by John Pabst
In the late 19th century, weather observation was a decentralized, non-standardized task performed by a vast network of disparate nodes. Data were collected by postmasters, lighthouse keepers, and railroad agents—individuals whose primary professional responsibilities were administrative or logistical, not meteorological. While these observers were often diligent, they operated within a system that lacked the metrological infrastructure we require today. Observations were recorded without the benefit of universally standardized shielding, precise, synchronized time-stamping, or consistent siting requirements. Consequently, the temperature was often logged based on the equipment and location that was practically available at each specific site, rather than through a uniform, instrument-grade protocol.
The observations were real, but the measurement system was highly heterogeneous. Instruments, exposure conditions, and recording procedures evolved and varied over decades. If you look at the raw data from that era, it is a patchwork of gaps, “round-number” biases, and inconsistent reading times. The challenge is not whether temperatures were measured, but whether those measurements can be combined into a century-scale record with the degree of accuracy often implied in public discussion.¹
The “Fix”
To turn that chaotic journal into a clean, global trend line, modern climate agencies have to perform a massive act of translation. The original thermometer reading goes in one side of the process; the final temperature record comes out the other. Between the two sits what most people never see: a black box of adjustments, reconciliations, and statistical corrections designed to account for station moves, changing instruments, urban growth, and missing data.
Here is the problem: while known biases can be estimated and adjusted, information that was never measured cannot be fully recovered. If a thermometer in 1890 was sitting on a sun-drenched porch in a growing town, no algorithm in 2026 can know with certainty how much that reading differed from the true ambient air temperature.² The black box may produce a cleaner record, but confidence in the final trend depends not only on the original measurements—it also depends on the assumptions embedded in the reconstruction process.³
The Hidden Uncertainty
We are told these records are accurate to within fractions of a degree—typically cited as an uncertainty interval of approximately ±0.15°C for the late 19th-century global mean. However, we must be precise about what this figure actually represents. This is not a measurement of physical reality; it is a calculated margin of error derived from the homogenization model itself.
In other words, this uncertainty value is a report card on the algorithm’s internal consistency, not an audit of the Earth’s temperature. It measures how well the model believes it has “corrected” the data, not the degree of truth in the original observations. This budget assumes that the errors are random and can be smoothed away, but it cannot account for systemic biases—such as the widespread conversion of land use—that the model itself may be blind to. Statistical methods can improve the model’s precision, but they cannot confer physical accuracy onto a century of heterogeneous, unverifiable data.
In 1979, everything changed. We launched the first global satellite sensing systems. For the first time, we had a continuous, globally consistent, and physically traceable way to measure the temperature of the atmosphere. We moved from an era of anecdotal, human-mediated estimates to an era of systematic, physical observation.⁴
The Question of Confidence
This brings us to the central tension in climate policy. Should we treat the “human-journal” record of the 19th century as if it were the same quality as the “satellite-sensor” record of the 21st?
This is where the public narrative falters. We are routinely presented with a single, seamless temperature graph that ignores the fundamental shift in measurement architecture that occurred in 1979. Treating the pre-1979 “adjusted” surface record and the post-1979 “satellite-sensor” record as a single, consistent data stream is a category error. They are two fundamentally different forms of evidence: one is reconstructed from heterogeneous historical observations whose original conditions often cannot be independently verified, while the other is produced by a globally consistent sensing system whose measurements, telemetry, calibration records, and processing steps can be audited and reproduced. When we conflate the two, we conceal the difference between a record that must be reconstructed and a record that can be systematically verified.
If this were any other high-consequence system—like the design of a bridge, the safety of a drug, or the management of a pension fund—we would demand a clear separation between “anecdotal estimates” and “verified facts.” We would insist on knowing exactly how much of our warming trend is observation and how much is statistical correction.
My argument is not that the historical record is “fake.” My argument is that it is imprecise. By pretending that a messy, century-old volunteer log has the same authority as modern satellite data, we are masking the true level of uncertainty in our climate models. We are making multi-trillion-dollar decisions based on a ledger that is far “fuzzier” than the public is led to believe.
The thermometer is an observation. The black box is an interpretation.
If this were any other high-consequence system—like the design of a bridge or the management of a pension fund—we would demand a clear separation between ‘anecdotal estimates’ and ‘verified facts’. We would insist on knowing exactly how much of our warming trend is observation and how much is statistical correction. We don’t need to throw the history away, but we must stop treating an interpreted ledger as if it were a direct physical observation. Until we differentiate between what we can verify and what we have ‘fixed,’ we are not managing a climate—we are ignoring the limitations of our own instruments.
Notes
¹ For a discussion on the limitations of 19th-century observational networks, see James R. Fleming, Meteorology in America, 1800–1870 (Baltimore: Johns Hopkins University Press, 1990).
² Matthew J. Menne et al., “On the Reliability of the U.S. Surface Temperature Record,” Journal of Geophysical Research: Atmospheres 115, no. D11 (2010).
³ Ross R. McKitrick, “On the Adjustment of Surface Temperature Records,” Energy & Environment 21, no. 8 (2010): 945–966.
⁴ John R. Christy and Roy T. McNider, “Satellite Bulk Tropospheric Temperatures as a Metric for Climate Sensitivity,” Asia-Pacific Journal of Atmospheric Sciences 53, no. 4 (2017): 511–518.
I examined historical surface temperature records both closely and globally in essay “When Data Isn’t” in ebook Blowing Smoke. The conclusion was very simple: not fit for climate purpose. Too many problems that no ‘black box’ could ever fix. Lack of coverage, siting, technology change without calibration,…
In fact, black box ‘fixes’ often make things worse. My favorite example is station 100900 in the BEST ‘black box’. Using regional QC comparisons, BEST removed 29 monthly extreme cold readings as ‘anomalous’, thereby changing no trend into warming there. Station 100900 is Amundsen Scott exactly at the South Pole, arguably the most expensive and best maintained weather station on Earth. The BEST QC comparison station was McMurdo, 1000s of km away on the Antarctic coast and thousands of meters lower in elevation!
My experience with San Diego County climate is that a few 10’s of km in distance from the coast and a few hundreds of meters in elevation can make a dramatic difference in temperature and temperature trends.
The nail has surely been hit on the head. How can anyone tell you a global temperature when the data points are not only distantly placed but are not evenly placed upon the globe.
We have had publications on this and other sites about the precision of the modern digital thermometer compared to the time smoothing properties of the mercury units. However, this minor blip pales into insignificance when compared to the declared global ocean temperature records, back calculated from four or five ships doing a couple of measurements whilst crossing defined routes on just a couple of the oceans.
It’s worse than we think, I just can’t see it ending whilst the existing education system remains in place, the brainwashing starts at primary school and they will eventually vote to keep the same system in place.
Some 70% of the globe.. ie the oceans, has very few scientific observations before ARGO in 2005.
And much of the land surface also is almost totally lacking in measurements outside of a few of regions. (USA, Europe, Australia)
What land measurements that do exist are now totally corrupted by urban warming, airport engines, data manipulation etc etc etc
Any “global” temperature record can only be pure fabrication.
Actually, there is one country that has made reasonable sea temperature measurements, and that is Canada.. in the Archipelago region.
Guess what it shows. 😉
Good points being made here.
But on this one, there is a misdirection.
“By pretending that a messy, century-old volunteer log has the same authority as modern satellite data, we are masking the true level of uncertainty in our climate models.”
None of the pre-stabilized, time-step-iterated, parameter-tuned-to-hindcast models ever had any diagnostic or prognostic authority at all concerning the influence of rising concentrations of CO2, CH4, N2O on the climate system. The rapid buildup of uncertainty through the iterated computation is far greater than the factor being investigated. It was implausible from the start that this program of modeling was anything but a circular exercise confirming a “warming” result from the assumption at the outset that those rising concentrations operate as a “forcing.”
More here in a formal comment last year to DOE on the “Critical Review” report.
https://www.regulations.gov/comment/DOE-HQ-2025-0207-0371
Don’t get me wrong here. There is value in satellite observation, and in analysis of the historical record of measurements. But it won’t help the models do anything to properly inform policy options.
Thank you for listening.
Many studies have shown that the uncertainties in climate models are 10-100x greater than the tiny effect they are trying to quantify. Pretending that those models represent an unequivocal determination of the human forcing is blatant fraud. “We don’t know” is the only scientifically valid conclusion about CO2’s impact on climate based on the uncertainty bands.
Observed Warming Rates Since 1979
• UAH Satellite (Lower Troposphere): ~0.16°C/decade → ~0.8°C per 50 years
• NASA GISTEMP (Surface): ~0.18-0.20°C/decade → ~0.9-1.0°C per 50 years
• NOAA GlobalTemp (Surface): ~0.18-0.20°C/decade → ~0.9-1.0°C per 50 years
Key takeaways:
Bottom line: NASA and NOAA’s flagship temperature records show warming rates only modestly higher than UAH. The principal disagreement is not about what happened over the last 50 years, but whether future warming should be expected to follow the historical trend or accelerate due to increasing greenhouse gas concentrations.
Key supporting Findings
Hermann Harde, German physicist’s work beginning in 2017, (rejected, of course, by the IPCC) argues that natural carbon sinks scale with atmospheric CO₂ concentration and therefore absorb an increasing share of emissions as concentrations rise.
Sulpis and coauthors, (2018) have been cited as evidence that the ocean’s carbonate buffering system remains far from saturation.
If both interpretations are correct, they suggest that Earth’s carbon sinks may have greater long-term capacity than is often implied. However, neither claim by itself overturns the mainstream conclusion that atmospheric CO₂ is increasing because emissions exceed current sink uptake.
Even UAH has a measurement uncertainty far greater than the ability to differentiate in the hundredths digit for a global average. Path loss for each measurement is *estimated* – meaning each and every measurement UAH makes has a built-in measurement uncertainty due to just that one factor. If we had a true, comprehensive measurement uncertainty for UAH it’s base uncertainty would become evident.
My opinion is that UAH has a true measurement uncertainty much like the land-based system has, at least +/- 1C and probably larger. That would mean it would be impossible to accurately determine a change of 0.8C per century let alone 0.16C per decade.
This doesn’t even begin to address the issue of what the average of an intensive property means physically. If you hold a two rocks in your hand, one rock at 60F and a second at 70F, are you holding a larger system consisting of two rocks whose temperature is 130F? (i.e. is your hand burning?) If not, then what does the average value mean physically? Yes, you can perform the mathematical operation but what does it mean physically?
Also beware the statistical fallacy of saying the measurement uncertainty of an average is the variance of the average values from a set of samples. The measurement uncertainty of an average is the variance of the component data. Combining a set of averages into a data set and finding the average of the averages is the same thing as finding the average value of a set of sample means – it is sampling uncertainty, it is *NOT* measurement uncertainty. The measurement uncertainty is the SUM of the variances of the component data sets. And a data set using “global” temperature measurements will have a large variance due to combining summer temps with winter temps – a truism climate science stubbornly refuses to admit.
My comment was about climate model uncertainties not observations. I’m not following how your comment relates to that?
Thanks for keeping the qualifier of iterative growth of uncertainty in play, David.
I have a manuscript under review now that takes the idea further. If it passes review, I’d hope you’d find it interesting.
Thanks for your reply, Pat. Keep up the good work. Looking forward to seeing what you’ve been working on.
Same here! Can’t wait!
So, the basic summation is that the Probity and Provenance of temps “data” make them appallingly unfit for scientific purposes.
I’ve been saying this for ages.
The worst of this false equivalence is the fraudulent claim that modern warming is unprecedented. They took the average rate of warming coming out of the last ice age over thousands of years from course proxies and compared that rate to the modern short term high resolution instrument record. Media propagandists still push that lie as if it were unquestionable fact.
Satellite temperatures of the lower troposphere derived from microwave sounders are subject to many more adjustments than any of the global surface temperature records. Instrument calibrations, Earth Incidence Angles and orbital decay, diurnal drift, layer extraction (different groups use different techniques) and, yes, homogenisation. Many of these adjustments rely heavily on models.
In any case, since 1979, the warming trends in the various satellite data sets and the surface data sets are statistically consistent with one another.
This chart shows two satellite lower troposphere sets (RSS and UAH) and three surface temperature data sets (HadCRUT, GISS and NOAA), 1979-2025, as monthly anomalies all set to the same 1991-2020 base.
All show statistically significant warming. The error margins in all their warming rates overlap.
The current rate of warming isn’t any greater than the two previous warming periods after the end of the Little Ice Age, the period from 1850 to the 1880’s, and the period from the 1900’s to the 1930’s, had warming rates equal to the current warming period.
After the Little Ice Age ended the temperatures warmed through the 1880’s at the same rate as today and the high temperature in the 1880’s, was just as warm as the current warming. There has been no additional warming since the 1880’s. The 1880’s, the 1930’s and the high point of current temperatures are all located on the same horizontal line on the temperature chart.
There is no unprecedented warming today. It was just as warm in the recent, recorded past. Which means CO2 has had no discernible effect on the temperatures.
Even the bogus Hockey Stick global chart shows the high points of the 1880’s and the 1930’s to be within a few tenths of a degree of each other. The bogus part of the Hockey Stick chart is it does not show the 1880’s and the 1930’s to be as hot as today. Hansen said 1934, was 0.5C hotter than 1998. That makes 1934, just as hot as 2024.
All three temperature high points since the end of the Little Ice Age should be placed on the same horizontal line on the chart, if you want to know the truth.
What evidence can you present to support this claim, please?
The period 1979-2025 is 47 years (inc.), and the average warming rate in all the global surface data sets (HadCRUT5, GISS, NOAA, JMA, BEST) over that period is +0.20C/dec.
Only the HadCRUT5, NOAA and BEST global temperature data sets cover the period 1850-1880s, which is at most 40 years (since you were unspecific about your end date). Using the annual average of the three data sets mentioned, the fastest warming rate for any period of at least 20 years between 1850 and 1889 is +0.04C/dec, 1850-78 (29 years). The fastest warming rate between 1850 and any year ending in the 1880s is +0.03C/dec (1850-81, 32-years).
Using the annual average of the same surface data sets for the “from the 1900’s to the 1930’s” (at least 40 years – this time you provide no specific end or start date), the fastest rate of warming for any period of at least 20 years is +0.14C/dec (1907-26, 20 years). The fastest warming rate from any year in the 1900s and ending any year ending in the 1930s is +0.13C/dec (1908-32, 25 years).
All of these are much slower and of a much shorter duration than the rate of global warming observed between 1979 and 2025. Therefore, you can see why I might enquire as to the sources that inform your claim. Links to the data I used are below:
BEST
NOAA
HadCRUT5
No answer did there ever come from Tom, about his imagined previous warming periods.
“The error margins in all their warming rates overlap.”
It is to laugh.
UAH is real, not a fabricated mess…
…. and it shows that there no human caused atmospheric warming.
It shows the same warming as all the other data sets, statistically.
“All show statistically significant warming. The error margins in all their warming rates overlap.”
How do you know? if the measurement uncertainty is +/- 1C you don’t know if the data being graphed is accurate or not!
Because you can calculate the error in a trend. As to the actual figures, you need to take that up with Roy Spencer at UAH. His figures.
Error in a trend is based on the assumption that the residuals are 100% accurate because the stated values used to develop the residuals are 100% accurate.
You can run but you can’t hide. If the measurement uncertainty is large enough that you can’t even tell if the slope of the trend at a point is positive or negative then you don’t know what the trend line actually is – and no amount of sampling can fix that.
With the satellite measurements, we know what the errors are and how to adjust for them.
With the historical records, we have no way of knowing exactly what the errors are or how big they may be.
“With the satellite measurements, we know what the errors are and how to adjust for them.”
Satellite temperature estimates are reliable.
Surface temperature estimates are consistent with the satellite temperature estimates.
Therefore, surface temperature estimates are reliable.
“Satellite temperature estimates are reliable. ”
To within ±0.3 C. At best.
Adjusted Surface temperature estimates are consistent with the satellite temperature estimates.
There’s a surprise.
And no physically valid uncertainty bounds on the historical record. The signature of good science, indeed, that lack.
If I calibrate my micrometer against an inaccurate gage block does that make my micrometer readings reliable?
The one qualifier you left out is:
“Inaccurate surface temperature estimates are consistent with satellite temperature estimates”.
Your analogy doesn’t apply because the two records are independent.
“Contrary to some reports, the satellite measurements are not calibrated in any way with the global surface-based thermometer records of temperature. They instead use their own on-board precision redundant platinum resistance thermometers (PRTs) calibrated to a laboratory reference standard before launch, embedded in high-emissivity targets that are viewed by the radiometer once each scan. “
https://www.drroyspencer.com/latest-global-temperatures/
Your criticism doesn’t deal with the issue. Independence is no guarantee that measurement uncertainty is reduce or eliminated. A micrometer can read to 10⁻⁶ meters. When you zero it, you assume the gauge block is accurate to that extent. If the gauge block is faulty, that adds a systematic error into each and every reading.
First, PRT’s platinum resistor may have a small value of change but is only part of the system. There are connections and other electronic devices that actually measure the change in temperature of the PRT element. These can all drift over time invalidating the calibration procedure performed before launch.
Second, more than one satellite is used, each having its own measuring devices whose uncertainty must be propagated. This similar to interlaboratory uncertainty procedures recommended by NIST and ISO.
In the end, this excuse is simply hand-waving. It indicates that no actual work has been done to investigate the uncertainty. It is just assumed to be small.
PRTs (used in satellites) exploit the fact that the electrical resistance of platinum changes predictably with temperature. MMTS surface thermometers exploit the temperature dependence of semiconductor resistance. These are fundamentally different measurement technologies, and platinum resistance thermometers are generally much more stable over long periods.
Yet despite relying on completely different physical principles, independent instrumentation, and independent processing methods, the global satellite and surface temperature anomaly records show remarkably similar long-term warming rates and month-to-month variability, as shown in TheFinalNail’s graph.
For your argument to be correct, you would have to assume that the systematic errors affecting satellite PRT measurements and those affecting MMTS surface thermometers just happen to produce very similar global anomaly time series over decades.
The mathematical odds, lol.
Such obvious desperation.
“platinum resistance thermometers”
The PRT’s are *NOT* thermometers. They are sensors that are a PART of the thermometer.
Even with a perfectly calibrated PRT the measuring station has all kinds of issues that contribute to a measurement uncertainty interval. Anything that disturbs the airflow into and out of the station compartment will affect the temperature measurement in the hundredths digit. Think insect detritus, ice, snow, even heavy precipitation. This is on top of station enclosure degradation due to UV exposure affecting the reflectivity of the paint on the enclosure.
The measurement uncertainty of the satellite measurements covers a LOT of components. Drift of the hot calibration source, spreading of uncertainty through inter-satellite cross-calibration and homogenization, no actual measurement of path loss for each measurement, and measurement uncertainty in the radiance-temperature models due to things like scan geometry, orbital decay, diurnal drift, etc.
Validation of the satellite data against land/ocean temperature data does nothing but propagate the measurement uncertainty in the land/ocean temperature data onto the satellite data.
If the reference dataset has uncertainty, then using it for comparison purposes just causes the data set being validated to inherit the uncertainty of the reference.
Since the measurement uncertainty in the land/ocean data sets are at least in the units digit and most likely in the tens digit, using the land/ocean data sets to validate the satellite data just causes the satellite data to inherit the same uncertainty magnitude, at least the units digit and likely the tens digit.
Validation is primarily a statistical alignment check, it is *NOT* an accuracy judgement.
“Your analogy doesn’t apply because the two records are independent.”
We’ve been down this road before. You had your nose rubbed in it then. How quickly you forget.
Independence is *NOT* sufficient criteria.
I can buy a tape measure from the hardware store. I can buy a tape measure of a different brand at the sewing store.
If their measurements agree, you imply that means the measurements they make must be accurate.
The truth is that BOTH can be wildly inaccurate and the fact they both give the same reading means nothing other than a highly inaccurate reading.
Independence solves nothing when it comes to accuracy.
“Are we really to believe that (at least) nine independent global temperature data producers all came up with more or less the same conclusion by coincidence?”
https://wattsupwiththat.com/2026/07/30/the-era-of-decentralized-observation/#comment-4224145
No.
Again: “Independence is *NOT* sufficient criteria.”
We have at least some indicators allowing us to make some guesses at the measurement uncertainty of the data sets using land/ocean based measuring stations. THEY ARE NOT VERY ACCURATE. Certainly not to the hundredth of a degree. The measurement uncertainty of these data sets is certainly in the units digit and, more likely, in the tens digit.
That means that any satellite based data sets that agree with the land/ocean data sets also have the same measurement uncertainty magnitude.
Until climate science and its supporters (like YOU) abandon trying to conflate sampling uncertainty with measurement uncertainty while assuming that all measurement uncertainty is random, Gaussian, and cancels AND seek to actually apply international metrology standards to their measurement protocols NO ONE can trust conclusions about climate put forth by climate science.
This is on top of the idiocy of assuming that temperature is a metric for climate.
You don’t actually know if the satellite measurements are accurate or not. One major component in the accuracy is the knowledge of the atmospheric path loss associated with each measurement. How do the satellite sensors measure path loss. If it is “parameterized” (like CO2 is well mixed) or guessed at then it can be big time inaccurate. And it *will* change between measurement intervals.
Not only that, but microwave sounders have an intrinsic resolution of ±0.3 C, which is never included in the reported satellite temperatures.
I’m sorry Pat, but an increase of 0.01° (C or F) is a NEW WORLD RECORD! (If you ignore the error bars.)
None of these actually provide a measurement uncertainty budget.
And model outputs are 100% accurate? You forgot to list path loss through the atmosphere is guessed at for each measurement – another component of measurement uncertainty.
“In any case, since 1979, the warming trends in the various satellite data sets and the surface data sets are statistically consistent with one another.”
“statistically consistent”? These are averages of averages – in other words sampling uncertainty. Having the same sampling uncertainty is meaningless as to the accuracy of the underlying data.
“We would insist on knowing exactly how much of our warming trend is observation and how much is statistical correction.”
If you insist, there is a simple way to know, which sceptics never seem to manage. Calculate the global average with the original raw uncorrected data. I do that every month (TempLS). And it makes very little difference. Here is a map of the global average since 1900, calculated with unadjusted GHCN data (TempLS) and by GISS with homogenisation (12 month running average):
You could be wasting your time every month.
Discussions on global warming often refer to ‘global temperature.’ Yet the concept is thermodynamically as well as mathematically an impossibility, says Bjarne Andresen, a professor at The Niels Bohr Institute, University of Copenhagen, who has analyzed this topic in collaboration with professors Christopher Essex from University of Western Ontario and Ross McKitrick from University of Guelph, Canada.
https://www.sciencedaily.com/releases/2007/03/070315101129.htm
Nick, the Probity and Provenance of temps “data” make them appallingly unfit for scientific purposes.
As a real-world comparison, any honest registered corporate auditor of a publicly-listed corporation would go ape-shit if the corporation’s financial transactions & situation presented were based on so many adjusted, made-up numbers.
The auditor’s qualifications statements of the annual accounts would be longer than the lodged annual report filing itself.
Heads would roll, resignations would be demanded, and fines & jail sentences would be in prospect for the perps.
But climate “science” carries on with their crappy numbers (to hundreds of a decimal point even), happy as pigs in shit in their fantasies.
This is a good example of why discussions at WUWT can never get anywhere. John Pabst asked a reasonable question – what is the effect of corrections etc to the old data. It’s often asked, but no-one seems interested in the fairly obvious answer. Meanwhile there is the usual chorus of, the data is rubbish, you can’t have a global average anyway, etc, which all makes nonsense of the Pabst question and any possible answer.
Says it all.
“It’s often asked, but no-one seems interested in the fairly obvious answer”
There is no obvious answer. Not unless you have a time machine.
Hubbard and Lin showed over a two decades ago that regional temperature adjustments to measuring station readings don’t work. Adjustments to readings must be done on a station-by-station basis and should be based on calibration applied to *future* readings, not to past readings.
Good luck on trying to “calibrate” readings in the past.
As the article states: “If a thermometer in 1890 was sitting on a sun-drenched porch in a growing town, no algorithm in 2026 can know with certainty how much that reading differed from the true ambient air temperature.” (bolding mine, tpg)
The issue here is that the measurement uncertainty of a “guessed at” adjustment must be ADDED to the estimated measurement uncertainty of the original reading!
You and climate science have *never* understood measurement uncertainty. You simply cannot decrease measurement uncertainty by “guessing” at an adjustment. The “guess” just adds in MORE uncertainty for the final value. And I don’t care if the adjustment is pulled out your backside or from an “adjustment algorithm”.
You apparently didn’t read what you wrote. The question is legitimate, but your answer is not. The question dealt with:
Your graph assumes that the temperatures have no uncertainty, either random or systematic. That is one underlying failure that the essay attempts attempts point out.
Let me add that 12 month running averages of monthly averages of daily averages hides so much it isn’t funny. Converting to anomalies hides even more.
According to that bogus chart of yours, alarmists in the past were wringing their hands over a new Ice Age coming in the 1970’s, which showed a mere 0.4C drop in temperatures from the 1930’s to the 1970’s.
now does any honest person think scientists are going to equate a 0.4C cooling as a harbinger of a coming ice age? Only a fool would do so.
Since the high temperature point in 2024, the temperatures have cooled by about 0.5C. Do you hear anyone saying there is an ice age just around the corner? Of course not. But you want us to believe scientists were fretting over a 0.4C cooling in the 1970’s.
Your chart is bogus and doesn’t represent reality. The truth is the temperature cooled by more than 2.0C from the 1930’s to the 1970’s. That’s why alarmist scientists were wringing their hands.
”the original, raw, uncorrected data”. Is just a Big Lie, and a fellow like you ought to know it’s a Big Lie. Are you lying to yourself? Is that how it works?
You’re not good with specific start or end dates, Tom; but taking Jan 1930 to Dec 1979, there is actually a slight warming trend in the GISS temperature data; +0.02C per decade, to be precise, or +0.12C in total.
Can you specify what start and end dates you are using, because I can find no evidence of “0.4C cooling” for any substantial period between 1930 and 1979.
(I’m sure you’re not just eyeballing a high point in the 30s and drawing a straight line between that and a low point in the 1970s… or are you??)
“taking Jan 1930 to Dec 1979, there is actually a slight warming trend in the GISS temperature data”
And GISS data is not “corrected” data?
As Nick showed above, the difference between the corrected and raw data is minimal.
But I’m asking where Tom found his “0.4C cooing” between the 1930s and the 1970s from. It doesn’t have to be GISS. Use the raw data if you prefer, but I don’t think you’re going to find it.
The point of the essay is to point out that you can’t make that conclusion.
ASOS stations have an accuracy uncertainty of ±1.8°F (±1.0°C). Assuming a uniform distribution the standard uncertainty is “1.8/√3=±1.0°F” (±0.6°C). That is just one component of an uncertainty budget. It is not unreasonable when other components are added that the uncertainty would be ±1.0°C.
Weather stations prior to the installation of ASOS would have had at least the same uncertainty and LIG studies indicated that far larger uncertainty is probable.
±1.0°C places all of your “temperatures’ well within that uncertainty range. That means you cannot ignore the fact that the actual temperatures could differ far from what you are examining. In other words, you don’t know and cannot know that your temperature rise is true or not. You can’t even judge statistical significance of the sign being plus or minus.
And the point remains unanswered by mrs. nail and stokes — if the adjustments and fake data don’t matter, why bother with them?
Hadn’t picked up on that! 100% right. Why bother?
From the book ‘The Complete Ice Age’ edited by Brian Fagan (2009)
“As Lowell Ponte (1976) summarized:
Since the 1940s the northern half of our planet has been cooling rapidly. Already the effect in the United States is the same as if every city has been picked up by giant hands and set down more than 100 miles closer to the North Pole. IF the cooling continues, warned the National Academy of Sciences in 1975, we could possibly witness the beginning of the next great ice age. Conceivably some of us might live to see huge snowfields remaining year round in northern regions of the United States and Europe.”
That’s specific to the US, by the looks of it. The data Tom was referring to was global.
By the way, the US data don’t show “0.4 cooling” between 1940 and 1975 either. There was cooling in the US of -0.08C per decade (-0.3C total, 36 years).
But that was just for the US. Globally there was less than -0.1 C cooling over the same period.
“The thermometer is an observation. The black box is an interpretation.”
The instruments aboard the satellites are not thermometers. Their readings are run through a second “black box” to insure that the “observed” microwave intensity corresponds to the thermometer reading for the same location and time. Think how embarrassing it would be if the satellite reading of “temperature” differed from the local thermometer reading by 2 or 3C consistently! Are we supposed to believe that the perfect symmetry of ground and satellite “readings” since 1980 are the result of two completely independent data sets?
There are far more that two global temperature data sets providers. There are at least three derived from microwave sounders on satellites for the global lower troposphere and at least six, probably more, covering global surface temperatures derived from thermometer readings.
There are variations over the short term but long term they all show statistically significant warming since 1979 of around +0.2-0.3C per decade.
Are we really to believe that (at least) nine independent global temperature data producers all came up with more or less the same conclusion by coincidence?
“There are at least three derived from microwave sounders on satellites “
And not a single one measures the path loss associated with any measurement.
“at least six, probably more, covering global surface temperatures derived from thermometer readings.”
And what is the MEASUREMENT uncertainty of each of these? +/- 3C? +/- 1C? Is this measurement uncertainty (NOT SAMPLING UNCERTAINTY) added into the measurement uncertainty of the satellite data?
You’re saying that they are all wrong even though they all come up with the same answer?
How would you do it?
You have never ONCE attempted to actually understand the use of measurement uncertainty.
There is simply no way to judge the accuracy of ANY measurement unless some genuine attempt is made to create a comprehensive uncertainty budget that can allow others to understand how accurate the measurements are. That means there is no way to judge any of these data sets for accuracy. We simply don’t know if they are all wrong or not! No competent physical scientist would just ASSume that they are all correct because they match. Independence is *NOT* a sufficient criterion for assuming 100% accuracy.
NONE of these global data producers you are putting forth give ANY kind of actual measurement uncertainty budget with corresponding intervals, instead trying to pass off sampling uncertainty as measurement uncertainty.
How would I do it? Path loss measurement is pretty straightforward. Each and every measurement should be done over a land/ocean based transmitter aimed skyward putting out a calibrated signal that the microwave sounder can use for measuring path loss. Since the sounders measure a wide area with each measurement, several of these transmitters might be needed for each measurement in order to develop a path loss profile that can be applied against the radiance measurement. The transmitters can operate at a slightly different frequency in order to differentiate them.
Think an upside-down GPS system. Even this, however, will have its own measurement uncertainty budget that will be necessary to judge the accuracy level of the measurements.
A system like this would also eliminate any use of inaccurate land based temperature measurements as a calibration metric. This alone would make the satellite data less uncertain.
Mean measurement uncertainty, 2σ = ±0.4 C.
Innocent oversight, that.
I see no error bars consisting of measurement uncertainty that has been propagated properly. ±1.0°C uncertainty would indicate that no one could conclude this warming trend even exists. The trend could be entirely spurious and based upon biases that are not included in the analysis.
Your very graph indicates that you assumed the “raw” data was entirely accurate just as the author pretty much said in his essay and as Ruf Istvan has also concluded. You need to answer how you know that 1900 anomaly wasn’t +0.5°C instead of –0.5°C. Or how you know it wasn’t -1.5°C instead of -0.5°C.
If you answer that you don’t know, then you have just confirmed the authors conclusion that the data is unfit for purpose.
“I see no error bars consisting of measurement uncertainty”
As usual, a mechanical response, off topic, and a day late. The problem John Pabst posed was the effect of adjustments. He posed it as a black box, which it isn’t, but OK, you can do the black box arithmetic. How much difference does it make if you do the same arithmetic with raw data instead of corrected data? I show that difference (small). Local nonsense about measurement uncertainty does not affect that arithmetic result.
As the article states: “If a thermometer in 1890 was sitting on a sun-drenched porch in a growing town, no algorithm in 2026 can know with certainty how much that reading differed from the true ambient air temperature.” (bolding mine, tpg)
Just how do you figure it is simple arithmetic to calculate the adjustment factor for this measuring device?
Do you have some “mathematical crystal ball” that lets you see into the past better than the rest of us?
All that happens is that the adjustment algorithm ADDS more uncertainty to the result, there is no way it can decrease it since it is nothing more than a guess.
Talk about off topic. You are considering the both the raw data and the “corrected” data as 100% accurate. The issue is the uncertainty of both the raw and the “corrected” values. You can only increase the uncertainty in the raw data by correcting it because you have no time machine to evaluate its uncertainty at the time the temperature was recorded.
You are just like the other mathematicians here. You can manipulate numbers, change them, perform adjustments to them and the end result is 100% correct arithmetic. But what you have lost is the meaning, the information, the value of the measurement that occurred in the past.
The author attempted to educate you on measurements and their uncertainty. Yet you correct him and say it doesn’t matter. To you the stated values are 100% so you can compare them. The author is trying to tell you that the uncertainties are such that you can’t compare them. I’m trying to tell you that comparing 1 ±0.5 versus 0.5 ±1.5 tells you nothing.
Start with calculate the global average.
Nope.
Your graph is not even temperature data.
First you start with an average. Then you subtract that average from other averages (+/-) then smoothe them over creating a 12 month running average.
You do not have the original raw uncorrected data. You have the tabulated numbers.
A satellite does not measure temperatures. It measures electromagnetic energy, converts that to an electrical current, passes through linearization circuits, and into a sampling analog to digital converter.
Acquisition angle limits what is seen. Bandwidth limits what it detects.
Certainly one can calibrate before launch and there can be calibration in mission, but the bottom line is, the final conversion from sensor to reported data is based on a model.
CERES acquires 99.95% of the total spectrum. It cannot acquire the entire 25km grid in the same instance. There are linearization tolerances and tolerances due to the slopes of band pass filters. Time variations increase inaccuracy. Then there is the published tolerances of 0.5% to 1.0% error in the readings. Per NASA.
There were earlier satellites that had to have data adjusted because of orbital decay. Recall EM field power density varies as 1/R^2. Repeating the word: Adjusted.
What seems to be missing from the documentation is how CERES adjusts for terrain. Mountains, land slopes, etc.
CERES is as good as it is, but treating it as if it is perfect data is a fools errand.
By the way, the word anomalies is abuse of a perfectly well defined engineering term, and does not mean what the climate graphs portray.
US Army Posts in Arizona were connected by telegraph lines in the 1870’s. The telegrapher at each post was to take a temperature reading at midday daily and transmit the figure to HQ. One such telegrapher admitted: I’d stand in the doorway of my shack and look at the post where the thermometer was hung, on the other side of the parade ground. “It feels like 113° – that’s what I’ll send.”
Very nice and very helpful.
Measuring the temperature of thermometers is just silly, and achieves nothing at all, except a sense of satisfaction for the measurers.
People generally like doing pointless things – just look at the number of people with a rain gauge! It tells you how much water went into your apparatus, and nothing more. Two gauges 50 m apart can collect wildly different amounts, and to what purpose?
Predict the next flood or drought?
And still people believe that recording thermometer temperatures will enable them to peer into the future.
Surface temperatures vary between roughly +90 C and -90 C. Measuring them won’t make the slightest bit of difference.
The TRUE metric for measurement uncertainty is the variance of the underlying data. Yet climate science never, at least as far as I can find, EVER calculates the variance of the underlying data components let alone add them up when combining samples into a larger data set.
Since variance is the metric for measurement uncertainty, why does climate science weight temperatures from locations with high variance equally withe temperatures from locations with lower variance? You would think it would be better to weight the more accurate, low variance data higher than the less accurate, high variance data. But not climate science.
Because the practitioners of “climate science” are either frauds or fools?
I attribute to most of them being non-physical scientists but, instead mathematicians, and statisticians whose basic meme is “numbers is just numbers”. And piss-poor statisticians at that.
Temperature, especially diurnal mid-range temperature, tells you almost nothing about actual climate differences. Precipitation is a far better metric for climate differentiation. So, of course, climate science emphasizes the worst possible metric for climate.
You left out computer programmers.
For a discussion on the limitations of 19th- and 20th-century observations, including SST, see – LiG Metrology, Correlated Error, and the Integrity of the Global Surface Air-Temperature Record
Nice article, the facts seem clear, uncertainty is systematically suppressed in particular for potential systematic errors.
The same is true for basically any other field regarding antropogenic global warming (I refuse to call this thing by any other name, no matter what the press and politics says, this is the question we need to solve and address first!)
For proxy reconstruction McShane and Wyner have made the published peer reviewed argument that the proxy selection causes unknown and undisclosed uncertainty problems (very clearly demonstrated by strip bark pines, their very name gives a hint that other factors besides temperature might have affected it’s growth, so how representative is any data set extracted from it? Or the Cape Ghir series discussed at Climate Audit years ago, a clear straight and strong warning trend was extracted contradicting anything we know about middle age and roman warm periods in that area)
Models have all kinds of undisclosed problems, from clear effects from low resolution in older models and incorrect physics to missing effects like the shape of the earth being oblate rather than spherical (which changes the average angle of solar radiation and might have up to 1% effect, which is about 7W/m2) or considering solid water in the atmosphere..
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2025GL120130
“However, because SON2 models are underrepresented in CMIP6, their benefits are muted in the multi-model averages.”
Then there are of course fields like the “attribution science” which never really discussed the fact that their old prediction relied on models we now know to be unrealistic and wrong (what I wrote above plus the extensively used RCP8.5 – to evaluate if their science is trustworthy, the older publications must be brought up to current knowledge!)
Linear least-squares regression trend lines are models, they are not reality.
And, they are models of time vs temperature. As if time causes the temperature to change. What a joke! The data are a time series. Time series analysis should be applied by finding a statistically stationary trend, then adding models of variables to achieve the resulting trend. That allows one to statistically claim knowledge of predictor variables.
Most importantly you (and Pat Franks, whom I was happy to see in this discussion! I am a big fan of his elephant in t he room!) are spot on and the way global temperatures are reported is mathematically incorrect!
I asked Gemini two questions
And got “””When researchers only report the raw regression output from a large dataset, they fall into the large-\(N\) statistical trap. Because standard error scales with \(1/\sqrt{N}\), a massive dataset drives statistical uncertainty near zero. This gives a false impression of extreme precision, while hiding the true accuracy limits imposed by systematic errors.”””
.. also offering methods and python code to calculate the correct uncertainty!
While seemingly trivial, this has very wide implications for global warming, basically any value used is wrong!
You are correct, the difference is not trivial. The “uncertainty of the mean” means exactly what it says. how close the mean has been calculated. It is based upon sampling error which results in the mean of each “sample’ being slightly different. In other words, the samples are not truly “Independent and Identical Distributions” or IID. The larger the size of each sample is from the population, the more likely you are to have IID samples and the “uncertainty of the mean” gets smaller and smaller.
The uncertainty of the mean has no relation to the dispersion of measured values that can be attributed to the mean which is where measurement uncertainty is derived. International agreement has declared that with a Gaussian distribution of values the variance and standard deviation are the base statistical descriptors to use in evaluating the measurement uncertainty.
Climate science has also ignored the fact that almost all temperature measurements are “single measurements”. The uncertainty in each reading must be determined from an uncertainty budget developed for at least a given measuring device, its housing (if it has one), and the microclimate surrounding the measurement system.
JCGM 100:2008 Paragraphs F.1.1.2 and H.6 provides recommendations on how to calculate uncertainty when different measurands and/or non-homogeneous materials are measured. These result in using both individual measurement uncertainty and the variance between measurements to find a combined uncertainty.
Here is a question I have about “global temperature” of land + ocean. It concerns how El Nino is treated.
It has been noted that UAH has a step temperature increase associated with El Nino. That seems odd to me.
El Nino is caused by trade winds “stacking” up warmer water into the western Pacific. When El Nino occurs the trade winds die or even reverse, letting this warmer water “slosh” to the eastern Pacific.
Question:
Does this spreading of heated water really warm the globe or is the warmth just spread over a larger area with less depth?
Question:
Is temperature a good proxy for stored heat? If the warmer water in the western Pacific is spread over a larger area toward the eastern Pacific, then the heat is also spread over a larger area but having less depth so that the total heat remains the same.
Question:
Heat loss gradient is increased when the area of loss is increased. This should result in lower temperatures, yet UAH shows no sustained cooling at the end of an El Nino. UAH simply shows little change after the step increase. Why?
We will let you know when the complexities of the problem are ravelled and each factor qualified and quantified.
I do not believe there is a simple answer. 🙂
I was trying to get some to think about what El Nino does to the “global average” temperature. Rising temperature over a larger area due to “spreading” existing heat isn’t really warming due to CO2. I think it’s another example of where enthaly should be used for heat. Spreading enthalpy across the ocean like butter on bread only raises temperature because of area, not increased heat.