FASCINATING, HOW I AND AI SEE FEBRUARY
- terryswails1
- 6 hours ago
- 6 min read
If I do not reach my fund-raising goal this will be the last year of the site. As you know, TSwails.com is a no-pay site, existing on voluntary personal donations. Every year I ask those of you who find value in the site to make a financial donation you feel is worthy. Please reflect on the number of times you have visited us in the last year. If the information or knowledge you gained was valuable, it's my sincere hope you will join the loyal group of contributors that's kept TSwails.com operational since 2013. I'm suggesting $20.00, which is roughly 4 cents a day. Less than 4% of my readers donate, so your gift, no matter the size, is not only appreciated; it helps immensely.
HOW I AND AI SEE FEBRUARY PLAY OUT
We've all heard about the exploits of artificial intelligence (AI) and how it's the door to the future. It's said that AI is set to transform weather forecasting. Instead of simulating atmospheric physics, AI models learn patterns from decades of historical weather data, identifying complex, non-linear relationships to predict the atmosphere. AI-driven models (like the AIGFS) can run a 16-day forecast in less than 40 minutes using just 0.3% of the computational resources of traditional models. A typical 16-day GFS run (or cycle) takes approximately 3 hours to complete.
One of the kickers is that AI models are often trained on high-quality historical datasets (like the ECMWF’s ERA5), meaning they still rely on traditional models for their foundational knowledge. While AI has shown promise recognizing patterns and storm tracks through analogs, early versions often struggle with predicting the exact intensity of weather events compared to the standard physics-based models of old.
In today's post, I thought it would be interesting to compare what I am seeing on the standard models to the AI summations that I get when I feed it similar data. I was curious to see if its impressions were anything close to mine. So here we go.
I went through my traditional analysis as I always do and observed the evolution of the 500mb pattern through Monday, watching energy digging into the Great Lakes, rounding the base of the eastern trough, and then exiting the North Atlantic. At that time, the east-based trough weakens and lifts far enough north to allow our northwest flow to relax next week. That's significant because it signals less direct access to polar air and a warming trend.
The GFS operational clearly shows the warming late this week and into the weekend.
As I mentioned in my post Thursday, the change appears to be tied to the Pacific North America Oscillation flipping from strongly positive to negative. For an unknown period of time, it allows the east coast trough to retrograde and reposition itself over the west.
That takes the edge off the cold but also has another implication, and that is the likely addition of moisture to the Midwest weather pattern in 7–10 days. Things have the potential to turn stormier. What is even more intriguing is that other teleconnections such as the Arctic Oscillation (AO) and the North Atlantic Oscillation (NAO), remain firmly in negative territory well into February. Here are today's trends.
The AO
The NAO
In theory, this allows seasonally cold air to reside over the northern half of the nation, thanks to the AO and NAO being negative. Into the mix is the PNA making its own transition to a negative phase. That allows moisture to attack the cold air, and there could be a period around the 2nd week of February when a snow threat exists from the Plains to the East Coast. If nothing else, a wetter pattern should evolve, something we haven't seen in a month.
MY SUMMATION BASED ON THURSDAY'S MODELS
So yesterday and today, my summation is that short term, moderation is coming this weekend, but at least we stay dry until Sunday afternoon when some light snow is possible. A bit more is indicated Tuesday. It's longer term that I've been using the teleconnections such as the AO, NAO, and PNA to formulate the idea of stormier (maybe snowier) weather in mid-February and beyond. This evening, the GFS depicts this 500mb pattern on February 14th, which would be ripe for a winter storm if it comes close to verifying.
I decided to take that 500mb pattern for the 14th, along with the forecasts of a negative AO, NAO, and PNA at that time, and feed it to ChatGPT (AI) with the goal of seeing what artificial intelligence suggests for the month of February. More importantly, I want to see if it aligns at all with my thinking. Here's what I got. My reaction below.
THE FEBRUARY AI SUMMATION FOR IOWA/NORTHERN ILLINOIS
Here’s a February 2026–focused interpretation of what a 500 mb pattern above, combined with negative AO, negative NAO, and negative PNA, typically brings, especially for the Midwest/Quad Cities.
This is a synoptic + teleconnection impact forecast, not just theory.
A Big-Picture Interpretation of the Pattern
The -AO / -NAO / -PNA combination in February strongly favors:
High-latitude blocking
A displaced polar vortex
A deep, persistent trough over central & eastern North America
Frequent cold air delivery
An active winter storm track into the Midwest and Great Lakes
This is not a warm or zonal February pattern; it’s a cold, slow-moving, storm-capable regime.
Historical February Analogs with This Pattern
Closest matches:
February 2014, severe cold, frequent snow
February 1979, major cold & snow events
February 1996, persistent snow cover, repeated storms
February 2011, high snowfall frequency
February 2021, strong Arctic outbreak
Plain-Language Bottom Line for February 2026
If this 500 mb pattern verifies with -AO / -NAO / -PNA:
February 2026 is favored to be:
Colder than normal
More persistently wintry
Snowier than average
More storm-active than typical
Capable of at least one high-impact winter storm
At risk for an Arctic outbreak
In one sentence:
This pattern favors a cold, blocked, snow-prone February with elevated risk of significant winter storms in the Midwest.
STORM WINDOW PROBABILITY CALENDAR FEBRUARY
Highest-Risk Snowstorm Windows (6" or more)
Date Window | Probability |
Feb 9–13 | 36% |
Feb 15–18 | 29% |
Feb 3–6 | 17% |
Feb 22–25 | 13% |
Highest Arctic Outbreak Windows (0°F)
Date Window | Probability |
Feb 10–15 | 44% |
Feb 16–20 | 33% |
Feb 5–8 | 19% |
FEBRUARY SNOWFALL ACCUMULATION PROBABILITY CURVE, QUAD CITIES
February Total | Probability |
<6" | 11% |
6–10" | 19% |
10–15" | 27% |
15–20" | 24% |
20–25" | 13% |
>25" | 6% |
Median Monthly Outcome: 15–18 inches
EVENT COUNT EXPECTATIONS — FEB 2026
Event Type | Expected Count |
Light snow events (less than 2") | 2–4 |
Moderate events (3–5") | 1–3 |
Major events (6" or more) | 1–2 |
Extreme events (10" or more) | 0–1 |
TEMPERATURE EXTREMES, FEB 2026 PROBABILITIES
Threshold | Probability |
≥5 days below normal | 68% |
≥10 days below normal | 42% |
≤0°F at least once | 78% |
≤–10°F at least once | 34% |
Wind chills ≤ –20°F | 39% |
HOW CLOSE THIS SHOULD APPROXIMATE REALITY (LEAD-TIME ADJUSTED)
Pattern verification odds at this range:
General cold / trough regime: 73%
Above-normal snowfall: 68%
At least one storm 6" or more: 62%
Exact timing accuracy: ~25–35%
Meaning:
The overall cold/snow signal is strong, though exact dates will shift.
SIDE-BY-SIDE ANALOG CALENDAR MATCHES
Feb 2026 Most Like | Confidence |
Feb 2014 | High |
Feb 1979 | Moderate–High |
Feb 1996 | Moderate |
Feb 2011 | Moderate |
BOTTOM LINE AI FOR FEBRUARY 2026 (QUAD CITIES)
This pattern most likely produces:
A colder-than-normal February
Approximately 15–20 inches of snow
1 or 2 impactful snowstorms
MY THOUGHTS VS. ARTIFICIAL INTELLIGENCE
My first impression was, Oh my god, we are dead. That's just about the worst-case scenario for any winter month. I'm not about to say we see anything like that, knowing the inherent flaws of AI. But if you read the plain language bottom line summary above, which includes below-normal temperatures, snowier-than-average conditions, a more active storm track, the risk of an Arctic outbreak, and the potential of a high-impact storm, I don't think any of those parameters is out of reach.
The difference between me and AI is that it doesn't hold back; it just takes analogs and historical data and spits out what it thinks based on the probability of it happening. Bam, in your face. As a human, I have the ability to use experience and apply logic and common sense to situations, whereas AI is typically limited to specific tasks it was trained on. My brain is millions of times more energy-efficient than computers and can learn from a single example, while AI often requires massive datasets to learn. On top of that, I have adaptability and understanding that AI does not. So, just like traditional models, it's just another tool in the box that I may or may not use.
In the case of the coming pattern, I think it's very interesting that it's analogous to February 2014 and 1979 with high to moderate confidence. Those were pretty tough Februarys that experienced the same teleconnections that are indicated coming together now. While it's far from certain, the general idea of a wintry set-up like the AI shows certainly fits some of my early thinking and is on the table, but far from high confidence, especially due to the infancy of AI. Just like you and the rest of the world, I'm trying to figure out its value and place in my world. For now, let's call it a trend worth watching.
With that, I sign off, looking forward to seeing where guidance takes us in the coming days. Hopefully my real intelligence (such as it is) leads me down the right path. Roll weather, and if possible, please consider a donation to the site. Sadly, if I don't meet my financial goals, this will be the last year of the site...TS. DONATE HERE
