mania that prevailed at that time. I did not use either the work of Jack Eddy or that of Theodor Landscheidt.

More than 16 years of my prior climate analysis and paleoclimatologist Tim Patterson’s career of scholarship contributed to our 2002 prediction of “global cooling starting by 2020 to 2030”. Another 11 years of my study thereafter enabled me to further refine and re-calibrate the start date in 2013 to “cooling by 2020 or sooner”.

Over the years, I analysed the solar records, the measured and proxy temperature records, etc. Of particular value in the 2013 re-calibration were the cold climate events starting circa 2008 at the end of SC23.

Tim Patterson’s publications are here: https://carleton.ca/timpatterson/publications/

I will live or die by the accuracy of our prediction – so far, it’s looking highly accurate. I’d really rather be wrong, because humanity and the environment suffer during cold periods.

Regards, Allan

Reply to  ALLAN MACRAE
April 24, 2021 11:09 pm

Allen, it’s funny how almost every single person who tells me they are sure that the sun rules the climate runs for the door when I ask them for the strongest evidence that they know of … me, I’ve analyzed dozens of studies claiming a relationship, and I have yet to find one that is solid evidence.

I’ve looked at paleo data, cloud data, lake data, surface temperature data, river data, rainfall data, sea level data, tropospheric temperature data, ice core data, beryllium deposition data …

… nothing.

Next, I didn’t ask what “led to this prediction”. I asked, “What do you think is the solidest, most unassailable evidence that any sunspot-related change affects surface weather?”

In response, you say you don’t have any such evidence?

Really?

I tried to find anything by Tim Patterson showing such effects. I found him claiming such a relationship … but no links to whatever data he’s using, or links to his year-by-year results, so that was useless. I need both the data and the analysis to replicate the results, and he hasn’t archived them.

Ah, well, I suppose I shouldn’t be surprised. This is exactly what the “It’s the Sun!” folks generally do … they claim certainty in their beliefs, but they are either unable or unwilling to provide DATA AND ANALYSIS to back it up.

Sadly, your friend regardless,

w.

Reply to  Willis Eschenbach
April 25, 2021 4:22 am

That is not correct Willis – you are asking me to re-create for your benefit decades of work that led me to my (our) global cooling prediction, work that was done from 1985 to 2002. I provided you with a general outline of what I recall using, and what I believe I did not use. I never formalized that work in a paper, and don’t have the time to re-create that detail for you now.
 
The important question is not “Are you convinced?” The important question is actually “Is the prediction correct?” That is the acid test – unlike the global warming alarmists, I don’t make ~80 failed predictions and hope one will be correct. I published one global cooling prediction in 2002, revised the timing slightly in 2013 based on new data, and that cooling is apparently happening now, exactly as predicted.
 
The details you seek will be filled in in the fullness of time by others. I am spending less time on climate science, and more time on the toxic political climate that seeks to destroy our society, based on the twin frauds of global warming and the Covid-19 lockdown, and the leftists Final Solution, the “Great Reset”. The USA, Canada  and the other few remaining free countries in the world are now failing, due in large part to policies that are justified by these huge twin frauds.
 
I was confident enough to publish a prediction in 2002 that is happening now, decades in the future. More and more major cooling events continue to occur all over the world. Update 1c of my recent paper will be published soon:
CONCLUSION: Dangerous global cooling will continue, it will be sporadic, moving from continent to continent with the seasons and the polar vortex, and could last for decades.
_________________
 
A caution: You like to write papers that dispute the conclusions of others, but you are not always correct.

In 2008 you wrote a paper calling my January 2008 discovery that CO2-changes-lag-temperature-changes “spurious correlation”. Your conclusion was wrong.

Here it is again – it is obviously NOT spurious correlation. Your attempt to disprove this important hypothesis failed – you should withdraw your 2008 paper.

https://www.woodfortrees.org/plot/esrl-co2/from:1979/mean:12/derivative/plot/uah6/from:1979/scale:0.18/offset:0.17

Regards, Allan

Reply to  ALLAN MACRAE
April 25, 2021 10:50 am

ALLAN MACRAE April 25, 2021 4:22 am

That is not correct Willis – you are asking me to re-create for your benefit decades of work that led me to my (our) global cooling prediction, work that was done from 1985 to 2002. 

No, good sir, that is not true in the slightest. I made a very simple request. I did not ask you to “re-create” anything.

I simply asked, “What do you think is the solidest, most unassailable evidence that any sunspot-related change affects surface weather?”

I’m asking for a link to ONE example of data and analysis that you think is the strongest evidence that the sun affects the climate at the surface of the earth. Not “decades of work”. Not your conclusions or your predictions. ONE example of data and analysis that you think is the best evidence for your case regarding the sun.

However, like most solar proponents, you find reasons to not post up what supports your claims. Your choice, my regret.

Finally, you say:

A caution: You like to write papers that dispute the conclusions of others, but you are not always correct.

In 2008 you wrote a paper calling my January 2008 discovery that CO2-changes-lag-temperature-changes “spurious correlation”. Your conclusion was wrong.

Here it is again – it is obviously NOT spurious correlation. Your attempt to disprove this important hypothesis failed – you should withdraw your 2008 paper.

https://www.woodfortrees.org/plot/esrl-co2/from:1979/mean:12/derivative/plot/uah6/from:1979/scale:0.18/offset:0.17

While you are 100% right that I’m “not always correct”, I’ve written over 900 posts here on WUWT, so I fear I have no idea which post of mine you are referring to. If you’d be so kind as to provide a link to my post, I’m happy to take a look and see whether I was wrong. If I was, I will of course admit it. However, I don’t “withdraw” any of my posts. Instead, I put a big notice on them that I was wrong. That way, my errors stay out there for folks to see, rather than being disappeared.

Best regards,

w.

Reply to  Willis Eschenbach
April 25, 2021 3:55 pm

Hi Willis,

It was a paper you published on Climate Audit in 2008. Found it:
https://climateaudit.org/2008/02/12/data-smoothing-and-spurious-correlation/
 
I have no recollection of a particular best evidence of solar influence – I recall liking the work of Nir Shaviv and Willie Soon but that is probably post-2002. I see if I can find some of it and post it here. I’ll also ask Tim Patterson. No promises.
 
I like and respect much of your work – especially your 21March correct call on the Covid-19 lockdown – the same day I independently published the same call. That was epic – you and I were among the very the first to make that call.

Others were the Swedish docs and some smart person in South Dakota. Other than that, it was all sheep being driven to the slaughter.
 
The most learned physicians I know say the damage from the lockdown was 10 to 100 times worse than the Covid-19 illness.

That does not include the damage from the new variants – some or all of which are a result of the lockdown measures. By extending the life of the virus (via masking and distancing), which should have died out in the summer, the variants were given time and gradient to develop. I’ve had long talks with one of my eminent physician friends and he agrees with this analysis.
 
Regards, Allan

Reply to  ALLAN MACRAE
April 25, 2021 5:04 pm

Thanks much, Allan. You’d said:

In 2008 you wrote a paper calling my January 2008 discovery that CO2-changes-lag-temperature-changes “spurious correlation”.

I see that my actual conclusion was:

And for the $64,000 question … is the correlation found in the Macrae study valid, or spurious? I truly don’t know, although I strongly suspect that it is spurious. But how can we tell?

This is an excellent example of why I ask people to quote the exact words under discussion … my actual statement is nothing like yours.

That post, however, strongly reinforces the idea that we should NOT use smoothed data in our analyses.

From memory, the CO2 levels lag the temperature levels by about 3 months. This is because of the annual cycle of plant growth and decay, which alternately remove and add CO2 from the atmosphere.

Thanks for the note pointing out that you called for the end of the lockdowns at the same time I did. I was unaware of that, very well done.

w.

Reply to  Willis Eschenbach
April 25, 2021 7:56 pm

Hi Willis,

Because of your 2008 paper, I added a case with no smoothed data. Same result – a bit more messy.

The lag of atmospheric CO2 after atmospheric temperature is ~9 months. This observation is an absolute disproof of the CAGW hypothesis: “The future cannot cause the past.”
Ed Berry cited my 2008 paper in his recent paper and book.

You requested a paper that I respected on solar-drives-climate. I only had limited time, so I sought one by my friend Willie Soon.

RE-EVALUATING THE ROLE OF SOLAR VARIABILITY ON NORTHERN HEMISPHERE TEMPERATURE TRENDS SINCE THE 19TH CENTURY
Willie Soon, Ronan Connolly, Michael Connolly
Last updated 11th August 2015
Paper:
http://globalwarmingsolved.com/data_files/SCC2015_preprint.pdf
Data:
http://globalwarmingsolved.com/data_files/SCC2015-SI.zip

Re Covid – I credited you in my latest paper – this update was published earlier today. On 21March2020 we nailed it.

CLIMATE CHANGE, COVID-19, AND THE GREAT RESET
A Climate, Energy And Covid Primer For Politicians And Media
By Allan M.R. MacRae, April 21, 2021 UPDATE 1c
https://thsresearch.wordpress.com/2021/04/25/climate-change-covid-19-and-the-great-reset/
 

Reply to  ALLAN MACRAE
April 25, 2021 9:27 pm

Thanks so much, Allan. I’ll take a look as soon as I get a chance and get back to you.

w

Reply to  Willis Eschenbach
April 25, 2021 9:37 pm

Willie Soon is a great friend. I spent time with him in Calgary and also in Boston. He is definitely one of the very best guys anywhere to drain a few beers with – brilliant, quick and funny.

Greg
Reply to  ALLAN MACRAE
April 26, 2021 2:36 am

This discussion about smoothing data is very pertinent. One thing I would certainly criticise in McRae’s original is the use of the damned “running average smoother”. That distorts data horribly. Willis’ use of a gaussian filter is a much better choice if low-pass filtering is desirable at all.

One thing not mentioned about filtering data is that you are reducing the number of independent data points. eg if you do a 12mo running mean you have 1/12 the number independent data points, even though you have (roughly) the same number of data. Gaussian is similar though because of weighting the reduction is less obvious. Since a triple R-M is close to a gaussian you could probably approximate a 12mo FWHM as 12+9+7mo RM so 1/28 reduction in independence. (To be verified ! )

This needs to be taken into account when assessing if the correlation coeff is significant. In Willis’ CA article we see the peak correlation initially increasing as the length of filter is increased. This needs to be related to the number of independent data points. It may be becoming less significant .

I would also caution against the idea that the temp. “anomaly” (∆T) and Mauna Loa CO2 change (∆CO2) can be called “raw data”. The idea of averaging several years to get an “average” annual cycle then subtracting this will also introduce noise into the data. Though typical faire in climatology this is horrible data mangling since there is no “average” year: it’s a chaotic system.

I would be surprised if even the 3mo lag is significant at 0.1 , though I have not checked the numbers.

The anomaly processing is a very crude notch filter which will have an appalling phase response curve and introduce all sorts of spurious phase shifts: lags and leads into the data. Maybe there was some clearer correlation before “anomaly” processing mangled it.

Since we know temp affects growth and growth affects CO2, and there is a heavy land/sea bias in NH, any sub-annual change is probably not that helpful.

Since you are both interested in exploring this maybe it would be worth working with temp and CO2 directly rather than anomalies, then ignoring sub-annual change and see what remains.

I tend to agree with Allan McRae, that there are more important issues in the current insanity. Climate stupidity will not be cured by scientific proof even if one could be found. Moral arrogance is not susceptible logical argument.

Please remember that being literate or numerate is simply racist and that proving something scientifically is just white, cisgender, patriarchal oppression.

Last edited 1 month ago by Greg
Reply to  Greg
April 26, 2021 10:43 am

Thanks, Greg. I’ve railed against the “running average smoother”, AKA “boxcar filter” before. It’s a crime against numbers.

However, I’d tend to disagree when you say:

“I would also caution against the idea that the temp. “anomaly” (∆T) and Mauna Loa CO2 change (∆CO2) can be called “raw data”. The idea of averaging several years to get an “average” annual cycle then subtracting this will also introduce noise into the data. Though typical faire in climatology this is horrible data mangling since there is no “average” year: it’s a chaotic system.”

First, this is generally described as “removing seasonality” rather than a “temperature anomaly”, although both are correct.

And although there is no such thing as an “average year”, somehow in the NH June on average is warmer than January. And we are all familiar with the concept of a “warm winter” or a “cold summer”, so these are not simply mathematical constructs.

For me, the issue is a bit different than “introducing noise”. Removing seasonality from the data introduces uncertainty in the results. This is often either overlooked or underestimated. One generally ignored issue is long-term persistence.

As an example, the Berkeley Earth global dataset starts in 1850, and it has 2021 data points. However, it is very strongly autocorrelated, as are the monthly datasets. Not only that, but the summers are more strongly autocorrelated than the winters. So when you take say all of the Junes, while there are 168 data points that are estimates of historical June temperatures, the Hurst exponent is way up at 0.88, and the effective number of data points is only 3.

This means that the standard error of the June mean is not 0.026°C as standard statistics would calculate. Instead, it’s seven times that, 0.186°C.

Thanks for the comment,

w.

Reply to  Greg
May 2, 2021 10:04 pm

Hi Greg – thank you for your thoughtful comments.

Re my Jan2008 paper::
 
Here again is the key relationship – it is obviously NOT spurious correlation. I dealt with it in Jan2008 by adding a case with no smoothing. Same result. 
https://www.woodfortrees.org/plot/esrl-co2/from:1979/mean:12/derivative/plot/uah6/from:1979/scale:0.18/offset:0.17
 
In the modern data record, the lag of atmospheric CO2 changes after atmospheric temperature changes is ~9 months. This is an absolute disproof of the CAGW hypothesis, which states that increasing CO2 drives temperature. “The future cannot cause the past.”
 
In my 2019 paper below I explained why the lag is ~9 months – it is basic calculus, the 90 degree (1/4 cycle) lag of the derivative and its integral, which is the ~3 year ENSO period.
 
My 2008 paper remains very important. My 2008 conclusion was duplicated and expanded by Humlum et al in 2013, for which I am grateful.
 
My 2008 paper has been cited by Ed Berry in his 2020 book and related paper, which is at the cutting edge of climate science.
“Climate Miracle: There is no climate crisis – Nature controls climate”
amazon.ca/Climate-Miracle-climate-crisis-controls-ebook/dp/B08LCD1YC3/
 
“Carbon cycle model shows nature controls CO2 level”
edberry.com/blog/climate/climate-physics/preprint3/
 
All warmists and most skeptics argue about the magnitude of climate sensitivity to increasing CO2, and whether the resulting CO2-driven global warming will be hot and dangerous or warm and beneficial. Both groups are probably wrong, because global cooling is happening now, even as CO2 concentration increases
 
There is a high probability that the mainstream debate is wrong – a waste of decades of vital time, tens of trillions of dollars of green energy nonsense and millions of lives. Vital energy systems have been compromised, damaged with intermittent, unreliable wind and solar generation – a debacle.
 
Cheap abundant reliable energy is the lifeblood of humanity – it IS that simple. The green sabotage of our vital energy systems, whether innocent or deliberate, has cost lives and could cost very many more.
 
Scientific details here:
CO2, GLOBAL WARMING, CLIMATE AND ENERGY June 15, 2019
https://wattsupwiththat.com/2019/06/15/co2-global-warming-climate-and-energy-2/
 
 
Repeating, “The future cannot cause the past.”

Reply to  ALLAN MACRAE
April 27, 2021 3:21 am

Hi again Willis,
 
My updated paper has now been published on several sites.

Our 2002 prediction of global cooling starting circa 2020 is happening. Unfortunate, and it’s not over.
 
MORE RECORD COLD EVENTS SINCE PUBLISHING – I will re-publish:
 
RECORD APRIL COLD SWEEPS THE UNITED STATES, WITH “VERY RARE” AND “BIZARRE” SNOW FALLING AS FAR SOUTH AS OKLAHOMA  April 22, 2021
https://electroverse.net/record-april-cold-sweeps-the-united-states/
 
WINTER ARRIVES EARLY ACROSS THE SOUTHERN HEMISPHERE  April 26, 2021
https://electroverse.net/winter-arrives-early-across-the-southern-hemisphere/
 
OVER THE PAST 7 DAYS, THOUSANDS OF LOW TEMPERATURE RECORDS FELL ACROSS THE UNITED STATES   April 26, 2021
https://electroverse.net/thousands-of-low-temperature-records-fell-across-the-united-states/
 
EUROPE BRACES FOR EXTREME MAY FREEZE  April 27, 2021
https://electroverse.net/europe-braces-for-extreme-may-freeze/
 
I also correctly predicted the timing of these cold events late last year, based on Nino34 temperatures.

“This will end badly.”
 
 
REGARDING COVID-19:
 
Actual USA deaths attributed to Covid019 were only about 1/16 of those reported – approx. 37,000 – not 586,000.
FAULTY COVID DEATH NUMBERS EXPLAINED
The American Thinker, April 9, 2021
 
My Fall 2020 estimate of same was about 1/9 (vs 1/16), based on Alberta per capita mortality. Average age of Alberta Covid-19 death here was ~82 to ~end 2020.

There were no “Excess Total Deaths” in Alberta or Canada to 30June2020. No Excess Total Deaths means no dangerous pandemic. Lockdown Fraud!
 
As we both published on 21March2020, the lockdown of the workforce and students was not justified. Total Alberta Covid-19 deaths of under-65’s to ~1Nov2020 was 13 in a population of ~4 million. We destroyed our economy for this?
 
THE BIG PICTURE:
 
Canada and the USA are failing under current corrupt administrations. The Covid-19 full-Gulag Lockdown was a scam.

Energy systems have been dangerously destabilized by green generation schemes at the very time we will need them.

Reply to  ALLAN MACRAE
April 27, 2021 2:39 pm

I just sent this to many journalists. I call this Treason. USA = same.
 
THE GREAT RESET – PRESENTATION FROM THE BANK OF CANADA
https://www.bankofcanada.ca/wp-content/uploads/2020/08/presentation-2020-08-20.pdf
https://www.bankofcanada.ca/profile/paul-beaudry/

Regards, Allan MacRae in Calgary

Reply to  ALLAN MACRAE
May 2, 2021 5:32 am

By extending the life of the virus (via masking and distancing), which should have died out in the summer, the variants were given time and gradient to develop. 
Update:
I’ve had long talks with TWO of my eminent physician friends and THEY BOTH agree with this analysis.
Attaboys all around for those who ordered the Covid-19 lockdowns – the deaths from the variants “are on you”.

Reply to  Willis Eschenbach
April 24, 2021 7:47 pm

“what do you think is the solidest, most unassailable evidence that any sunspot-related change affects surface weather?”

I know of this recent paper also to consider. I think WUWT covered it too.

https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2020EA001223

Reply to  Eric Frei
April 24, 2021 8:35 pm

It was posted on WUWT. I showed that the data is inadequate to support their conclusions. See here and here.

w.

Greg
Reply to  Willis Eschenbach
April 26, 2021 2:51 am

Willis, there was post in the last couple of days where you were discussing the presence or not of a 120y “cycle” with Lief Svalgaard but I can’t find it now.

As a complement to what you posted then, here is the power spectrum from Leif’s 9000 year “multimessenger” solar data set ( source on graph ).

comment image

It shows a fairly strong peak at 114y and another at 67y.

As shown in the legend I split the dataset as you often do to see whether the cycle is persistent.

Last edited 1 month ago by Greg
Reply to  Greg
April 26, 2021 12:16 pm

Greg, as I commented on your other comment with this dataset:

Greg, I took a look at that dataset. The problem with it is that it’s calculated at 10-year intervals. That means that we get a spurious frequency at 10 years time the 11+ years of the sunspot cycle … which is the 114 year frequency shown in the spectral power density graph above.

See the comment here.

w.

John Bell
April 23, 2021 7:19 am

What is A.I. but just a computer program?

Carlo, Monte
Reply to  John Bell
April 23, 2021 8:42 am

It is using neural net programing/learning, calling it A.I. is a false but popular notion.

DHR
Reply to  John Bell
April 23, 2021 8:51 am

I understand that “AI” is what you call a computer program if you want to seem really smart and modern to others. Otherwise they are “programs” or “algorithms.”

chris
Reply to  DHR
April 23, 2021 10:40 am

yes, in this case, but not in general.

AI is the science of using heuristics to enable a program make useful predictions or decisions in problems where the optimal solution is computationally intractable to compute. Intractability is not an issue for smallish versions of some intractable problems, but (typically) when there are lots of variables certain instances of the problem can take an unbounded amount of time (or memory or both) to compute the optimal solution.

Its “artificial intelligence” in the sense that using heuristics is similar to how humans solve intractable problems (where the heuristics are usually based on expertise in the domain).

Note that AI programs do not necessarily use the same heuristics as do people. That’s good, because “quick and dirty” solutions can easily be wrong due to misapplication of cognitive heuristics that can result in biases (e.g., given Y, choose X because often Y is most associated with X).

John Dueker
Reply to  chris
April 23, 2021 4:25 pm

No AI is Hal and Data. Call me when we get there.

Clyde Spencer
Reply to  DHR
April 23, 2021 6:32 pm

I’ll go along with “program,” but not algorithm:

  • A finite set of unambiguous instructions that, given some set of initial conditions, can be performed in a prescribed sequence to achieve a certain goal and that has a recognizable set of end conditions.
  • a precise rule (or set of rules) specifying how to solve some problem; a set of procedures guaranteed to find the solution to a problem.

Those are not the ways that a neural net works.

Peta of Newark
April 23, 2021 7:37 am

Quote:“”If we are right, our electricity grids and our satellites should be safe until then.

Don’t count on it.
‘tween now and then, Yours Truly is gonna ‘pass away’
If, big if but how d’ya kno, YT has any say in what subsequently gives, YT is gonna re- appear with a horny head and a pointy tail

Sputniks will be history from that point onwards.

Consider it a bit like what Prince Phillip was misquoted as saying about ‘returning as a virus’ and wiping out a big portion of humanity.

I’ll do ‘wiping out’ alright, except to computers & Sputniks and actually give humanity their lives (back)

Alan the Brit
Reply to  Peta of Newark
April 23, 2021 9:59 am

He was ALL heart after all!

Sara
Reply to  Peta of Newark
April 23, 2021 2:10 pm

Well, we are due for a major change in everything, anyway, so d’you mind if I join you, Peta? I’ll be a bit behind you, but it could be bodacious fun.

S.K.
April 23, 2021 7:43 am

I become concerned when I hear about computer programs correcting historical data. Example is the homogenization of the temperature data is out right altering to promote the alarmism dogma.

Not a mention of why it will get cooler. Is that because Dr. Soon does not believe in the galactic cosmic ray particle of cloud formation? If it’s not cloud formation linked to solar magnetic field activity then what is driving the cooling trend?

Richard Page
Reply to  S.K.
April 23, 2021 10:05 am

I think the theory held up during experimental tests at Cerne, but observational data was inconclusive, I believe anyway – I think the overall conclusion that there were other factors at play that hadn’t been fully considered.

iflyjetzzz
Reply to  S.K.
April 23, 2021 7:45 pm

Yes, I find the altering of past data very problematic paper. I believe I read that they first filled in the ‘missing’ solar activity of the past and then ran the forecast with the ‘corrected’ data. That, in my mind, makes it garbage in, garbage out.
While I hope that the forecast is correct and that we see significant cooling in the next two decades, this use of altered data to form the prediction is a very big problem for me. It’s no better than what the Australian BOM did when they altered Acorn 1 data to show a 23% increase in warming with just the stroke of a pen in Acorn 2. Both now suffer from confirmation bias.

The big reason why I hope this forecast turns out correct and the earth cools is that it would show that CO2 levels have a de minimis impact on global temperatures. We’re already wasting inordinate amounts of money on trying to fix global warming. To the point that we are building extremely toxic EVs which are far worse for the environment than ICE vehicles.

oebele bruinsma
April 23, 2021 7:51 am

It seems that two different approaches of using logic came to a similar conclusion: https://www.tandfonline.com/doi/full/10.1080/23328940.2020.1796243

Reply to  oebele bruinsma
April 23, 2021 8:54 am

Zharkova’s AL and here source data are far more creditable than one which changes the data inputs. Her findings of a dual solar dynamo will also be verified based on some forthcoming findings about magnetic fields. She should win the Aristotle Martyr Award for the century.

Reply to  oebele bruinsma
April 23, 2021 8:57 am

should read Zharkova’s AI and her source data

Coach Springer
April 23, 2021 7:59 am

Not clear on the difference between model and AI in the algorithm. Still, it’s assumptions in, assumptions out – as acknowledged by Dr. Soon. “Place your bets.”

Ron Long
April 23, 2021 8:02 am

How convinient, the minimum is right now, and the next cycles are shown/predicted/guessed to be more active, so stronger solar output in future and more warming? Show me the money?

ScienceABC123
April 23, 2021 8:11 am

“For the first time, scientists have used artificial intelligence not only to predict sunspots but also to correct the incomplete record of past sunspot activity.”

The Adjustocene continues…

leowaj
Reply to  ScienceABC123
April 23, 2021 8:58 am

I saw that whopper of an opener and thought to me self, “How does one confirm the correction is correct?”

Drake
Reply to  ScienceABC123
April 23, 2021 9:33 am

The battle of the Little Big Horn, Custer’s Last Stand had a similar issue. Historians had much of the battle wrong, only taking the i put from soldiers and scouts who survived. The native input was ignored.

30 years or so ago there was a fire and the grass burned to the ground. After the fire, the location discovered shell casings verified the accounts of natives that fought and lived through the event, and “history” was corrected to match the physical facts.

The case of assuming that those keeping the sunspot records at the time were incompetent and therefore incorrect is really troubling. The eye witnesses should be PROVEN wrong, not computer modeled wrong.

I would have expected better from Dr. Soon

Gary Kerkin
Reply to  Drake
April 23, 2021 4:05 pm

You should read the paper before you make that comment.

B Clarke
April 23, 2021 8:23 am

1760s, early in the modern sunspot record (the gray band below), sunspots appear to have been under-recorded: as the 120-year cycle approached its maximum amplitude, sunspots should have been more numerous than reported at the time.”

Possibly because they were not observable with the methods of the time, its also true during the maunder-minimum less sun spots were observed because there were less sun spots to observe and what was missed had very little effect on the maunder-minimum, they would be perhaps relevant for statistics but not effect.

I’m dubious AI can predict in any detail what the sun will and won’t do in any cycle,sure we know in a minima cycle less likely spots / maxima cycle more likely spots, we can’t really see whats going on in the limb out of view to any certainty, how would AI predict in a 400 year solar minima cycle (gsm) .

dk_
April 23, 2021 8:30 am

“It’s tough to make predictions, especially about the future”
-Yogi Berra

It is good to see someone make a prediction using the words may and might instead of will or must. This is how models should be treated.

Sad that we still worship at the over hyped silicon feet of AI.

Timo Soren
Reply to  dk_
April 23, 2021 11:09 am

Here is a link, Yogi might have said it, but he didn’t invent it.
It’s Difficult to Make Predictions, Especially About the Future – Quote Investigator

DHR
April 23, 2021 8:43 am

Our electricity grids will be safe until then? Perhaps from solar activity but not from Biden. His CO2 control plans could easily have it destroyed well before.

Doonman
April 23, 2021 9:20 am

My solar panels are guaranteed to produce at least 80% of their original power output for 20 years. This was a governmental requirement before they could be sold Nobody ever said a solar flare could damage them or void the guarantee. So I am thrilled that scientists think the sun will remain quiet for the next few decades.

Joseph Zorzin
Reply to  Doonman
April 23, 2021 9:56 am

Was your original power output equal to what was expected? I’ve had several people who put solar on their roofs say they never got what the salesmen told them.

Doonman
Reply to  Joseph Zorzin
April 23, 2021 10:32 am

My system was rated as 4.5kW when installed. On a clear day just afternoon in June the first year I got 4.2 kW once. However, my panel orientation is not perfect for my latitude. Five years later and I never can exceed 4 kW. But clouds, fog and dirt all conspire against me anyway, so the rated power is never practically achieved or should be expected. It will pay off eventually in the next couple of years, but that’s mostly because of utility rate increases.

BTW, I knew all this before I installed the system. But it has other advantages. Whenever greenies lecture me about CAWG, I point to my system and then ask how theirs is doing. Turns out for the most part they don’t have one, and after the excuses are made, the lecturing stops.

Joseph Zorzin
Reply to  Doonman
April 23, 2021 11:41 am

in my ‘hood in north central Mass. pretty much a working class ‘hood of modest ranch houses- about half the homes have solar- but most houses don’t have a south facing roof- most are facing east and west- I keep wondering how much less effective are such roofs compared to a south facing roof? And, many have big trees partially shading the panels- I bet the salesman didn’t explain that shaded panels aren’t so productive- I happen to have an L shaped house and one side is south facing with little shading by trees- so, if I can get panels for free or almost free, I’ll take them- since this state passed a net free bill- and there’s so much talk about climate justice, I’ll just say, “hey, give me some, I live on SS”

Clyde Spencer
Reply to  Joseph Zorzin
April 23, 2021 6:41 pm

And because most cities in the USA have a street system aligned N-S, E-W, on average about half the homes will have roofs that are not oriented optimally. Even those that have a south-facing roof will have less output than potentially possible because they are fixed and don’t track the sun through the day and the seasons. This rarely gets mentioned!

markl
April 23, 2021 9:22 am

It’s becoming more commonplace of the reporting media to accept mathematics computation as scientific proof.

Tom in Florida
April 23, 2021 9:27 am

Which sunspot numbers and counting system did they start out with? The original Wolfe numbers, the Hoyt & Schatten counting method numbers or the new corrected numbers put out by the Sunspot workshop lead by Dr S.? One must be concerned that GIGO may apply.

Vuk
Reply to  Tom in Florida
April 23, 2021 11:37 am
Looks like SIDC, the new corrected Sunspot workshop lead by Dr S., peak in 1957 was 269, see my link, two comments below) 
Last edited 1 month ago by Vuk
bluecat57
April 23, 2021 9:27 am

And despite hundreds of years’ of observations, can’t get even close.

Vuk
April 23, 2021 9:43 am

Awaiting Doc Svalgaard’s comment, till then here is the most accurate (within a decimal point) prediction ever, in this case 10 years ahead of the SC24 max, not by the AI; but simple extrapolation, coincidence, chance or the way the nature operates, pick your choice.
All details are here
http://www.vukcevic.co.uk/SSN.htm
so you are welcome to go and check it if you are so inclined (and if you wish you can calculate the SC25 peak, having in mind that the values are in the classic Wolf numbers).
btw. I think their prediction for the SC25 is contrary to what Dr. S is predicting, but in my view it is on the high side.

OweninGA
April 23, 2021 9:48 am

I fear in the future the sun activity as with the monthly temperature summaries will be models all the way down. Sad. No observed data should ever be “corrected” by a model!

April 23, 2021 9:48 am

AI seems to be working out quite smashing for Tesla cars. 😬
…better hire a co-pilot before retiring to the rear lounge seat.

That said, I’m currently committing all my volunteer crunching resources to Machine Learning and Comprehension @Home, a small distributed supercomputer run by he Cognition, Robotics, and Learning (CORAL) Lab at the University of Maryland, Baltimore County (UMBC). The field of AI is still in its infancy judging from what I’ve been crunching.

April 23, 2021 9:55 am

If I remember well, I have seen a comparable cycle graph from T. Landscheidt, working that out with his own intelligence and based on data.

Brian Bellefeuille
April 23, 2021 10:00 am

“… first half of the 21st century would be a period of unusually few sunspots, potentially slowing the rate of global warming.”
 Does this infer that the sun/active sun causes the globe to warm? Or is this an underhanded way of saying that man is warming the planet and a cool sun is saving our butts for awhile.

Lurker Pete
April 23, 2021 10:06 am

I hope it’s not related to the AI that’s ‘moderating’ social media! The Sun would be in the sin bin!

Olen
April 23, 2021 10:07 am

Refreshing to see credit given to the influence of the Sun on the Earth’s weather and electrical grid, communications and sunburn. And having the objectivity to write If we are right.

Ab Mix-O'Lydian
April 23, 2021 10:14 am

However, the sword has two edges. The other: increased exposure to cosmic radiation for pilots and air travelers.

chris
April 23, 2021 10:29 am

Note that what this article terms “Artificial Intelligence” is more accurately termed – in the words of Patrick Winston, founder of the MIT AI Lab – “computational Statistics”.

This application is AI only in the sense that the pattern recognition algorithm is in the class of algorithms that use computational heuristics (aka short cuts) to approximate a full Bayesian classifier. Learning a full joint probability distribution from data is computationally intractable in the number of features. (doing inference using a full JPD is also intractable – NP Complete – but does not need to be re-learned every time and in practice can be run overnight).

There are standard methods to estimate the possible error in the posterior distribution (the PDF of the target feature after incorporating evidence). But scientists who are not deeply experienced in the computational methods rarely have the competance to do such a calculation.

Bottom line: nothing new here, possibly excepting that data was substituted for expert knowledge.

Editor
April 23, 2021 11:27 am

Despite my respect for my friends Dr. Soon and Dr. Legates, I have some issues with this study.

First, I’m always concerned about the use of AI. The issue is not the results, which are sometimes spectacular. It’s that in general, how the AI got the results is totally opaque … meaning that we have no way to verify if it is AI or it is actually AS (Artificial Stupidity).

Second, we have annual sunspot data from 1700 to 2020, a total of 321 years. In my opinion, diagnosing a putative cycle that is 120 years, 40% of the length of the dataset, is … well, let me call it “wildly optimistic”.

The problem is what I call “pseudocycles”. These are variations that look like regular cycles, but instead they appear and disappear without warning. My rule of thumb is that you need at least three complete cycles to make a tentative claim that a real cycle exists, and even then you can get fooled. I once analyzed a sea-level dataset that appeared to have a connection to no less than five sunspot cycles … but once I looked at the full dataset, there was no sign of the connection before or after the five apparent cycles.

Third, a CEEMD analysis shows no sign of a putative 120-year “Gleissberg Cycle”. I’ve never believed in the actual existence of this cycle, in part because various researchers have strongly declared that it has a certain length … but the length has varied in each study:

In different studies the length of the period of the secular variation was determined to be equal to 95 years, 65 years, 55 years, 58 years, 83 years, 78.8 years, 87 years [Siscoe, 1980; Feynman and Fougere, 1984]. That situation is understandable, because the longest record of direct observations of solar activity was and still is the sunspot numbers which provides more or less reliable information since 1700 (see below). That gives one only 300 years of time span by now which encompasses ~3.4 periods of Gleissberg cycle which is quite low for its statistical analysis.

Note they make the same point I make above, that 300+ years of data is too short to diagnose an ~100-year cycle.

The study discussed in the head post purports to find a 120-year cycle …

In any case, here’s the CEEMD analysis:
comment image

As you can see, there’s almost no power in the 120-year range. There is a very broad peak around 90 years or so (Empirical Mode 6) … but the width of the peak indicates it’s not any kind of solid cycle. And indeed, when we plot up that Empirical Mode 6, here’s what we find:
comment image

Remember how I said that “pseudocycles” appear and disappear? It looks like there’s a strong ~ 90-year cycle in the early part of the record … but it fades away entirely.

So … those are my objections to the study.

My very best regards to Drs. Soon and Legates, and to everyone,

w.

PS—For further discussion of these issues, see my posts The Tip Of The Gleissberg and The Effect Of Gleissberg’s Secular Smoothing.

Reply to  Willis Eschenbach
April 23, 2021 12:31 pm

I see two problems with the AI claim:
1) AI is just glorified curve fitting with no physics involved. And further mystified by the use of ‘hidden layers’.
2) The assumption that there are ‘cycles’ running on for centuries [millennia?]. The record is too short to demonstrate that, and there is no explanation of where the ‘memory’ of such cycles is stored.

Even if the ‘prediction’ turns out ‘right’ there are at least 20 registered ‘predictions’ of lower activity for SC25, see slide 25 of https://leif.org/research/How-to-Predict-SC-25.pdf so most will be right for the wrong reasons.

Reply to  Leif Svalgaard
April 23, 2021 2:16 pm

Thanks, Leif, your opinion on matters solar is always greatly appreciated.

For those who don’t know of Leif, he’s the man behind one of my many rules of thumb. This one is:

“Be very cautious when disagreeing with someone who has an effect in the field under discussion named after them.”

In this case, see the “Svalgaard-Mansurov effect” …

w.

Gary Kerkin
Reply to  Leif Svalgaard
April 23, 2021 4:42 pm

<blockquote>1) AI is just glorified curve fitting with no physics involved. And further mystified by the use of ‘hidden layers’.</blockquote>
At 80+ I’ve been getting to grips with machine learning or ‘AI’ (which is just a mystifying title!) and I am not expert in the subject. However, as I see it, the “hidden layers” are not really hidden, per se, and each has a function associated with it which modifies input to the element before it is passed on as output. This function, as far as I can see, might contain physical relationships. So, in my opinion, physics can be involved—I’m working on understanding that. As far as I can tell it appears to be a modern development of Richard Bellman’s concept of dynamic programming from the 60’s and 70’s. I used that method successfully in complex process control systems. I see no reason why AI can not be used similarly.

Not all AI methods use neural networks. Some use the CEEMD techniques Willis talks about.

There are a couple of points which might be confusing. In AI terms a “model” refers to an implementation of one of the methods used in the technique—which might be a neural network. It doesn’t refer to a GCM model. The other is that a particular implementation of an AI solution might involve processing many different models on the same data searching for most accurate result from the training data. Which is why AI approaches tend to be computing resources hungry. Still it is surprising what can be achieved on a laptop! Haven’t tried it on my iPad. Yet!

Greg
Reply to  Leif Svalgaard
April 26, 2021 3:03 am

The record is too short to demonstrate that,

Agreed, would you prefer a 9000 year “multimessenger” record ??

comment image

This power spectrum from the auto-correlation fn shows peaks at 114y and 67y .

The data is 10y resolution so the alleged 5.5y modulation is not detectable.

Last edited 1 month ago by Greg
Greg
Reply to  Greg
April 26, 2021 3:51 am

Correction periods marked are 87.6, 150y also 67y

Reply to  Greg
April 26, 2021 12:30 pm

Greg, I took a look at that dataset. The problem with it is that it’s calculated at 10-year intervals. That means that we get a spurious frequency at 10 years times the 11+ years of the sunspot cycle … which is the 114 year frequency shown in the spectral power density graph above.

In addition, what I find is that there isn’t much in the way of clear longer-term cycles.
comment image
comment image
comment image

This shows the advantage of CEEMD over Fourier analyses. You can actually see the cycles in question. In this case, as in many cases, they fade into and out of existence, with nothing solid like the 11-year sunspot cycle.

w.

Joel O'Bryan
April 23, 2021 11:31 am

The accurate Solar Cycle-sunspot number observational record of ~400 years is far too short to properly train an AI on the pseudo-chaotic flaring behavior of a 4.6 billion year star. The proxy reconstructions (10Be, etc) going back thousands of years can give us a general picture, but the low resolution doesn’t tell us much about details of a cycle or two on extremes. Additionally, 10Be can be significantly affected by solar energetic particle (SEP) fluxes, further “muddying” the major flaring frequency and confidence in that resolution.
https://www.swsc-journal.org/articles/swsc/full_html/2018/01/swsc170075/swsc170075.html

As more large data set continue to grow from stellar surveys on G-class stars, the one thing that is becoming clear — our Sun is a well-behaved G-class star with limited to no-super energetic flaring (energy > 10^33 erg) like what seems more common. G-class stars have surface temps of 5,200 K to 6,000 K, and our G2V class Sun is teff = 5,772 K.
By common, I mean superflares every few millennia or shorter, producing a major flare that would fry our pretty blue ball. Our Sun doesn’t seem to do that, evidence: we’re still here.

An important discussion on superflaring frequency can be found in Wikipedia on the subject of Superflares and G-class stars observed so far:
https://en.wikipedia.org/wiki/Superflare#G-type_stars

Age, rotation rate, specific surface temp above or below a teff = 5600K, and metallicity are all apparently important factors in categorizing frequency of superflares statistical distribution. Cooler G-class stars actually appear in the Kepler data to have more superflaring frequency than those above 5600K, like our own Sun (at teff = 5772 K).
The role of orbiting exo-planets, like close, hot gas-giants is still unknown, just as our own gas-giants role in the Sun’s 11-yr magnetic cycle is still a matter of much debate, even where we have 400 years of data, and 66 million years of biological evolution since the last great biosphere extinction event (from a bolide impactor).

So training an AI it is absolutely critical that the learning algorithms get “truthful” data if one is to expect anything that is not garbage on the output. And we simply do not have a long enough, high resolution record to understand the frequency and magnitude of large solar flares (energy > 10^32 erg), especially during periods of low SC-magnetic activity. Train an AI on a combination of truth and some amount of unknowable “lies”, in other words, and it will lie to you on the output. Guaranteed.

Last edited 1 month ago by joelobryan
William Astley
April 23, 2021 11:31 am

OK. The sun might as well have a sign on it… That flashes ….Danger! Danger! Danger!

30 years of stupid pointless discussions about atmospheric CO2 levels and planetary temperature.

Super large, Earth affecting weird solar events are coming.

As solar cycle 24 progressed, it is an observational fact, tha the large long-lasting sunspot in sunspot groups that have an average group lifetime of 22 days (in all of the other solar cycles which humans observed) have been replaced by a sunspot group of tiny sunspot ‘pores’ that has a lifetime of less than 10 days.

Why is it there are no real scientific presentation about this in your face solar crisis.

The following papers are few of the dozens of papers that find that solar cycle 24 is anomalous ….

Based on current detailed solar observations and astronomical observations of other solar like stars,…

The conclusion is based on observations of other stars, the sun is changing and will no longer produces an 11 year sunspot cycle.

What we did not find out is these stars (which do not have a solar magnetic cycle) about every 3000 years of so….

Get a very a single super, super, large sunspot, on their surface) which has a lifetime of about a year. The super, super large single sunspot on these stars ejects stuff which we assume is a normal sunspot flare. These flare like ejections are roughly 10 to a million times larger than solar flares that have been observed.

Are the sunspots really vanishing?

Anomalies in solar cycle 23 and implications for long-term models and proxies

Conclusions: Our results and interpretation show the necessity to look backwards in time, more than 80 years ago. Indeed, the Sun seems to be actually returning to a past and hardly explored activity regime ending before the 1955–1995 Grand Maximum, which probably biased our current space-age view of solar activity.

https://www.swsc-journal.org/articles/swsc/full_html/2012/01/swsc120019/swsc120019.html
The Sun in transition? Persistence of near-surface structural changes through Cycle 24

….. we confirm earlier findings that there appears to have been a change in the frequency response to activity during solar cycle 23, and the low frequency shifts are less correlated with activity in the last two cycles than they were in Cycle 22.

At the same time, the more recent cycles show a slight increase in their sensitivity to activity levels at medium and higher frequencies, perhaps because a greater proportion of activity is composed of weaker or more ephemeral regions.

This lends weight to the speculation that a fundamental change in the nature of the solar dynamo may be in progress.

https://arxiv.org/abs/1705.09099v1

“The Maunder minimum may represent one manifestation of the Sun beginning to enter this flat-activity phase.”

https://iopscience.iop.org/article/10.3847/2041-8205/826/1/L2/pdf

STELLAR EVIDENCE THAT THE SOLAR DYNAMO MAY BE IN TRANSITION

The source of disagreement between the age scales from asteroseismology and gyrochronology came into focus after scrutinized Kepler targets with precise ages from detailed modeling.

They confirmed the existence of a population of field stars rotating more quickly than expected from gyrochronology. They discovered that the anomalous rotation became significant near the solar age for G-type targets, but it appeared earlier in F-type stars and later in K-type stars.

Perhaps the most significant indication that the Sun is in a transitional evolutionary phase is its 11-year magnetic cycle, which falls between the active and inactive branches established by other stars (see Figure 2).

Even the slightly younger solar twin 18 Sco exhibits a normal cycle on the I-branch, while the Sun is the only star with a rotation period between 23 and 30 days that shows a cycle at all.

There were several Mount Wilson targets with rotation periods in this range, but they are all flat-activity stars. The Maunder minimum may represent one manifestation of the Sun beginning to enter this flat-activity phase.

 The fact that all of the slower rotators with cycles are K-type stars is perfectly understandable, since magnetic braking ceases in G-type stars before they reach these long rotation periods.

Joel O'Bryan
Reply to  William Astley
April 23, 2021 12:10 pm

Where they write: “Such a transition may represent the shift from a dominant αΩ dynamo to an α^2 dynamo, but this remains to be demonstrated.”

The problem is, “How many hundred millions of years must we wait for that transition to be confirmed or refuted?” At a minumum, any transitions are certainly in the range of many millions of years.

William Astley
Reply to  Joel O'Bryan
April 23, 2021 2:12 pm

You are missing the point. Big surprise. Have you looked at the paleo data and wondered what makes the cyclic changes in temperature? There is an imagination issue.

Our science assumed the sun was a simple fusion engine, a conventional star. That assumption appears to be incorrect.
 
The sun is a complex active object which is gaining mass… if it was a simple conventional fusion star, it would be losing mass.

An observation to support the assertion that the sun is a complex active object is the fact that based on a half dozen independent satellite radar observations the planet’s orbits are changing exactly as if the sun was gaining mass. This is an observational fact, not a theory or something that could be explained by an error or limitation in measurement. The paradox is real.
 
“If the reported increase is absorbed into a solar mass increase, and not into a changing gravitational constant G, the inferred solar mass increase is (6.0 ± 1.6) × 10 ^18 kg yr −1 .”
 
Astrometric solar-system anomalies
 
 https://www.cambridge.org/core/services/aop-cambridge-core/content/view/7FB105AC2EC0518570DF502BF3D00A17/S1743921309990378a.pdf/astrometric-solar-system-anomalies.pdf
 
“Astrometric solar-system anomalies
 
 However, rather than increasing, the AU should be decreasing, mainly as a result of loss of mass to solar radiation, and to a much lesser extent to the solar wind. The total solar luminosity is 3.845 × 10 26 W (Livingston 1999).
 
This luminosity divided by c ^2 gives an estimated mass loss of 1.350 × 10 17 kg yr −1 .
 
The total mass of the Sun is 1.989 × 10^30 kg (Livingston 1999), so the fractional mass loss is 6.79 × 10 −14 yr −1 . Again with the factor of three from Eq. 2.1, the expected fractional decrease in the AU is 2.26 × 10−14 yr −1, or a change in the AU of − 0.338 cm yr −1.
  
If the reported increase is absorbed into a solar mass increase, and not into a changing gravitational constant G, the inferred solar mass increase is (6.0 ± 1.6) × 10 18 kg yr −1 .
 
 This is an unacceptable amount of mass accretion by the Sun each year. It amounts to a fair sized planetary satellite of diameter 140 km and with a density of 2000 kg m −3 , or to about 40,000 comets with a mean radius of 2000 m.”
 
It can do what it does to serially force the geomagnetic field and the forcing of the geomagnetic field causes the abrupt climate change.
 
An observation to support that assertion is the discovery of geomagnetic excursions at all of the past abrupt large climate changes and recently at the abrupt small climate changes.
 
 Based on the paleo evidence there is a serial forcing event that appears every 16 million years, every 6000 to 8000 years, every 2,200 years, every 1400 years (beat of plus minus of 200 years) and every 500 years. If the sun is the cause then the sun changes on that frequency to cause what is observed.
 
Based on observations and systematic paradoxes (paradoxes which are structured) … It is appears it is an observational fact, that we live in a ‘steady state’ universe that has active complex objects which make atoms and shoot out stuff like baby complex objects which become ‘dwarf’ galaxies and which become large galaxies, like the Milky Way and Andromeda.
 
 An example of the paradoxical complex structure are globural clusters which have been found to occur in every galaxy and the number of these strange objects correlates with galaxy luminosity, except the baby dwarf galaxies have 30 times more and the large Cd galaxies which there is one of in every large super old cluster also have 30 times more globural clusters.

As discussed in the paper below, the globural clusters have a complex structure that cannot be produced by collapsing gas clouds or by nucleosynthesis. Absolutely physically impossible. No chance. Did not happen.

About 30 years ago a German astronomer Halton Arp discovered observational evidence that active galaxies (these are galaxies that have a quasar like object in their core that is producing massive amounts of dust and shoot stuff out of the poles of the massive object) were ejecting quasar like objects which…

… as there is a line of these objects… turned into quasars by seeing the connecting gas cloud using a combination of radio telescope observation and optical.

30 years later using new more powerful radio telescopes…which can see detail in the radio spectrum. Observations that Halton Arp has conceptually correct. Matter is being made ejected from complex objects.

There are now pictures of a massive electric current (light years long) that is travelling in the emitted gas from the ejected object back towards the AGN towards the ejected quasar (powerful radio telescopes now can ‘see’ the line of radio emission which is light years long pointing at from the AGN to the ‘knot’ which is emitting x-rays.

The ejected knot will become a baby complex object which will turn into a quasar and which will later become a dwarf galaxy which will then grow up to be a large galaxy like Andromeda or the Milky Way… which produces (the massive electric current from the AGN towards the knot) the most powerful emission at radio frequencies in the universe…. Emitting radio waves just like an antenna.

https://arxiv.org/abs/1503.03071v1
 
https://arxiv.org/pdf/1510.01330.pdf
 
 
Astronomy about 10 years ago that all globural clusters older than 1 billion years have two or more generations of stars in them. And all globural cluster younger than 1 billion years had only one generation of stars in them.

The big bang theory cannot make/explain the elemental patterns found in the second generation. These complex objects (and our sun if it is gaining mass has a complex object in it) is new physics … which is the new physics of a ‘steady state’ universe.
 
The second generation has exactly the same elemental composition except for elements lighter than calcium which are more and less than the first generation of stars.

The second generation is the majority. About 60% based on number. Gas clouds will not collapse and do not collapse in regions where there are clusters of tightly packed stars.
 
Ignoring that fact, it is impossible to make more light elements to explain the increase in some light elements and decrease in others in the second generation of star like objects

The globural clusters star like objects at tightly packed in a closed cluster with stellar density (the distance between star is roughly about the distance between the sun and Pluto) similar to that in the core a galaxy where there is a massive object.
 
 

Vuk
April 23, 2021 11:58 am

There is a fundamental problem in their assumption that half cycle is 5.5 years. Sunspot cycle duration alternates about every 100 years between ~10.4 and 11+ years

SS.gif
n.n
Reply to  Vuk
April 23, 2021 4:04 pm

Assumption, assertions, an AI model (i.e. hypothesis), but is it plausible? Is it politically congruent? That is, can it be exploited for political, social, and economic leverage?

April 23, 2021 5:01 pm

Was that picture of the Sun done by Vincent van Gogh?

Vuk
Reply to  nicholas tesdorf
April 24, 2021 8:04 am
Last edited 1 month ago by Vuk
Editor
April 23, 2021 5:23 pm

Like some other commenters, I don’t buy it. 1. Changing past data. 2. Opaque logic (how was the AI trained?, has the changing of the past led to circular logic?, eg.). But at least the predictions are out there for testing. It seems there is high divergence between scientists now, which is a good basis for progress. Pity it will take at least 5 years for the testing, but that’s a lot better than the totally-protected-from-testing climate models.

April 23, 2021 5:34 pm

> That interaction between the two periodicities led the algorithm to indicate that from the 1730s to the 1760s, early in the modern sunspot record (the gray band below), sunspots appear to have been under-recorded

Not an entirely new discovery 🙂

https://link.springer.com/article/10.1007/s41116-020-0023-y

It so appears that this subject, as far as I can judge especially according to the Paris memoirs, was cursorily very rarely observed since 1719, and since 1722 in particular, rather out of amusing speculation than out of true scientific zeal.

“Coverage of decades by days with observations, adapted from Vaquero et al. (2016). The numbers include all sunspot observations, also the ones without drawings, and zero-spot detections”

:

SunspotCounts.jpg
April 23, 2021 6:40 pm

“AI” is just another term for “software”.

Greg
April 25, 2021 2:50 am

That interaction between the two periodicities led the algorithm to indicate that from the 1730s to the 1760s, early in the modern sunspot record (the gray band below), sunspots appear to have been under-recorded: as the 120-year cycle approached its maximum amplitude, sunspots should have been more numerous than reported at the time.

OH dear. More models being used to “correct” observations. If your model is contradicted by observation you review the MODEL not the DATA.

If this indeed reflects the content of paper rather than BS added by press release, I’m disappointed in Soon and Legates.

OK paper says:

It also suggests the possibility of missing sunspots during Sunspot Cycles1, 0

That gets translated into “appears to have been under-recorded”. More chinese whispers by uneducated media studies students writing pressers.

Last edited 1 month ago by Greg
Greg
April 25, 2021 2:51 am

The three scientists taught a machine-learning algorithm how to recognize underlying patterns and cycles in the past 320 years’ sunspot record.

I thought the whole point of AI was that you did NOT train it how to do something.

Ulric Lyons
April 25, 2021 11:55 am

Centennial solar minima occur on average every 107.9 years, but varying from around 80 to 130 year intervals. So a 120 year envelope is too long and cannot capture the variability either.
I can discretely map every centennial minimum by these correlations, and this centennial minimum is as short as they ever get, affecting only solar cycles 24 and 25.

https://www.linkedin.com/pulse/schwabe-cycle-variability-ulric-lyons

%d bloggers like this: