chubbs

On 8/17/2026 at 11:11 AM, bluewave said:

 

 

Climatologist49

 

@climatologist49.bsky.social

 

Follow

The first 14 days of August have blasted through the previous record for the Aug 1-14 period globally. 

 

 

 

In El Nino mode.  2026 started #5 in GISS in Jan and Feb, but reached #1 in July. August on track to easily top 2024. Can expect monthly records well into until 2027. The one exception is Sept, the "gobsmacking" warmth of 2023 will be tough to beat. 

This study provides an estimate of transient climate response (TCR) using CERES net radiation estimates and updates temperature-based estimates through 2025. The new TCR estimates for temperature are a little higher than previous studies (orange bars) and the CERES TCR estimates (red bars) are higher still, inline with TCR estimated with the warmer CMIP6 models. 

https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2026EF008356

21 hours ago, GaWx said:

 Per the following reliable and objective source, year 2026 as of week 32 has had the lowest area burnt in North America since 2012 by a good margin vs other years as of the same week (by a third). Also, globally it has been by a large margin the lowest for this point in the year since 2012 as I’ve been posting despite being near the warmest globally along with 2025 being 2nd lowest. My point in posting about 2026 for N.A. and 2026/5 globally is not to say CC is causing a decrease in wildfire activity because I’m not saying that. Instead, my point in posting this is to show that that there isn’t a proven simple relationship between CC and wildfire related activity and that different kinds of wildfire related data from various sources from various periods can be presented in various ways:
 

I don't think area burned is telling you much about CC; because, other human factors are driving the long-term trends. Per Our World in Data, the vast majority of wildfires occur in grassland and savanna in the sub-tropics and tropics, roughly 50% in Africa. Forest burning is a small portion of area burned in all regions. Our World in Data cites  a Science paper that attributes the long-term decrease in area burned to the spread and intensification of agriculture. In Europe and North America, reduction in cropland burned is another important long-term trend.

Have to look at other stats to see the effect of climate change. The blog I linked makes the point that the intensity of the largest fires is increasing due to climate change. Big, intense fires are the most problematic but aren't well captured by the area burned stats. Here's a recent paper that finds that a small number of extreme fires are having an out-sized impact on the forest ecosystem in western and boreal North America.

https://www.science.org/doi/full/10.1126/sciadv.aeg5802

Blog on wildfire smoke in the US. There's been an increase in smoky years since the late 2010s. 

https://www.theclimatebrink.com/p/smoke-gets-in-your-eyes

49 minutes ago, Mike Cycle said:

AI can do that easily.  I will put it to work.  Should we look at daily, monthly, or annual?  Tavg, or Tmin and Tmax?

I would look at Tavg, annual and monthly, to start. 

17 hours ago, Mike Cycle said:

Chubbs, I did try to expand your manual breakpoint detection process by asking AI to look at surrounding stations and identify "islands" where multiple stations are moving together.  I explored various settings, such as the minimum amount of synchronized temperatures before declaring an island, minimum stations to make an island, rules for stations leaving an island and new stations joining, and new islands forming.  The geographical area grew too large, pulling in Allentown etc. It will work, but its not a clear demonstration of using semi-local stations to identify breaks.  

However, Reading and West Chester do synchronize from 1929 to 1960 and make what looks like a solid reference against which breaks can be detected.  I have not looked at the individual breaks, but these breaks should be those that would be found by your approach.

Edit:  this does not rule out the case where Reading and West Chester have overlapping breaks.  

 

Yes, Reading and West Chester are stable in the 1929 to 1960 period and I am going to add Reading to my analyses. A couple of comments. As I said above looking at shorter time periods significantly expands the list of reference stations for individual station breaks as all stations are stable between the station breaks. Second, there is no need to restrict to short distances from Chester County. Weather station data is well correlated over hundreds of miles. An interesting first step would be to check how far good correlation with Chester County extends.

9 hours ago, ChescoWx said:

So let's address Charlie @chubbs post above 

Charlie, of course no dispute that Phoenixville stepped down at the 1948 move. But you've written the very objection into your own post: "Not surprising because the bias adjustments are calculated from the raw data." Charlie, your test is pairwise comparison of a target against raw neighbors across a break date. But so is the PHA!  You've actually hand-implemented the method NCEI automates. Agreement demonstrates reproducibility, but of clearly not correctness as you are simpling showing us the same method, same data. It's the same issue as Berkeley and NCEI agreeing being offered as independent confirmation - they are of course not!

On your "bullet proof" claim - the four available estimates of this one documented move:

  Berkeley -1.23 F / yours vs Coatesville  -1.49 F
  NOAA  -1.88 F / yours vs West Chester   -2.03 F

Four pairwise estimates of the same event spanning 0.80 F; the largest is 65% bigger than the smallest. Same sign and rough scale but the real question here as always has been the magnitude is what's in dispute here, and that spread isn't what bullet proof looks like!!. Also, your 2 references disagree by 0.54 F. If both were stable through 1947-50 they should agree on that Phoenixville's step. They don't, so they moved relative to each other in your window - and you date West Chester to 1970 and Coatesville to 1946, before your "before" period even opens.

So you also say "all of the major county COOP stations moved in the same direction and roughly the same magnitude after the war". If so, doesn't that actually represent a real problem for the method, not support for it??. Pairwise comparison sees only RELATIVE change - a shift common to the whole network differences out and is invisible by construction. The more synchronised those moves were, the less your test and the PHA can detect, in exactly the era carrying the county's largest adjustments. So Charlie while none of this says your step estimate wrong. It means the method you are using is simply being validated against itself, the spread across implementations exceeds the confidence claimed, and the scenario you offer as clinching is the one the method handles worst.!!

Paul

Lets look at the big picture. All four estimates you've listed agree that Phoenixville experienced spurious cooling in 1948. Better to recognize that Phoenixville had a cooling move vs ignoring the move and using the raw data.

However things aren't as uncertain as the four numbers make it seem. First the BEST estimate is 1.58F on Mike's chart not 1.88F, closer to NCEI. Second  If you look at Mike's charts there are many small differences between NCEI and Berkeley that cancel out over a longer time period. I don't see a big difference in the NCEI and Berkeley local climate history. Third, I wouldn't expect my estimates to be that accurate. I haven't taken the time to optimize my approach and more importantly there are many other stations besides West Chester and Coatesville that can be used to estimate the impact of Phoenixville's move.  

I say the NCEI method is bullet-proof because it has a strong scientific foundation.  Science says that there is a strong correlation in year-to-year temperature changes over a wide area and that's exactly what I have found. This chart is an example. These stations are well correlated at the time of the Phoenixville move indicating that they all can be used to estimate the timing and impact of the Phoenixville move.  Moreover, I have only scratched the surface, there are just the ones that had the time to check. 

The only way to estimate the magnitude of station breaks is from raw temperature data, by comparing stable and changed stations.  As documented in this thread, my tests using raw temperature data in Chester County agree very well with the NCEI bias adjustments. Not surprising because the bias adjustments are calculated from the raw data. My experience in Chester County has convinced me that the NCEI method is bullet proof. 

Below is raw temperature info on the Phoenixville 1948 move. NCDC has monthly means for Phoenixville for Janaury through July 1948, partial data from August through November but no monthly mean until December 1948. Phoenixville cooled relative to Coatesville and West Chester when the monthly mean started again in December 1948.  Phoenixville cooled 1.6F relative to Coatesville, and 2F relative to West Chester, due to the move. The stats using the monthly data are highly significant for both Coatesville and West Chester. 

The post-war moves at Coatesville (1946), West Chester (1970) and Phoenixville (1948) are all highly significant statistically. A coincidence that all of the major county COOP stations moved in the same direction and roughly the same magnitude after the war. Clearly not weather because the cooling occurred at different times at each station.

7 hours ago, TheClimateChanger said:

I wanted to take a closer look at the claim that NOAA/NCEI’s July 2026 temperature ranking is largely the product of “tampering” or adjustments to the historical temperature record. Tony Heller has advanced this claim on X/Twitter, and we've seen similar claims made by members here (cough, cough, ChescoWx).

So I downloaded the raw USHCN Tmax and Tmin monthly data directly from NOAA, without applying NOAA’s homogenization adjustments, and tried to see how far one can get using the raw observations alone. The results were pretty illuminating.

The first issue is that simply averaging the absolute temperatures of all available USHCN stations each year is not a very good way to construct a national temperature series.

Although USHCN contains 1,218 designated stations, nowhere near 1,218 stations report in every July. The number of stations with both Tmax and Tmin data increased from only about 528 in 1895 to roughly 1,200 by the 1930s–1960s, then began declining sharply in recent decades:

• 1990: 1,180
• 2000: 1,115
• 2010: 980
• 2020: 779
• 2025: 730
• 2026: 458 so far in the August 12 archive

That creates a compositional problem. If the stations disappearing from the network are climatologically warmer or cooler than those remaining, the simple national average can change even if temperatures at every individual station do not. Note that 2026 data is provisional, as a lot of stations are not yet reported in the dataset file.

Step 1: Calculate station anomalies

To deal with this, I calculated a 1951–1980 July climatology for each individual USHCN station. This was a particularly useful reference period because station coverage was near its peak; 1,217 of the 1,218 stations had sufficient data to calculate a usable climatology.

For every station and every July, I then calculated:

July anomaly = observed July temperature − that station’s own 1951–1980 July mean

This removes most of the changing-station-composition problem, because a hot station disappearing from the network no longer mechanically lowers the national average merely because its absolute climatological temperature was higher.

That made a surprisingly large difference in recent decades.

For July Tmax from 1990–2025:

• simple raw station-average trend: +0.35°F/decade
• station-anomaly trend: +0.55°F/decade
• composition effect: about −0.20°F/decade

From 2000–2025:

• simple raw Tmax trend: only +0.09°F/decade
• station-anomaly trend: +0.34°F/decade
• composition effect: about −0.25°F/decade

Why? Because the climatological July Tmax of the stations actually reporting has declined substantially:

Year	Mean July climatology of reporting Tmax stations
1990	87.39°F
2000	87.43°F
2010	87.31°F
2020	86.96°F
2025	86.88°F

In other words, the modern USHCN reporting network has increasingly become composed of cooler stations. Simply averaging their absolute temperatures therefore introduces an artificial cooling tendency.

The same phenomenon appears in Tmin.

Using an equal-station anomaly calculation, July 2026 came out approximately:

• Tmax: +1.66°F
• Tmin: +3.71°F
• Tavg: +2.68°F

On that basis, 2026 ranked roughly:

• 14th warmest Tmax
• 1st warmest Tmin
• 5th warmest Tavg

But there was still another problem.

Step 2: Geographic weighting

A station in a densely sampled part of Ohio should not carry the same national weight as a station representing a huge area of Nevada, Montana, or Wyoming.

So I took the raw USHCN station anomalies and placed them on a fixed 0.25° CONUS land grid. Each grid cell was assigned the anomaly of its nearest reporting USHCN station, and the resulting cells were area-weighted using the cosine of latitude.

This is essentially a simple Voronoi-style geographic weighting.

Importantly, it does not use NOAA homogenization. It does not alter the raw station observations. It merely prevents regions containing many stations from receiving disproportionately large weight in the national average.

The effect was substantial.

After rebasing our resulting national series to the same 1901–2000 reference used by NCEI, several important Julys looked like this:

Year	Equal-weight raw USHCN	Area-weighted raw USHCN	NCEI
1901	+3.25°F	+2.55°F	+2.60°F
1934	+2.99°F	+2.87°F	+2.72°F
1936	+3.32°F	+3.29°F	+3.14°F
1980	+2.02°F	+2.12°F	+2.09°F
2000	−0.59°F	+0.02°F	+0.22°F
2006	+2.32°F	+2.59°F	+2.76°F
2012	+3.15°F	+2.95°F	+3.10°F
2022	+2.02°F	+2.56°F	+2.76°F
2023	+1.25°F	+1.76°F	+2.02°F
2024	+1.47°F	+1.77°F	+2.07°F
2026	+2.67°F	+3.15°F	+3.26°F

The 1901 result is especially revealing.

Under a naïve equal-station average, 1901 appeared about 0.65°F warmer relative to NCEI. Once the raw observations were geographically weighted, the discrepancy collapsed to about 0.05°F.

The same thing happened at the other end of the record.

Our equal-station calculation put 2026 at only about +2.67°F, compared with NCEI’s +3.26°F. Geographic weighting alone moved the raw-USHCN result to +3.15°F.

No homogenization adjustment was necessary to explain most of that gap.

Final raw-USHCN ranking

Using the geographically weighted raw station anomalies, the leading Julys were:

1. 1936: +3.29°F
2. 2026: +3.15°F
3. 2012: +2.95°F
4. 1934: +2.87°F
5. 2006: +2.59°F
6. 2022: +2.56°F
7. 1901: +2.55°F

NCEI places 2026 slightly ahead of 1936 instead:

• 2026: +3.26°F
• 1936: +3.14°F

So after controlling for station composition and geographic weighting, the difference between our deliberately simple raw-data-only method and NCEI is not some enormous discrepancy.

Our raw calculation has 1936 about 0.15°F warmer than 2026.

NCEI has 2026 about 0.12°F warmer than 1936.

That leaves only about a 0.27°F swing in their relative difference between our raw-data method and the full NCEI analysis.

Now compare the entire records

This was perhaps the most surprising result.

Once the raw USHCN observations are converted to station anomalies and geographically weighted, the resulting curve lies almost directly on top of the official NCEI series.

Across 1895–2026, the mean absolute difference between the two annual July anomaly series is only about 0.11°F.

And remember what went into our series:

• raw USHCN observations
• individual station climatologies
• geographic weighting

That's it.

No NOAA homogenization adjustments were applied.

What does this actually tell us?

This certainly does not demonstrate that every NOAA adjustment is perfect, nor does our simple experiment precisely reproduce NOAA’s methodology. NOAA uses a larger station network, homogenization, and substantially more sophisticated spatial interpolation.

But it does demonstrate something important.

Most of the apparent disagreement between a naïve “raw USHCN” calculation and NOAA/NCEI can be reproduced without changing a single thermometer observation.

Two mundane methodological problems explain an enormous portion of it:

1. Changing station composition
2. Unequal geographic station density

The raw USHCN network has lost hundreds of reporting stations in recent decades, and the stations disappearing have, on average, been climatologically warmer than those remaining. That artificially suppresses recent temperatures if one simply averages absolute station readings.

Meanwhile, equal station weighting gives densely sampled portions of the country far too much influence. Correcting that spatial bias pushes recent hot years upward and, in some cases, early hot years downward.

The transformation of 1901 is particularly instructive:

Naïve raw USHCN: +3.25°F
Geographically weighted raw USHCN: +2.55°F
NCEI: +2.60°F

And for 2026:

Naïve raw USHCN: +2.67°F
Geographically weighted raw USHCN: +3.15°F
NCEI: +3.26°F

So when someone produces a graph of “raw USHCN temperatures” and argues that the difference from NOAA must therefore have been manufactured through data adjustments, there is a very large omitted variable: how the raw data are aggregated in the first place. A simple station-anomaly calculation plus basic geographic weighting — using the unaltered raw observations themselves — gets astonishingly close to the official NCEI record.

Nice analysis. Zeke Hausfather had a blog recently on how the 1930s warmth gets overestimated. Here's one of his plots. US stations were concentrated in the Great Plains in the 1930s, exactly where the warmth of the 1930s was focused. That is why it is important to grid the station data as you did. 

https://www.theclimatebrink.com/p/raw-data-wrong-answers

13 hours ago, ChescoWx said:

 As you know homogenization remains nothing more then a hypothesis-driven statistical correction, it is of course not a measurement.

Nope.  Homogenization is accepted science. Thoroughly tested over decades. Proven to obtain the correct climate information from raw temperature when station changes have occurred. The raw temperature data in Chester County is contaminated by myriad station moves and changes. Assuming that Coatesville and other Chesco COOPs stayed in exactly the same spot with zero equipment and method changes over decades will give you the wrong climate answer every time.

The cooling of Coatesville between 1945 and 1949 is well documented in this thread. The regional raw temperature data is 100% conclusive on this point. Coatesville was much cooler in 1949 than it was in 1945 relative to all stable regional stations. Comparison of Coatesville to Phoenixville over the period 1941-75, is immaterial; because, as the bottom chart shows, Phoenixville had a station break in 1948. The post-war  Phoenixville and Coatesville station breaks occurred at different times proving that they are not weather-related.

Mike's analysis and the NCEI bias adjustments are completely independent, derived from 2 different sources of information. Mike uses the station records. Bias adjustments use raw temperature data. The fact that they agree so well shows how powerful and accurate homogenization is in Chester County.

Your AI slop is wasting our time if it doesn't focus on the station breaks themselves. Regression over 35 years is not the right tool to identify station breaks. You need much more focused analyses to look at station breaks. The 1945-49 period for the Coatesville post-war move/changes for instance. 

Good quick explainer. 

July was a record for GISS 1.23 vs 1.20 in 2024. NOAA's July tied 2024. First of many monthly records over the next year. 

3 hours ago, Mike Cycle said:

Interestingly, there is pre-1894 data available. From 1873 forward there is daily temperature.  I am somewhat awed by the level of scientific competency shown in these (and other) old records.  

Observer is hard to read, but not Jesse Green.  Almost certainly there are usable archives at West Chester that would provide more information, including photos.  Tracking this stuff down and building a story line would be a perfect cross-disciplinary project for a college intern.

Would you mind if I had Claude run your method above on Chester County and nearby stations?  The output would be a complete human-readable timeline for how reference stations are used to arrive at adjusted records for these stations--something the OP has requested numerous times.  As I understand it the actual NCEI process involves matrix math, too complex for easy communication if you ask me; a manual method using your approach will be easy for anyone to understand.  

 

No I don't mind running Claude as you describe. Should be useful as a learning tool. 

Berkeley Earth has a West Chester 1 station with data back to Dec 1845, link for West Chester 1 below. The Berkeley Earth write-up for West Chester 2NW says that West Chester 1 is an alternative name for the station. So West Chester 1  may be the older data for West Chester 2NW. Worth checking anyway.

https://data.berkeleyearth.org/stations/35128

22 hours ago, Mike Cycle said:

If I understand what you are showing, the two pairs find a shift of -1.67 and -1.44 respectively for West Chester 2NW's April 1970's move.   Averaging 12 station to station pairs before the move, and 16 after.  

It would be straightforward to identify the non-move segments of local stations, and use those to test the others iteratively, building up a timeline for all stations.   Assuming all of the station moves are at HOMR or on the 530-1s, that would still leave the TOB changes.  Coatesville 1 SW switched from 8 PM to 8 AM & 8 PM at the end of October 1921, effectively eliminating the double counting of daily highs, shifting the time series down by  1.4 deg F. 

You understand what I did correctly. Yes it would be easy to use the stable portions of the Chesco station data to test the others. That's essentially what NCEI does (and other groups). The NCEI adjustments are based on raw temperature data not station information. NCEI also uses data from outside of Chester County to determine Chesco bias adjustments, taking advantage of the fact that weather data is correlated over hundreds of miles.  That's one of the reasons I found your Coatesville analysis interesting. The NCEI adjustments match the detailed site information you developed very well; and. for the periods where your site information was more complete and accurate than NCDCs, the match between the site info and the bias adjustments improved.

Interesting that the West Chester site has data from before 1894. Can you obtain the data? 

17 hours ago, Mike Cycle said:

Paul, using other stations as references will only work if those stations are reference grade themselves.  If there are station moves or other major changes, they are no longer "references".  The stations you have chosen as "references" show 14 detected breaks in the 1941 to 1975 windows alone.  I was working on this graph anyway, see below.

The fact that known good references are so scarce is a major motivation for developing homogenization processes or alternatives like BEST's scalpel.  These methods presume the raw data is replete with breaks and work to overcome this challenge.  Your approach does not.

The September 1921 TOB for Coatesville shows up clearly on the kinds of comparisons seen above. 

 

 

Yes stations contaminated with station changes are not suitable references. Myriad station changes are readily apparent on Paul's chart. Just look for a single station with-year-to-year movement different than the others. Between 1941 and 1975, West Chester, Phoenixville, Holtwood and Ephrata all had large station changes making them bad references for Coatesville over that period.

However, there's an easy way to turn these stations from bad into good references. Use shorter time periods without station changes. The station changes only effect a small portion of any one station's 35-year record. The other years are untainted.

Several examples are posted upthread. Here's another : West Chester's 1970 move using monthly data from Phoenixville and Coatesville as the reference. Both stations were stable in this 3-year period. The West Chester station closed at the end of 1969 and re-opened at a new site in May 1970. West Chester cooled immediately on reopening. The stats show a statistically significant 1.4-1.7F cooling relative to Coatesville and Phoenixville due to the move.

This simple analysis only uses 2 reference stations. Many more are available outside of Chester County to further improve the power of the statistics. It is clear that the most important NCEI bias adjustments are highly statistically significant.

When you focus on periods just before and just after station changes, the changes are have very strong statistical significance. Below is an example for Coatesville, using the monthly difference between Coatesville and West Chester.  On average, Coatesville was 2.5F cooler relative to West Chester in 1949-50 vs 1944-45, before the 1946 move. The stats show that the move-related change was highly significant statistically. 

https://dashboard.theclimatebrink.com/#global

1 hour ago, Mike Cycle said:

Observer Pyle made a remark about the station move on the January 1946 record.  Then there is a record gap through to October 1948, when the first PECO observation record appears.  I don't know if this is a real gap.

During the same time, observer Pyle is also taking stream gage measurements.  

The earliest address is shown as 547 E. Main, across from the high school, but observer Gordon owned 574 E. Main, and there is no 547 E. Main on current mapping.  https://arcweb.chesco.org/LR_INFO/lr_info.aspx?PARID=1606 03640000

 

The bulk of the post-war Coatesville cooling occurred in 1946 and 1947 which is consistent with a 1946 move. The cooling relative to West Chester and ABE also matches the timing and roughly 2F magnitude of the NCEI and Berkeley Earth bias adjustments that you posted above.

19 hours ago, Mike Cycle said:

Hi All, occasional lurker and Chester County resident here.   

I have done a deep dive on Coatesville 1 SW's history, working through the 530-1, the monthly paper records and property deeds for observers Gordon, Pyle and PECO.  I then compared the record of station changes with the breakpoints identified by NOAA/NCEI and Berkeley Earth Surface Temperature (BEST).  Claude helped with the graphic.

There is remarkable agreement between the history of changes at 1 SW and the breakpoints detected.   

edit:  The HOMR record for this station, (presumably pulled from the 1956 530-1 (itself likely pulled from the B44s)) is flawed.  

Very interesting. More evidence supporting bias adjustment in Chester County. The detailed documentation describes many more station changes than the NCDC documented. Indicates that the move from the City of Coatesville to Doe Run Road occurred in 1946. Per NCDC documentation Doe Run road site wasn't used until 1949.

Most of the Arctic Ice caps are having a record or near record melt year. The exceptions are Greenland which has above average losses and Ellsmere Island the which has below average losses. The loss distribution is consistent with the average temperature map since May 1. The Atlantic side of the arctic has been warm, while the Pacific side has been cool. Other than Ellesmere, the arctic ice caps are on the Atlantic side.

https://www.climato.uliege.be/cms/c_5855702/fr/climato-arctic

On 7/30/2026 at 5:04 PM, ChescoWx said:

You limited your local review to only 3 chesco stations - My review with AI had more local stations than you including Sadsbury, Devault 1W, Kennett Square and Morgantown- I focused on only local stations in or on the immediate borders of chesco like Morgantown/Elverson to better identify if there is any actual support based on the raw data that was able to identify if any significant drifts occurred at any stations near Coatesville 1SW to suppor the NCEI chilling adjustments. The statistical analysis clearly does not support the NCEI adjustments for Coatesville 1SW
 

Nope. Your station selection and AI's method are both faulty.  The key period for evaluating the Coatesville post-war moves is 1945-1950, the period immediately after the war when the moves occurred. I didn't include Sadsbury, Kennett Square, and the other Chesco stations you listed, because they didn't operate in that period. After reading the AI analysis more closely, I see that AI used regression analysis to make its determination. As I said above, that is a poor choice because the station moves are focused in time not spread out over a long period of time. AI based its conclusion largely on regression results from Devault and Morgantown both of which didn't start operating until 1951. AI is drawing a conclusion from two stations that don't provide any information on the Coatesville post-war moves; because, they weren't operating at the time. A clear indication of a faulty analysis approach. 

With 2 of the 3 Chesco stations, Caotesville and Phoenixville, both experiencing spurious cooling between 1945 and 1950, It is important to include non-Chesco stations in the analysis.  That's why my analysis included 10 non-Chesco stations. The non-Chesco stations and West Chester agreed completely with West Chester in identifying the common weather signal between 1945 and 1955. Having the common weather signal makes identification of the Coatesville and Phoenixville moves easy; and, with plenty of stations is highly significant statistically.

A faulty analysis does not disprove a good one. To disprove my analysis, you need show that the significant difference between the move and non-move stations in the 1945-55 period changes when more data is added. Good luck with that, the P value is 0.0001.

Extended the stat analysis to the 1970s to include the West Chester move. Two comparisons were made: individual years using 1972 minus 1945 as a metric, and a 5-year average using 1971-75 minus 1941-45 as a metric. The results are the same for both metrics: the Chester County COOPs are significantly different statistically from the non-Chesco stations.   As was the case for the 1950s there is no overlap between the move and non-move stations. The Chester County COOPs have spurious cooling of over 2F. This analysis and the one above provide strong evidence that the larger NCEI bias adjustments are very robust statistically.

This analysis also shows the problems with AI. As Don stated above AI is handy for an initial screening; but, AI is unlikely to give a complete and accurate answer to a complex problem.

Here's a closer look at the statistics of the Coatesville and Phoenixville post war moves. Differences between stations was removed by setting 1945 to zero to facilitate comparison between stations.

Coatesville and Phoenixville behaved differently than the other stations, cooling relative to the other stations at the time of the moves; and, remaining cooler after the moves..

Using the temperature difference between 1945 and 1952 as a metric, Coatesville and Phoenixville are significantly different from 11 stations without station changes. The P value is 0.0001 which means that there is virtually no chance that Phoenixville and Coatesville are the same as the other 11 stations. Furthermore Phoenixville and Coatesville are significantly different from the other stations in every year after the moves in this period.

Finally note also the good agreement among the 11 stations without changes. Chester County shares the same weather with a much larger region. Many stations can provide useful information on the Chester County COOP stations; which makes the identification of Chester County station changes very robust.

On 7/26/2026 at 12:19 PM, ChescoWx said:

@chubbs below is some more detailed research and some solid evidence below to find if the single station NCEI applied cooling adjustments to Coatesville 1SW are supported by surrounding stations. For site moves and TOB changes.  The final finding is that the clean evidence points to a real, still-unexplained cooling signal during at least the 1952-1982 portion of this period -- not caused by a location move or TOB change (neither was validated), Also Station relocation, across the 88-year period with usable NCEI-adjusted data (1895-1982) spanning the station's six relocations, never shows up as an independent, validated driver of any adjustment on its own. Once the confounded stations are set aside, the clean evidence actually points toward a real, unexplained cooling trend during at least the 1952-1982 portion of that period — just not one caused by a location move or TOB change. Take a look at the below and let me know what this analysis is missing? This is only for Coatesville 1SW raw vs the NCEI adjusted for that individual station. Based on this I will look to see if the other NCEI station adjustments to the raw data are valid and supported by the data. I have much more supporting data so feel free to ask and I will provide more data.

Coatesville 1SW: Overall Testing Approach and Findings
Overall Testing Approach
1. Pulled Coatesville 1SW's exact location and observation-time history from NCEI's HOMR database -- six site relocations and five distinct observation-schedule eras spanning 1888-1982, independently verified against the live HOMR page and a user-provided satellite image of the final site. 2. Initially tested using NCEI's own adjustment as the yardstick (comparing how the adjustment changed before/after each documented event). This is where standard time-of-observation-bias (TOB) theory first looked promising, then broke down: an all-morning schedule (1946-1982) should need little or no adjustment under TOB theory, not the sustained cooling actually applied. 3. Switched to raw-data-only testing -- the methodologically correct approach, since testing against NCEI's own output can't establish whether that output was justified in the first place. For each of Coatesville's 7 documented events, compared Coatesville's raw record against every available nearby raw-data station in clean, non-overlapping before/after windows. 4. Required cross-validation across independent references -- an event only counted as a real, validated discontinuity if multiple neighbor stations agreed on the direction of the shift, not just one. This is the key filter separating genuine breaks from ordinary weather noise. Expanded from 2 to up to 4 reference stations (West Chester 2 NW, Phoenixville 1E, Kennett Square, Sadsburyville) depending on era. 5. Result: only 3 of 7 documented events passed. 1916, 1922, and 1946 showed real, multi-station-confirmed shifts. 1902, 1910, 1930, and 1948 did not -- either unconfirmable (no second reference available) or references actively disagreeing (the signature of noise, not a real break). 6. Built a raw-justified adjustment chain using only those 3 validated steps, anchored to the stable, unchanged 1946-1982 era as a zero baseline, and compared it decade-by-decade against NCEI's actual adjustment, with formal significance testing at each step.
Overall Finding
NCEI's adjustment ran colder than the raw evidence supports in every single decade from the 1890s through the 1980s, with no exceptions. Restricting to the 8 complete decades (1900s-1970s), 6 of 8 reach conventional statistical significance (1900s, 1910s, 1930s, 1940s, 1950s, 1960s); the other two (1920s, 1970s) don't reach significance individually but still point the same direction. Pooled across all 88 years, the effect is overwhelming (p < 0.0001). The size of the over-cooling does not trend over time -- it is a persistent, roughly steady bias, not something that worsened or improved across the record. See "Raw vs NCEI vs Corrected Trend" for the full year-by-year data, decade table, and significance tests.
Adjustment Broken Out by Rationale (all 7 documented events)
Of NCEI's stated PHA rationale categories, only station relocation and observation-time change could be directly tested here. Only 3 of the 7 documented events show real, cross-validated evidence.
Event	Rationale Category	Validated?	Contribution to the Adjustment Chain
1902	Location move	No -- single reference, unconfirmable	0 (no independent step; folded into the 1902-1915 era)
1910	TOB change	No -- 2 references disagree in direction	0 (no independent step; folded into the 1902-1915 era)
1916	Location move + TOB shift (simultaneous)	Yes -- 4 references agree	+0.36°F, partially OFFSETTING (applies to 1902-1915 only; can't attribute to either cause alone)
1922	TOB change (location unchanged)	Yes -- 3 references agree	-1.28°F, the entire step applied to 1922-1945
1930	Location move	No -- 2 references disagree in direction	0 (no step between 1922-1945 and 1946-1982)
1946	TOB change (location unchanged)	Yes -- 2 references agree (thin sample)	Combined with 1922: the entire -2.11°F applied to 1916-1921
1948	Location move	No -- 2 references disagree in direction	0 (1946-1982 treated as one continuous baseline era)
PRACTICAL ANSWER: of all 7 documented events, only 3 (1916, 1922, 1946) show real, cross-validated raw discontinuities -- 2 of those 3 (1922, 1946) are TOB changes with no location change; the third (1916) is an inseparable combined event. The 4 location-related events (1902, 1910, 1930, 1948) show no validated evidence at all. Station relocation, across the 88-year period with usable NCEI-adjusted data (1895-1982) spanning the station's six relocations, never shows up as an independent, validated driver of any adjustment on its own.
Important: What "TOB" Explains Here, and What It Does Not
The "TOB" label in the table above explains why the RECONSTRUCTED adjustment chain steps DOWN between eras (e.g. why 1922-1945 sits 1.28°F below the 1946-1982 baseline) -- it describes validated ONE-TIME transitions in Coatesville's own raw record relative to its neighbors. It does NOT mean TOB explains or justifies NCEI's actual, continued adjustment behavior within any era, including 1946-1982. If anything, classic TOB theory says a single MORNING reading (used throughout 1946-1982) should need a mild WARMING correction, not the sustained cooling NCEI actually applied that whole period (-0.82°F in the 1950s down to -0.21°F in the 1980s). That makes NCEI's continued cooling during 1946-1982 LESS justified under TOB theory, not more -- see "No Basis Found for Any Cooling Adjustment During 1946-1982" below, which is a separate, additional finding from the step-transition analysis above, not an extension of it.
Important Caveat: What "Corrected" Does and Does Not Establish
The "Corrected" series treats the 1946-1982 era as a zero-adjustment baseline because it is the final, most stable era with no further validated events. This is a standard homogenization convention (adjust the past to match the present), but it is NOT the same as proving 1946-1982 is itself free of bias. Under classic TOB theory, a single MORNING reading (used throughout 1946-1982) should carry a mild COLD bias, not be bias-free -- meaning the true fully-corrected record might need the entire chain shifted warmer, including the baseline era, which would make the required cooling for the earlier eras SMALLER than calculated here, not larger. This analysis establishes the RELATIVE size of each validated step; it does not establish an absolute, bias-free zero point.
No Basis Found for Any Cooling Adjustment During 1946-1982
Separately from the baseline-anchoring question above, a cleaner and more direct claim: within the 1946-1982 period itself, location and observation schedule were BOTH stable throughout (no validated events of either kind). An additional check for undetected drift WITHIN this period (not just at its boundaries), using every available overlapping reference station, found no consistent evidence either: Devault 1W and Morgantown each show a significant COOLING drift relative to Coatesville (-1.15 and -0.26 F/decade), while West Chester 2 NW shows a significant WARMING drift in the OPPOSITE direction (+0.52 F/decade), and Phoenixville 1E shows no significant drift at all. By the same cross-validation standard used throughout this investigation, this disagreement means no validated drift exists either. CONCLUSION: with no validated location change, no validated TOB change, and no validated drift anywhere in this 36-year window, there is no raw-data basis found for ANY of NCEI's cooling adjustment during 1946-1982 -- not the varying decade-by-decade amounts NCEI actually applied (ranging from -0.82°F in the 1950s to -0.21°F in the 1980s), and not a flat adjustment either. This covers what could be directly tested here -- station moves and TOB, two of NCEI's four stated PHA rationale categories. No evidence of an undocumented instrument or land-use change was found either, but the available data could not specifically rule those out.
Reference Station Confound Check (added later)
The cross-validation method above assumed the reference stations (West Chester 2 NW, Phoenixville 1E, Kennett Square, Sadsburyville, Devault 1W, Morgantown) were themselves stable during the windows used to test Coatesville. Checking each one's own HOMR history directly found this was not fully true.
Reference Station	Own HOMR History	Confound Found?
West Chester 2 NW	Moved slightly in early 1921; further moves in 1936, 1955, 1970, 1998, 2009, 2016	YES -- 1921 move overlaps the 1922 test's "before" window (1918-1921); 1955/1970 moves fall inside the 1946-1982 interior-drift window
Phoenixville 1E	Stable 1915 to 1948-05-01, then moved	PARTIAL -- clean for 1922/1930/1946; its 1948 move falls inside the 1946-1982 interior-drift window
Kennett Square	Moved between 1916-11-30 and 1919-01-01	YES -- overlaps the 1916 test's "after" window (1917-1920)
Sadsburyville	Single location, entire 1910-1922 operating life	None -- clean
Devault 1W	Single location, entire 1951-1989 operating life	None -- clean
Morgantown	Single location, entire 1951-1987 operating life	None -- clean
REVISED CONCLUSIONS: (1) The 1916 finding still holds -- Sadsburyville (fully clean) and West Chester (clean for this specific window) both independently confirm it; only 1 of its 4 supporting references (Kennett Square) has a confound. (2) The 1946 finding is unaffected -- both its supporting references were stable throughout that specific window. (3) The 1922 finding is WEAKER than originally presented: it only ever had 2 supporting references, and the West Chester one is now known to be confounded by West Chester's own 1921 move. Phoenixville alone remains clean and still shows the same direction, but that is one clean reference, not the two-station cross-validation required elsewhere in this analysis -- treat 1922 as a lower-confidence, single-reference finding, not a fully cross-validated one. (4) MOST IMPORTANTLY, this partially reverses the "No Basis Found for Any Cooling Adjustment During 1946-1982" conclusion above: the earlier interior-drift check found 4 references disagreeing and read that as noise, but 2 of those 4 (West Chester, Phoenixville) turned out to have their own undocumented moves inside that exact window, while the 2 clean references (Devault 1W, Morgantown) both agreed on a real cooling drift (-1.15 and -0.26 F/decade). Once the confounded stations are set aside, the clean evidence points to a real, still-unexplained cooling signal during at least the 1952-1982 portion of this period -- not caused by a location move or TOB change (neither was validated), but a genuine physical signal nonetheless, not simply noise as previously stated.

A good first step in using the raw data to identify station changes, with several station changes identified. What's missing are most of the station changes in Chester County. A couple of issues cause station changes to be missed: 1) It appears that only station moves and TOB  changes described in NCDC documentation were investigated. There are two problems: NCDC documentation is incomplete and there are many other causes of station changes besides moves and TOB. 2) The common weather signal across stations is not identified. Having this signal facilitates the identification of station changes.  and 3) Stations outside of Chester County are not used. Comments 2 and 3 work together. It easier to identify the common weather signal with more stations. The fact that many station changes are missed causes station comparison to break down.  In summary the analysis is much less thorough than NCEI's

A couple of comments on write-up itself. First like most AI its very wordy and wishy-washy. Too long to review carefully in detail. More importantly, there are no charts or quantification of any kind. Impossible to verify or compare to other results without hard numbers.

11 hours ago, Typhoon Tip said:

 it’s tentative but we may be seeing a spring 2023-like global thermal surge beginning. Air and sea presently lurching in tandem.     

That's the $64 question for this year. Will we see the same type of "gobsmaking" global warming we saw in 2023 or will this year behave more like a typical strong/super el nino. So far I'd put this year in the typical camp, with global air temperatures lagging the rise in ENSO and global SST.  

One aspect that has held back global temperatures in the past 2 months has been very cold temperatures in Antarctica associated with a strong AAO+ regime.  In 2023 AAO was strongly positive in January (see AAO table below), reversing by March, and staying negative through September. So the AAO boosted global temperatures in the first half of 2023 but are holding them back this year. One of the many factors that contributes to a different outcome in 2026 vs 2023. If the AAO reverses in August there will be an added boost in global temperatures on top of ENSO. We will see.

2023  2.304  0.554 -0.258 -0.921  1.452 -0.438 -0.818 -0.038 -1.050  0.535  0.097  1.510
2024  0.922  1.043 -0.058  1.005 -0.073  0.210 -0.597 -2.150  0.098 -0.567  1.020 -0.826
2025 -0.080 -0.271  0.733  1.138  0.509  0.209  0.753  0.357 -0.709 -1.236 -1.324 -1.136
2026  0.848  0.489  0.492 -0.125  0.544  2.506