Junk Grade Models: Even Short-Term Climate And Weather Modelers Get It All Wrong

Reposted from NoTricksZone

By P Gosselin on 19. February 2021

Many climate policies are based on scenarios generated by models. Depending what these models churn out, actions and costly regulations get enacted to mitigate the worst consequences. So we hope that the modelers get it right. Unfortunately they are still shooting in the dark. Even short term models!

TWC  forecast a basket case

It turns out most models are junk grade when it comes to forecast quality. For example, The Weather Channel here not long ago issued it’s winter outlook up through March:

The area of the US now being hit by one of the most vicious cold waves in the last 100 years was indeed forecast to be “way above normal” or “much above normal”. So far the exact opposite has happened.

We have to wonder what it takes to be issued a license to practice weather forecasting nowadays, and we have to feel sorry for those businesses and industries that relied on these botched forecasts.

Yet, these are the type of outlooks that policymakers insist we need to heed and take immediate (costly) action. “Follow the science,” activists and policymakers like to say.

Potsdam Institute’s El Nino debacle

Another example of short to midterm forecasting involves the El Nino events, which have a global impact. Having the ability to predict these events accurately would be a very valuable tool.

And not long ago the Potsdam Institute for Climate Impact Research (PIK) in Germany claimed to have developed a model that could predict the events with 80% probability. A PIK November 2019 press release in fact boasted that its team of renowned researchers had developed a new, far better model – which in November 2019 said was capable of forecasting a late 2020 El Niño event a year in advance.

Today in February, 2021, the results are in and they are ugly: The equatorial Pacific 3.4 region is near La Nina conditions, thus in complete contradiction to the warm forecasts of the Potsdam Institute. The “pioneering” PIK model, which in part was developed by Prof. Hans-Joachim Schellnhuber, is a complete failure and totally wrong:

The ECMWF graph above shows ENSO forecasts for the period July 2020 to January 2021 (many thin red lines) compared to measured SSTA in Nino area 3.4 (blue dotted line).

The ECMWF forecasts were on average about one degree Celsius too warm. There have been La Niña conditions since August 2020 with measured values of -0.5°C deviations and colder.

“Model rubbish”

“The ‘pioneering’ PIK model produced model rubbish and disgraced German ‘climate science’ worldwide. Moreover, it currently looks as if the cold La Niña wants to cool us down further until 2022,” reported Schneefan here.

4.8 25 votes
Article Rating
Notify of
Newest Most Voted
Inline Feedbacks
View all comments
Dave Fair
February 19, 2021 10:11 pm

Well, the solution is obvious: Raise the blue dotted line. [h/t Karl and Mears for the idea.]

February 19, 2021 10:28 pm

Modelling has been the death of proper, old style forecasting which relied on the experience and intuition of experts in the interpretation of observations based on earlier precedents.
It was by that method that the decision was taken as to when to invade France in WW2. The forecaster made a difficult judgment call and was proved right.
Unfortunately, that approach gives only about 3 days ahead with any utility so models were supposed to do better and they do up to about 5 days out.
Beyond 5 days they become overwhelmed by weather system instability and despite vast amounts of investment over decades have failed to improve on that.
Look at any modelled forecast for 7 days ahead and it will be adjusted substantially as one approaches days 6 and 7.
Proper recognition of that sad fact would destroy the weather modelling industry.

Reply to  Stephen Wilde
February 20, 2021 1:29 am

Channel 7 in Melbourne Australia has a Meteorologist/Presenter, Jane Bunns, who does an 8 day forecast. Originally she did try for 14 days but the vast majority of the viewers did not take much notice of the 14 day forecast. This will give an indication of why that is the case:

Alan Robertson
Reply to  RickWill
February 20, 2021 4:30 am

I was hoping that she’d give the forecast for Bundanyabba, but it looks like she didn’t get that high up on the map.

Walter Sobchak
Reply to  RickWill
February 20, 2021 12:10 pm

I think she has a lot of assets. I am not sure I would care about the correctness of her forecasts if she wore the right dresses.

Hugh Abercrombie
Reply to  Stephen Wilde
February 20, 2021 8:02 am

Modelling is the death of any science. Empirical (observation based) data and observations are discounted and model results are presented as God-given truths, commonly in pretty colours to seduce the eye.

Gunga Din
Reply to  Stephen Wilde
February 20, 2021 10:19 am

I think the problem is that some have forgotten that the model is a tool.
They rely on it too much then they forget what they’ve learned.
Sort of like people using calculators so much they forget how to do the math themselves.

Reply to  Stephen Wilde
February 20, 2021 10:35 am

They do 5 days instead of only 3 because of the increased amount of data observations

Bob Hunter
Reply to  Stephen Wilde
February 20, 2021 10:48 am

The Farmer’s Almanac is more accurate predicting the weather for a 12-16 month period than the computer models.

Chris Nisbet
February 19, 2021 10:29 pm

1) If they had to bear the cost of failed predictions would they predict so many abnormal conditions?
2) I wonder if all they need to do to improve the accuracy of their predictions is multiply predicted temperature deviations from ‘normal’ by -1, and change their % likelihood values by subtracting them from 100%.

February 19, 2021 11:11 pm

This is far more important than blaming the power generators for lack of capacity in Texas..how could they have gas plants on standby when the forecast is for higher than average temperatures.
Same thing happened in my city, the models said the rainfall was going to be average or above average over summer and did it again the next year, the reality was lowest rainfall over 6 months for the first year was was exceeded by even lower for the second year. The water storage was marginal for a time but OK in the big picture . This only happens because the models can’t be useful 3 months out, yet they are being used to predict decades ahead at a scale that is far to fine to be worthwhile 2 weeks ahead . Overuse of around average or above or below average is letting them get away with forecasts that exibit no skill what so ever.

Reply to  Duker
February 20, 2021 10:48 am

A few years ago, climate warmists convinced local councils that snow and ice would be less and less. They reasonably reduced the budget for snow removal and supplies. Of course, they got far more snow and ice than average in the following year.

Roads went unplowed and unsalted. Accidents went up, and people died. Relying on these forecasts have real life consequences, and those who demand we act on them should be held accountable for the damages they may cause.

Reply to  jtom
February 20, 2021 10:48 am

The above took place in the UK.

Antero Ollila
February 19, 2021 11:41 pm

We are far, far away from the models capable of predicting events shorter than one year. The coldness of this year is a weather phenomenon.

What is for me much more interesting is what are the temperatures of IPCC’s simple model from the year 2001 to 2020. They are pretty ugly.

Reply to  Antero Ollila
February 20, 2021 1:18 am

You can take your pick. The Chinese FGOALS will keep you as you were in 2001 if you are not warmed by homogenisation. The EU MIROC6 is a tad closer to hell.

All data from KNMI Climate Explorer apart from CMRW-21, which is a reliable physics based model of the ocean/atmosphere energy balance.

These are all forecasts for the CMIP6 ensemble of models that will be used for AR6 due in 2021. So right up to date.

Richard M
Reply to  RickWill
February 20, 2021 5:39 am

Too bad the actual value on your chart stops in 2019. We are probably back to starting value with the big drop the last few weeks.

Reply to  RickWill
February 21, 2021 7:13 am

Why are they using USHCN which is crap instead of the USCRN which is unmodified to match their agenda?

February 20, 2021 12:40 am

You will never see last week’s forecast back on any channel: keeping the model-forecasting illusion alive for $$$$$$$ $ake.

David Stone CEng
Reply to  AntonyIndia
February 20, 2021 1:13 am

The British MET officee has added lots of new stuff based on Climate in place of weather to its model, and the results are now usually wrong more than 2 or 3 days out. Previously they were fairly good up to a week out. Temperatures are particularly miles out, similar to climate models. All very foolish.

Ron Long
February 20, 2021 1:59 am

I live in west-central Argentina and my twin brother lives in Eugene, Oregon. We converse every day via skype. There is a large difference between the accuracy of Argentina and Oregon weather forecasts, with forecast events a week ahead of time common. Whereas, in Oregon the forecast even a few days out is mixed results. So, given that both forecasts have access to the same data, it appears to me that some combination of ocean and atmospheric currents produces more even flow of weather events than other combinations which are more chaotic. Now stretch weather time-wise out into climate and the predictions get to a ridiculous stage, might even call it a model.

Reply to  Ron Long
February 20, 2021 8:02 am

Forecasting in Argentina is probably easier than in Oregon. The weather systems in the southern hemisphere move much more predictably since they are much less affected by topography. Essentially they just circle endlessly round and round Antarctica:


it is way more irregular at the other end:


February 20, 2021 3:44 am

Weather models are improving. The best we are able to forecast into the future with and skill is about 7 to 10 days. 30 years ago, the best was 5 days.

Reply to  Loydo
February 20, 2021 8:37 am

7 to 10 days? Complete nonsense. You made that up. Studies have shown that weather forecasts degrade significantly in accuracy beyond 5 days and by 14 days have dropped to random for even simple things like will future temperatures be above or below normal. Forecasting based on finding a historical weather pattern that best matches current conditions and progressing the predictions based on the historical patterns that followed the matched pattern routinely beat weather models.

Reply to  Loydo
February 20, 2021 12:25 pm

Loy-dodo with his little evidence -free fantasies, as always

So funny hat he/she/it continues to expose the complete idiocy and ignorance that comes with ACDS (Anit-CO2 Derangement Syndrome),

… making a mockery of the AGW cult in the process.

Reply to  fred250
February 20, 2021 3:46 pm

No, I’ve hear this said by a couple of different meteorologists.

The best we are able to forecast into the future with and skill is about 7 to 10 days. 30 years ago, the best was 5 days.”

An innocuous enough statement but you go straight to the acrimonious abuse.

This one happens to be a direct quote of Anthony Watts from yesterday: https://wattsupwiththat.com/2021/02/19/friday-funny-nature-makes-a-mockery-of-month-ahead-model-forecasts/

He’s exposing his “little evidence -free fantasies and complete idiocy and ignorance” too is he? Knee-jerk much?

Reply to  Loydo
February 20, 2021 9:38 pm

“heard it said”


That is Loy-dodo’s idea of EVIDENCE… roflmao. !!

You are a MORON, Loy !!

You make a mockery of all ACDS sufferers.

Reply to  Loydo
February 21, 2021 3:57 am

You are correct! Improving at abject failure, just like you.

February 20, 2021 3:56 am

Thankfully many industry sectors and even municipalities have quit relying on climate and weather forecasts produced by government agencies and instead have switched to paid private services from forecasters who have a demonstrated track record for accuracy much superior to that of the government agencies.

If this were not the case, companies like Weather bell Analytics, of which there are several, would not be in business.

Reply to  rah
February 20, 2021 7:36 am

The private weather services have skin in the game. If they had the accuracy record of the government agencies they’d go out of business.

Reply to  Yooper
February 20, 2021 9:48 pm

Proving once again that capitalism works and is superior to socialism.

Giordano Milton
February 20, 2021 3:57 am

People forget that a model is nothing more than a hypothesis put into code. It can be no better than the understanding the modeler (or modelers) have, and—as we see—that understanding is flawed.

Until a model is validated, repeatedly, it should be treated as no better than someone’s opinion.

Richard M
Reply to  Giordano Milton
February 20, 2021 6:43 am

You’ve hit on the big reason climate models are so bad. They don’t correctly model the interface between the surface and the atmosphere. There are probably reasonable atmospheric models and oceanic models but when they try to tie these together they fail.

It is at this interface that the greenhouse model diverges from reality. The difference in heat capacity allows the added energy generated by greenhouse gases to be radiated to space every night.

Reply to  Giordano Milton
February 21, 2021 7:21 am

You are absolutely right. Moreover, when the predictions fail to match observations, the hypotheses MUST be discarded or modified – it has been disproven. One more thing – the model must be able to answer past observations. Clearly they don’t (MWP, RWP in particular) and fail at that also.

February 20, 2021 4:10 am

Tie their paychecks to their failures and you would see A. far fewer people on TV/Print spewing this stupidity and 2. far more accurate short term weather forecasts. As long as telling lies pays lies are all they will tell.

February 20, 2021 4:17 am

The divergence of all of those red lines occurs very quickly, therefore the model is utterly worthless. A coin toss would be just as good and much cheaper.

Peta of Newark
February 20, 2021 4:50 am

I have posted this before, its about Covid and yet another pathetic excuse to go after Mr Trump.That’s a large part of this climate thing also, school-playground politics

Anyway, what about my ‘fix’

Covid-19: Climate Models: Social mruder, they wrote—elected, unaccountable, and unrepentant

Is anyone here going to object?
Would something like it, wake a few folks up?

Last edited 5 days ago by Peta of Newark
Reply to  Peta of Newark
February 20, 2021 3:50 pm

Unmitigated blather.

February 20, 2021 4:51 am

IF — and I use that IF advisedly – IF these twits actually spent a little time in the area where I live, the WEATHER is normal winter AND the TEMPERATURE has dropped from COLD to NEARLY BITTER COLD. This is rather normal around here.

I would love to pile them into a vehicle and drag their ridiculous personages up to one of my favorite hiking spots – dunes area near restless, testy Lake Michigan shore – and just leave them there overnight. But that’s just me.

My snowfall total so far is just shy of 20 inches. Measured it, took pictures, and assessed it as normal winter. Now the National Weather Service forecast for the end of next week is mid to upper 20s and maybe some low to mid-30s: NORMAL WINTER WEATHER here in the Upper Midwest. Chicago got hit harder than we did, which is fine with me.

I have some really cool photos and the birds have returned to my feeding station: many sparrows, some juncos and two cardinals and a gorgeously marked female red-bellied woodpecker. My life is pretty darned good.

Reply to  Sara
February 20, 2021 7:42 am

Something is really out of wack here in northeast Indiana, it was 9F this morning and I saw a robin red breast outside. Today I have to go up on my roof to move the heat tape so the ice dams don’t cause a leak into the house.

Reply to  Sara
February 20, 2021 11:29 am

Here in central Indiana I have a couple red bellied woodpeckers at my feeder. Also Hairy and Downy woodpeckers, Nuthatches, Tufted tit mice, several species of sparrows, cardinals, blue jays, cow birds, black capped chickadees and too many damned European starlings. Had a red headed woodpecker last summer but haven’t seen it since then.

I put out two large seed blocks in cages hanging from a double shepherds hook outside a dining room window. During the winter months I also have a squirrel proof feeder that holds 8 lb. of seed hanging from a low branch on the spruce tree, The really flock to both and especially so when there is snow covering the ground or very cold weather is coming in.

Reply to  rah
February 20, 2021 12:01 pm

One blue jay showed up this afternoon. I expect to see the Hairy and Downy woodpeckers before long. Have not had the starlings show up at all. Grackles will come along in the spring, as will the redwinged blackbrids. Looking forward to Spring!!!!

Joseph Zorzin
February 20, 2021 4:55 am

Are any climate models described in detail along with the code? I suppose it could all be described well in a flow diagram- but I’ve yet to see one anywhere. It’ll be easier to deconstruct them if we can see the logic or failure of logic.

Reply to  Joseph Zorzin
February 21, 2021 7:34 am

A quote from the end of the reference:
“But until modelers figure out how to confidently forecast changes in the winds, Smith says, “We can’t take the models at face value.” “

Even with this, they still cling to their CO2 obsession.

February 20, 2021 5:03 am

One can understand why owners of Texas electricity generating plants did not freeze protect them. Why spend the money when tomorrow is going to be warm anyways? This is a prime example of very bad public policy created by bad science.

Gerald Machnee
Reply to  DHR
February 20, 2021 6:39 am

They should have got the hint during the ice storms around the 2011 Superbowl.

Richard M
February 20, 2021 5:47 am

It will be interesting to see where we go from here. CFSR has planet Earth close to the late 20th century average.
comment image

Models might only be missing a few key elements that cause them to fail spectacularly.

Reply to  Richard M
February 20, 2021 6:08 am

Assuming this trend continues with the La Nina in place, think of all the papers that were set to be released, but will not be released before the Scotland COP in November. What a waste of resources. I am prepared for a slew of reports claiming that cooling was caused by the global shutdown to combat COVID.

Reply to  Richard M
February 20, 2021 1:33 pm

Looks very like a long-term cooling trend to me.

Reply to  Richard M
February 21, 2021 7:37 am

No, they have so many degrees of freedom that get “tuned” separately to match the adjusted temps of the past that they can’t separate the coffee beans from the rat turds.

February 20, 2021 5:58 am

Heavy frost in the Great Lakes region causes them to freeze quickly.comment image

February 20, 2021 6:05 am
February 20, 2021 6:09 am

The forecast of a warm winter was an easy one to make. A strong La Nina had formed. This usually means a warm, dry southern US. The further west the bulk of the cold, the warmer it is. And the bulk of the cold was closer to Australia. That was one big signal of warmth there. Then there is the PDO, Pacific Decadal Oscillation. PDO- amplifies La Nina (and mutes El Nino) but PDO+ mutes La Nina (and amplifies El Nino). PDO- is a warm blob surrounded by a horseshoe shaped cool pool near Alaska. Well, the PDO is currently negative.

So, all the major weather drivers said warm and dry. But, there are many weather drivers. You can research all these: Madden-Julian Oscillation, Eastern Pacific Oscillation, Arctic Oscillation, North Atlantic Oscillation, Atlantic Multidecadal Oscillation, Global Atmospheric Angular Momentum, Pacific/North American teleconnection, Pacific Meridional Mode, North Pacific Oscillation, North American Dipole. And I’m sure I left some out. Weather and climate are very complex. These other drivers had other ideas.

There will always be exceptions. I have noticed this winter that even the European model is having a very hard time with it. You just need to be humble enough to admit that 150 years of weather data ain’t enough to fully understand the weather. I’m not sure 5000 years of weather data would be enough either. The best thing you can do is not pretend you and your complex computers are all-wise but be willing to admit there is much to learn.

Reply to  Wade
February 20, 2021 8:17 am

Still, by January 19 when that forecast was made the SSW over the Arctic was already in full swing:


And that almost always means cold weather, I certainly foresaw it in late January, it is actually one of the few situations when a multi-week forecast is possible.

February 20, 2021 6:22 am

How do the folks who live and die by models want the output to be used?

February 20, 2021 8:30 am

I think the myth of German engineering was exploded by the T34 tank.

Reply to  Terry
February 20, 2021 9:45 am

It’s arguable the lack of speciality materials caused German manufacturers and consequently the Wehrmacht etc. to significantly underperform. Blithering bureaucracy and cronyism also had a serious role in that underperformance.

You might find some enlightenment in this webminar

February 20, 2021 8:41 am

Perhaps the modelers should consult with Farmers Almanac.

Reply to  Renee
February 20, 2021 9:39 am

Yep. So far, in my AO, they’ve been quite close to the real results.

Reply to  Sara
February 20, 2021 6:02 pm

Farmers almanac actually incorporates solar influences into their model predictions. IPCC should include their model as a base case.

Last edited 4 days ago by Renee
Gunga Din
February 20, 2021 10:50 am

If it hasn’t already been done, why doesn’t someone develop a “model” that wouldn’t forecast the weather but search for past weather patterns that match the current patterns then report what happened next?
Seems to this Layman that such a tool might prove useful.

Reply to  Gunga Din
February 20, 2021 3:47 pm

Joe Bastardi does that, manually + computer files.

Gunga Din
Reply to  EdB
February 20, 2021 3:58 pm

I’m sure the good ones do.
We have facial recognition software to match a face in a crowd with the name of a face on file.
Why not software to match a current weather pattern with a past weather pattern?

Earl Rodd
February 20, 2021 2:12 pm

Some years ago, I tracked the Weather Channel 5 day and 10 day forecasts for a year and compared it with actual temperatures. For comparison, I took the spread of record low/record high for each day and randomly picked a temperature. The 5 day was a little (but not much) better than my random approach. The 10 day was actually inferior that year to a random number.

When I looked at the data, one feature stood out. Often, the 10 day forecast was off a LOT, but the reason was subtle. The error was more often in timing than in kind of event. For example, the 10 day correctly picked that a strong cold front was coming but missed it by 2-3 days. This throws the long term temperature forecast way off even though the model had the basic idea of coming weather systems.

Reply to  Earl Rodd
February 21, 2021 8:00 am

There is what is called “Persistence Forecasting” – the weather tommorw will be the same as today. Depending on how “tight” you want to measure (temp, wind, cloudiness), you’ll probably be right more often than not.

February 20, 2021 5:40 pm

Anyone know what The Old Farmer’s Almanac predicted? I’d bet it’s prognostications are at least, if not more, accurate.

Doubting Rich
February 21, 2021 12:11 pm

Correct me if I am wrong, but didn’t they basically forecast every every possible temperature variation from the given starting temperature?

%d bloggers like this: