The US Blows Hot And Cold

Guest Post by Willis Eschenbach

I got to thinking about the raw unadjusted temperature station data. Despite the many flaws in individual weather stations making up the US Historical Climate Network (USHCN), as revealed by Anthony Watts’ SurfaceStations project, the USHCN is arguably one of the best country networks. So I thought I’d take a look at what it reveals.

The data is available here, with further information about the dataset here. The page says:

UNITED STATES HISTORICAL CLIMATOLOGY NETWORK (USHCN) Daily Dataset M.J. Menne, C.N. Williams, Jr., and R.S. Vose National Climatic Data Center, National Oceanic and Atmospheric Administration

These files comprise CDIAC’s most current version of USHCN daily data.

These appear to be the raw, unhomogenized, unadjusted daily data files. Works for me. I started by looking at the lengths of the various records.

Figure 1. Lengths of the 1,218 USHCN temperature records. The picture shows a “Stevenson Screen”, the enclosure used to protect the instruments from direct sunlight so that they are measuring actual air temperature.

This is good news. 97.4% of the temperature records are longer than 30 years, and 99.7% are longer than 20 years. So I chose to use them all.

Next, I considered the trends of the minimum and maximum temperatures. I purposely did not consider the mean (average) trend, for a simple reason. We experience the daily maximum and minimum temperatures, the warmest and coldest times of the day. But nobody ever experiences an average temperature. It’s a mathematical construct. And I wanted to look at what we actually can sense and feel.

First I considered minimum temperatures. I began by looking at which stations were warming and which were cooling. Figure 2 shows that result.

Figure 2. USHCN minimum temperature trends by station. White is cooling, red is warming.

Interesting. Clearly, “global” warming isn’t. The minimum temperature at 30% of the USHCN stations is getting colder, not warmer. However, overall, the median trend is still warming. Here’s a histogram of the minimum temperature trends.

Figure 3. Histogram of 1,218 USHCN minimum temperature trends. See Menne et al. for estimates of what the various adjustments would do to this raw data.

Overall, the daily minimum temperatures have been warming. However, they’re only warming at a median rate of 1.1°C per century … hardly noticeable. And I have to say that I’m not terrified of warmer nights, particularly since most of the warmer nights are occurring in the winter. In my youth, I spent a couple of winter nights sleeping on a piece of cardboard on the street in New York, with newspapers wrapped around my legs under my pants for warmth.

I can assure you that I would have welcomed a warmer nighttime temperature …

The truth that climate alarmists don’t want you to notice is that extreme cold kills far more people than extreme warmth. A study in the British Medical Journal The Lancet showed that from 2000 to 2019, extreme cold killed about four and a half million people per year, and extreme warmth only killed a half million.

Figure 4. Excess deaths from extreme heat and cold, 2000-2019

So I’m not worried about an increase in minimum temperatures—that can only reduce mortality for plants, animals, and humanoids alike.

But what about maximum temperatures? Here are the trends of the USHCN stations as in Figure 2, but for maximum temperatures.

Figure 5. USHCN maximum temperature trends by station. White is cooling, red is warming.

I see a lot more white. Recall from Figure 2 that 30% of minimum temperature stations are cooling. But with maximum temperatures, about half of them are cooling (49.2%).

And here is the histogram of maximum temperatures. Basically, half warming, half cooling.

Figure 6. Histogram of 1,218 USHCN maximum temperature trends.

For maximum temperatures, the overall median trend is a trivial 0.07°C per century … color me unimpressed.

Call me crazy, but I say this is not any kind of an “existential threat”, “problem of the century”, or “climate emergency” as is often claimed by climate alarmists. Instead, it is a mild warming of the nights and no warming of the days. In fact, there’s no “climate emergency” at all.

And if you are suffering from what the American Psychiatric Association describes as “the mental health consequences of events linked to a changing global climate including mild stress and distress, high-risk coping behavior such as increased alcohol use and, occasionally, mental disorders such as depression, anxiety and post-traumatic stress” … well, I’d suggest you find a new excuse for your alcoholism, anxiety, or depression. That dog won’t hunt.

My very best to everyone from a very rainy California. When we had drought over the last couple of years, people blamed evil “climate change” … and now that we’re getting lots of rain, guess what people are blaming?

Yep, you guessed it.

w.

As Always: I ask that when you comment you quote the exact words you’re discussing. This avoids endless misunderstandings.

Adjustments: This raw data I’ve used above is often subjected to several different adjustments, as discussed here. One of the largest adjustments is for the time of observation, usually referred to as TOBS. The effect of the TOBS adjustment is to increase the overall trend in maximum temperatures by about 0.15°C per century (±0.02) and in minimum temperatures by about 0.22°C per century (±0.02). So if you wish, you can add those values to the trends shown above. Me, I’m not too fussed about an adjustment of a tenth or two of a degree per century, I’m not even sure if the network can measure to that level of precision. And it certainly is not perceptible to humans.

There are also adjustments for “homogeneity”, for station moves, instrument changes, and changes in conditions surrounding the instrument site.

Are these adjustments all valid? Unknown. For example, the adjustments for “homgeneity” assume that one station’s record should be similar to a nearby station … but a look at the maps above show that’s not the case. I know that where I live, it very rarely freezes. But less than a quarter mile (1/8 km) away, on the opposite side of the hill, it freezes a half-dozen times a year or so … homogeneous? I don’t think so.

The underlying problem is that in almost all cases there is no overlap in the pre- and post-change records. This makes it very difficult to determine the effects of the changes directly, and so indirect methods have to be used. There’s a description of the method for the TOBS adjustment here.

This also makes it very hard to estimate the effect of the adjustments. For example:

To calculate the effect of the TOB adjustments on the HCN version 2 temperature trends, the monthly TOB adjusted temperatures at each HCN station were converted to an anomaly relative to the 1961–90 station mean. Anomalies were then interpolated to the nodes of a 0.25° × 0.25° latitude–longitude grid using the method described by Willmott et al. (1985). Finally, gridpoint values were area weighted into a mean anomaly for the CONUS for each month and year. The process was then repeated for the unadjusted temperature data, and a difference series was formed between the TOB adjusted and unadjusted data.

To avoid all of that uncertainty, I’ve used the raw unadjusted data. 

Addendum Regarding The Title: There’s an Aesop’s Fable, #35:

“A Man had lost his way in a wood one bitter winter’s night. As he was roaming about, a Satyr came up to him, and finding that he had lost his way, promised to give him a lodging for the night, and guide him out of the forest in the morning. As he went along to the Satyr’s cell, the Man raised both his hands to his mouth and kept on blowing at them. ‘What do you do that for?’ said the Satyr. ‘My hands are numb with the cold,’ said the Man, ‘and my breath warms them.’ After this they arrived at the Satyr’s home, and soon the Satyr put a smoking dish of porridge before him. But when the Man raised his spoon to his mouth he began blowing upon it. ‘And what do you do that for?’ said the Satyr. ‘The porridge is too hot, and my breath will cool it.’ ‘Out you go,’ said the Satyr, ‘I will have nought to do with a man who can blow hot and cold with the same breath.’”

The actual moral of the story is not the usual one that people draw from the fable, that the Man is fickle and the Satyr can’t trust him.

The Man is not fickle. His breath is always the same temperature … but what’s changing are the temperatures of his surroundings, just as they have been changing since time immemorial.

We call it “weather”.

4.8 48 votes
Article Rating

Discover more from Watts Up With That?

Subscribe to get the latest posts sent to your email.

239 Comments
Inline Feedbacks
View all comments
March 11, 2023 4:38 pm

Nice overview, thanks for posting

March 12, 2023 1:35 am

As Anthony Watts wrote, the nighttime upward trend comes also from relocation of stations to inappropriate sites, like close to buildings, on concrete slabs, asphalt or rocks which radiate stored day heat well into the night, which raises the nightly measured temperatures. This effect “disappears” when measuring day temperatures, and that’s exactly what you pick up.

rbabcock
March 12, 2023 7:41 am

About the only two “unadjusted” datasets are probably UAH and temperature.global.

http://temperature.global/ (not secure)

The latter one sucks in readings from METARS, buoys and others as explained on the site as they occur and posts them to a database. METARS only report integers for the most part so posting a final number to the hundredth of a degree has some error built in. But when you are polling tens of thousands of these stations, you get the benefit of big numbers. You can see this database is basically flat since the super Niño year of 2016.

Reply to  rbabcock
March 12, 2023 10:56 am

Have you any idea how many models and adjustments are required to produce the UAH data?

Reply to  rbabcock
March 12, 2023 11:45 am

How exactly does temperature.global work? According to the site every year since 2015 has been below normal. With no definition of what period they are using for normal.

The site gives no indication of the methodology or who is running it.

Reply to  Bellman
March 12, 2023 12:34 pm

It’s a nonsense site. It claims to be using NOAA data but states that 2020 was 0.00C ‘below normal’ [sic], while 2016, the warmest year in the NOAA data, was -0.27C colder than that!

Reply to  TheFinalNail
March 12, 2023 1:25 pm

Indeed. But it seems odd to have this one site constantly pointed out as the most accurate, when there is zero indication on the site or anywhere of what method they use. Literally the only thing mentioned in the description of it’s process is to say they use “data functions”.

http://temperature.global/tg_dataflow.png

Reply to  rbabcock
March 12, 2023 2:58 pm

But when you are polling tens of thousands of these stations, you get the benefit of big numbers.”

A lot of people quote this without checking the assumptions behind the “big numbers”. If you check your statistics books you will find something like this.

The sample size (n) is the number of observations drawn from the population for each sample. The sample size is the same for all samples.

Conditions of the central limit theorem

The central limit theorem states that the sampling distribution of the mean will always follow a normal distribution under the following conditions:

The sample size is sufficiently large. This condition is usually met if the sample size is ≥ 30.

The samples are independent and identically distributed (i.i.d.) random variables. This condition is usually met if the sampling is random.

The population’s distribution has finite variance. Central limit theorem doesn’t apply to distributions with infinite variance, such as the Cauchy distribution. Most distributions have finite variance.

Central Limit Theorem | Formula, Definition & Examples (scribbr.com)

In actuality, Tmax and Tmin are highly correlated which means they are not independent. You can not translate two correlated variables into uncorrelated variables by simply calculating an arithmetic average.

Additionally, since Tmax is from one distribution (sinusoid) and Tmin is from another (exponential decay) they are not from identical distributions. Again, you can not translate samples into identical distributions when the piece parts of an arithmetic average is done.

Both the weak and strong Law of Large Numbers and the Central Limit Theory (CTL) require what is known as IID variables (samples). IID means “independent” and “identical distribution”. As you can see the LLN assumptions are not met.

In any case, significant number rules need to be applied correctly so that fantasy resolution is not reported. If you need, I can supply numerous examples of university laboratory instructions that make it plain it is not science to report results well beyond the resolution of the measured quantities.

Denis
March 12, 2023 9:36 am

Just how is the “average temperature” for each day from the Historic network determined? Do these stations consist of a single thermometer like the one on my house that I look at say twice a day at random or fixed times to get a high and low? Or are they min/max thermometers that stick at the days high and low values until reset? Whatever instruments they consist of, have they all been always so equipped? For min/max instruments, what does time of observation have to do with the min/max values reported unless the data handlers are assuming a steady transition between high and low during the day? If they are making such an assumption, what is the basis for it? We all have surely noted sudden temperature changes as weather fronts pass over at any time of day. I understand that for the Reference network temperature data are reported electronically many times per minute so an average of all of a days readings should be pretty close to a true value. It then seems to me that the Historic network numbers, now reported as nClimDiv data which are nearly identical to the Reference network data, must be simply forced to be the same with only politics involved, no science or logic.

bdgwx
Reply to  Denis
March 12, 2023 2:33 pm

Just how is the “average temperature” for each day from the Historic network determined? 

(Tmin+Tmax)/2

Do these stations consist of a single thermometer like the one on my house that I look at say twice a day at random or fixed times to get a high and low? Or are they min/max thermometers that stick at the days high and low values until reset?

They were primarily Min/Max LiGs until the switch to the MMTS and the like in the digital age.

Whatever instruments they consist of, have they all been always so equipped?

No. It is not uncommon for a station have a dozen or more instrument changes.

For min/max instruments, what does time of observation have to do with the min/max values reported unless the data handlers are assuming a steady transition between high and low during the day?

In a nutshell what happened is that the TOB changed from PM to AM in an effort to make precipitation observations more accurate because there would be less evaporation. Unfortunately this had the unintended consequence of switching from occasionally double counting highs to occasionally double counting lows. The double counting problem occurs because the TOB isn’t at midnight. For example, consider the TOB being at 8am on the Nov. 13th when the region is experiencing warm air advection. When the instrument is reset the min marker is set and never changes because the site warms over the next 24 hours. On the 14th when the instrument is read again at 8am it is reporting the min from the 13th which is lower than the min for the 14th. Refer to [Vose et al. 2003] for more details.

If they are making such an assumption, what is the basis for it?

The assumption is that min/max markers are for the day of observation. As illustrated above that assumption is not always true. The basis for making the assumption is that it is unreasonable to force observers to wake up every single night at midnight to record the daily values.

March 12, 2023 2:15 pm

Willis, a great essay. I think you are on the right track. You seldom see frequency charts from the trendologists. I have included a recent graph from a twitter friend that illustrates the unique distributions of temperatures during the day versus at night. An “average” of these two really don’t provide even a semblance of reasonable statistical parameter for analysis.

I have come to the same conclusions that the only legitimate way to examine temperature is by examining Tmax and Tmin separately. The first hint of this was an agricultural paper examining climate changes for crops. It did not use temperature directly, but instead using temperature days, growing days, first frost, last frost, etc. This study should be referenced often when arguing with climate alarmists.

From U.S. Agro-Climate in 20th Century: Growing Degree Days, First and Last Frost, Growing Season Length, and Impacts on Crop Yields | Scientific Reports (nature.com)

It has a long title.

U.S. Agro-Climate in 20th Century: Growing Degree Days, First and Last Frost, Growing Season Length, and Impacts on Crop Yields
Very good reading and provides a lot of info on how climate changes are being seen in the agricultural community. This study concludes that warming has not been a bad thing.

“Overall, we find that the observed changes in agroclimate, were beneficial for crop yields in the CONUS, albeit some crop and region specific exceptions.”

Edited: Corrected your link. (SUNMOD)

temperature distributions.jpg
TBeholder
March 13, 2023 7:04 am

The underlying problem is that in almost all cases there is no overlap in the pre- and post-change records. This makes it very difficult to determine the effects of the changes directly, and so indirect methods have to be used.

Technically? With refined statistical methods could handle this reasonably well. Even a layman like me can think at least of meta-statistics for the entire pool of instrument changes from A to B and running correlation analysis in a brute force iteration to find what values cause the minimal “bumps” in the data series. For the old stuff not available for direct experiments, at least.
Realistically? I would be very surprised if good old Frankenstein style stitching used in good old ozone hollering campaign turns out to be completely absent this time around.

Mike Shearn
March 15, 2023 2:06 am

How about re-doing figures 2 and 5 showing how “far” red or white they are. I would expect the Urban Heat Island effect to jump out. Thanks.