GaWx

^How well did Jamstec’s May forecast for JJA verify in the CONUS? May ‘26 Jamstec forecast for summer ‘26: Actual for summer ‘26: Jamstec forecast excellent in E US; overall CONUS pattern excellent with warmest in SW US and coolest Midwest although it wasn’t quite warm enough in sone of the SW US:

Here’s my analysis of the new Jamstec forecast for winter: hint SE cold lovers will enjoy this as it is progging the coldest winter in the SE/Mid-South since 2014-5!

-Sep Jamstec winter forecasts going back to 2017-8 are available at the site to compare to -Sep ‘26 is vs all Sep forecasts back to ‘17 by a good margin the coldest DJF in the CONUS overall, by a very wide margin the coldest DJF in the E 1/2 overall, and is easily the coldest for the SE, midsouth, S Plains, Ohio Valley, and Mid-Atlantic. I couldn’t even find a colder one than this for the corridor from Chic to NYC. -Sep ‘26 is the wettest overall for the E US vs the other Sep DJF forecasts -June-Sep ‘26 are all very similar in the CONUS for DJF ‘26-7 and are the most consistent of any DJF forecasts back to 2017-8! Here’s the ‘23 Sep forecast for DJF for comparison (the 2nd coldest since ‘17): coldest is -0.5 to -1C: Compare that to ‘26: like night and day with coldest -2 to -3C! This would easily be the coldest winter for the SE, Mid-South, and S Plains since 2014-5. Also, note how much colder is the Arctic vs ‘23-4. Keep in mind that the Arctic N of 80N from late May through summer was easily the coldest since 1958 for unknown reasons: How did Sep JAMSTEC verify in the E US overall? ’17: too warm NE/close SE ’18: close NE/too cold SE ’19: too cold ’20: close NE/too warm SE ’21: N/A ’22: too cold ’23: too cold ’24: too warm ’25: too warm So, for these 8 winters, Jamstec: -in NE was too warm 3 times, too cold 3 times, and close twice…so neutral bias averaged out -in SE was too warm 3 times, too cold 4 times, and close once…so, very slight cold bias but pretty neutral overall averaged out -Keep in mind that this cold E US forecast is with Jamstec progging an E based super Nino.

I and others have been posting what the model trend has been for later this autumn for the various Nino regions (clearly downward), but am not suggesting anyone “trust” what the CFSv2 (or any model for that matter) shows since model progs sometimes verify poorly. Regardless of model imperfections though, a good portion of this and other forecasting discussion threads typically centers around the trends of model output even though the output isn’t typically “trusted”. Models are very useful tools even though they’re not at all crystal balls and thus can bust badly. I’ve found over the many years that I’ve been following them that model prog trends are quite worthy of discussion since they often give good hints on what’s actually liable to occur down the road regardless of their imperfections. A good example of another current forecasting discussion thread that follows model trends is this one in the Mid-Atlantic forum that I and some others here also post in and which I recommend others check out:

Today’s weekly relative update (which has been NOAA’s official way of measuring these for quite awhile now) for last week cooled in Nino 1+2 from +3.9 to +3.7 and in Nino 4 from +0.1 to -0.1. Meanwhile, Nino 3 warmed from +2.7 to +2.8 while Nino 3.4 warmed from +1.8 to +2.0: 02SEP2026 3.9 2.7 1.8 0.1 09SEP2026 3.7 2.8 2.0 -0.1 This is the comparable week in 1997: 10SEP1997 3.8 2.6 1.8 0.5 This tells me that 2026 remains more E based than 1997 and was thus for another week the most E based on record for this time of year.

Today’s SOI is at ~-26. That’s the 33rd straight -SOI. The 30/90 day averages are at -17.5/-20.3: https://www.longpaddock.qld.gov.au/soi/

The AAM yesterday finally for the first time this month exceeded an impressive +3 per the following image after several early Sep runs had had that occurring by Sep 7-8th. The delay is typical due to a + bias of the CFSv2. This is saying it will hit an extreme level of +4+ in only ~5 days (~Sep 19th), which is the earliest in its forecast that I’ve ever seen it reach it, and that it will stay +4+ at least through Sep 24th. I’ll keep following these graphs to see if that actually occurs:

It was initialized at +4.6 for the full Sept averaged out. The mean then has +3.9 for Oct averaged out. There’s no requirement that the highly volatile 1+2 fall immediately for both of these to verify well.

Why do you say it’s initialized too low?

1+2 also dropping:

That was an earlier update. Check out the latest: down to +2.9 peak/blues still lower than reds thru NDJ:

Why is it overwhelming the signal on RMM now but in 97-8 it was nearly opposite (majority in MC/IO)?

And here’s the 2 week fcast below. Does anyone know why the MJO is avoiding the IO & MC? I looked at the other 16 Ninos since 76/7 and found only two others only somewhat similar in JAS: 2015: still had good # in IO 1994: still had large # in MC

The MJO continues to avoid the Maritime C and IO like no other time on record:

Per this the min of 3.6 may have already been hit. If so, that’s well above the worst: