$10 Million AI Grant to Improve Climate Modelling of Clouds

Guest essay by Eric Worrall

A Team of Scientists plans to use AI to improve climate modelling of sub-grid scale phenomena such as turbulence and clouds. But there is no reason to think an AI will have any more luck than human climate modellers.

International Collaboration Will Use Artificial Intelligence to Enhance Climate Change Projections

MARCH 23, 2021 10:18 PM AEDT

A team of scientists, backed by a $10 million grant from Schmidt Futures, will work to enhance climate-change projections by improving climate simulations using artificial intelligence. 

$10 million effort, backed by Schmidt Futures, to be led by NYU Courant Researcher

A team of scientists, backed by a $10 million grant from Schmidt Futures, will work to enhance climate-change projections by improving climate simulations using artificial intelligence (AI).

Led by Laure Zanna, a professor at New York University’s Courant Institute of Mathematical Sciences and NYU’s Center for Data Science, the international team will leverage advances in machine learning and the availability of big data to improve our understanding and representation in existing climate models of vital atmospheric, oceanic, and ice processes, such as turbulence or clouds. The deeper understanding and improved representations of these processes will help deliver more reliable climate projections, the scientists say.

“Despite drastic improvements in climate model development, current simulations have difficulty capturing the interactions among different processes in the atmosphere, oceans, and ice and how they affect the Earth’s climate; this can hinder projections of temperature, rainfall, and sea level,” explains Zanna, part of the Courant Institute’s Center for Atmosphere Ocean Science and a visiting professor at Oxford University. “AI and machine-learning tools excel at extracting complex information from data and will help bolster the accuracy of our climate simulations and predictions to better inform the work of policymakers and scientists.”

Due to the complexity of the atmosphere, ocean, and ice systems, scientists rely on computer simulations, or climate models, to describe their evolution. These models divide up the climate system into a series of grid boxes, or grid cells, to mimic how the ocean, atmosphere, and ice are changing and interacting with one another. However, the number of grid boxes chosen is limited by computer power; currently, climate models for multi-decade projections use grid box sizes measuring approximately 50 km to 100 km (roughly 30 to 60 miles). Consequently, processes that happen on scales that are smaller than the grid cell–clouds, turbulence, and ocean mixing–are not well captured.

Read more: https://www.miragenews.com/international-collaboration-will-use-artificial-532909/

Top marks for admitting model temperature projections struggle to capture important processes. Clouds, storms, ocean mixing and turbulence are likely the reason open ocean surface temperatures in the tropics are capped at 30c.

But why do I think the AI approach will struggle to improve on human efforts?

The reason is decades of effort to improve understanding the global climate has not answered basic questions, like how much does global temperature change in response to adding more CO2, and human brains are far more powerful than any AI.

AIs work best when the solution is easy to approach, when a gentle slope of improving results provides a strong indication to the AI that it is making progress.

A gentle slope guides the AI towards the optimum solution.

But I do not think this is a good description of the climate system. The lack of progress over the last three decades, despite thousands of intelligent people dedicating years of their lives to the effort, implies the solution to better climate modelling is very difficult to find. Outside the narrow range of correct solutions there is likely a vast wilderness of poor quality answers, with very little indication of which direction the AI needs to travel to discover a high quality solution.

Either that, or there is something fundamental missing from the theory, and a high quality solution will not be possible until the missing piece of the puzzle is found.

Even a powerful AI struggles to search a multi-dimensional problem space when the correct answer is poorly signposted – there are many more ways to be wrong than right.

4.9 15 votes
Article Rating
Newest Most Voted
Inline Feedbacks
View all comments
March 24, 2021 10:03 am

Pardon the brevity but:


John Tillman
Reply to  Rob_Dawg
March 24, 2021 10:32 am

Actually to model clouds would require at least 10,000-fold increase in computing power. An average cumulus cloud is about a kilometer in width, length and height. Present GIGO GCM grid cells measure around 100x150x1 km. Thus, getting cells of 1×1.5×1 km needs 10,000 times more power. But many clouds are smaller than a cubic km, so 100-meter or 10-meter grid cells would be better. Hence ideally 10 million to 10 billion times present computational crunch.

Paul Chernoch
Reply to  John Tillman
March 24, 2021 11:51 am

Not necessarily. If you use an adaptive grid size that is tighter where the phenomena are more extreme (thunderstorms, hurricanes) and looser where weather is more uniform, you can make tremendous improvements in accuracy with less than that 10,000 fold increase in resources. You might need only 100x more computing power to generate smaller gridded-garbage from your garbage input.

Tim Gorman
Reply to  Paul Chernoch
March 24, 2021 4:30 pm

Hmmmm, just how do you decide where to use a tighter grid and where to use a lower grid? Thunderstorms happen almost everywhere on the globe except the Arctic and Antarctica. Hurricanes are temporary phenomena, how do you decide where to insert one over 100 years of time?

There simply aren’t very many places that I know of where the weather is “more uniform”, especially over the long haul.

You need the smaller grids in order to determine where weather phenomena are *going* to happen, where the non-uniformity might occur.

Reply to  Tim Gorman
March 24, 2021 5:23 pm

Back in the early 1970s when I programmed at NOAA in Boulder, Colorado, meteorologists who were just beginning to grasp chaos theory told me that any forecasts beyond two weeks would have to take into account (among many other phenomena) all the little dust devils occurring on the planet — and the grid size for that is very small!

Reply to  Paul Chernoch
March 24, 2021 6:21 pm

and where they like in space and time?

Pat Frank
Reply to  John Tillman
March 24, 2021 11:52 am

And then one would need a very good physical theory of aerosol coalescence around the variety of nano-particulate solids found aloft.

John Tillman
Reply to  Pat Frank
March 24, 2021 12:57 pm

Hence, more computing power to model each type of CCN at level in the atmosphere, class of cloud, water v. ice, etc. Many parameters. So needed increase in computing power more likely on the order of a billion or higher.

Reply to  Pat Frank
March 24, 2021 2:19 pm

Missing links are a minor detail, which can be inferred, infilled, smoothed, and created from whole cloth when politically congruent.

Reply to  Pat Frank
March 24, 2021 4:30 pm

“aerosol coalescence around the variety of nano-particulate solids found aloft”

It is always a pleasure to see the correct and precise terminology

Robert W Turner
Reply to  John Tillman
March 24, 2021 1:00 pm

Not to mention start with the correct principles of physics.

Reply to  John Tillman
March 24, 2021 2:04 pm

Then there is the problem of initial conditions for those 10000 times as many grid cells. They are already interpolating most of the initial conditions as most of the grid cells have no weather observations at all. Let’s not even get into the whole aspect of whether the parameters are complete on the grids with observations.

Jeff Alberts
Reply to  OweninGA
March 24, 2021 3:43 pm

Interpolating=making stuff up.

Alan M
Reply to  Jeff Alberts
March 24, 2021 5:17 pm

Extrapolating=making even more stuff up

Reply to  John Tillman
March 24, 2021 4:29 pm

Only the bulk behavior of clouds can be efficiently modeled. You can’t model clouds independently and hope to get an answer. The reason is that clouds are chaotically self organized where each cloud is codependent on every other.

Cloud behavior can be modeled as driving towards a strange attractor, which is to maintain a constant ratio between the SB emissions of the surface and the emissions at TOA. This ratio converges to about 1.62, where it takes 1.62 W/m^2 of surface emissions to result in 1 W/m^2 emitted at TOA. This ratio is remarkable constant from pole to pole, varying by only a few percent, while the averages per hemisphere are identical to within the error margin of the data used to calculate it. In addition, this ratio show no statistically relevant trend whatsoever.

There is some compelling math that explains why this particular ratio emerges. When you consider the RADIANT transform of the atmosphere as a chaotically varying 2×2 transform, this ratio has infinitely more ways to be achieved than any other, providing the chaos spans this ratio. The inputs are solar input at TOA and the energy replacing surface emissions at the surface. while the outputs are planet emissions at TOA and the SB surface emissions corresponding to its temperature.


John Tillman
Reply to  co2isnotevil
March 24, 2021 7:04 pm


Strange attractor:
comment image

Reply to  John Tillman
March 24, 2021 7:44 pm

Hmmmm… Strange indeed, but I’m not attracted… at all.

Anyone feel a tug in his direction? Anyone? Buehler? Buehler?

John Tillman
Reply to  H.R.
March 25, 2021 10:09 am

The culprit attracts media attention.

Reply to  John Tillman
March 24, 2021 9:47 pm

You should be cancelled for posting that. Please don’t do it again.

John Tillman
Reply to  philincalifornia
March 25, 2021 8:22 am

Know your opponent!

Reply to  John Tillman
March 24, 2021 10:13 pm

more like an immediate gag response. !

Reply to  John Tillman
March 25, 2021 5:15 am

There is no such thing as an average cumulus cloud or any cloud for that matter. In the lower levels you get cumulus, around 10,000′ you encounter alto cumulus, around 30,000′ and above you get strato cumulus and with the right condition for storms you get cumulonimbus reaching from ground level to 50,000′ and upwards. Clouds continually change shape and area coverage at all altitudes and are primarily formed by the pressure patterns which give rise to the frontal systems. Sometimes the winds can be 180 degrees in the opposite direction only 5000′ apart and this will change cloud cover and formation. There are not clouds existing in small one cubic km areas because the conditions of pressure and temperature for cloud to form, don’t just exist in such a small area. Other considerations can involve fog and temperature inversions and what the pressure systems are doing at height. With 56 years in aviation and a long airline career I have seen a line of thunderstorms 1300 klms long and on some occasions 400 mile lines of storms with tops at 70,000′ so any theories of clouds forming in any sort of average way is not what happens in reality. Nor can it be modelled.

John Tillman
Reply to  R K
March 25, 2021 7:31 am

Of course many clouds are bigger than a kilometer.

Regardless of any average, the fact is that most clouds are far smaller than the grid scale of presenr GCMs. The resolution needed to model them is orders of magnitude greater than now available.


“Cumuliform clouds typically have diameters roughly equal to their depths, as mentioned previously. For example, a fair weather cumulus cloud typically averages about 1 km in size, while a thunderstorm might be 10 km.”

Modelling wouldn’t assume any average, but grid size needs to be small enough to model little as well as large clouds.

Last edited 1 year ago by John Tillman
Reply to  John Tillman
March 25, 2021 2:07 pm

In answer to your reply, the guy in the link you gave does not know what he is talking about. As a Professor from British Columbia he works in an office and the weather does not exist as averages produced by formulas as he suggests. He is mentioning fair weather cumulus in that link but when they form they develop in all sorts of shapes, sizes and numbers. He states ” not all clouds are created equal “. Well I have news for him, none are. Cumulus clouds exist as different forms as I have stated because of height and develop over wide areas. He obviously knows nothing about thunderstorms – very few would just develop to 10 klm and exist by themselves. Whilst isolated thunderstorms do exists from time to time where they are formed by convection, nearly all thunderstorms form from frontal systems and continue to build and join together in lines.The article states that they will model turbulence as well as cloud – well that shows they really are delving into areas they know nothing about.
To give you some idea of how deep and wide ranging cloud can be and why you can’t model it, I have taken off from Brisbane for Singapore and gone into cloud shortly after take off and been in continuous cloud at 38,000′ until Darwin, around 2000 miles. A fair bit of cumulus cloud in that situation. You only have to look at the frontal systems moving across the globe all the time and realize all the changing pressure and temperature that is occuring to conclude that you can’t average it or model it.

Reply to  Rob_Dawg
March 24, 2021 2:17 pm

Perfect, some would say catastrophic, in its simplicity, while capturing the inclusive spectrum of this hard problem, special and peculiar conflicts of interest, and the forward-looking wicked solution(s).

Jay Willis
Reply to  Rob_Dawg
March 25, 2021 1:30 am

GIGO indeed. But imagine if this AI actually worked. It puts me in mind of a quote of a Dutch friend: “if you make a sufficiently complex model of a river, they’ll only be two things you don’t understand. The model and the river.”

Reply to  Rob_Dawg
March 25, 2021 2:57 am

Itsa da cloud computing!

March 24, 2021 10:13 am

More wasted tax monies. AI is only as good as the programmers’ knowledge base. The models can’t even replicate passed ice age cycles correctly.

Jeff Alberts
Reply to  John Shewchuk
March 24, 2021 3:45 pm

Does AI really exist? Is there a system that knows when it has made a mistake, learns from that mistake, and doesn’t make it again? Is there a system that betters itself as it learns, without continual poking and prodding from programmers?

Reply to  Jeff Alberts
March 24, 2021 5:30 pm

Yes, AI involving neural networks with highly accurate input and feedback from data taken directly from meteorological instruments during its training stages should theoretically work. What won’t work is biased human feedback during its training stages.

Jeff Alberts
Reply to  noaaprogrammer
March 24, 2021 6:42 pm

I’m skeptical.

Captain Climate
Reply to  noaaprogrammer
March 25, 2021 5:26 am

No it won’t work. The inputs are not granular enough.

Reply to  John Shewchuk
March 24, 2021 5:12 pm

I tried to use neural nets to predict credit card fraud in a past life, and it a real eye-opener. Training killed our effort because there wasn’t enough fraud. It was interesting how sensitive it was to inputs, and I’d bet a donut they will train it to get the results they want. What a waste!

Reply to  Felix
March 24, 2021 7:58 pm

You’re saying things along the lines I’m thinking, Felix.

If you wind-up your AI and set it down saying, “It’s CO2 wot dunnit. Go figure out how and how much,” imagine where that AI will wind up.

Quelle surprise! It’s CO2 wot dunnit, and we’re all going to die!

You could save enormous amounts of money if the climate models were just

For any variable:
Do while
CO2 Exists
Print “We’re all gonna Die!”
End Do

Reply to  Felix
March 25, 2021 12:03 am

“I’d bet a donut they will train it to get the results they want. What a waste!”. Maybe it will be a lot worse than that. They can already safely pump out model predictions as if they were gospel, knowing that the models are too complex for any criticisms to be effective. With AI they can do it in spades.

I do agree with other comments on AI, that this is not suitable for AI. But if someone offers $10m ……. I won’t be betting against your donut.

Pillage Idiot
March 24, 2021 10:15 am

The analogy I use for my non-science friends, as regards climate modelling:

Shuffle a deck of cards, let your “model” and the supercomputer predict the card on the top of the deck.

For the next model prediction, replace the card and fully re-shuffle the deck. Let the model again predict the top card.

If you have a process that is random OR for which you do not understand the underlying phenomenon, then you cannot predict the results by adding more computing power or fine-tuning your model!

March 24, 2021 10:16 am

Somebody check on Willis. I think I heard something loud.

Paul Johnson
Reply to  ResourceGuy
March 24, 2021 1:50 pm

What happens if the AI proves that Willis is right?

Reply to  Paul Johnson
March 24, 2021 1:58 pm

Will that be allowed? One must find ways to tell AI what its boundaries are with some heavy assumptions on the front end of the process to keep the climate gods happy. Call it the cloud rhythm method.

Steve Case
March 24, 2021 10:20 am

Kip Hansen put this one up a year or two back:

Besides that:

Due to the complexity of the atmosphere, ocean, and ice systems, scientists rely on computer simulations, or climate models, to describe their evolution.

Golly, why wouldn’t you also check against the empirical record?

Reply to  Steve Case
March 25, 2021 2:54 am

Steve, that’s a perfect example of a built in bias ( in this case it’s gravity).
It appears to show random chaotic effect from two parameters (a coupled non-linear chaotic system) …
but the result is always the same … when the energy is exhausted =
equilibrium, vertical at rest.

Now alter the lengths & weights of the pendulums, have more or less friction on pins
… initially the patterns will be wildly different
but the result is always the same … equilibrium, vertical at rest.

Then try a triple pendulum – https://www.youtube.com/watch?v=dDU2JsgLpm4
but the result is always the same … equilibrium, vertical at rest.
because the bias is built in.

That’s why climate muddlers get it consistently wrong, they don’t see the bias, only the answer they seek.

Neal in Texas
March 24, 2021 10:21 am

Since AI systems require a truth model to tune against, the key question is “who gets to determine the truth?” Whoever is selected to pick this model automatically inserts bias into any answer the AI system ultimately spits out.

Reply to  Neal in Texas
March 24, 2021 2:21 pm

It is common practice, past, present, and progressive, for science to take a knee to truth(s).

Reply to  Neal in Texas
March 24, 2021 3:56 pm

we have a winner…..

March 24, 2021 10:21 am

Eric, here in the real world, in the time that I took to open WUWT and read this post, the sky outside my window changed from sunny to cloudy.

How do they plan to capture these changes with models?

UPDATE – in the time it’s taken me to write this comment, the sky outside my window is back to sunny. And I reckon the temperature here went up as well.

(This effect may be unprecedented. If I could get some grant $$$s, I’d be happy to sit here for as long as it takes to determine if this is a regular event or not. I accept PayPal)

Reply to  Mr.
March 24, 2021 10:55 am

I remember to have seen a picture about contrails, a region south of Himalaya, if I remember well. There was a region, formed like a circle full with contrails of all directions , around that “circle” the air was clear..The surrounding air had other conditions as the air in the circled region, differnt humidity at least. So cloud or contrail forming is weather dependent, possible forecast max. a week. Will not be easy to project it on climate behaviour over years, just when considering the differnt hight of different clouds.
Best guesses will be the mainstream of these models.

Reply to  Mr.
March 24, 2021 1:42 pm

“How do they plan to capture these changes with models?”

I assume they’ll do something similar to how they “capture these changes” when infilling temperatures in grids without weather stations for weighted averaging of the global “average” surface temperature… how do they know whether it was sunny or cloudy 300 miles away from the nearest station? They don’t.

Reply to  BobM
March 24, 2021 2:16 pm

Exactly Bob.
I have a friend who lives about 60 miles direct line from me, but over the opposite side of a 5,000 ft mountain range, where he gets the unimpeded exposure to the westerly atmospheric flow from the eastern Pacific.
I’m on the “sheltered” side.

We agree that we live in very different climates.

The indigenous people figured out through observations aeons ago that ‘my’ side of the range was the most comfortable place to live. And so they did.
In contrast, my friend’s locale is very sparsely inhabited (by humans at least).

See, even just 60 miles away, it’s a different world, climatically speaking.

Globally model that, climate carpetbaggers.

Tim Gorman
Reply to  Mr.
March 24, 2021 4:38 pm

My brother and I have the same thing on opposite sides of the Kansas River Valley, about 30 miles apart. You don’t require a 5000ft mountain to cause a difference.

Gerald Machnee
Reply to  Mr.
March 24, 2021 3:40 pm

As one of my co-workers said whole on the phone to media “but clearing very rapidly”.

March 24, 2021 10:26 am

“But why do I think the AI approach will struggle to improve on human efforts?”

Because the Climate Change Scientists don’t want the answer to be known.

Because it would show that the processes, that are involved in the formation and dissipation of clouds and that increase exponentially with temperature, represent, at temperatures above ~ 20C, an increasingly strong, negative feedback that destroys the whole theory of CO2’s influence on climate.

Reply to  dh-mtl
March 24, 2021 12:47 pm

I would guess if you gave AI access to all of the raw temperature data and all of the possible theories (eg, including solar perturbation) and asked it to combine all of the theories to produce the best possible fit with temperature observation, the carbon dioxide hypothesis would be kicked to the curb (along with Michael Mann).

Tim Gorman
Reply to  Anon
March 24, 2021 4:42 pm

The so-called AI’s of today don’t really do “theory”. They are data miners, looking for relationships between data.

What the AI’s will likely find that the use of mid-range temperatures from every station is a misuse of time series data. The *real* mid-range temperature of the globe has to consider half of the globe being in sunlight and half in dark. That means you must use time consistent data – which is not the practice in with the climate models today.

March 24, 2021 10:27 am

So the team of scientists are planning on using AI to codify GIGO?

March 24, 2021 10:28 am

And what makes people think artificial intelligence is any better than human intelligence? It may, or may not, make mistakes quicker than those that program it.

Sweet Old Bob
March 24, 2021 10:32 am

. “AI and machine-learning tools excel at extracting complex information from data and will help bolster the accuracy of our climate simulations and predictions to better inform the work of policymakers and scientists.”

So , the projections will only be 99% wrong vs 100% ?


Schrodinger's Cat
March 24, 2021 10:33 am

Cognitive bias is a powerful influence.

Thousands of climate scientists are never going to admit to their employers that the climate is remarkably stable, extreme weather is an exaggeration and the absorption bands of the greenhouse gases are already saturated.

Steve Case
Reply to  Schrodinger's Cat
March 24, 2021 11:51 am

And whatever they are claiming has happened before.

Reply to  Schrodinger's Cat
March 24, 2021 4:14 pm

“the climate is remarkably stable”

Until it reaches a tipping point then flips, then flops back.
comment image

” the absorption bands of the greenhouse gases are already saturated.”

That old chestnut? You misunderstand GHG.

Last edited 1 year ago by Loydo
Tim Gorman
Reply to  Loydo
March 24, 2021 4:45 pm

You don’t consider the graph you showed to depict a stable condition? A pendulum *is* a stable condition. It doesn’t all of a sudden fly out of the grandfather clock and hit someone!

Reply to  Tim Gorman
March 24, 2021 4:51 pm

I don’t think Schrodinger’s Cat was thinking of a pendulum when he said “remarkably stable”.

Reply to  Loydo
March 24, 2021 10:20 pm

I don’t think you have the remotest clue what “stable” means

You cretinally are not.

Reply to  Loydo
March 24, 2021 6:25 pm

stop whining! prove or disprove his point with evidence.

Reply to  Lrp
March 24, 2021 9:50 pm

Evidence, ha ha. You’re talking to Loydo, This site’s village idiot.

Reply to  Lrp
March 25, 2021 2:17 am

But, but, but Loydo supplied the evidence! A remarkably stable oscillation he graphed there, don’t you think?
But then again, evidence is not proof… Like his retort about Schroedinger’s cat: Clear evidence the man in insane, but actually he’s just trolling!
Although his trolling is no proof of his sanity… just evidence of his loneliness.
Ya’ll have to admit, his posts near always leads to long and involved side conversations?
Look how long he kept us busy now, you and me both!

Reply to  Loydo
March 24, 2021 10:20 pm

How would a clueless clown like you have the vaguest idea about GHGs?

You have made it totally clear that you have ZERO understanding.

CO2 radiation mean free path in lower atmosphere is 10m or so.

So, Yes it saturates very quickly.

Remain CLUELESS,loy-dodo.. its all you can do.

And as you can clearly see, PEAK CO2 is followed by COOLING always

Why do you walk around with a target pointed on your forehead, loy-dodo.

Are you so DUMB that you forgot you painted it there ?

John Garrett
March 24, 2021 10:33 am

“Schmidt Futures” ????

…as in Eric and Wendy Schmidt ???

…as in the Schmidt Family Foundation ???

…as in Google ???

…as in http://www.theschmidt.org/ ???

“…In addition, The Schmidt Family Foundation’s investment portfolio has been divested of virtually all investments in fossil fuel related industries and businesses…”

“…The Energy Program focuses on two, linked goals: challenging the development of fossil fuels and accelerating the adoption of renewable energy.
Climate science indicates that in order to avoid catastrophic climate change, the majority of known fossil fuel reserves cannot be produced. But beyond just the climate necessity for such action, a rapid transition away from fossil fuels is critical…”

Wendy Schmidt:
“…Wendy Schmidt is a philanthropist and investor who has spent the past 14 years creating innovative non-profit organizations to address challenges facing communities around the world, working for clean, renewable energy, healthy food systems, healthy oceans and the protection of human rights. The critical interconnections between human activity, the land we live on and the ocean we depend upon are the central drivers of Wendy’s philanthropic work.

Wendy is president of The Schmidt Family Foundation, which she
co-founded with her husband Eric in 2006. She leads the foundation’s two grant-making and investment programs—The 11th Hour Project, which works to create a just world where all people have access to renewable energy…”

Last edited 1 year ago by John Garrett
Reply to  John Garrett
March 24, 2021 12:33 pm

HMMMMM…seems to be same goal as Billy & Melinda Gates Foundation….to Zuck up the entire world….sooner rather than later.

Rory Forbes
Reply to  Anti_griff
March 24, 2021 2:21 pm

Their goal appears to be inserting expectation bias (of the globalist/authoritarian slant) wherever they can.

Curious George
March 24, 2021 10:38 am

AI will be a huge improvement. It will double an average IQ of climate researchers.

Reply to  Curious George
March 24, 2021 10:57 am

Triple at least for three-digit.

Reply to  Krishna Gans
March 24, 2021 12:04 pm

only just, then they have to figure how to use it.

Frank from NoVA
March 24, 2021 10:39 am

It’s not clear if they’ll be using AI to improve their understanding of the GCM inputs, in which case more power to them, or the GCM outputs, in which case it’s just more GIGO. In either case, until there exists enough computing power to significantly reduce the size of the grid cells, the GCMs will remain fundamentally flawed.

March 24, 2021 10:42 am

What you describe as AI is in fact just an optimisation, like a levenberg Marquardt. AI, or deep learning has its place, but to my knowledge you have to know the answer to train the AI to get it right. It requires a large amount of data, of high quality for training. The measurements in the small cells will not be sufficient for them to use any kind of deep learning algorithm to correctly know what is happening, regardless of the chaos involved with a non linear system.

Robert of Texas
March 24, 2021 11:02 am

I spent my entire life in computers, programming, modeling, architecture. The most likely outcome is an AI system that appears to produce more lifelike cloud formations but it likely will have nothing to do with how nature really works. There also will be a complete lack of understanding – no “ah-ha! That’s how that works, just a bunch of probabilities and numbers.

When you teach an AI system to recognize faces, you only care that it produces a better (more accurate) output than a traditional computer system. It has to be taught somehow – whether before hand or on the job. You do not really care “how” it works, only that it works. Facial recognition using AI likely has NOTHING to do with how a human brain performs the same task.

Now apply that to clouds. You actually DO care how it works or you learn nothing. The computer system can produce more likely cloud formations in more likely places, but it isn’t based on anything you understand – it just happens to work. It may have nothing to do with how nature works…so it’s use in prediction is perilous. If it has never seen certain conditions before, it may produce an unexpected outcome. If not biased to prevent it, an AI system can become chaotic. If biased against this, it can never return correct results under unexpected boundary situations or if the natural system behaves chaotically.

Last edited 1 year ago by Robert of Texas
Rory Forbes
Reply to  Robert of Texas
March 24, 2021 2:29 pm

That was one of the best explanations of the drawbacks that AI systems are burdened with and why real world climate conditions are out of their range.

March 24, 2021 11:05 am

The same limitation will remain in that any climate modelling needs to demonstrate ability to backcast known historic conditions and clouds are unique in that there is no historic data for cloudiness that correlates with the time frames of other inputs.

March 24, 2021 11:08 am

AI will not fare any better simply because AI is a human contrivance. Just like the models are.

I can think of better things to throw $10 million at even if they can’t.

Reply to  fretslider
March 24, 2021 1:40 pm

This is stimulus money 😉

Jean Parisot
March 24, 2021 11:16 am

They will spend years trying to figure out why the AI is wrong; hopefully the CO2 hysteria will pass and we will be worried about asteroid impacts or alien invasions to generate funding.

Reply to  Jean Parisot
March 24, 2021 11:37 am

If only you could trust AI, but you can’t, apparently…

Microsoft shuts down AI chatbot after it turned into a Nazi

the company launched “Tay,” an artificial intelligence chatbot designed to develop conversational understanding by interacting with humans. Users could follow and interact with the bot @TayandYou on Twitter and it would tweet back, learning as it went from other users’ posts. Today, Microsoft had to shut Tay down because the bot started spewing a series of lewd and racist tweets.


Reply to  fretslider
March 24, 2021 11:44 am

Ha ha – have they been arrested yet for murdering a poor little chatbot.

chatbot’s lives matter.

Reply to  philincalifornia
March 24, 2021 11:50 am

They do to Bill

Reply to  fretslider
March 24, 2021 4:18 pm

They are as good or bad as their teachers are 😀

Last edited 1 year ago by Krishna Gans
Chris Nisbet
March 24, 2021 11:16 am

“Despite drastic improvements in climate model development, current simulations have difficulty capturing the interactions among different processes in the atmosphere, oceans, and ice and how they affect the Earth’s climate”.
Isn’t the whole purpose of a climate model to capture interactions among these different processes (in order to predict the result of them)? What are they doing (or failing at doing) if not that?
Isn’t ‘climate’ one massive pile of interactions between different processes and entities?

Reply to  Chris Nisbet
March 24, 2021 7:10 pm

And, that’s after the “drastic improvements”. Imagine how bad they must have been about capturing the interactions between the different processes prior to said “improvements”.

That got me wondering though, how drastic have these improvements been? And, how would they measure the “drasticity” of these improvements? Perhaps you could use the range of ECS …. wait, nevermind.

Reply to  Chris Nisbet
March 24, 2021 9:17 pm

These “drastic improvements” gave us CMIP6. 😉

March 24, 2021 11:21 am

I can write the last sentence of the Abstract of the first paper right now:

The robust conclusion of the ensemble modeled simulations showed that clouds represent a positive feedback to climate forcing by anthropogenic carbon dioxide emissions.

Just gimme a million of that and I’ll shut up. Promise.

Reply to  philincalifornia
March 24, 2021 7:22 pm

Do you add the “further research is required” bit here, or does that go in the Conclusions section?

E. Schaffer
March 24, 2021 11:28 am

Spoiler alert: 42!

Serious, how do you want to model clouds, if you don’t even understand what their effect on climate is? All satellite “data” on cloud forcing (CF) so far are utter non sense. Let me tell you why.

SWCF is defined as difference between all sky and clear sky albedo. Accordingly LWCF is defined as difference between all sky and clear sky emissions. This erroneous approach means, that while clouds do reflect some ~70W/m2 of solar radiation, their SWCF is only ~50W/m2. The sun light reflected by clouds, which counterfactually would be reflected by the surface in their absence, is not allocated to SWCF (thus the difference).

With LWCF we have the same problem, but it is far worse! GHGs and clouds are largely overlapped in their emission reducing effect. Both together reduce emissions from 355W/m2 (surface at 288K with 0.91 emissivity) to only 240W/m2. With clear skies emissions are ~270W/m2 btw.. Of a total GHE of 115W/m2 30W/m2 are exclusively allocated to clouds. With the remaining 85W/m2 about 50-60W/m2 are overlapped GHGs/clouds and only the remaining 25-35W/m2 are exclusive to GHGs.

One of the many mistakes in assessing the GHG-effect is to say, the overlapped part was only due to GHGs. It is not. Another one obviously is to claim the surface was a perfect emitter. Anyhow, the overall effect of clouds is warming, not cooling the planet. And the GHE of GHGs is actually quite small.

Charles Fairbairn
March 24, 2021 11:50 am

Unless these scientists extract their brains from the current GHE/Radiation mindset and include the basic science of enthalpy exchange and movement, particularly of water, other than by radiation, They will get nowhere. Other influences are involved and should not be omitted.

The AI programme should have the capability to calculate the way the evaporation process works and how the energy involved is transmitted UP through the atmosphere for dissipation with some to space. Much of this information is available when considering the workings of the Rankine Cycle which provides the basis for understanding how the Hydro Cycle works.

For instance: If the program considers Water provides a positive feedback to the GHE then they will continue to chase a wild goose; just more expensively. The trouble here is that the Groupthink Mindset is preventing thinking outside the box. GIGO comes to mind.

Information gleaned just from data sets merely tells what has happened over a chosen period of time. Who chooses this? It can never tell you WHY it has happened. That requires a return to the basic science of each and EVERY influence involved and how thay interact. Indeed the recipe for chaos, putting prediction to the level of conjecture.

Pat Frank
March 24, 2021 11:51 am

Everyone is looking for a deus ex machina in AI, to export their decision-making.

It’s a dangerous fantasy, as though pure truth will come from a dispassionate computer, uncorrupted by politics.

AI can correlate inputs and look for recurrent patterns. But it won’t do conjecture and refutation in our lifetimes.

No AI will predict the climate any better than climate physical theory allows. And climate physical theory presently allows no predictions of climate states.

Reply to  Pat Frank
March 24, 2021 11:58 am

Everyone is looking for a deus ex machina in AI, to export their decision-making.

We have that now, the political elite have transferred policy/decision making to the Experterati

If it goes wrong….

Last edited 1 year ago by strativarius
March 24, 2021 12:03 pm

Even though Schmidt futures is technically private (looks like it is tax exempt), NYU is not. Remember the Johnny 5 principle: “It just runs programs.” An AI (or any) software can never be smarter than the IQ of the smartest programmer divided by the number of programmers on the team. Then half that number for government funded software. But, being the product of a government funded team, there is no one responsible for failure. Winning all around!
In a sane world, the grant proposal and all documentation, and source code, should be subject to FOIA and third part audit.

March 24, 2021 12:08 pm

“Climate scientists™” were never going to be able to do it with REAL intelligence..

… but they think “artificial” will work…

Sorry guys, in both cases, its the “intelligence” part that is missing.

Reply to  fred250
March 24, 2021 1:43 pm

C’mon man, you don’t think Griff, Simon, Big Oil….are intelligent 🤓

Reply to  Derg
March 24, 2021 10:22 pm

not in the least bit…… Artificial, maybe, but NEVER the intelligent part.

David Dibbell
March 24, 2021 12:12 pm

“The reason is decades of effort to improve understanding the global climate has not answered basic questions, like how much does global temperature change in response to adding more CO2, and human brains are far more powerful than any AI.
AIs work best when the solution is easy to approach, when a gentle slope of improving results provides a strong indication to the AI that it is making progress.” Good point Eric. The bottom of the bowl in your animation is at ECS = Zero C. Or rather, ECS is some value not reliably distinguishable from zero C. But the modelers have stubbornly refused to consider this as the most likely result. So whatever this “team of scientists” accomplishes will be like colorizing a stick-figure cartoon and calling it more realistic.

In this image I have plotted the hourly “vertical integral of total energy” in the atmosphere for one year at a single gridpoint near where I live. I have used units of Watt-hours per square meter on the vertical scale. It is commonly taken that the direct warming effect of a doubling of CO2 from pre-industrial times is about 3.7 Watts per square meter, or 3.7 Watt-hours per hour per square meter. Notice how that value disappears in the vertical scale as the total energy in the atmosphere varies rapidly and constantly as weather happens.

The point? No “team of scientists” using “AI” will ever find that 3.7 Watt-hours per hour by computation. We are blind to it because the atmosphere is the only authentic model of its own performance, and it operates so as to completely obscure our view of the ECS of such a small influence as the doubling of CO2.

I apologize to those readers who may have already seen this from comments on other postings.

This data is from the ERA5 reanalysis product by ECMWF (The European Centre for Medium-Range Weather Forecasts.)

Last edited 1 year ago by David Dibbell
Jim Gorman
Reply to  Eric Worrall
March 24, 2021 4:50 pm

Just a note. The curve is not “noise”. It is representative of the true value of the component being measured. Noise is extraneous information that is not part of the generated component and interferes with determining the true value of the component.

Reading this a lot of folks have been indoctrinated into the thought process that the data being measured and recorded IS THE SIGNAL. It is not. Temperature is a continuous function just like the sound from a violin string. All the other parts of the climate are continuous functions also.

Trying to use entries in a database of discreet measurements of various items is a waste of time if you don’t even have a clue what the physical continuous waveforms are in a synchronous basis.

Using one temperature per day which is then averaged until it screams stop is not going to help with continuous variations in clouds on a moment to moment basis. We have a pretty good mathematical basis for handling continuous waveforms such as speech by digitizing them and processing them. We don’t even have a clue how to define a temperature waveform mathematically. Can you use Fourier or wavelet analysis. How about on a global basis? The same applies to humidity, wind, convection, everything you can think of.

My fear is that these folks will train an AI with hosed measurements to give the response they want then claim see, we can insure you about GAT. It will basically be used to perform multiple regressions on temperature and spit out what it was trained to do.

David Dibbell
Reply to  Eric Worrall
March 24, 2021 6:23 pm

I think you “get” the problem visually, which is why I posted this. But I also agree with Jim Gorman here that it is really not “noise” per se that drowns out the signal. It is the power of the measurable and rapid changes and reversals of the energy state of the atmosphere itself. I just can’t understand how a climate scientist can claim to have detected the “signal” of human influence on the climate via greenhouse gas emissions.

Jim Gorman
Reply to  David Dibbell
March 25, 2021 5:19 am

In statistical terms it is the “variance”.

March 24, 2021 12:22 pm

AI is the trendy approach to programming. But like all computer programs, they will only say what they are programmed to say.

Steve Z
March 24, 2021 12:32 pm

So are the recipients of the $10 million going to send up balloons to measure how much IR radiation is trapped by clouds at various altitudes, and how much sunlight is reflected? If not, AI by itself cannot improve the modeling of clouds without valid input data.

Robert W Turner
March 24, 2021 12:59 pm

The PEs are taking information from the artificially intelligent to create an artificial intelligence – yeah that’ll work.

Walter Sobchak
March 24, 2021 1:00 pm

You realize that if they solve the problem, they will all be out of jobs, don’t you?

March 24, 2021 1:03 pm

It seems to me, since I’ve photographed a lot – a LOTTT!! — of clouds that cloud formation is more oriented toward chaos than anything else. What might appear to be a thunderhead promising a whopper of a storm can peter out to nothing, and at the same time, a puffy mop that looks like a dandelion gone to seed can turn into a derecho that will do more damage than a semi-truck on an icy road…. or NOT. It is chaos at work there, not prescribed mathematics, and this seems to be something the people who make these proposals fail to understand.

It’s the reason we seldom get forecast further ahead than 7 days maximum, and those forecasts frequently change within 12 hours.

You can’t do this stuff in a laboratory or pretend that you can make predictions cast in stone, not when you’re addressing a system that changes on a whim, something over which they have absolutely zero control.

This should be interesting.

Corky the cat
March 24, 2021 1:08 pm

AI is a fancy tag for neural networks. These can work well if you throw a lot of training data and right and wrong answers at them. An awful lot.
So, can we expect real world climate data to find it’s way into modelling that has resisted verification from the outset? Don’t wait up.

March 24, 2021 1:48 pm

“The lack of progress over the last three decades, despite thousands of intelligent people dedicating years of their lives to the effort, implies the solution to better climate modelling is very difficult to find.”

Doubtful. In reality it means the climate is not doing what the “thousands of intelligent people” and their programmers want it to do.

March 24, 2021 2:00 pm

Or, there is something fundamentally wrong with the theory, and no amount of computer power will produce an observationally valid result, until that theory is modified to reflect observed reality – or abandoned all together.

Jim Gorman
Reply to  Davidf
March 24, 2021 5:05 pm

You’ve hit the nail on the head. They are trying to define continuous functions with some digital sampling, i.e., one Tavg per day and think they can define how that function is varying through time.

March 24, 2021 2:07 pm

The “I” in AI is the wrong choice. Computers are not “Intelligent”.

It’s like humans who have learned stuff by rote. They don’t really know what they are talking about.

The problem with AI is that you are expected to trust the answers. It cannot explain its methodology and reasoning.

And another thing. Use of AI is completely against the Scientific Method.

I can just about condone the use of computer models in some science and engineering situations – e.g. meteorology (where no other method can yield the required forecasts in the required time frame) or structural engineering (where the prototype testing would be prohibitively costly or completely impossible, and anyway the constitutive laws are quite simple and well-understood). But computerised numerical models are still a very poor second to real testing.

But AI isn’t modelling – and how is AI even close to being equivalent to real testing?

Last edited 1 year ago by JCalvertN(UK)
Chris Hanley
March 24, 2021 2:10 pm

The climate frenzy started in the mid-seventies with Hansen’s computer model and wouldn’t exist without them.
Climate computer models are ‘the disease of which they purport to be the cure’.

March 24, 2021 2:14 pm

Because adaptive smoothing functions perform better than human deduction inside a limited frame of reference (i.e. scientific logical domain) and inference over a marginally expanded scope. Same characterization deficits. Same computational limits. Same known and unknown unknowns. Same brown matter to infill the missing links.

F. Ross
March 24, 2021 2:24 pm

My opinion: If the AI models run colder than existing models they will be “disappeared”; if the AI models run hotter than existing models they will be touted as better than mom’s apple pie (which is an impossibility – but there ya go)

Peta of Newark
March 24, 2021 2:33 pm

Good grief, $10M will barely cover the first week’s leccy bill for any decent computer.

Assuming they can lay hands on one

10M would have got a nice little seafront abode in Aus, what about that instead?

Am sure anyone with even a modicum of intelligence, artificial or not, could learn a stack more about climate by shacking up there than a box, any box, of lectronix could teach them

Last edited 1 year ago by Peta of Newark
Gerald Machnee
March 24, 2021 3:43 pm

They need to study cloud behaviour first. Models are only as bad as the modellers.

March 24, 2021 4:45 pm

Put Willis in complete charge and control to choose his team and that may be the best $10 mil ever spent!

michael hart
March 24, 2021 5:12 pm

It can be instructive to read researchers’ words when they talk about what they intend to do, or hope to achieve, when they are seeking funding.

They tend to be a lot more honest about the failings in their field.

I don’t have much objection to govt spending an extra $10 million or so on improving cloud modelling. God knows, they need it. AI might even help a little bit, though not as much as they hope/claim.

And a extra million might possibly have the inadvertent effect of drawing attention to just how bad they are at it, and how we are wasting $Trillions on political initiatives based on currently inadequate modelling.

Reply to  michael hart
March 25, 2021 2:48 am

So, what you are saying, is we should welcome this initiative, promote it like it’s the latest half-naked teenager dancing to a synthesized backbeat, add some brawny boys in the background? You know, advertise the whole thing to the point where everyone in the whole world knows about this marvelous new Thing, then we sit back and wait for the inevitable to happen: A huge spout of garbage to flood the collective consciousness. That would finaly prove the worthlessness of this kind of project, and we can return to normal life?
Tempting, tempting.
Unfortunatley, most people will fail the Turing Test, so whatever garbage this AI produces, they will gleefullly dance to the discordant tune, like there’s no tomorrow (literally).
In military terms, Artificial Intelligence translates as Fake Information…
And just to be sure, modelling a cloud perfectly, would require resolution down to the smallest eddy current, which is smaller than the naked eye can see. Approximations will improve over time, though…always with the approximations.
But hey, at least the poor bastards are recognising those fluffy things in the sky now…

March 24, 2021 5:32 pm

If AI really works then in lightening speed it will take over everything. The reasoning is that if AI works then that AI can be used to create a better AI and that better AI can be used to create an even better AI and so on ( until some physical limit hit like chip speed ).

Reply to  Stevek
March 25, 2021 2:52 am

I think that be heuristics you describe there, the recipe for hubris, which is the word you use in polite conversation to say “that guy’s head is full of his own shitty preconceptions”.
Back to GIGO…

michael hart
March 24, 2021 5:51 pm

The .gif showing the computational energy well is instructive. For a first year undergraduate.

One of the most significant difficulties is that it is unknown where you are starting from on an energy landscape. For one quick example found on the web, look at a bigger picture here:comment image
You simply don’t know where your calculation is starting from and where it might finish. Are you climbing to the main summit, or some little bump on a ridge far removed from the peak of interest?

The program typically focuses on climbing a local ‘energy hill’. But it is likely not the biggest most important hill. It only hill climbs to the peak which the program randomly finds or is arbitrary assigned to. In a cloud you may think you’ve climbed Everest, when you actually took a wrong turning at the South Col and are standing on the summit of Lhotse.

Of course, people are long aware of this problem, but there is no encompassing solution that I have heard of. It would be front page news if they had solved it. For some background:
“M. Mitchell, J. Holland, S. Forrest
When will a genetic algorithm outperform hill climbing?
J. Cowan, G. Tesauro, J. Alspector (Eds.), Advances in Neural Information Processing Systems, Morgan Kauffman, San Francisco, CA (1994), pp. 51-58″

Gunga Din
March 24, 2021 6:57 pm

I thought the science of CAGW was settled?
How will $10 million for yet another model settle it more?

Mickey Reno
March 24, 2021 7:00 pm

Oh hell, here comes Sky-net.

Gunga Din
Reply to  Mickey Reno
March 24, 2021 7:02 pm

Don’t you mean “Cli-net’?

March 24, 2021 7:05 pm

Parameterizing clouds is one of the lessor deficiencies in GCMs. Other deficiencies include:
Not using measured water vapor
Failing to account for thermalization
Bogus application of feedback control theory
Failure to account for ocean cycles
Failure to account for variations in solar influence 

March 24, 2021 8:34 pm

I think Joni Mitchell had this all figured out a few years back.

I’ve looked at clouds from both sides now
From up and down, and still somehow
It’s cloud illusions I recall
I really don’t know clouds at all. . .

https://youtu.be/aCnf46boC3I <– Have a listen

March 24, 2021 9:36 pm

As I recall, in a previous life in a galaxy far, far away, The IPCC was tasked with finding all the bad things that would happen due to rising CO2 in the atmosphere. Chief among them was runaway Global Warming.

So Boatloads of stupid money was thrown at people to come up with all the bad things that would result from an increase in atmospheric CO2.

The way I see it, we can quit spending money on all those useless models and studies that find… sunnuva gun! Only bad things happen, including acne and halitosis, from rising atmospheric CO2.

So the good news is we can quit funding all the studies that, as requested, seek to find the negative effects of rising CO2.

Instead, for a measly $100 or $200 million or so, we can get some AI programmers to write a program that will do the same.

It will spit out all the bad things caused by CO2 (there is nothing good that will come about from rising “caaahbon”) and the AI program will reach the conclusion that we’re all gonna die.

It seems that, true to funded task, every study has shown .that it’s worse than we thought! So use AI instead to discover that it’s worse than the AI programmers thought.

Now a computer, using indisputable AI!, will say it’s so. How cool is that? And thrifty, to boot.

(Sadly, I can’t put my usual winky emoji here.)

March 25, 2021 12:27 am

To manipulate CMIP CAGW models to make them SScary (sic) enough to justify wasting $100’s of trillions of taxpayers’ money, Leftist alarmist MUST assume clouds have a net global warming effect…

All physics and empirical evidence show global cloud cover has a net global cooling effect…

If this $10 million AI grant would correct Leftists’ delusional assumption that clouds cause net global warming, it would be the best investment in human history, but we know that would NEVER happen…

Reply to  SAMURAI
March 25, 2021 2:56 am

Tax the clouds!!! and another “!”.

Reply to  paranoid goy
March 25, 2021 3:19 am


Bill Gates’ brilliant idea is to seed chalk dust in the ionosphere to block solar irradiance in an effort to cool the earth from ravaged of CAGW….

What could go wrong that? LOL!

I’m going long on chalk-dust futures….

Captain Climate
March 25, 2021 5:15 am

AI can’t solve theory error. What a waste of money.

Coach Springer
March 25, 2021 7:27 am

$10 million? A team of scientists can use that up pretty quickly to produce a need for further work.

Gordon A. Dressler
March 25, 2021 7:35 am

One simple question: What kind of “intelligence” creates the ability for “artificial intelligence” to function?

One simple follow-up question: Given the answer to the above question, why do people have any faith in the future abilities of “artificial intelligence”?

March 25, 2021 9:19 am

I’ve looked at clouds from both sides now
From up and down and still somehow
It’s cloud’s illusions I recall
I really don’t know clouds at all

March 25, 2021 4:45 pm

A generalisation was made a few years ago, you would need the whole Top500 list of supercomputers to operate more than 24hrs to model 1 day into the future. It could take another 20 years before we could get the top 10 together to make simulation/model runs practical with required detail.

%d bloggers like this:
Verified by MonsterInsights