2025-2026 ENSO

[Stormchaserchuck1]

Atlantic (NAO) has slight inverse correlation from October to the Winter

Pacific has a pretty strong direct correlation from October to the Winter, a lot of that is around the PDO, which is strongest in Oct out of PNA/ENSO. 

[donsutherland1]

59 minutes ago, michsnowfreak said:

They are very biased towards enso and also biased towards a warmer climate. That is why periods of cold are never seen far out, you will see them grow colder as the timeframe nears. 

I suspect that its seeming inability to see cold anomalies except at shorter timescales has a lot to do with the idea that boundary conditions drive seasonal averages. ENSO, PDO are prominent conditions. The oceans overall are warming. Therefore, the model forecasts are tipped toward the warmer idea more broadly than is realistic. Worse, the coefficients of determination for such variables related to boundary conditions and actual seasonal outcomes are very low. 

These weak relationships reveal that other important factors are involved, including synoptic scale events that cannot be reliably forecast beyond 10-14 days. Some of these additional variables may not yet be known. Synoptic scale events i.e., large snowstorms, Arctic blasts, etc., can have a great influence on the overall seasonal outcomes. Thus, even a warm winter can be much snowier than normal or a cold winter can lack snowfall. On account of these other variables, every La Niña or El Niño event is not alike.

The seasonal models are not yet at a stage where they can even begin to consistently resolve the actual events that ultimately produce the seasonal outcome. A similar situation applies to subseasonal forecasting. Not surprisingly, beyond two weeks, model skill on the weekly guidance largely disappears. There also seems to be a larger deal of persistence in the two week or longer forecasts than what actually occurs.

AI may improve some of these outcomes. But even then, big challenges could still persist. For example, even as some experiments with random forest models have shown a degree of improved skill in forecasting ENSO, those models are constrained by their knowledge base. Hence, when it comes to forecasting extreme events e.g., super El Niño events, they have great difficulty.

Perhaps the combination of AI and quantum computing might produce some significant breakthroughs. But that's still in the future and perhaps a decade or more away, assuming society values science and basic research to make the investments necessary to arrive at that improved state of forecasting. That's an open question in some areas and it will become even more relevant as major states grapple with the costs of aging populations, rising debt relative to GDP, etc., and the trade-offs involved in making budget allocations.

[PhiEaglesfan712]

4 hours ago, roardog said:

I’ll take Nina over Nino every time around here in the winter. Hopefully we do have a “warm” October. It feels like a lot of times a cold October turns into a mild December. Although, that’s  probably due to a cold October being more common in a Nino. A cold October is kind of useless anyway except for chasing the first flakes of the season or some sloppy early season slush accumulation that melts in a few hours. 

Classic case is 2011. That was a cold October, with a snowstorm at the end of the month. Pattern flipped soon thereafter, and it was blowtorch from November through March.

[Mercurial]

9 hours ago, roardog said:

I’ll take Nina over Nino every time around here in the winter. Hopefully we do have a “warm” October. It feels like a lot of times a cold October turns into a mild December. Although, that’s  probably due to a cold October being more common in a Nino. A cold October is kind of useless anyway except for chasing the first flakes of the season or some sloppy early season slush accumulation that melts in a few hours. 

same