The 2000 National Climate Misassessment Transcript
Hello to our friends in the United States of America. You are actually very lucky people because every few years your government publishes a National Climate Assessment which gets a lot of media attention and presents lots of colorful charts giving you useful information about all the terrible things global warming is going to do to you over the coming decades.
At least, it’s useful if you think the models are right. And since they’ve been publishing these reports since the year 2000, which is now over 20 years ago, we figured it’s time to go back and check how the predictions worked out. How many of the terrible things it said were going to happen, did global warming actually do to you over the past two decades?
I’m John Robson, and this is a Climate Discussion Nexus Crystal Ball Check on the year 2000 US National Climate Assessment.
In 1990 the United States government established an agency called the Global Change Research Program, or GCRP, and passed a law requiring it to produce a National Assessment of the potential impacts of climate change on the US on a regular basis.
It took 10 years for the GCRP to put out the first National Climate Assessment. Since then there’s been no looking back. As bureaucracies will, the GCRP has built itself quite the empire, with 13 affiliated agencies all dedicated to informing Americans about the grave perils of climate change in dozens of assessments, reports and special studies since 2000.
This prodigious output includes four full-scale multi-volume National Assessment Reports, with the fifth currently in preparation. And while National Assessments include information about the current and historical US climate, they tend to focus heavily on projections about what is going to happen.
When it comes to climate, projections of course mean models. And in case you might be thinking that it’s not fair for us to think anyone was expected to take climate models seriously way back in the year 2000, well, for one thing scientists certainly expected us to back then and plenty of people did take them seriously, including politicians and activists.
For another, it’s not very sensible to say we should certainly believe their later predictions which we can’t yet test because their early ones that we can test were obviously rubbish. And for a third, I have to remind you that by that time the Intergovernmental Panel on Climate Change - the UN body - had already issued two Assessment Reports of its own which touted their model projections, and was on the eve of releasing its Third.
And while much of the attention on #3 was on the Michael Mann hockey stick which they emphasized so heavily in that Report, it was also in that report that they announced to the world that their so-called “coupled” climate models had reached the point where they were reliable for projecting the future (TAR Ch 8 p. 473).
“Coupled models have evolved and improved significantly since the Second Assessment Report. In general, they provide credible simulations of climate, at least down to sub-continental scales and over temporal scales from seasonal to decadal… We consider coupled models, as a class, to be suitable tools to provide useful projections of future climates.
If they’ve claimed that their models were certainly ready for prime time in the past, or others have insisted that they were, and they’ve been wrong, it’s one more reason to be skeptical this time. And since the GCRP was using similar models to the IPCC’s back in 2000, and was just as confident about them, it’s absolutely fair to check how that first major National Climate Assessment did.
Right up front, the first prediction in the National Assessment was that, assuming global greenhouse gas emissions continued to increase, the United States would warm by between 3 and 5 degrees Celsius over the 21st century.
That works out to point-three to point-five degrees C per decade. And the world hasn’t stopped emitting greenhouse gases, or even slowed down the pace, so we should be seeing the US warming somewhere in that range, which would mean between point-six and 1 degrees warming between 2000 and 2020.
The data to evaluate that prediction comes from NASA’s Goddard Institute for Space Studies, which publishes an annual estimate of temperatures over the continental United States, derived from a large network of weather stations all across the country.
The NASA data shows that from the year 2000 to 2020 the US warmed by just under 0.2 degrees per decade, less than 0.4 degrees in total, which is well below the low end of the model projections.
Even that 0.2 per decade over twenty years is not a particularly robust trend because 2000 was unusually cold and 2020 unusually warm. If we back the sample up just one single year and look at the change from 1999 to 2019, the figure is half that, or 0.1 degrees per decade.
Now in case you suggest that we might be cherry-picking, we note that the trend is not accelerating. If we start the sample in 1980 and look at the 40 years up to 2020, it’s a little over 0.2 degrees per decade over those four decades, which means it actually slowed down slightly towards the end.
If, that is, you believe the figures to a decimal place. Which you shouldn’t given all the problems in the surface temperature record related to urban heat islands and all the fiddling around with the data to make the past look colder. The bottom line is, the US probably is warming very slightly, but far less than even the low end of those 2000 model projections. (And again, if you’re going to suggest that the predictions we can’t test are way more reliable than the ones we can, they too predict fairly rapid warming that we know did not happen.)
Climate, and climate projections, aren’t just about temperature. The 2000 National Assessment also had a lot to say about precipitation. For each region of the country they presented forecasts based on two different climate models: one from Canada and one from the UK.
And we can check out how the models did by comparing them to the records shown at the RainSphere website run out of the University of California-Irvine, which summarizes satellite-measured rainfall trends from 1983 to 2020 for most locations around the world.
In the Northeast, the Canadian model predicted slight positive trends, while the UK model predicted large positive trends. As in a lot more rain and/or snow in a region that, the leftmost part of the chart shows, had seen an irregular increase in the 20th century.
What really happened? On the Rainsphere map, blue indicates an increasing trend, while red indicates a decreasing trend, and if there’s also diagonal shading it means whatever trend was present wasn’t statistically significant in that location. From 1983 to 2020, the data show the northeast mostly had negative trends in rainfall, but they weren’t statistically significant.
That’s a swing and a miss. Now we move to the Southeast, where once again the Canadian model predicted neutral or negative trends and the UK model predicted big increases in a region that, again, had seen increases in the 20th century.
Here the data show generally increasing trends, except Florida got a bit drier. Once again, the trends are not statistically significant. But we’ll give the models a point for that one. At least the UK model. So no strike, but no hit either.
In the Midwest, the UK model again predicted large increases, while the Canadian model predicted relatively small increases, and mostly in the northern areas.
The actual data show the region mostly getting a bit drier, especially in the northern part. But once again the changes aren’t significant.
So that’s another called strike. Especially because the models also predicted that water levels in the Great Lakes would start going down and got that one wrong too. In point of fact, and I recognize that when “Change” is part of your name you’re meant to find it regardless, climate change doesn’t seem to have changed anything.
As we discussed in our video on the Great Lakes Climate Crisis, no sooner had the experts all agreed that climate change meant falling Great Lakes water levels than the water levels shot up — which the experts then blamed on climate change and claimed they’d predicted it all along. In point of fact water levels continued to fluctuate within normal historical limits.
We now move to the Great Plains. The Canadian model said Texas, Oklahoma and Kansas would dry up but it would get a lot wetter as you move west to Arizona and Utah. The UK model said it was going to get a lot wetter everywhere, again especially the further west you went.
The data show … drum roll please … every part of the region got drier, and the drying was more intense the further west you went. In fact Arizona, New Mexico and Colorado had statistically significant drying trends. And no doubt some climate expert somewhere is telling the local news station not only that it is a sign of anthropogenic climate change but that they predicted it all along.
Moving further west and south, both models predicted Nevada and California would get a lot wetter. But according to the satellite data, nothing much changed. California got a bit drier and Nevada a bit wetter, but the changes weren’t significant.
Finally we fly up to the Pacific Northwest where both models said they’re in for more rain, especially in the southern part of the region, again predicting a continuation of 20th-century patterns due to dramatic changes in climate.
And the satellites say… Nope. Except for a bit extra in Montana, the northwest got a bit less rain overall.
So out of 6 regions the models got the precipitation changes dead wrong in five of them. They’d have done better by flipping a coin.
Which, it turns out, they knew at the time. Dr. Patrick Michaels, now with the Competitive Enterprise Institute, is formerly a Research Professor of Applied Climatology at the University of Virginia, and served as an expert reviewer for the 2000 National Assessment. As part of his analysis back then he took the Canadian and UK model runs for the continental US over the previous decades and compared them to observed data. And to his surprise he found the models fit the data worse than a table of random numbers. So he notified the report’s Lead Author, a US government scientist named Tom Karl.
We spoke to Dr. Michaels about what happened next.
[Interview with Pat Michaels]
You spotted a problem with the computer models and you did communicate this to Tom Karl. What happened there?
Well, they chose two computer models. They had a choice of fourteen models and the model they used for temperature, which was the Canadians model, has the largest temperature changes in the United States for the 21st century of all the climate models that were available for the report. And the model that predicted the greatest changes in precipitation, the British model, was the one that they chose also. So they chose the extreme models.
And of course I wanted to see how well they worked. So I asked to see an incredibly simple test to see if the models could produce the ten year running means of US average temperature back to the year 1900. I mean that's about as simple a thing as you can get. What I mean by that is 1901-1910, 1902-1911 etc. ... all the way to when the report came out.
Not to be too technical about it but the model added noise to the data. In other words the data had a certain amount of year to year noise in it and when you applied the model to it — pretend you’re running means — your result got more noise. So the model created negative knowledge. This is the exact analogy to giving students a four part multiple choice test and having them somehow get less than 25% which seems impossible but it can happen. They were worse than random numbers.
So I sent that result to Tom Karl who was the head scientist for the first assessment report and he wrote back to me and said, “Yes, you were right, but we didn't just look at ten year running means. We looked at 1, 5, 10, 20, and 25. And in each case the models produced more variance than was in the raw data.”
So the head scientist for the report acknowledges to me that yes you were right, in fact it's worse than you thought, it's everywhere. To go ahead with that report — and I’ll use a word, not a pleasant word — that is scientific malpractice. It is exactly equivalent to medical malpractice in prescribing something that you know either is not going to work, or is going to harm the patient, or is going to work in the opposite way that it was supposed to. That is precisely what happened with that report.
Are you willing to read his email back to me by the way?
Let me read what I wrote to him.
[From: Pat Michaels
To: Tom Karl
“All implied effects including the large temperature rise are therefore based upon a multiple scientific failure. The National Climate Assessments continued use of those models and that of probes is a wilful choice to disregard the most fundamental scientific rules. And that they did not find and eliminate such an egregious error is astounding. For that reason alone the NCA should be withdrawn from the public sphere until it becomes scientifically based.”]
[From Tom Karl
To: Pat Michaels
“One has to look at time averages, in the assessment we were most interested in decadal century trends, not annual averages. We would not be inclined to perform the test you did. Nevertheless we ran the tests you did but changed the averaging period.”]
And then he included the results for all the averaging periods - 1, 5, 10, 20, 25 years - and he provided graphics. And you can see that the variance increases when you apply the model instead of decreases. That's the hallmark of a failed model that you shouldn't use to project anything.
So is this a smoking gun?
No, it's a mushroom cloud obviously!
That was just the beginning. We had the second report that was so bad that I could find something wrong in every paragraph and I wrote a multi-thousand page document that looks just like it, and the flow of that one is all the same and it goes on and on and shows exactly what’s wrong. The third one was simply blatant. The National Oceanic and Atmospheric Administration wrote in their introduction that this is a key deliverable in President Obama’s climate action plan. Okay, just tell us it's political because you just did!
I could go on and on about that but let’s talk about the major prediction from the first assessment report. After all that report was a 2000 report released in 2001. It’s 2021, 20 years after that. It predicted that US temperatures in this century would rise between 5 and 9 degrees Fahrenheit and most of the rises in temperature that are predicted by these models are linear, they're not exponential. So if you make that linear assumption we should have had a rise of about a degree Fahrenheit by now in the US record.
One interesting thing happened after the report came out. The National Climatic Data Centre put out a new temperature monitoring record called the Reference Climate Network. A relatively small number of stations - very good ones - without interfering things like buildings, the same instrumentation and everything calibrated. It went online in 2004 and you know how much warming it shows in the US between 2004 and now? None. Not a lick. And by the way that matches their other long-term record, the historical climate network, which doesn’t show anything since 2004. I don’t know how many people know that but it's kind of obvious looking at the data.
And these have got to be about the best surface temperature measurement that anyone is doing anywhere?
The Reference Climate Network is the best surface record that we have because the instruments are all standardized, they’re calibrated. They were designed because of criticism of the older climate networks, that they were maybe fussed around with or something like that. Tom Karl, the same guy who was head of the National Assessment, was the head of the National Climatic Data Centre and he pushed for this. So yeah, they produced a very good record. The only problem is it doesn’t show a lick of warning.
Groups like the US National Assessment and the Intergovernmental Panel on Climate Change like to boast that their reports go through extensive peer review prior to publication. But what they don’t tell you is that the report authors are free to ignore any comments they don’t like, even when they know the reviewers are right. In the end these reports are not true summaries of what the scientists think, but what government bureaucrats want you to believe the scientists think.
Despite all these flaws, the report did get a couple of things right. They said agricultural output was likely to increase as farmers adapted to climate change, which would benefit consumers. And they said output of the forest sector would also likely increase in the coming decades.
We hope nobody got fired for slipping those things in. Especially since they were about the only projections the report did get right and, of course, they were based on classical economics not computerized climate, on human adaptability not paralysis in the face of looming disaster, and on good news not bad.
Speaking of economics, it’s not surprising that in the years since issuing its first report the US Global Change Research Program has, in classic bureaucratic fashion, massively expanded in size, scope and budget, and is now hard at work on its fifth National Assessment. But in all the time it’s been in business, and with every new report, it has never gone back and asked how its past model simulations turned out, discussed why anyone should take its new predictions at face value given its high rate of failed past predictions, or analyze how it might correct various biases in its modeling and other procedures to correct for those past failures.
Not to worry, that’s what we’re here to do. And if you want to help us continue going where climate bureaucrats fear to tread, subscribe to our weekly newsletter, our videos on YouTube and Rumble and our podcast, and become a supporter.
For the Climate Discussion Nexus I’m John Robson. Just as I said 20 years ago as well.