ChescoWx

Back to the NCEI adjustments Station Breaks Code 1 (documented move) Code 2 (gap) Code 3 (documented TOB) Code 4 (no documentation) Coatesville 1SW 14 0 0 0 14 West Chester 2NW 10 2 0 1 7 Phoenixville 1E 9 0 1 0 8 Devault 1W 5 0 0 0 5 West Chester 1W (Hadley) 5 0 1 0 4 Kennett Square 4 0 1 0 3 Coatesville 2W 2 1 0 0 1 Morgantown 2 0 0 0 2 Sadsburyville 0 — — — — West Grove 1SE 0 — — — — TOTAL 51 3 3 1 44 Four of 51 breaks (8%) have a documented physical cause. Forty-four (86%) are statistically inferred with no documentation. And yes — the two systems agree at r = 0.9986, mean difference 0.030 °C/decade across the five stations NCEI adjusted. That's the bridge I failed to draw last message. Put together, the argument runs: NCEI's files publish no breakpoint information at all — I checked every field; there's no changepoint list, no dates, no reasons. But NCEI's adjustments are near-identical to Berkeley's, and Berkeley does publish the reasons. So Berkeley's inventory is a defensible proxy for what the PHA is doing to the same stations. And that inventory says: overwhelmingly undocumented, statistically inferred corrections. Coatesville 1SW is the clearest case — fourteen breakpoints, every one empirical, against six documented relocations in HOMR. The algorithm found fourteen discontinuities and matched none of them to a known event. Meanwhile at West Chester it did the reverse error: classified the April 1921 and Nov/Dec 1927 documented observer-and-site changes as empirical, and produced no breakpoint at the documented March 1936 relocation. So the metadata is being used badly in both directions — real events missed or mislabelled, inferred events unsupported. Now "undocumented" is not the same as "wrong" — we know the PHA is designed to catch undocumented changes, and station histories are genuinely incomplete. But the specific issue at Coatesville 1SW remains that 86% is a high proportion on which to rest a correction that reverses the sign of Coatesville's trend, and it's a fair thing to put in front of them: on what basis were fourteen breakpoints identified at a station with six documented relocations, none of which they coincide with? I have drafted an email that I have sent to them this AM see below and I will keep you updated. Jared -Thank you again for the QCU and QCF files — they were exactly what I needed. Working through them raised a question I don't think the distributed product can answer, and I'm hoping you can point me to whoever handles PHA methodology. The GHCN-M v4 records carry ID, YEAR, ELEMENT and twelve repetitions of VALUE/DMFLAG/QCFLAG/DSFLAG. As far as I can tell there is no changepoint list, no adjustment magnitudes, and no indication of what supports any given correction. For USC00369464 (West Chester 2NW) the QCF series applies roughly -2.37 C to the 1855-1869 period — a stretch covered by a single observer at a single site, per WB Form 530-1, Index No. 36-9464-3. For USC00361589 (Coatesville 1SW) and USC00369464 (West Chester 2NW), could I request for the entire POR: 1. The changepoint dates the PHA identified for each station. 2. The adjustment magnitude applied at each changepoint. 3. Whether each was supported by documented station-history metadata or detected statistically from neighbor comparison and if so what station(s) were used for this neighbor comparison? I understand the PHA is designed to detect undocumented discontinuities, so I'm not assuming documentation exists for every break — I'd simply like to know which are which. Berkeley Earth publishes exactly this per station (a per-month code distinguishing documented moves, time-of-observation changes, gaps, and empirical detections), which is what prompted the question. If this is better directed elsewhere within NCEI, I'd be grateful for the referral. Thanks very much, Paul

Our muggy warm weather looks to continue through tomorrow with chances of showers and thunderstorms before a cold front crosses the area by later Thursday. Following that front we get an almost autumnal sneak peek with highs Friday through Monday in the upper 70's to around 80 degrees and lows by Saturday morning in the refreshing 50's.

Our muggy warm weather looks to continue through tomorrow with chances of showers and thunderstorms before a cold front crosses the area by later Thursday. Following that front we get an almost autumnal sneak peek with highs Friday through Monday in the upper 70's to around 80 degrees and lows by Saturday morning in the refreshing 50's.

Below is the climate summary for West Chester for the decade of the 1870's with data now back to 1871

Thanks again Mike I am in the process of having the data reviewed it looks like I can add 21 years of raw data to West Chester 2NW to now include 1871-1892 thank you!! It looks like that's the only Chesco site on there....I know there were older obs at least back to 1888 for Coatesville 1SW on a summary basis not daily but if you find any other stations let me know. Really appreciate your help! Paul

You are correct the processed data is listed as Berkeley station 35128, "WEST-CHESTER-1W" — monthly means, sole ID hadcru 720461, no COOP number - that is processed not raw data! This is a different animal entirely. The CSV matching the handwritten sheet says nothing about that file, because the CSV isn't its source this indeed looks to be West Chester 2NW old data! Below is the QC Check and of note only those 2 days in June 2 and 3 were flagged

Interesting and it matches the CSV? I will have another look - thanks again!! Great find!!!

Interesting in reviewing the data it is apparently an already altered and processed product and not raw! What about 1855–1892? That's the honest attraction — 462 months, 35 complete years, entirely predating your Source File. But it fails your own standards on three counts: Monthly means only. Your Source schema is daily: High, Low, Avg Temp, Precip, Snow, Sn Cover. There's nothing to put in those columns. It's a processed product, not raw. Your standing rule is raw observations only, no estimated or filled values. This has passed through CRU's pipeline and carries an undocumented adjustment. The 1.32 °C offset almost certainly extends backward. The bias is present throughout the checkable period and drifts at +0.126 °C/dec. Pre-1893 values are probably ~1.3–1.5 °C too cold relative to your network, and there's nothing to calibrate against.

From the analysis by Claude - Coatesville 1SW trends, °C/decade, 1894–1982 - Key headline the adjustments are questionable.... 1894–1903, mean adjustment applied: −0.72 °C (−1.30 °F) — early years pushed down 1972–1981, mean adjustment applied: +0.79 °C (+1.42 °F) — late years pushed up Net swing across the record: +1.51 °C (+2.72 °F) Cool the beginning, warm the end. That's the whole mechanism, and it's why the adjusted series warms while the raw series cools. Why I put that column in the table rather than the raw trends. The r² of 0.671 is the striking part: 67% of the variance in Berkeley's correction is explained by a straight line in time. These corrections are supposed to be a sum of independent step functions, each fixing an unrelated site or instrument problem at an unrelated date. There's no mechanical reason fourteen such steps should add up to a straight line. That they do — at five of seven stations, all with p < 10⁻⁴ — is the observation worth putting in front of NCEI. Every single breakpoint is code 4 Fourteen breakpoints: 1896-12, 1902-02, 1917-08, 1920-09, 1922-08, 1930-10, 1936-05, 1941-06, 1946-04, 1959-11, 1968-05, 1971-04, 1973-07, 1976-02. Not one is code 1 (documented move), 2 (gap), or 3 (time-of-observation change) — all fourteen are empirically inferred from neighbor comparison. Given that your HOMR work documents six actual site relocations at this station between 1888 and 1982, Berkeley detected none of them as documented events. Worth cross-checking your Coatesville 1SW History sheet against those fourteen dates: any that coincide with a real move is a defensible correction; any that don't is a correction inferred purely from the neighbors disagreeing. Of note only Coatesville 2W had a documented flagged station move across the entire Coatesville lineage - in April 2001. That's the first and only documented move Berkeley has flagged anywhere in the Coatesville lineage, against six relocations in your HOMR research and fourteen all-code-4 breaks at 1SW.

Thanks Mike!! It looks like there is t-max and t-min starting pretty consistently for West Chester starting late in 1873. I will try to add this to my data set and run some reports! This is great!! If you can find any other stations from before 1894 let me know!

I had a request to update the "heatwave" stats to something potentially more meaningful. Which is 3 or more consecutive days with daily maximum temperatures above the 90th and 95th percentile for all Chester County Stations. The below rank summary makes the absolute-vs-relative divergence concrete. Sorted by absolute rank, the biggest movers are the 2010s (11th → 4th) and 2020s (13th → 6th), both rising 7 places once heat is judged against each station's own norm — meaning recent decades have been locally extreme even though they weren't absolutely record-hot. Going the other way, the 1960s (−4) and 1990s (−5) fall under the relative lens, and the 1940s slip from 1st to 3rd. The 1930s hold firm near the top under both definitions, and the 1950s actually take the #1 spot under P90

I had a request to update the "heatwave" stats to something potentially more meaningful. Which is 3 or more consecutive days with daily maximum temperatures above the 90th and 95th percentile for all Chester County Stations. The below rank summary makes the absolute-vs-relative divergence concrete. Sorted by absolute rank, the biggest movers are the 2010s (11th → 4th) and 2020s (13th → 6th), both rising 7 places once heat is judged against each station's own norm — meaning recent decades have been locally extreme even though they weren't absolutely record-hot. Going the other way, the 1960s (−4) and 1990s (−5) fall under the relative lens, and the 1940s slip from 1st to 3rd. The 1930s hold firm near the top under both definitions, and the 1950s actually take the #1 spot under P90

I had a request to update the "heatwave" stats to something potentially more meaningful. Which is 3 or more consecutive days with daily maximum temperatures above the 90th and 95th percentile for all Chester County Stations. The below rank summary makes the absolute-vs-relative divergence concrete. Sorted by absolute rank, the biggest movers are the 2010s (11th → 4th) and 2020s (13th → 6th), both rising 7 places once heat is judged against each station's own norm — meaning recent decades have been locally extreme even though they weren't absolutely record-hot. Going the other way, the 1960s (−4) and 1990s (−5) fall under the relative lens, and the 1940s slip from 1st to 3rd. The 1930s hold firm near the top under both definitions, and the 1950s actually take the #1 spot under P90

Continued warm and humid through Tuesday before we start a welcome cooling trend which by Saturday morning will see some locations with lows in the 50's with highs by the end of the week in the pleasant 70's. Best chances of any rain will be Monday night through Wednesday

Continued warm and humid through Tuesday before we start a welcome cooling trend which by Saturday morning will see some locations with lows in the 50's with highs by the end of the week in the pleasant 70's. Best chances of any rain will be Monday night through Wednesday