Debunking: "Climate Change Model Predictions Are Unreliable!"


Thumbnail photo: Edited version of graph from: "Testimony of John R. Christy." Feb 2, 2016.


Many climate change deniers argue that the predictions made by climate change models are unreliable, don't match up with the observations, and oftentimes vastly overestimate the rate of warming and the associated consequences. As we'll see here, this is simply not the case.

The land and ocean temperature predictions have been extremely accurate. Projections have actually underestimated the rates of sea level rise and Arctic ice melting—precisely the opposite of the "climate science alarmism" that deniers love to rant and rave about.

Only in the very specific case of middle tropospheric warming do predictions outpace the observations. Much of this data, however, has since been corrected to account for various forms of bias, with the revised data showing only a very slight deviation between the models and the measurements. The real irony is that the most inaccurate predictions are made by so-called climate change "skeptics," who absurdly project a decrease in temperatures over time.

In addition to making these points, I also argue here that deniers often misleadingly present their data and wildly overstate their conclusions. I also examine some of the reasons for models differing from observations, I discuss the classic mistake of conflating weather with climate, and I debunk a ridiculous argument made by Jordan Peterson about whether we can measure the impact of our climate-related actions.

It is a damn good video, if I might say so myself, so I hope you all enjoy.

We read an example of this argument on the climate change denialism website (Oh, look at that! They've got a little "Donate" button on the side of their website. Yeah, the only thing I'll be sending you people is a lump of coal. They're like: "Really? 'Cause we would genuinely appreciate that. How do you think we keep the servers running over here?")

As Mike Jonas writes,


"There are dozens of climate models. They have been run many times. The great majority of model runs . . . have produced global temperature forecasts that later turned out to be too high."


Mike Jonas, I think that's one of the Jonas brothers, right? He's like: "When I'm not singing on stage, I'm busy spreading climate change misinformation online." His groupies are like: "Eww, not interested."

The key basis for claims like this is the following graph, created and presented by climate scientist John Christy, co-developer of the UAH satellite temperature record. As we can see, this graph compares satellite and balloon observations against the average temperature increase predicted by many different climate change models.

Starting around 1998, there's a clear divergence between the model predictions and the actual temperature recordings. Here's another commonly-cited John Christy graph which shows the same basic trend. And CMIP5, in case you're wondering, is described on their website as:


". . . a collaborative effort involving 20 climate modeling groups from around the world . . . as part of the World Climate Research Programme (WCRP)."


So is this proof that climate models do, in fact, overexaggerate the future warming trend? While on first glance, such a conclusion might be justified, once you actually carefully investigate the subject, you find that there are many different problems with this position.

The first thing to point out is that some of the data used by Christy in these graphs has since been revised upwards to correct for various forms of bias. As I'm sure you all know from your extensive experience collecting data from space, measuring temperature via satellite is no small undertaking. As they write in a article,


". . . there are a number of challenging biases in the satellite record which must be corrected and accounted for.  For example, the orbital drift of satellites, the fact that they have to peer down through all layers of the atmosphere but need to isolate measurements from each individual layer, etc.

. . . contrary to Christy's assertion that TLT measurements are made by a single satellite, they have actually been made by several different satellites over the years, because the measurement instruments don't have lifetimes of 34 years.  Splicing together the measurements from various different satellite instruments is another of the challenging biases which must be addressed when dealing with satellite data."


Yeah, not to mention the fact that all the NSA spy satellites keep getting in the way!

"Move! I'm trying to record the temperature here!"

"You move, bitch! I'm trying to spy on this guy watching porn!"

One set of measurements used in this Christy graph—from RSS—has since been revised upwards by RSS to account for bias caused by orbital drift. As Zeke Hausfather writes on,


"Researchers from Remote Sensing Systems (RSS) . . . have released a substantially revised version of their lower tropospheric temperature record.

After correcting for problems caused by the decaying orbit of satellites, as well as other factors, they have produced a new record showing 36% faster warming since 1979 and nearly 140% faster (i.e. 2.4 times larger) warming since 1998."


And note that 1998 is right around the time period when the divergence began in that Christy graph. Here we see this data correction depicted graphically in that same article.

Remote Sensing Systems, on their website, compares their most recent set of measurements (version 4) against the CMIP5 model predictions, and as we can see, the difference is significantly less pronounced than it appears in the Christy graph, with many of the data points during the original period of divergence being well within the 90% envelope—with the 90% envelope basically describing the area within which 90% of model projections lie.

Benjamin Santer et al, in a 2017 publication, take it a step further and show us the average of all of these different satellite datasets after they've been recently corrected to account for various biases. As we can see, these observations track very closely to the CMIP5 predictions. Compare this against the John Christy graph and you see that the difference is night and day.

I also have a few problems with the Christy graph that are purely a matter of visual presentation. One that they mention on Skeptical Science is the lack of uncertainty ranges. When presenting data in a graph where uncertainty exists, you want to illustrate the range within which the true statistic likely lies. So when you provide a 95% confidence interval, for example, you're basically communicating that we're 95% confident that the true statistic lies somewhere within this range. When you provide a 95% envelope, you're showing that 95% of the models lie within this particular region.

No such uncertainty ranges are presented in this famous John Christy graph. Skeptical Science provides a re-made version of his graph that does include a 95% envelope for the model predictions. They also separate out the different sources of data, rather than simply providing the averages, to show that there's a considerable spread in the data points depending upon which satellite you're getting your information from. And finally, they include the most recent, updated versions of the RSS and UAH data sets.

When this is how the data is presented, the divergence between model predictions and observations appears much less pronounced than in the Christy graph. So how you present the data can make a significant impact on what conclusion people walk away with.

You'll also notice that Christy's UAH data set provides the lowest temperatures. Perhaps that's not a big surprise considering that his co-publisher of this data, Dr. Roy Spencer, is the author of such timeless classics as Global Warming Skepicism For Busy People and An Inconvenient Deception: How Al Gore Distorts Climate Science and Energy Policy.

So yes, bias might be partially at play here, but I think I know what's really responsible for these low UAH data points: Christy's enormous moustache is somehow interfering with his instruments. Look at this guy! Is he a climate scientist, or the police chief of Vicksburg, Mississippi?

And this might be kind of a minor gripe, but it does bother me that the model predictions continue to be plotted on the graph past the point where the observations are being plotted, past the year 2015. This makes the difference appear even more significant—but we have no idea what the data points during those future years would actually be. What if they would've actually caught up to or even surpassed the model predictions? Indeed, more recent data from RSS shows that after 2015, there was a fairly significant temperature spike—significant enough to exceed the average of those model projections.

With all of that said, for the sake of argument, let's assume that deniers are correct and let's say that the Christy graph is a completely accurate description of how the models compare against the observations. Even if that was the case, the most we could do from this is draw a very narrow conclusion because this is very specific data that Christy is looking at.

These measurements and these model predictions specifically focus on temperature changes within the middle of the troposphere, between 20,000 and 50,000ft above the surface—as Christy made absolutely clear in his Congressional testimony on the subject.

I'm not sure why the graph is labeled surface to 50,000ft, because he writes in the description of this graph that it depicts "global bulk (termed 'mid-tropospheric' or 'MT') temperature," and he defines the mid-troposphere in that same section as being "between 20,000 and 50,000 ft." Maybe he was trying to deceive people into believing that this data includes surface temperatures, or maybe this is just some weird atmospheric science terminology that I'm not familiar with—but this is no small difference right here.

It's like your pilot says: "Alright, everybody. We've now reached our cruising altitude of 20,000ft."

And you're like: "Dude, we haven't even taken off yet!"

By the way, why does every pilot on the planet seem to have the exact same voice and style of speaking: *rapidly and in monotone airplane-pilot voice*: "Good evening, ladies and gentlemen, this is your captain speaking. We'll be arriving in Phoenix in about two and a half hours, so sit back and enjoy your flight!"

Just once I want to have a pilot who's like: "Hey, uhh, what's up, guys?... I'm your pilot. Hope you're doing well. Um, so, let's see here... Looks like we are headed to Cleveland today. Huh. Ya know, I got family in Cleveland. My mother in law. She's a real bitch! Anyway, let's get this show on the road, I'll try not to crash this thing into the ground and, uh... yeah. That's basically the gameplan.

Enjoy your flight—or should I say my flight. I'm the one doin' all the hard work up here; you're just gonna sit on your ass and eat peanuts the whole time! You people are lazy. Entitled. It sickens me! Anyway, I'll let you know when we're close. Captain out."

So like I said, the Christy graphs look specifically at the middle of the troposphere. His reasoning for doing so is explained in a lecture he gave:


"Now the reason we picked tropical atmospheric temperatures is a very good reason: This is a cross section of how the temperature of how the earth is supposed to behave, the atmosphere. . . . So if you go to the layer that's about 10–50,000ft, in the tropics, you see that's where the biggest signal is. If you want to find, according to climate models—these are climate model average simulations—if you want to find the response to globally increasing greenhouse gases, that's the place to find the biggest response."


I think his reasoning is pretty straightforward and fair, but many deniers go way too far and conclude that since these models were wrong about temperature changes in this particular section of the atmosphere, that therefore means that all climate change predictions all across the board are equally as inaccurate. Here's a great example of a climate change denier leaping to that unjustified conclusion. As Larry Hamlin writes on,


"Mr. McIntyre provided expert findings of his review of the statistical results of Dr. Christy’s work as showing that climate models were indeed  'over warm' in their projections as follows: a model run will be warmer than an observed trend more than 99.5% of the time."


Now if that first sentence reads to you like completely incoherent gibberish, that's because that's exactly what it is: poorly-written garbage. But here's what I took away from what he wrote here: Climate models run warmer than the actual temperature changes 99.5% of the time. When I first read this, I kept saying to myself: "There is no fucking way that that's correct," and that's because it's not correct.

When you follow the link to the source for his claim, you find that he's not talking about climate models generally, here; this statistic comes from an analysis of how these CMIP5 models perform specifically in their predictions of the middle-troposphere.

What Hamlin is doing here is using a very narrow finding to reach a very general conclusion. This would be like performing a study on what percentage of South Korean surgeons over 60 who smoked cannabis that day made a mistake during a coronary bypass—and then reporting the results as the percentange of any kind of doctor who makes a mistake during any kind of surgery. These are just not the same things.

When you broaden out your analysis from the mid-troposphere to the many other areas where data is collected, you find that time and time again, the climate model predictions are right on the money.

Let's begin with the land surface temperatures, which I would argue are much more important for living creatures—because we, as well as most other land-based organisms, live on the surface of our planet, so this is where most of the changes that actually affect us will be taking place. Unless you're a migrating goose crossing the Himalayans, or unless you're some kind of freak that lives in a hot air balloon, temperature changes from 10,000 or 20,000 to 50,000ft just aren't going to affect you as directly as temperature changes near the surface will. So how do these models compare against the observations?

A great analysis of this question is provided by Zeke Hausfather in a 2017 article. He compares a variety of climate change predictions going all the way back to the 1970s against the observed temperature changes, and the pattern is one of strikingly accurate predictions.

Wally Broecker, 1975: As we can see, his climate forecast matches up perfectly with temperature changes up until about 2005 when it starts to outpace the observations. The temperature data that his model is being compared against comes from NASA, the NOAA, HadCRUT, Cowtan & Way, and Berkeley Earth—or as climate change deniers call them, "The Axis of Evil."

Hansen et al 1981 matches up almost perfectly with the observations, whereas Hansen et al 1988 mostly coincides with the observations up until about 2005 when it starts to slightly surpass the observations.

The IPCC's First Assessment Report in 1990 tracks right along with the observations, as does their second report in 1995, their third report in 2001, their fourth report in 2007, and their fifth report in 2014.

So going back as far as over 40 years ago, you see that climate model predictions have been very accurate. The IPCC's predictions, in particular, have been exceptional—and this is worth highlighting because of how influential the IPCC is.

I should also point out that when climate change models diverge from the observations, this oftentimes isn't a tremendous mystery to the climate scientists who create and use these models. No, usually they can look at the assumptions made in the prediction and see exactly where they went wrong. Here's an example of this, provided in that same article. In the IPCC's First Assessment Report,


"Their featured business-as-usual (BAU) scenario assumed rapid growth of atmospheric CO2, reaching 418ppm CO2 in 2016, compared to 404ppm in observations. The FAR also assumed continued growth of atmospheric halocarbon concentrations much faster than has actually occurred.

. . . the FAR overestimated the rate of warming between 1970 and 2016 by around 17% in their BAU scenario, showing 1C warming over that period vs 0.85C observed. This is mostly due to the projection of much higher atmospheric CO2 concentrations than has actually occurred."


Compare this against the anti-scientific defeatism of the climate change denier. When they see that certain models got it wrong, they don't say: How interesting; let's figure out what we got wrong and learn from this mistake to improve our models going forward. No, the climate change denier—or at least certain climate change deniers—see an inaccurate model prediction and irrationally conclude that virtually all such models are completely untrustworthy and aren't worth paying attention to or taking seriously. This is not how science is done and this is not a serious response to faulty predictions.

How do the models do at predicting the rate of ocean warming over time? Pretty damn well, as we see here in this graph provided by Cheng Li-Jing et al in a 2015 paper. The black line tracks the mean predicted rate of ocean warming, and the red line shows us the observed rate of ocean warming. As we can see, both lines follow a very similar upward pathway.

Contrary to what the climate change denier would lead you to believe, there are also several areas where the models have substantially underestimated the rate of warming. Brysse et al write about this extensively in a paper entitled "Climate change prediction: Erring on the side of least drama?"

One example they give is the following:


"In the TAR, released in 2001, the IPCC predicted an average sea level rise of less than 2 mm/yr, but from 1993 to 2006, sea level actually rose 3.3 mm/yr—more than 50% above the IPCC prediction . . . The underestimate in sea level rise can be traced in part to under-projection of ice loss from Antarctica and Greenland."


And here, in a graph provided in The Copenhagen Diagnosis, we see this trend illustrated. As we can see, the observed sea level rise since 1990—as measured by tide gauges and satellites—is at the very highest range of the IPCC projections.

"Pfft, fucking alarmists!"

We see a similar trend with the rate of Arctic ice melting. As Brysse et al continue,


"Summertime melting of Arctic sea-ice has 'accelerated far beyond the expectations of climate models'. Indeed, using unusually vivid language, [Allison et al 2009] note that the record for previous Arctic sea ice summer minimum extent was 'shattered' in 2007, 'something not predicted by climate models . . . This dramatic retreat has been much faster than simulated by any of the climate models assessed in the IPCC AR4'—with summer sea ice now well below the IPCC worst case scenario."


"Summer sea ice is now well below the IPCC worst case scenario? That's not so bad! It's only... worse than the worst-case scenario. Could be worse!"

Here, once again, we see a Copenhagen Diagnosis graph depicting the sharp divergence between IPCC model predictions and observations of Arctic sea ice melting.

You wanna talk about being an alarmist? If I was a climatologist who saw the ice melting this rapidly, the title of my next paper would literally just be "SHIT YOUR PANTS!" in all capital letters!

Brysse et al, in their paper, suggest "some possible causes of this directional bias, including adherence to the scientific norms of restraint, objectivity, skepticism, rationality, dispassion, and moderation."

Now that we've looked at a much broader set of comparisons between the observations and the predictions made by the models, notice how different a conclusion we reach from the climate change denier. When it comes to land surface and ocean temperatures, the models have been very accurate in their predictions. When it comes to sea level rise and the melting of sea ice, the models have actually underestimated the rate of change. Only in the very narrow area of middle-tropospheric temperature do the models overestimate the warming trends.

So looking at the models collectively, we find two key areas where they're very accurate, two where they underestimate the warming, and only one where they overestimate it—and that's assuming the correctness of this data which has since been revised upwards.

To focus only on this one particular area where the models slightly exceed the observations is no way to reach an accurate conclusion about the overall veracity of climate change model predictions—and to argue that climate models generally are biased in the direction of alarmism and overstating the threat of global warming simply doesn't match up with the facts.

Why would the models get it wrong when it comes to middle-tropospheric temperature changes? I don't know, but many different things could potentially explain this. Maybe there's just something about this particular region of the atmosphere that we don't yet fully understand. Maybe a faulty assumption was made in these models. Maybe Al Gore, hell-bent on world domination, hacked into the IPCC and modified the code? As RSS writes on their website,


". . . Why does this discrepancy exist and what does it mean? One possible explanation is an error in the fundamental physics used by the climate models. . . . There [could be] errors in the forcings used as input to the model simulations (these include forcings due to anthropogenic gases and aerosols, volcanic aerosols, solar input, and changes in ozone), errors in the satellite observations . . . and sequences of internal climate variability in the simulations that are different from what occurred in the real world.

. . . these four explanations . . . are not mutually exclusive. In fact, there is hard scientific evidence that all four of these factors contribute to the discrepancy . . ."


I should also point out that you don't need to be able to predict every single outcome of climate change to a perfectly precise degree to be able to understand the basic trends that are taking place. Sure, perhaps the middle of the troposphere warms slightly slower than scientists initially thought that it would; that doesn't mean that global warming isn't taking place, and that doesn't mean that the impact of global warming won't be rapidly felt in other areas of the planet—whether we're talking about the land surface, the ocean, the polar ice caps, or wherever it may be.

Just take a step back for a minute and ask yourself: What is going to happen if we continue pumping enormous amounts of greenhouse gases into the atmosphere? Obviously the temperature will continue to increase. To overlook this basic truism and painstakingly scrutizine only the middle tropospheric temperature record strikes me as an extremely silly example of missing the forest for the trees.

It's like I start filling up a bathtub with water and ask: What is going to happen in the future? The answer is very obvious: If I leave the faucet running, the bathtub will continue to be filled with an increasing amount of water.

"Oh yeah? Well a small percentage of your predictions about how quickly the bathtub would be filled with water in certain areas were slightly inaccurate, so whattaya think about that?" What I think about that is that you're missing the big picture here. Also, who are you and what are you doing in my bathroom? Can't you see that I'm trying to masturbate in here?

One of the most laughable versions of this argument makes the classic mistake of conflating weather and climate. As Jose Manuel Flores writes in a Tweet of his,


"That isn't the govt job. You guys and the climate change, global warming, or cooling initiative needs to get off of your high horse. We can't even predict the weather next week accurately, and you expect us to believe we can predict it for the next 30 years?"


Yeah, you tell 'em bro! Get off your high horse, climate scientists!

They're like: "There are no more horses because they're dead because of climate change."

Look, if you can't distinguish between the short-term, regular variation known as weather, and the long-term, global trend known as climate, I don't know what to fucking tell you. It's like not being able to tell the difference between drinking a few beers on one particular night and being a lifelong alcoholic. (Christ, I'm about to become a lifelong alcoholic if I read one more argument like this.)

Skeptical Science writes the following on this point:


"A common argument heard is 'scientists can't even predict the weather next week - how can they predict the climate years from now.' This betrays a misunderstanding of the difference between weather, which is chaotic and unpredictable, and climate which is weather averaged out over time. While you can't predict with certainty whether a coin will land heads or tails, you can predict the statistical results of a large number of coin tosses."


The alleged unreliability of climate change projections is something that Jordan Peterson talks about, and he's so incredibly wrong about the subject that he borders on incoherence when he talks about it.


"So, as you project outwards with regards to your climate change projections, which are quite unreliable to begin with, and the unreliability of the measurement magnifies as you move forward in time, obviously, because the errors accumulate. And so, if you go out 50 years, the error bars around the projections are already so wide that we won't be able to measure the positive or negative effects of anything we do right now. So how in the world are you gonna solve a problem when you can't even measure the consequence of your actions? How is that even possible?"


Man, looks like that meat-only diet is really starting to take its toll! He's like "Speaking of climate change, I haven't eaten a single piece of fruit in over 6 years!" We're like "Uhh... what?"

So first he makes the claim that climate change projections are "quite unreliable to begin with." He cites no specific examples of this, and as we've already seen, this is just not true; most of them are pretty damn accurate. Where is he getting this information from? Is he just lapsing into some kind of vegetable-deprived delerium?

Where he really achieves peak Galaxy Brain is when he says that increasingly wide, future error bars make us unable to measure the positive or negative effects of anything we do that's climate related. This is so hilariously wrong that it's almost brilliant. Forget climate change denialism; this is climate change nihilism right here.

Ask yourself this simple question: If we sharply increase our greenhouse gas emissions, will this increase the temperature on the planet? The answer is an obvious yes. So cutting back our greenhouse gas emissions will have the impact of minimizing this temperature increase. Peterson is simply obfuscating that very basic point when he pontificates about error bars and future uncertainty.

According to Peterson's logic here, switching the entire planet to clean energy over a period of 50 years will have an unknown impact on future climate when compared against using nothing but oil to fulfill our energy needs. This is obviously nonsense.

Peterson clearly doesn't understand the way error bars actually work in climate science. Here's a graph from a paper by J.R. Hunter et al which shows future projections for sea level rise. The shaded orange area is the 90% envelope, meaning that 90% of projections lie somewhere within that area. As we can see, this projection goes all the way up to 2100, and while yes, the range of projections does get wider the further into the future you go, it doesn't become so wide 50 years from now as to make the projection meaningless—which is precisely what Peterson seems to think takes place.

Even though the projection range does widen as you go further into the future, there's still a detectable and obvious upward trend taking place. And our actions aren't disconnected from what happens in the future; the degree to which we decarbonize around the world will have a direct impact on sea level rise and temperature in the future. When it comes to climate change, Jordan Peterson is simply confused and misinformed.

Finally, there's one additional category of climate change projections which get it very, very wrong—and these are the projections put forth by so-called climate change "skeptics." Skeptical Science compares several of these predictions against the observed temperature changes, and the results are frankly embarrassing. Especially that last one by McLean. What the fuck was that guy thinking? Did he predict that some diabolical supervillain would block out the sun or something?

Basically, in the midst of worldwide global warming, these people are predicting that global cooling should be taking place.

And in case you're wondering, these aren't just a bunch of random dumb-fucks; Richard Lindzen, in particular, is seen as some sort of demi-god by climate change deniers. He's featured in article after article on denier websites and he even starred in a misinformation-filled PragerU video on climate change where he bragged about his number of publications. Quality over quantity, Dr. Lindzen!

Here's another graph from Skeptical Science which compares Hansen's 1988 predictions against Lindzen's 1989 predictions, and again, the result is abject humiliation for Dr. Lindzen. (Yes, another Skeptical Science reference. I don't mean to be such a Johnny One-Source in this video; they just have a lot of great resources on their website.)

I think it's funny that deniers will make this faulty argument about the unreliability of climate change models, then they'll turn around and cite Richard Lindzen as an authority who backs up their climate change denialism—despite the fact that his predictions have been grossly and laughably inaccurate.

So as we've seen here, most climate change projections are actually impressively accurate. Where they get it wrong, they're more likely to lean towards the conservative side than the alarmist side. And in the often-cited case of the middle troposphere, more recent, bias-corrected data shows a much more modest trend than the famous John Christy graphs. The real embarrassment comes from predictions made by so-called climate change "skeptics," who project global cooling in spite of the obvious upward temperature trends.

When the models do get it wrong, climate scientists very often can determine why they got it wrong—and the answer to faulty science is better science, not a dismissal of science. Finally, future uncertainty about the exact changes that will take place don't prevent us from understanding the basic, ongoing trends, nor does is turn taking action to combat climate change into a futile and pointless exercise.