UAH v6.1 Global Temperature Update for July, 2025: +0.36 deg. C

Oral traditions passed from generation to generation among the indigenous residents that have been here for centuries. My guess is that you have no idea of the fact that native bands followed the buffalo long before the introduction of the hirse.

“Then it is even more hopeless to reconstruct it from oral tradition, especially if we want great time depth and geographical extent”

Why do you need more depth or geographical extent for CLIMATE? Have you ever even bothered to look at a Koppen map of climate?

I’ve given you two widely separated areas of large extent, the High/Great Plains of the US and the pampas of Argentina, where the climate has been the same for literally thousands of years. Both with wide variation in annual temperatures, rainfall, etc but the *same* climate. Both with the same climate today as 10,000 years ago as judged by the animals alive then and now as well as by the oral traditions of those humans living there.

And all you can do is whine. Just like a six year old.

The only ones that don’t understand anomalies are you and climate science.

1. measurement uncertainty totally subsumes the differences climate science is trying to identify. In other words anomalies in even the tenths digit let alone the hundredths digit are part of the great Unknown.
2. Anomalies do nothing but shift the data along an axis. They do *not* decrease the variance of the data. And it is the variance of the data in relation to the absolute values that accurately size the variance. It’s why for the anomalies climate science has to refuse to even admit that variance exists.
3. Anomalies cannot increase either accuracy or resolution. Basic metrology. Climate science simply can’t get this into their head.

Anomalies supposedly help identify “trends”. Yet it was agricultural science that found the increase in growing season length using absolute temperatures, not climate science with their “anomalies” based on daily mid-range temperatures. Which is more important to *climate*? Anomalies that can’t identify what is happening in real world or actual reality?

“Wrong, and it’s hilarious that you can’t comprehend this.”

Why do you insist on coming on here and lecturing people on things you have *NO* understanding of? The global pre-industrial average temperature is based on measurements with at least a +/- 0.5C measurement uncertainty and more likely a +/- 1C interval. This is not only the dispersion of the values that can reasonably be attributed to the values but also a limit on the resolution the measurements offer. You can *NOT* increase the resolution with averaging and you cannot decrease the measurement uncertainty by averaging either because you can’t separate out random measurement uncertainty from systematic measurement uncertainty in measurements that are approaching 150 years old.

“What the fokk is “accurately sizing the variance” supposed to mean? https://s.w.org/images/core/emoji/16.0.1/svg/1f609.svg This is one of those Gormanisms you cough up from time to time.”

Once again you are demonstrating your absolute ignorance concerning metrology. Variance of measurement data is a metric for the measurement uncertainty of the measurements. For experimental measurements of the same thing using the same instrument under the same environmental conditions the standard deviation (sqrt[Variance])is considered to be the measurement uncertainty. For single measurements of different things using different instruments under differing conditions the measurement uncertainty (i.e. the variance of the combined dataset) is considered to be the sum of the measurement uncertainties of the data components (e.g. the Type B measurement uncertainties).

The relative measurement uncertainty of each component is the measurement uncertainty interval divided by the absolute estimated value of the measurement. If you scale the absolute estimated value of the measurement by creating an anomaly you no longer know what the relative uncertainty is – it becomes hidden. Thus comparing the anomaly is no longer representative of reality. You can longer judge the significance of the Δ value by comparing it to the absolute value of the anomaly.

This is all available in literally *reams* of literature on metrology on the internet. If you don’t study this literature then you are consciously deciding to be willfully ignorant, the worst kind of ignorance.

“Flatly wrong”

Again, you are showing your absolute ignorance of metrology concepts. The variance of temperatures is wider in colder temperatures than in warmer temperatures. When combining data with differing variances you can’t simply average the estimated values without doing some kind of weighted calculation for the combined variance. Climate science NEVER does this. They just throw the estimated values together and calculate an average without ever considering the impact of this on the mean itself. In essence the SEM goes up, it’s a form of sampling error. So not only does the accuracy of the mean go up (i.e. the measurement uncertainty) the determination of the population average gets more uncertain.

“They are not based on that.”

Of course they are! Daily (Tmax + Tmin)/2. This gets propagated into monthly averages and then into annual averages. It’s a pyramid built on a garbage foundation – meaning the pyramid is garbage all the way up!

“FYI anomalies make change comparable.”

They actually don’t. If you can’t identify the Δ’s because of measurement uncertainties then you can’t compare anything! You don’t know if the slope of the change is negative, positive, or zero!

“Oh, so that clumsy bs was about relative uncertainty.”

It’s about the Δ and its relationship to reality. Of course that doesn’t matter to you at all apparently.

“Why do you want to use relative uncertainty? “

Relative uncertainty is *always* important. It’s how you determine significance. You can’t just change the scale willy-nilly without losing the significance. But for someone not concerned with variance I’m sure the level of significance isn’t of any concern either.

“Measurement uncertainty in the range we are interested in is essentially a constant, independent of the magnitude of the measurand. Relative uncertainty just fokks up this.”

Can you point to ANY reference that says this? I can’t find one. Measurement uncertainty has two components – random uncertainty and systematic uncertainty. Systematic uncertainty can definitely be different for different measurands being measured by different devices under different environments. If the systematic uncertainty cannot be quantified then neither can the random uncertainty.

The entire concept of relative uncertainty stems from having to handle functional relationships where the components have different dimensional units and/or differently scaled magnitudes. Again, it’s no different than the same ΔT at 10C and at 20C. Different significance for different magnitudes. If ΔT is 0.1C then for 10C the relation is .1/10 = 1%. For 20C it is .1/20 = 0.5%. Big difference in significance.

You’ve never once in your life had to develop a measurement uncertainty budget, have you?

“You can’t even get this right. Weights are independent factors, the formula is s^2=Sum(wi^2 si^2) if the measurements are pairwise independent (which they are w/r/t measurement uncertainty).”

You have no idea of what you are talking about. If you have two independent variables with different variances then when you combine them into the same data set the variance of the mean becomes

Var(new) = { n1[Var1 + (μ_1 – μ_new)^2] + n2(Var2 + (u_2-u_new)^2] } / (n1 + n2)

where n1 and n2 are the sample sizes of the independent variables.

Again, this is all over the internet. There are reams of literature on this. You and climate science totally ignore all of this just like you ignore the fact that systematic measurement uncertainty exists.

“No, and I have actually asked this”

Malsrky. I have no idea who you asked but they don’t know any more about the subject than you do. Climate science uses what they call a daily mean temperature (which isn’t actually a mean but a mid-range value) to calculate monthly averages. That’s why climate science totally missed the fact that minimum temps are driving any change in mid-range values. Most other disciplines, such as agricultural science and HVAC engineering, have moved to using integrative degree-day values. And climate science has *still* not moved away from using the mid-range daily tempeature.

You have absolutely no understanding of metrology at all.

If you are creating a new dataset by concatenating two independent random variables, which is what you are doing when you combine NH and SH temperatures into one data set, the formula I gave you is correct.

Var(Z) = [ n1Var(1) + n2Var(2) ] / (n1 + n2) +  [n1n2/(n1+n2)^2 ] * (μx – μy)^2

where μ denotes the mean of the variable.

Suppose you are combining the monthly average temperature in Topeka, KS (NH) with the monthly average sea temperature in the middle of the Pacific Ocean (SH) into one dataset used to calculate the global average temperature. This is creating a concatenated data set. The Topeka monthly average has 30 daily entries and the sea temperature from an Argo float in the Pacific has only 20 daily entries. n1 = 30 and n2 = 20.

Now think about doing this for 1000 measuring stations. It gets to be a pretty involved equation although it should be straightforward to calculate once the algorithm is written in code.

The issue is that it would require climate science to get much deeper into the process of developing the data to do the actual calculation. So climate science does exactly what climate science always does – it just totally ignores the variance of the data. And you do the same. 

It should be noted that for measurements the values Var(1) and Var(2) are the measurement uncertainties of the estimated measurement values. Ignoring the variances is part and parcel of the climate science garbage assumption that all measurement uncertainty is random, Gaussian, and cancels thus it can be ignored. 

It should also be noted that it is not as simple as just adding the measurement uncertainties and dividing by 2 even if the sample sizes are equal. There is also a component of (μx – μy)^2/2. If the means are different this component is a significant contribution to the total variance. This is a major reason for the use of absolute values rather than a garbage anomaly which is always scaled to be a small value. And when adding the mean value of a warm temperature with the mean value of a cool temperature you will always get a significant contribution to the total variance. 

In fact, if μx – μy) is greater than the standard deviation of either variable, metrology rules would tell you that the measurements are not even valid for combination. Metrology would tell you that one measurement is not correct and re-measurment is necessary. Yet climate science gets around this by just scaling the difference away through the use of garbage anomalies.

What are you talking about? Temperature measurements are given as “estimated value +/- measurement uncertainty”. The random variable *is* the measurand. The measurement uncertainty is the dispersion of the values that can reasonably be attributed to the measurand – i.e. the standard deviation of the measured indications for an experimental data set made on the *same* measurand.

Correlation is typically determined from the estimated values, not the measurement uncertainty.

The measurement error is accumulative, especially when the data set consists of measurements of different measurands jammed into the same data set.

And measurement uncertainty is CERTAINLY dependent on the measurand as well as the measurement device and the environment n which the measurements are made.

Have you researched the word “hysteresis” yet? Have you researched how digital logic electronics work yet? Do you have even the faintest of clues about LIG thermometers having a different measurement uncertainty for temps going up than for temps going down?

“We measure the measurand at a certain point in time.”

Once again – not for global average temperatures. The base of this is daily (Tmax + Tmin)/2 for each measurement station used in the data set.

This is *NOT* the measurement at a certain point in time.

“Correlation is a general term describing relationship between random variables. Measurement error is uncorrelated between measurements.”

Error is not uncertainty. Go study JCGM 100:2008.

“No, and this is the thing we have been talking about.”

Once again, measurement uncertainty is not error. Measurement uncertainty is the dispersion of values that can reasonably attributed to the measurand. Error and true value are unknowable. You do *not* propagate error. You propagate measurement uncertainty. And measurement uncertainty always accumulates as you add more random variables into the data set.

“Measurement uncertainty at 10 am and at 11 am are independent of each other.”

The measurements are independent. The measurements, including the measurement uncertainty associated with each independent measurement, *IS* dependent on the measurand, however.

“Yes, and these are marginal for the end result. They are usually taken into account with a bit higher uncertainty. BTW they don’t use LIG anymore.”

How do you know they are marginal at the hundredths digit? Can you provide *any* backup in the literature for this? Without this it is just your opinion – an UNINFORMED opinion. Since the resolution limit of an ASOS weather measurement station is 0.1F just how do you get a reliable value for the hundredth digit?

BTW, did they have PRT sensors in the 1850-1900 IPCC baseline for pre-industrial average global temperatures?

There are two ways to combine random variables. You can convolve the probability density functions or you can concatenate the data sets. Climate science works by concatenating the values of the measurements into a single, new data set. When this is done the formula I’ve given you is the correct way to calculate the result.

go here: http://www.emathzone.com/tutorials/basic-statistics/combined-variance.html

Each temperature measurement is a different data set. It is *not* a combination of similar measurements of the same measurand. And since each can actually have a different number of elements you can’t just assume everything cancels.

I simply do not know why so many self-professed statistical experts (like you) get on here and try to lecture everyone when they really have no idea what they are actually speaking of.

“No, and I know it regardless of the fact that I’m not a climate scientist.”

The problem is that you don’t know *anything*. When you stick the daily mid-range temperature in Topeka with the daily mid-range temperature in Oklahoma City to determine an “average” temperature you have CONCATENATED the two values into a single data set.

“Each temperature measurement is a reading, not a set”

No, each temperature measurement is *NOT* a reading. It is the daily mid-range temperature formed from (Tmax + Tmin)/2. The fact that you have two measurements forming the mid-range value makes that mid-range value a random variable with an “average” and a standard deviation.

That standard deviation is important since two different sets of Tmax and Tmin can result in the same daily mid-point value. Say 50F-90F (Des Moines) vs 65F-75F (San Diego), middle of a continent vs a coastal location. Yet what does climate science do with the standard deviation of the random variable? THROW IT AWAY.

“This reading is an outcome of a random variable (the error) added to the “true value” “

First, error is not uncertainty. Go read the GUM. You haven’t studied how measurements are made and specified at all. Here is the second paragraph in the GUM:

“0.2 The concept of uncertainty as a quantifiable attribute is relatively new in the history of measurement, although error and error analysis have long been a part of the practice of measurement science or metrology. It is now widely recognized that, when all of the known or suspected components of error have been evaluated and the appropriate corrections have been applied, there still remains an uncertainty about the correctness of the stated result, that is, a doubt about how well the result of the measurement represents the value of the quantity being measured.” (bolding mine, tpg)

from 2.2.4: “Although these two traditional concepts are valid as ideals, they focus on unknowable quantities: the “error” of the result of a measurement and the “true value” of the measurand (in contrast to its estimated value), respectively. Nevertheless, whichever concept of uncertainty is adopted, an uncertainty component is always evaluated using the same data and related information.” (bolding mine, tpg)

You don’t even know the very basic concepts of metrology as it is formulated today.

“And this is why relative uncertainties do not make sense for error propagation. “

Relative uncertainties are REQUIRED for those situations where the components of the measurement have different units or magnitudes. A relative uncertainty has no units, it is a percentage. This makes summing them a simple task. This typically occurs in a functional relationship of different types of components combined in a multiplicative/quotient relationship. In a straight sum, relative uncertainties do not work and this applies to temperatures. The uncertainties associated with temperature measurements do add. You need to teach climate science this, they just assume that the addition of measurement uncertainties always add to 0 (zero).

“The good news is that we know the distribution of the error from calibration.”

As I keep telling you, calibration only involves the measurement device itself. The measurement device is only ONE COMPONENT of the measurement uncertainty budget. Even for the measurement device you only know the distribution for the laboratory environment. Once that device leaves the laboratory environment that distribution goes out the window. For instance, you can do a calibration run on a measurement device using a Stevenson screen in the laboratory. After that device has been in use for a year the paint on the screen will have deteriorated due to UV exposure – meaning the laboratory calibration run is useless since the internal environment of the screen will be different than it was in the lab. Even one day after leaving the calibration lab, the screen will be placed in a different environment than it had in the lab – it might be placed over concrete, sand, gravel, brown grass, green grass, dirt, or etc instead of over a tile floor in the lab. Each of these will change the systematic measurement uncertainty associated with the measurement device.

I’ll repeat: “You don’t even know the very basic concepts of metrology as it is formulated today.
“

“Okay, so the data sheet of a modern sensor says that in the range of (-50C, +50C) the characteristics of the result are the same, just as I have claimed. Thx.”

Once again you show your complete lack of knowledge of metrology. The uncertainty intervals shown are for measurement device in a calibration lab. Nothing is included for differences in the microclimate involved in a field installation. Nothing is included for weathering of the housing paint from UV exposure, for insect detritus in the air intake/output, for thermal drift from continuous heating of the sensor and other components, or even for different shades of color in the grass below the sensor due to seasonal changes. It’s why Hubbard and Lin found in 2002 that regional adjustments simply don’t work for “correcting” measurement station calibration. Any adjustment has to be done on a station-by-station basis using a calibrated reference.