TheClimateChanger

Icy on my back patio. Watch for slick spots tonight.

Fortunately, it's the UKMET, so it's probably way overdone.

Maybe I'm wrong to call it "rigged" but it just so happens that this incontrovertible fact favors the spin that the original poster wanted. And Grok even cited climate as an example, I didn't even bring it up in my query. All I'm saying is that explains a significant portion of why records - both highs and lows - tail off later in the dataset. Of course, lows are dropping more rapidly than highs because the mean is not constant, but rather is slowly rising. As the famous saying goes - often attributed either to Benjamin Disraeli or Mark Twain - "there are lies, damned lies and then there are statistics."

I just knew someone would call me out for simply stated an incontrovertible fact. Run a random number generator with a certain degree of random variability around a, more or less, constant mean. If you credit the record (i.e., highest and lowest value) only to the first occurrence, more "records" will occur early in the dataset. This is simple mathematics/statistics - no tricks.

Data not truncated, it's from the perspective of today (i.e., records still in existence). A tied record value is still a record value. Regardless, what I said is true, showing only the first occurrence necessarily favors more records in the earlier years. That is not always the convention... NOAA lists all years with a record high in its climate data, typically emphasizing the most recent occurrence where multiple ties have occurred.

And these are the corresponding temperatures at hour 84. Should be good SWE ratios, I would think.

NAM looking very juicy at 84.

@GaWx Revisiting December's anomaly, I think the biggest surprise is UAH's satellite record for lower tropospheric temperatures had December 2025 as the SECOND most anomalously warm month on record in the satellite era. Again, not second warmest December, but second warmest of any month for both US48 & US49 [Conus + AK]. For CONUS (US48), it was behind only March 2012 & for US49, it was behind only February 2017. Since the satellite measures across a certain depth of the atmosphere, it makes me wonder if surface inversions and the like prevented us from fully realizing just how warm it was last month at the surface (2m temps). I know California was plagued by surface inversions for much of the month, so that might be a big part of the divergence between the surface record and the LT anomalies.

iPhone app is up to 16-20 inches, 15-18 of which is forecast to fall on Sunday.

Yes, there must be a 10% chance of something other than one of those three scenarios occurring. Maybe a few percent chance of a cutter, and the balance on a northward curve along the coast that is too far east to produce measurable snowfall (i.e., hybrid of cases 1 & 2).

Some good information from the AFD:

I don't agree with you all that often, but this is very true. Ratio moving rapidly in the direction one would expect. Although I must note that the chart itself is also "rigged" in a few ways to produce an incomplete picture. I'm assuming, of course, it's accurate. Who knows about station selection, have to assume all stations meeting the criteria were selected. (1) Martz only posts the first time a record was set and ignores later ties. This, of course, produces a bias for both record highs and lows towards the early years - which we can plainly see. The dropoff in highs (and, to an extent, lows) would not be as dramatic otherwise. (2) The data is unadjusted. Most of the GHCN stations switched from afternoon/evening observation times (5/6 pm) to morning observation times (~7 am), which is the time in which the instruments would have been manually reset. While maybe not a huge factor, this favors occasional "double" counting of record highs up through the mid 20th century and occasional "double" counting of record lows thereafter. (3) Unclear what direction this bias goes. But it should be noted that a more complete picture would be rendered by indicating the percentage of total data available for each year. The criteria was for stations that reported at least 85% data for 100 years - but where is the missing data. If there is station dropoff towards the end, then that also contributes to the general "softening" of the records. Like I said, this information wasn't presented so I can't so for sure what effect, if any, this has. But I can say for certain that presenting that data would be helpful for presentation. Regardless of the rigging, the data is fully consistent with a warming climate - record highs are clearly outpacing record lows. The apparent loss of extremes in both directions from the earlier decades is largely (maybe not entirely, but significantly) affected by the convention to list only the first occurrence of a record high or low. That choice necessarily will produce a dynamic where earlier decades show more record-breaking temperatures than later decades.

Most anomalously warm months for the CONUS, per UAH lower tropospheric data [expressed as deviation from 1991-2020 mean] (1) March 2012: 2.24°C (2) December 2025: 2.10°C (3) February 2017: 2.06°C (4) April 1981: 1.97°C (5) October 2024: 1.89°C (6) November 1999: 1.87°C (7) November 2016: 1.84°C (8) December 2021: 1.81°C (9) May 2018: 1.75°C (10) March 2007: 1.57°C

UAH satellite temperature record confirms incredible lower tropospheric US warmth in December 2025. (1) CONUS average a whopping +2.10°C above the 1991-2020 mean, which was the SECOND warmest of any month on record (since December 1978). Only March 2012's +2.24°C was warmer. That month is sometimes referred to as "Morch" of 2012, as a portmanteau of March and torch. (2) CONUS + Alaska average checked in at an incredible +1.77°C. This was also the second highest value of record, being narrowly edged by the incredible torch of February 2017 ("Torchuary") which came in at +1.80°C. This one is even more incredible given the noteworthy cold in parts of Alaska. Source: nsstc.uah.edu/data/msu/v6.1/tlt/uahncdc_lt_6.1.txt

I just figured I'd take a look and see what it shows for Pittsburgh, after seeing Clay Travis's post on X/Twitter saying it was showing 15-18" for Nashville on Saturday.

Just took a peek at the iPhone app, and it says 9-11" on Sunday for Pittsburgh? Anyone seeing this?

CPC nails it again.

Eyeballing, but it looks like close to an inch of snow last night. Didn’t realize it was supposed to snow.

Hat tip to user @PhiEaglesfan712 for noting this over in the ENSO thread. This proves definitively that this has been, by far, the worst decade ever to be a Mid Atlantic snow weenie.

As we approach the 10-year anniversary of the Snowmageddon of 2016, the 10-year moving sum of snowfall is set to absolutely plummet to levels never seen before. Effective January 23, 2026 (covering January 24, 2016 to January 23, 2016), the moving 10-year total snowfall observed at Washington, DC and Baltimore, MD is set to plummet to 75.6" and 95.3" respectively. This will obliterate the prior records of 92.8" and 123.1" respectively, which occurred from the last year on all dates between the last snowfall in February and the first snowfall in December.

I think the most surprising thing is even with that 30" + storm on the books, they've been straddling the record low 10-year snow. That's why once it drops off, it doesn't just a set a new record low - it absolutely obliterates the prior record.

I suppose the Baltimore tally which is set to obliterate the record by nearly 30" is probably even more unbreakable. Really bad time to be a Mid-Atlantic snow weenie - that's for sure!

I confirmed manually that no additional snows fell in January 2016 after snowmaggedon at any of the sites, so these values would be the rolling 10-year sum [even on a daily resolution] effective January 23, 2026 (covering January 24, 2016 to January 23, 2026). A rolling 10-year sum of 75.6" [equivalent to an average of just 7.56" each 12 months] at DC, assuming no measurable snow through Friday, seems like it would be difficult to ever break. Maybe an unbreakable record?

And at Philadelphia. The record sorted by month end is 144.1" from March 1, 1922 to February 29, 1932. Last year saw the second lowest at 146.1 inches. Looks like 15.6" is needed to avoid the moving 120-month total record low. Not sure about 10-year daily moving totals - probably similar to these. Almost certainly, the prior record belongs to one of these periods, but might be a little bit lower than these totals, since it's on a daily metric (and this analysis is limited to discrete days - end of each month).

For Baltimore, the rolling 120-month total is currently at 95.3" (February 1, 2016 to January 18, 2026). The record is 123.1" for every month from March 31, 2025 to November 30, 2025. Baltimore would need nearly 30" of snow the rest of the month to avoid a new record.