Guest Post by Willis Eschenbach (@WEschenbach on X, my personal blog is here)
The IPCC describes computer climate models as follows:
“Climate models are mathematical representations of the climate system, expressed as computer codes and run on powerful computers. One source of confidence in models comes from the fact that model fundamentals are based on established physical laws, such as conservation of mass, energy and momentum, along with a wealth of observations.”
“The Atmosphere–Ocean General Circulation Models, Earth System Models and Regional Climate Models evaluated here are based on fundamental laws of nature (e.g., energy, mass and momentum conservation).”
And it’s a good thing that these models are “based on fundamental laws of nature” and are “run on powerful computers”, because they are being used to tell us to totally restructure the global energy system, change our personal lives, jack up our taxes, warn us that some weather event is made eleventeen times more likely because of “climate change”, and to require that we make endless costly sacrifices to an unknown climate god.
We are constantly assured that these models are the real deal, totally lit, not to be questioned. Zeke Hausfather says, “We found that climate models – even those published back in the 1970s – did remarkably well, with 14 out of the 17 projections statistically indistinguishable from what actually occurred.” What’s not to like?
So I thought I’d take a look at the latest crop of climate models, those used in the Climate Model Intercomparison Project 6 (CMIP6). I mean, if the 1970s models were as good as Zeke says, the more recent models must be bang-on, especially since they are ” based on established physical laws, such as conservation of mass, energy and momentum, along with a wealth of observations.”
To start with, Figure 1 shows what thirty-six of those recent CMIP6 models say about the global surface air temperature from 1850 to 2024. Note that for the period 1850-2014 they are fed with what we believe are the actual concentrations of greenhouse gases and aerosols, the solar and volcanic changes, and all the other input variables, so that period is a hindcast. The forecasts with predicted concentrations begin only after 2014.
Given that the models all have exactly the same inputs, we’d expect at least their hindcasts to be quite similar. However …
Figure 1. Absolute global average surface air temperatures from 36 CMIP6 climate models.
ZOWIE! Talk about a dog’s breakfast—it would be hard to be more jumbled than that. I wonder how many of these Zeke would claim are “statistically indistinguishable from what actually occurred”.
There’s a further oddity about Figure 1. I have no clue why the models that start out colder warm up more compared to the models that start out warmer. I thought it might be the result of the relationship between W/m2 and temperature, which involves temperature to the fourth power, so I checked. Turns out, it’s not from that—the temperature change is far too small for the fourth-power relationship to make a meaningful difference. So … what is causing it?
Next, note the huge differences in overall shape of the individual model outputs. In some there is a relatively steady increase in temperature from 1850 to 2024. Others run about level or even cool down from 1850 to around 1970 and then rise rapidly to 2024. And others are constantly going up and down over the entire period.
Now, I’m not expecting them to get individual years right. They’re not designed for that. But I certainly didn’t expect that variation in results in a 175-year run.
Call me crazy, but it seems to me that either the models or the modelers have very different ideas about what constitutes “established physical principles”. From my perspective, I’d say that these models are “based on established physical principles” in about the same way that a Hollywood blockbuster filled with CGI, flying dragons, and special effects is “based on a true story”.
Having been gobsmacked by the results in Figure 1, I moved on. Given the endless fascination with the warming since “pre-industrial times”, I thought I’d see what the models say about that. People keep saying that if we warm more than 1.5°C or 2°C since “pre-industrial times”, it will be Thermageddon™ and the oceans will boil. Well, perhaps not literally, but the UN Secretary-General António Guterres said, “the era of global warming has ended; the era of global boiling has arrived” … be very afraid.
Now, the average temperature from 1850 to 1899 is what is generally used by mainstream climate scientists as the “pre-industrial temperature”. (Although actually, it’s nothing of the sort. The Industrial Revolution is generally thought to have started around 1760, which was immediately preceded by the very cold depths of the Little Ice Age … but I digress …)
So following that standard practice, I took all the model outputs shown in Figure 1 above, expressed them as anomalies around their 1850-1899 average temperatures, and graphed the results. Figure 2 shows that stunning example of basic physics …
Figure 2. What the models say about the warming since “pre-industrial” times.
Have we surpassed the dreaded 1.5°C threshold? Fifteen models say yes, twenty-one say no. How about surpassing the 2°C threshold? Four say yes, thirty-two say no. Heck, you can find a model that tells us that it’s warmed since “pre-industrial times” at anything from 0.6°C all the way up to almost four times that, 2.3°C. Almost a four-fold variation in results, with a number of them saying the temperatures in ~1960 – 1980 were below the “pre-industrial” average … basic physics at its finest.
When I put on my hip boots and started wading through this digital disaster, I was planning to compare the model results to reality. But upon consideration, I realized that it would only get me embroiled in arguments about how models can’t be compared to reality because models are showing purely surface air temperatures, but historical reconstructions like Berkeley Earth or HadCRUT use ocean surface temperatures as a proxy for air temperatures, and the like.
I also realized that wasn’t really the issue. The real issue is the following:
If all of these wonderful models are “based on established physical laws, such as conservation of mass, energy and momentum, along with a wealth of observations”, then why do their results differ so greatly?
And of course, the corollary to that:
Why does anybody pay the slightest attention to this inaccurate, ludicrously unfit-for-purpose climate model farrago?
As Chester Nimitz said in his message to Bull Halsey … “The world wonders”.
My very best to everyone,
w.
As Is My Wont: I ask that when you comment, you quote the exact words you are discussing. It prevents endless misunderstandings.
Further Reading: In my post yclept Meandering Through A Climate Muddle, I engaged in haruspicy regarding the GISS climate model. TL;DR version? Lots of kludges and tunable parameters, not much in the way of “established physical laws“.
Did comments all-in-bold get fixed yet?
A downvote for inquiring on a serious (continuing, annoying) issue? Have you LOST your MIND!!???
Get your sanity checked …
I never saw that. Only Narcissist Neil’s comments were all bold, to my knowlegde.
Climate models are now the first resort, simulations are being described as ‘experiments’ and the output as evidence. The World Weather Attribution outfit is the pinnacle of lawfare dressed up as science: it’s all your fault and their models say so.
Good as always Willis! I know it’s a bit of an ask, but could you give a link to your R code that generates the two graphs?
I think a lot of readers here would love to be able to regenerate them. When we cross-post them, we would be able to defend them more easily.
Mmm … I use a lot of my own functions, but I’ll give it a shot. I’ve put the data files (a zipped folder called “CMIP6 TAS.zip” in my Dropbox here. Download and expand it and put it in your R workspace. From there:
require(stringr)
require (Rlibeemd)
I think that’s all the packages needed, if not let me know. Then the fun begins:
thepath=”CMIP6 TAS/”
filenames=dir(thepath)
(goodfiles=grep(“245″,dir(thepath)))
goodnames=filenames[goodfiles]
(thelocs=str_locate(goodnames,”_ssp”))
justnames=left(goodnames,thelocs[,1]-1)
i=1
alltemps=ts(matrix(NA,2100,length(goodfiles)),1850,frequency=12)
rawtemps=ts(matrix(NA,2100,length(goodfiles)),1850,frequency=12)
allfits=ts(matrix(NA,2100,length(goodfiles)),1850,frequency=12)
rawfits=ts(matrix(NA,2100,length(goodfiles)),1850,frequency=12)
for (i in 1:length(goodfiles)){
print(i)
whichfile=goodfiles[i]
(testfile=paste0(thepath,dir(thepath))[whichfile])
(theskip=last(grep(“#”,readLines(testfile))))
rawdata=read.table(testfile,skip=theskip)
# head(rawdata)
testts=window(residual(tabletots(rawdata[,2:13],rawdata[1,1])),end=c(2024,12))
rawtemps[,i]=testts
rawfits[,i]=xcfit(testts,4)
testts=testts-mean(window(testts,end=c(1899,12)))
alltemps[,i]=testts
allfits[,i]=xcfit(testts,4)
}
save(alltemps,allfits,rawtemps,rawfits,file=”cmip6 245.tab”)
Now, that uses some of my own functions:
hardwrap=function(thelen,thetext){
require(stringi)
thetext=trimws(thetext)
if (nchar(thetext)<=thelen){
newtext=thetext
} else {
ptr=0
while(nchar(thetext)>thelen){
ptr=ptr+1
(thebreaks=stri_locate_all_fixed(thetext,” “)[[1]][,1])
(thesplit=thebreaks[max(which(thebreaks<thelen))])
thestart=thesplit+1
(textchunk=substr(thetext,1,thesplit-1))
if (ptr==1){
(newtext=textchunk)
} else{
(newtext=paste0(newtext,”\n”,textchunk))
}
(thetext=substr(thetext,thestart,nchar(thetext)))
}
if (nchar(thetext)>0) newtext=paste0(newtext,”\n”,thetext)
}
newtext
}
and
residual=function(tser,remove_mean=F){
tser-seasonal(tser)-
ifelse(remove_mean,mean(tser),0)
}
and
xcfit=function(thedata,howmany=3){
testxc=ceemdnu(thedata)
getxcfitted(howmany,testxc,thedata)
}
and
getxcfitted = function(x, testxc, orig,doprint=F) {
thelm=lm(orig ~ testxc[, (ncol(testxc)-x+1):ncol(testxc)])
if (doprint) print(summary.lm(thelm))
if (class(orig)==”ts”){
ts(
thelm$fitted.values,
start = start(orig),
frequency = frequency(orig)
)} else {
zoo(
thelm$fitted.values,
time(orig)
)
}
}
and
ceemdnu=function(somedata,titlename=”Dataname”,datasource=””,scale=F,
doplot=F,dospectra=F,specstart=4,marktime=F,devoff=F){
if (marktime) print(ceemdtime(somedata))
if (scale==T) somedata=scale(somedata)
theans=cbind(somedata,ceemdan(somedata))
then=ncol(theans)-2
colnames(theans)=c(“Raw”,paste0(“C”,c(1:then)),”Resid”)
thewid=ncol(theans)
theans[,2:(thewid-1)]=theans[,2:(thewid-1)]+mean(somedata)
if (devoff) dev.off()
if (doplot) {
resetplot()
plotxc(theans,titlename=titlename,datasource=datasource)
}
if (dospectra) {
resetplot()
plotspectra(theans,thistitle=titlename,minval = specstart)
}
theans
}
From there, I run:
# adj temps —————————————————————
plot.zoo(allfits,plot.type=”s”,col=rainbow(36,alpha=.8,end=2/3),lwd=3,
xlim=c(1850,2050),ylim=c(min(allfits),2.5),
xlab=”Year”,ylab=”Surface Temperature Anomaly (°C)”)
abline(h=0)
abline(v=2014,lty=”dotted”)
texty=-.3
text(2017,texty,”Forecast –>”,adj=0)
text(2011,texty,”<– Hindcast”,adj=1)
rightx=2042
arrows(x0=rightx,y0=min(allfits[2100,]),y1=max(allfits[2100,]),angle=90,code=3,length=.25)
thediff2=max(allfits[2100,])-min(allfits[2100,])
boxed.labels(rightx,mean(c(min(allfits[2100,]),max(allfits[2100,]))),paste0(round(thediff2,1),”°C”),
ypad=thepad)
offset=.1
text(rightx,min(allfits[2100,])-offset,paste0(round(min(allfits[2100,]),1),”°C”))
text(rightx,max(allfits[2100,])+offset,paste0(round(max(allfits[2100,]),1),”°C”))
text(mean(par(“usr”)[1:2]), 2.4, hardwrap(60,”(NOTE: To enhance visibility, all model outputs have been smoothed using a CEEMD smoother. Anomalies are expressed around a base period of 1850-1899)”),
font=1,cex=.9)
logo()
sigline()
title(sub=paste0(“DATA: https://climexp.knmi.nl/select_knmi23.cgi“),cex.sub=.9,line=3.1,font.sub=1)
title(main=paste0(“36 Climate Model Surface Air Temperature Anomalies, CMIP6 SSP245\nOne Member Per Model, 1850-2024″),cex.main=1,line=.8)
ablineclipv(h=c(1.5,2),x1=1850,x2=2028,lty=”dotted”)
write_clip(“CMIP6 SSP245 since 1850”)
# __ ———————————————————————-
Give it a shot. If I’m missing some function I’ve written or you need some part explained, let me know.
w.
Climate models are based on false assumptions and circular reasoning. Physics and data that contradict the agenda are simply ignored.
The core problem with climate models is that there isn’t enough computer power available to model all of the relevant physics. That makes the IPCC description of climate models a half truth at best.
Correct, but you understate the actual situation. I just posted details below.
Probably more accurate to say that I didn’t mention how many orders of magnitude of computer power needed for a true first principles analysis. As you mentioned in your post, there is little chance of adequate computer horsepower being available in the foreseeable future.
All of the relevant physics can’t be measured precisely enough to establish truth. Even if we had the grid cell resolution, climate is a dynamic chaotic system in continuous flux so measurements are a moving target.
In other words we don’t have the initial conditions known well enough to do a high resolution model.
No, in a chaotic deterministic system, knowing initial conditions exactly doesn’t help. It might be counter intuitive, but seems to be true.
Einstein refused to believe even the uncertainty principle, saying that God does not throw dice!
Einstein has been proven wrong.
The issue with a chaotic deterministic system is getting the initial conditions with sufficient precision as small errors can produce dramatically different outcomes.
The word Heisenberg used for “uncertainty” better translates into indeterminate. The position/momentum “uncertainty” is closely related to time/frequency with respect to an amplitude modulated radio signal, any change in amplitude will produce frequency components (sidebands) other than the carrier.
Isn’t this why climate models cannot hindcast from the known knowns at a specific point in time?
No measurement can be quantified precisely enough because of inherent errors in instrumentation. Also, there are errors involved in the conversion of analogue data to digital data and the time needed to complete and store that converted data , by which time the system has changed even if it’s only milliseconds later.
Aside from grid size, analog to digital is key.
The earth energy systems are analog.
Quantization to produce digital introduces errors at every interface.
Top that off with Nyquist sampling theory.
Many of the factors involved have frequency spectra of unknown complexity.
However Deep-thought will provide the answer at the end of the universe.
But not the question….
Nice post, WE. The claim that climate models are relying on basic physics may be true in theory, but not in practice—as you show. The underlying problem is intractable.
If climate models really were relying on basic physics, their grid scales would be on the order of 2-4km, similar to the best regional weather models. 2km resolution, for example, can resolve a thunderstorm convection cell as is done in ECMWF. Good for a ‘small’ region with known regional boundary parameters, forecasting reasonably well out just a few days.
But the CFL theorem mathematical constraint to numeric solutions of PDEs means such grid scales for global climate models are about 6 ORDERS OF MAGNITUDE computationally intractable. At the average CMIP6 grid scale—100km at the equator—they still take about 60 days per run (per NCAR) on today’s best supercomputers. At, say, 3km a single run would take about 60 million years—by definition computationally intractable. And there is no super duper supercomputer solution even remotely on the horizon.
So climate models have to be parameterized, whereby any pretense of actually using basic physics is lost in practice.
Per CMIP protocol, model parameters are tuned to best hindcast 30 years. And CMIP hides the resulting ‘tuned’ hindcast discord you vividly illustrate (your CEEMD smoothing understates the actual discord) by comparing only modeled 30 year hindcast anomalies.
To summarize, climate models pretend to use basic physics, then CMIP hides the awful results when they cannot.
They do guarantee transfer of funds from one bank account to another. Then, on the receiving end, these financial assets can get converted to physical assets or just plain fun.
Model to bank account is the global teleconnection that counts the most.
Thanks, Rud. The underlying problem is that the climate has relevant phenomena on a physical scale ranging from molecular to planet-wide, and on a temporal scale from picoseconds to millennia. As you point out, there’s no way to cram that into a computer.
Best to you, my friend.
w.
Likewise, and also to your gorgeous ex-fiance.
I always was fascinated how the models simulated individual molecules on a 25 km^2 grid.
Never mind that regional weather models are often wrong because we lack the grid cell resolution to measure precisely what is going on even at those scales. A pretend world is being represented as unequivocable evidence.
Actually, present weather forecasting models show surprisingly good ‘skill’ out to between 5-7 days. Something like 95% ‘right’ for tomorrow to about 70% ‘right’ five days out. Beyond a week, they are miserable for the reason you give.
Having until end of last August lived directly on the Atlantic Ocean in Fort Lauderdale for 25 years, I was particularly sensitive to Hurricane forecasts. Tracks have become very reliable 3 days out. Much better than when we ‘unexpectedly’ got slammed by Cat 3 Wilma in 2005—eyewall moved inland only 10 miles south of us. Intensity, not nearly as good yet—only OK within about 24 hours.
But never fear, while weather forecasting might be struggling for better accuracy > 7 days out, weather ATTRIBUTION has arrived with “settled science” relibility.
So there’s that . . .
🙂
‘So there is that’…
I have long ‘believed’ in better weather attribution systems, such as Puxatawny Phil’s February groundhog shadow. Heck, it must be right since been a Pennsylvania tradition for hundreds of years—always now broadcast live—plus also a best selling movie.
Define “right.” Weather models are usually “right” at synoptic scale, but they are notoriously poor at precise short term local predictions. Will it rain? — maybe. Exactly where and how much? Exactly when will local wind shift / pressure trough occur — commonly + or – 6 hours. Will we get snow or just cold rain or drizzle? Where exactly will a tornado occur and what intensity? Who gets the 2 inch hail and a new roof, and who sees nothing more than pea size hail and big rain drops?
Synoptically, the NWS and various models can reliably say conditions may be favorable for any of the above and roughly when, usually over a large multi-county watch area, which lets people be generally prepared and watchful for inclement weather. Beyond that, an old-school synoptic weather forecaster with good data feed and a manual paper map table can do almost as well.
I took my synoptic meteorology courses in summer in Texas, when weather is generally boring and same day after day. I vividly recall my professor saying that in summer, a forecaster can be 95% accurate by assuming continuity with yesterday’s weather, with only slight adjustments. BUT — it’s the 5% he gets wrong that can be the difference between life and death.
Advanced radar and finer scale and more timely observation networks help greatly, but single-point timing and severity predictions are notoriously inaccurate, sometimes laughable and occasionally tragically wrong.
You make a point I cannot logically argue with. I did define ‘right’ in my comment. I focus on the 95%, you focus on the 5%. That 5% wrong killed 26 people last year in Texas Hill country.
But they did have still some hours of flash flood warning, so the basic problem was still not the very short term weather forecast.
Despite being born in Chicago, I spent most of my youth in Northern California. I became complacent about weather forecasts because basically California only has two seasons — hot, dry summers and cool, wet winters. I lived in Vermont for a couple of years and got disillusioned by how frequently the rain forecasts were wrong; temperatures weren’t an issue — summer came on July 4th and left on July 5th. I rationalized that the problem was the Green Mountains. I’m now living in Ohio and dismayed that even with geostationary satellites, Doppler radar, and computer models, which didn’t exist when I was young, the forecasts usually change by about a day from the initial forecast. Sometimes, even the day of forecast rain, it doesn’t rain, and I think it is much greater than 5%. It appears to me that the error rate for false-positives is much greater than for false-negatives. If there aren’t any clouds, I don’t expect rain. I suspect that there is no statistical difference between forecast daily temperatures and model forecasts.
The expansion of doppler and satellite imagery has improved forecasting dramatically short term. There are still great differences in the modeled forecasts between the US, Canada and Europe. The professional service I use looks at all of them and tries to guess which one will be the most accurate. Pretending climate forecasts represent a reality that hasn’t happened was my main point.
If you live in a temperate zone, “forecasting” that tomorrow’s temperatures will be within a standard deviation of today’s will likely be right 85% of the time.
I tried it where I live (downloading a year’s temps, offsetting them in spreadsheet columns by one, and calculating the difference today/tomorrow, and got better than 95%, probably because of proximity to the sea and a tropical location.
So unless the “model” can consistently outperform the naive forecast that can be done by a 12 year old, colour me unconvinced. Who would fly with an airline who says that 95% of their passengers will survive the trip – on average?
In my experience “right” is about 50/50.
Much smaller grid scales and constant data updates is how they are achieving useful forecasts.
I recall, not that long ago, a local radio station received phone calls from listeners immediately after the weather forecast for those in the radio studio to go look out the windows.
Ah! The good old “Weather Rock”.
The “skill” of weather forecasting models depends on the geographic location to which the model is applied. I live in Northern Virginia, about 30 miles south of DC. This area is evidently on the cusp of two major weather cells, and while the weather further away from the cell boundary in either direction can be forecast pretty well, the location of the boundary moves unpredictably. Predictions at this specific location are just guesses.
But aren’t the parametric relationships incorporated into the models based on physics themselves?
Nope. Not even close. Read my post linked to at the end of the head post.
w.
Climate models are mathematical representations of the climate agenda.
If you want a model output that follows CO2.. just make CO2 a main driver, and hindcast to data that has been already adjusted to represent the “desired” warming by CO2..
Doesn’t matter if ANY of it is actually real.. it is the model output that matters.
Clearly, the transformation of temperatures to temperature anomalies, as shown in Figures 1 and 2, is necessary for the modelers to maintain any pretense of scientific reality.
Their real problem, however, is that their models are not based on ‘the fundamental laws of nature’, but on the conjecture that the phenomenological physics of radiative transfer theory explains the transfer of energy through a highly convective troposphere.
The current definition of anomaly is a total revision of the definition.
Laws of physics?
More like lies of physics.
GreenHouse Effect theory says with it Earth’s Global Mean Surface Temperature would be 288 K or 15 C, while without it tGMST would be 255 or -18 C, a difference of 33 C (not K) and Earth would become a -18 C ball of ice.
That is simply wrong.
There is no consensus that 288 K is the GMST & 255 K is the result of a 30% albedo also no consensus. GMST was 15 C in Arrhenius’ 1896 paper (walked back in 1906) and no GHE means no 30% albedo. Without GreenHouse Gas water vapor there is no GHE and Earth would become much like the Moon, barren, 400 K lit side, 100 K dark.
The GHE energy balance graphic found in TFK_bams09 and a plethora of clones does not in fact balance and violates both 1st & 2nd laws of thermodynamics.
The kinetic heat transfer processes of the contiguous atmospheric molecules render “extra” surface energy upwelling as a calculated 396 W/m^2 black body impossible.
The GHE is failed science and Catastrophic Anthropomorphic Global Warming a fear mongering political and social power grab.
I don’t go along with your Moon-like assumption.
Mars has pretty much 0 humidity, yet its temperature swings are not as violent as the moons one – not even close.
The reason for wild swings is an extremely slow rotation of the Moon.
No. The maximum surface temperature on the Moon is reached after the same exposure time as on Earth.
The Moon’s surface simply receives the maximum radiation from the Sun, while the Earth’s atmosphere prevents about 30% of the Sun’s radiation from reaching the surface.
However, in the absence of sunlight, a longer cooling time will result in a lower temperature. Hence, in the desert, the minimum temperature generally occurs just after sunrise.
Remove the Earth’s atmosphere or even just the GHGs and the Earth becomes much like the Moon, no water vapor or clouds, no ice or snow, no oceans, no vegetation, no 30% albedo becoming a barren rock ball, hot^3 (400 K) on the lit side, cold^3 (100 K) on the dark. At Earth’s distance from the Sun space is hot (394 K) not cold (5 K).
That’s NOT what the RGHE theory says.
EVIDENCE:
RGHE theory says “288 K (15 C) w – 255 K (-18 C) w/o = a 33 C colder ice ball Earth.” 255 K assumes w/o case keeps 30% albedo, an assumption akin to criminal fraud. Nobody agrees 288 K is GMST plus it was 15 C in 1896. 288 K is a physical surface measurement. 255 K is a S-B equilibrium calculation at ToA. Apples and potatoes.
Nikolov “Airless Celestial Bodies”
Kramm “Moon as test bed for Earth”
UCLA Diviner lunar mission data
JWST solar shield (391.7 K)
Sky Lab golden awning
ISS HVAC design for lit side of 250 F. (ISS web site)
Astronaut backpack life support w/ AC and cool water tubing underwear. (Space Discovery Center)
Nikolov “Airless Celestial Bodies”
Kramm “Moon as test bed for Earth”
How ’bout their moon like assumptions?
The same old twaddle.
Ages ago, the earth without atmosphere and black all over in the Goldilocks zone was calculated to theoretically have a certain temperature using simple physics, I think -15 C by some bloke trying to understand the differing temperatures on Mars and Venus compared to Gaia.
Then, an atmosphere was added and the same calculation was made.
Hey presto, having an atmosphere like earth allowed the surface to be significantly warmer, like, suitable for life with water present as solid, liquid and gas.
End of. Nothing more. Finis.
A guess. Back of an envelope. Proof of concept and NOTHING more.
Everybody thinking about the planets in our solar system moved on.
It was a ridiculous model, first with naked rock surface planet and then with atmosphere. Legacy intellectual exploration of reality – a tiny snippet of understanding in a sea of ignorance.
Catch up or shut up.
PS Do you actually have any idea what we all mean by and understand about the phenomenon “black-body radiation”?
Having an atmosphere makes the Earth cooler.
Demonstrated a BB by experiment.
Because of the significant (60% per TFK_bams09) non-radiative, i.e. kinetic, heat transfer processes of the contiguous participating atmospheric molecules the surface cannot upwell “extra” energy as a near Black Body.
EVIDENCE:
As demonstrated by experiment, the gold standard of classical science.
For the experimental write up see:
https://principia-scientific.org/debunking-the-greenhouse-gas-theory-with-a-boiling-water-pot/
Search: Bruges group “boiling water pot” Schroeder
A BB must absorb and emit all its energy by radiation alone.
Participation by kinetic processes negate being a BB.
Wanted this graphic.
Do understand a BB??
The history of science is one of acrimonious debate and ad hominem rude behavior mostly by the old guard who turned out wrong: caloric, phlogiston, luminiferous ether, plate tectonics, the calculus, longitude clock, etc.
I earned my BSME from CU Boulder in 1978 and proceeded to apply that over 35 lucrative years in the power generation field where heat transfer and thermo run 24/7.
So, yeah, I understand BB and modeling the Earth’s surface as a 396 W/m^2 BB is just flat wrong.
The Climate Modelers Handbook* tells them they can have their own facts but they all have to have the same opinion.
{An apology to Schlesinger, Moynihan, and others.}
[*Anyone else remember the CRC Handbook?]
I have a 1947, 1960, and 1995 copy of The Handbook.
Really? One copy (purchased in 1979) worked for me. 😊
I bought the oldest one in a used book store while a poor graduate student; I couldn’t pass up the price and was curious how it compared with the 1960 version that I had received as a prize for being the best science student at my high school. When I was visiting the University of Arizona in 1996, I decided to browse in the campus bookstore to see what texts they were using in geology. They were having a sale on the previous year’s ‘handbook’, which by then had grown to a full-sized book sufficiently heavy to subdue a charging bull. Again, I was curious to compare it to the two books I had acquired previously. One can get by with a single book, but sometimes there is more to a book than just the tabulated information inside.
Of course – every engineering student had one, or knew where to find it in the library. Is it still published, or has it given way to the internet?
I used Marks handbook.
Each model has a unique suite of adjustable parameters. Parameters cover climate processes that cannot be modeled (e.g., clouds). Each parameter set has been tuned to reproduce the historical anomalies.
Parameters are fixed at their tuned values. Christopher Essex calls them unphysical ad hoc conservation laws.
Successful hindcasts come when the parameter set tuning has produced off-setting errors.
The result is models with climate sensitivities ranging across a factor of about 3, but that can, each and all, reproduce historical anomalies (but not absolute air temperatures).
As soon as the projection runs past the calibration region, the tuning loses force, the errors no longer offset, and the individual climate sensitivities cut in. The models then project a whole set of unequal trends in future air temperature. Just as you found, Willis.
Jeffrey Kiehl pointed out this problem back in 2007, and it’s been conveniently ignored ever since.
That’s a very important point and I’m surprised Willis didn’t mention it. Even a concensus climate modeller once described paramaterisation as their dirty secret. When the IPCC says the models are based on physical laws they are fibbing as usual. Some people would call these parameters fudge factors. It means they can get any answer they want.
I once defined what I called a “pure” computer model. Such a marvel would have three elements:
1. The initial conditions.
2. The physical laws.
3. Absolutely nothing else
A pure computer model is impossible and probably will be for a very long time. Maybe quantum computers will make it possible but I’m not holding my breath….
I modeled distillation columns in grad school. We used the buckingham pi theorem that specified the number of dimensions for the model.
Do the number of independent variables employed in each model vary?
Playing in the wrong dimension vector space is wasting your time.
We at least had a chance to get it right, whereas the climate modelers add as many adjustable parameters as they wish. Who was it who criticized model “over specification” (too many variables) by saying “I could model an elephant with five parameters”?
Solving those models used a lot of mainframe time. I used a 300 baud modem to connect from my apartment to run FORTRAN jobs at night. Early ’80’s. Linear Algebra has not changed since then.
cw, I didn’t mention parameters in this post because I discussed them in endless detail in the post linked at the end of the head post, Meandering Through A Climate Muddle.
Best regards
w.
Where you say the models are hindcast “1850-1914″ I think you mean 2014. The next sentence says that forecasts only begin from 2014.
Thanks, Paul. Fixed. That’s the thing I like the most about writing for the web instead of the journals. My errors rarely last more than a few hours.
Best to you and yours,
w.
“My errors rarely last more than a few hours.”
No, Willis, that is simply another lie. Your complete ignorance of what “radiation” and “power” mean has somehow lasted more than 15 years, hasn’t it?
Now that’s not strictly true – I’m not giving you enough credit. You told us that “radiation is energy” and “power is the rate of doing work”. Those definitions are correct. But everything else you have written using these terms, like that nonsensical “Steel Greenhouse”, is nevertheless still wrong. Because you ignored your own definitions. In a genius-level display of cognitive dissonance.
Pass. All you do is driveby posting and nasty snark.
w.
“All you do is driveby posting”
Another lie. I’m not the one hypocritically refusing to defend my words, Willis. Wasn’t it you who told us that a man is expected to be able to do that? Yes, yes it was.
“Pass.”
You can’t “pass” your way to scientific understanding and knowledge, Willis. It doesn’t work that way. But you wouldn’t know anything about any of that, would you?
Rave on. It’s amusing.
w.
You have fallen off the bottom of your own pyramid of argument, as usual. Move up.
Steve, you don’t seem to get it. You’ve burned your bridges with me.
There’s no point in whining about me and my eeeevil ways, I’m done with you. Go bother someone else, there’s a good fellow.
w.
Still nowhere near even the bottom rung of your own pyramid of argument, right, Willis the hypocrite? Just keep lying and hope no one notices. And you still can’t defend your own words. But then, you did say that rule only applies to men.
Pass. But hey, keep raving, it’s hilarious.
w.
“it’s hilarious.”
Maybe to a six-year-old. But it’s never too late to grow up. I’m sure you can manage it if you try. Even now.
More insults. How surprising!
w.
If you want compliments, WIllis, you’re going to have to start behaving like an adult. Why not start now? For example, by defending your words. Or was that rule only for everyone else? Hypocrite.
Steve, you seem to be laboring under the misapprehension that I have the slightest interest in discussing science with you.
It goes nowhere, and you endlessly insult me along the way.
Puts me in mind of the old joke.
Q: What’s the difference between a rat running a maze, and a college student running a maze?
A: If you take the cheese out, the rat stops running …
Sorry, but I stopped running the stevekj maze long ago, despite the fact that you are focused on getting me started running again.
Never gonna happen. Go bother someone else. You’ve burned your bridges with me.
Have a great life.
w.
“burned your bridges”
You seem to be under the misapprehension that I care the slightest bit about your “bridges”.
No, I am here to stamp out bad physics, Willis, and you are full of it. Perhaps you should peddle that somewhere else. Try The Conversation, or Skeptical Science. They love that stuff over there.
As a sideline, I am also here to call out lies and hypocrisy, and you are full of those too. This is WUWT. Try to respect Anthony’s site policies. Thanks.
“Never gonna happen”
Yes, I know, but I keep trying…
“you endlessly insult me”
No, hypocrite, you don’t get to complain about insults until you apologize for calling me an ankle-biter. And a pig. And a horse. And pond scum.
“running a maze”
No, Willis, physics lessons are not a “maze”. And you wouldn’t recognize the “cheese” if you tripped over it, so that’s a terrible analogy for you.
But since you insist on using it anyway, you actually already found the “cheese”, didn’t you? And being a polite physics teacher, I congratulated you when you did, since very few (if any) of your fellow non-physicists have managed this. Here it is, in your words, since you probably forgot:
1) Radiation is energy
2) Energy is the capacity to do work
3) Work is what happens when a force acts across a distance
4) Power is the rate of doing work
That’s all the “cheese” you need. Of course, this “cheese” tells you that your entire Steel Greenhouse is nonsense, as is any waffle about “downwelling longwave infrared radiation power”. So you should probably retract all of that now, post-haste. Since you said that your errors only last a few hours on the internet. And you wouldn’t want to be a liar. Right?
Which is what had me wondering why the graph of models didn’t agree on the “starting point” of 2014, as all the modelers had the same data about that year. The spread of “hindcasts” would tell much about the utility of the models.
From the article: “but historical reconstructions like Berkeley Earth or HadCRUT use ocean surface temperatures as a proxy for air temperatures, and the like.”
Yeah, that’s the method they use to make it appear that today is 1.5C warmer than in the past. That’s their lie.
There is no credible sea surface temperature data. They make it up out of whole cloth.
The actual written, historical land temperature records from around the world show it was just as warm in the recorded past as it is today. It is not 1.5C warmer today than in the past.
So the climate change charlatans had to change that benign temperature profile into a scary “hotter and hotter and hotter” Hockey Stick chart temperature profile and they did it by including bogus sea surface temperatures in the mix.
They made it all up. The scary Hocky Stick chart profile is a figment of the imagination. It does not represent reality. It is the BIG LIE of alarmist climate science. Without this bogus temperature chart, the alarmists would have nothing to talk about.
And they have duped a lot of people into believing this lie.
There is no “hotter and hotter and hotter” temperature profile in the written, historical land temperature records, so where did the Data Mannipultors get a Hockey Stick profile?
The charlatans who did this won’t even tell us how they arrived at their conclusions yet they expect us to just accept what they have done. Not scientific.
Unfortunately, too many people look at the world through a Hockey Stick chart.
Don’t be fooled! You are being lied to.
Make them prove a Hockey Stick chart temperature profile is legitimate. They can’t do it! They want you to take their word for it.
TA, some factual embellishment to your astute SST observation. Historical SST has two fundamental problems. First, it was strongly trade route dependent. Most ocean was never sampled until ARGO. Second, methods have changed without calibration. As just one example, ‘Modern’ SST depends on measuring engine cooling water intake. But that depth depends on whether the vessel is laden or unladen, neither being the actual sea surface as was true with the old bucket overboard thermometer method.
‘…neither being the actual sea surface as was true with the old bucket overboard thermometer method.’
I’m sure those were extremely accurate readings, too. /s
It’s worse than that. Stevenson’s 1964 SST experiments on the Velero IV research vessel showed that the keel disrupted the near-surface thermocline.
He found that valid SSTs could only be obtained if the sensor was mounted on a forward boom, and the ship was sailing into the wind.
Stevenson ended with a very negative appraisal of the reliability of the SST record.
Also, all the at-sea ship-borne SST calibration experiments gave very poor results, indicating very large measurement uncertainties.
All quantitatively evaluated here.
A vessel being propelled through the water disturbs its environment? Who would have thought about that one?
As always, Pat, your contributions are valuable.
Best to you and yours,
w.
“But that depth depends on whether the vessel is laden or unladen”
African or European vessel?
Oh, very good!
w.
The Darth(sp) Valley Tmax record still stands after 113 years.
Some years ago, I visited the KNMI site and ran all 35 models. Apparently you could run the models using absolute temperature or the anomalous T. I chose absolute and got a result that looks much like Willis’ Figure 1, except my result showed what each model was predicting for the temperature today! The range nearly exactly matched the roughly 12-15 C that Willis got for “preindustrial”. Clearly the models were STARTING OFF with a wide difference in what they thought the global temperature was.
All models are wrong, some models are useful. That’s how the saying goes. I’m certain of the first part only, and very doubtful about the second part.
Incidentally, the title is an example of what the British call a ‘spoonerism’, named after the reverend Spooner who was famous for muddling his sermons in a similar fashion.
Maybe he was just lysdexic.
Trick or trout!
Spoonerisms FUNNY?
https://www.capsteps.com/lirty/harispilton.html
Use the “listen” button.
Yes, Willis, the modelers have discovered that simply throwing the spaghetti on the wall improves the flavor!
Before a model is published, the modeler is aware that simply adding random variables while keeping the sample data constant can improve system correlation even as the degrees of freedom, and thus the correlation’s reliability, suffer with each manipulation.
(And thus the non sequitur opening the post)
Which is to say that if there were only more earths, these models could be tested more effectively.
Always ask about the degrees of freedom of a model, even though you know the answer will be zero (because there are too many variables versus population data points).
More evidence that models only say what they are told to say.
Willis, I want to dwell on your point about model hindcasts and forecasts (projections) of actual temperature. Clearly, they are and have always been at great odds across the model ensemble, with a range exceeding the alleged temperature anomalies. Many of the models hindcast “pre-industrial” temperatures that are higher than other models simulate at the present time.
Would it be safe to say these modelers start with the assumption that CO2 added to the atmosphere will cause catastrophic runaway global warming? If that is the case don’t you think that they should provide up front the levels of CO2 being used in their models and the assumed effectiveness of each added unit of CO2 added.
What? You want the modelers to provide parametric sensitivity analysis of their equations?
Aside from identifying and challenging every assumption, sensitivity analysis was required or the whole kit and kabootle was rejected.
Basic stuff in my line of work. Partial differentiation and all that.
With exact partial differential equations replaced with parameterizations (i.e. best guesses for average behavior) for cloud calculations. Much of the models calculate correctly, and are then modified by the ‘best guesses,’ which are of unknown veracity. The outputs are also modified when a step calculation becomes un-physical. (I would think that if the equations and parameterizations were correct the calculations would never become un-physical.) Lastly, propagation of uncertainties are ignored. With a pedigree like that, why would anyone not have confidence?
>> “We found that climate models – even those published back in the 1970s – did remarkably well, with 14 out of the 17 projections statistically indistinguishable from what actually occurred.”
Sentences like that are often used to claim skill for the climate models, but reality is that while it is necessary that models are based on physics and need to be tuned to match reality, these two points are not sufficient!
A necessary but not sufficient condition is a requirement that must be present for an event or result to happen, but having it alone does not guarantee that the event will happen
It is essential that for these models the full possible parameter range (given from real world measurements) is explored and the range of results used to estimate uncertainty.
The common praxis to only use models tuned within a parameter range the modelers believe in is religion, not science!
“One source of confidence in models comes from the fact that model fundamentals are based on established physical laws, such as conservation of mass, energy and momentum, along with a wealth of observations, though the models have never actually been validated with observations. We just threw the word ‘observations’ in there to trick you into thinking that they had. Really they’re just computerized guesses and our statement of confidence is just mumbo jumbo designed to disguise that fact.”
Fixed it.
For the record, anyone who checks the models’ RCP projections for future temperature and sea level rise at specified CO2 levels and compares them to measured trends will quickly realize they’re all junk. Every single one of them. They’re guesses dressed up to look sciencey, but still just guesses.
But they “observed” the graphics spewed forth.
That doesn’t count as observations?
/sarc
Although they were careful to keep a copy out of the SPM the IPCC did actually show the results of a similar exercise in Figure 1.11 of the AR6 WG-1 assessment report (2022, on page 190).
Backup URL in case the direct hotlink to the PNG file stops working :
https://www.ipcc.ch/report/ar6/wg1/figures/chapter-1/figure-1-11/
They also included my favourite “cop-out clause” of the entire report in section
1.5.4, “Modelling techniques, comparisons and performance assessments”, on page 221 :
The final sentence above should be the preface of any climate model presentation.
But how far from “perfect” is acceptable? To climate realists? To Climate Scientists?
“Perfect” is just short of being ‘over-fitted.’ 🙂
Slipping in Chester at the end. Nice!
Who knew that the ‘settled science’ could produce such variation.