I've updated a new project to get a forecast of the spatial anomalies using a mix of ERA5 and the superensemble for the rest of the month (for reference, the super ensemble is weighted as I usually do, and regridded cells (bilinear interpolated to the same grid as ERA5) are adjusted from the superensemble calculations (at least for this run) and is on the order of +0.1 to +0.15 C (usually closer to 0.1 C -- since EPS is the largest weight and is usually biased globally on the cold side). It's fairly speculative given its nature but especially so since the March forecast at the end of the month is the hardest for the dynamic models...

The main reason for doing so is the following: I checked the latest GHCNm today and they still have not revised/fixed NW Russia. For last month (February data) we were missing stations from Vologda Oblast, Arkhangelsk Oblast (including Nenetsia and Novaya Zemlya, but not Franz Josef Island), and the Komi Republic. This time however most of the cold anomaly is over the Komi Republic, Perm Krai, Tomsk Oblast, most of Krasnoyarsk Krai, and Novaya Zemlya.

latest GHCNm for Feb (filtered by stations actually included in the gistemp calculations):

Overlaying them and zooming in gives a better picture of what might be infilled in and how in that scenario..

It looks like other than Novaya Zemlya (and its nearby ocean subboxes) (which would amount to ~ 5 subboxes?) might be being infilled a tiny bit on the warm side (there is another island station we got last month that is in the cold blob so it likely won't be too far off), and so it shouldn't be a large issue like last month. The west coast of Greenland I looked at also closely and where the coastal stations are should also be fairly representative this time (last month they were warmer than most of Greenland).

So based on this I don't see any particular reason to expect any other further biases if we end up with the same missing station set as in last month's scenario.

~

There is a small divergence between the two notebooks I rely on but the center point prediction between them is 1.211 and 1.196 unadjusted (and a usual +0.01 for own past forecast errors), suggests around 1.21 as a center point. (Based on this the highest bin still looks too high at 38%; if this was the end of the month value it would imply around 22%; the probabilities that I am using at the moment that includes past variance, but with a slightly sharper distribution suggests between 29%-37%, so a mid point of 33%; the ERA5->GISTEMP model I have though still sucks for March.)

...

The steep rise should start showing up in the (prelim copernicus) ERA5 2m temps tonight so the point center predictions might change quite a bit over the next 3 days depending on the residuals.

Compared to yesterday's run, GEPS today across the US has shifted to a more zonal flow for the 28th+, more in line with EPS yesterday&today, and this results in (today's run) relatively less cold air overall for the region at the end of the month.

Left and right is yesterday's run vs today's run single day (28th) snapshot of the super ensemble (note its not a usual daily anomaly for that day but with respect to the whole month's climatological mean):

GEPS yesterday and today's run for the 28th:

This results in a slightly warmer SE US compared to the month's mean for yesterday (and a slightly warmer Atlantic all the way upwards to the Labrador Sea), and a slightly higher rise globally for the end of the month:

MARCH 2026

🌡️ Forecast blended (unadj.):

02/03 - 1.067

03/03 - 1.072

04/03 - 1.112

08/03 - 1.097

09/03 - 1.097
10/03 - 1.093

11/03 - 1.113

12/03 - 1.142 (forecast starting to look reasonable)
13/03 - 1.164

14/03 - 1.160
15/03 - 1.155
16/03 - 1.153
17/03 - 1.170
18/03 - 1.178

19/03 - 1.186

20/03 - 1.199
21/03 - 1.198

✨ Model output [MODEL-a/MODEL-b] (unadj.):
08/03 - 1.167 / 1.175

09/03 - 1.175 / 1.189

10/03 - 1.177 / 1.184

11/03 - 1.199 / 1.198

12/03 - 1.240 / 1.250 (lol wtf)
13/03 - 1.267 / 1.270
14/03 - 1.255 / 1.253

15/03 - 1.240 / 1.243
16/03 - 1.255 / 1.249
17/03 - 1.249 (models merged)
18/03 - 1.246
18/03r2 - 1.242

19/03 - 1.265

20/03 - 1.278
21/03 - 1.276 (oceans temps are climbing fast, El Nino is coming 🔥 https://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso_advisory/ensodisc.pdf)

*Update 10/3*

Still in testing but provisionally forecast a=+0.013, model a=+0.024
Forecast blend is quite unreliable, my algo model will be closer to final figure.
Anything before 17th is a lot of guesswork too.
//new account decided to get my mana up for the swag

UPDATE 14/3
(after adjusting for gistemp release date and station availability at the time of release: forecast is probably safe to be ignored and only used as a rough estimate/referece, model adj +-0.01)

Update 16/3
Something is either seriously wrong with my model or we're getting an unusual high positive ERA5 to GISS variance this month. I will re-run different scenarios later today to see if i can get a lower result at all

Update 16/3 v2
Nope, lowest I got was 1.24

Update 17/3

Forecast tool upgraded, should be more accurate at the start of the month from next month

@zenarxy You'll get 25 mana a day if you just find another market to bet in if you don't want to update daily like I do.

x

@chaitea I still don't have a good ERA5->GISTEMP monthly model for March.

You are quite a bit lower than I am though but this may be all from this model adjustment (I'm blending a couple different statistical and the dynamical+statistical, with only one of the statistical slightly favoring the 1.20-1.25 bin now, but they are all sensitive to the last day's temperature).

Last day of medium range in the super ensemble has increased disagreement today (last day is only weighted GEPS/GEFS so its more sensitive to a warm biases which tends to push up the statistical extrapolation, however ARIMA complicates interpretation: the recent ERA5 forecast error for the unadjusted has been marginally worse (model warmer than observed) so its possible the ARIMA adjustment so far out is a contributor but its also possible GEFS is over doing the China and relatively less cold in east Antarctica -- hard to analyze it visually.