Three Scenarios for Winter Gas When a Historic El Nino Meets Near-Record Storage

The seasonal signal that will shape North American natural gas pricing this winter is already on the table, and it is unusually loud. The Climate Prediction Center's ENSO diagnostic discussion issued 13 August 2026 carries an El Niño Advisory, places the July Niño-3.4 anomaly at +1.4°C, and assigns a greater than 90% chance of a very strong event during the Northern Hemisphere fall and winter of 2026-27. For the October–December 2026 season it goes further, giving a 69% chance of an event that would exceed the strength of every El Niño back to 1950, defined there as a three-month RONI value of +2.5°C or more.

The familiar framing says a signal that strong should reprice the winter curve weeks before the first heating degree day is recorded. The August 2026 data complicates that. Lower 48 working gas stood at 3,153 Bcf for the week ending 7 August 2026, and the Short-Term Energy Outlook released 11 August 2026 expects Henry Hub to average $2.87 per MMBtu in the third quarter and to stay below $3.00 in the coming months, attributing that to near record-high storage heading into October. Marketed production is forecast at 122.5 Bcf/d for 2026, above the 118.5 Bcf/d record set in 2025. When the cushion is that thick and the tilt that widely published, the marginal price information in the next outlook update is smaller than the headline implies.

The August Numbers, Before Any Interpretation

Every argument below rests on a small set of published figures, separated here so the narrative can be checked against them later.

Input	Latest published value	Source and date
ENSO state	El Niño Advisory; July Niño-3.4 at +1.4°C	CPC, 13 Aug 2026
Event strength odds	>90% very strong for fall/winter 2026-27; 69% chance of a record event in Oct–Dec	CPC, 13 Aug 2026
Lower 48 working gas	3,153 Bcf	EIA weekly, 7 Aug 2026
Regional storage split	S. Central 1,093; Midwest 829; East 693; Pacific 300; Mountain 239 Bcf	EIA weekly, 7 Aug 2026
Henry Hub outlook	$2.87/MMBtu average in 3Q26; below $3.00 in coming months	EIA STEO, 11 Aug 2026
LNG exports	16.5 Bcf/d average in 3Q26	EIA Short-Term Energy Outlook, 11 Aug 2026
Marketed production	122.5 Bcf/d forecast for 2026 vs. 118.5 Bcf/d record in 2025	EIA, 12 Aug 2026

Two structural facts frame them. The withdrawal season runs November through March. And heating is a minority of consumption: in 2023 the residential sector was 14% of U.S. natural gas use and the commercial sector 10%, against 40% for electric power and 32% for industry, with roughly 60% of American homes using gas for heating, cooking or drying. Winter moves about a quarter of the demand stack directly, plus an indirect share through power burn.

How a Seasonal Outlook Becomes a Price, Step by Step

Step one: the outlook is probabilistic and regional, not a temperature

The Climate Prediction Center publishes thirteen overlapping three-month outlooks at a time, updated monthly on the third Thursday; the most recent full set was issued 16 July 2026. Each map expresses a tilt in the odds toward the upper or lower third of a region's historical distribution. It states no expected temperature and no national number. A tilt toward above-normal temperatures across the northern tier is a statement about probability mass, not a forecast that any particular January will be warm in Chicago.

That distinction is where the misreading happens. A headline saying a strong El Niño "means a mild winter" collapses a probability shift into a deterministic claim. The published outlook does not support that, and the variance it leaves unresolved is wide enough to contain the entire range of price outcomes below.

Step two: degree days convert weather into a demand number

Planners do not trade temperature anomalies. They convert forecast temperature into heating degree days, aggregate them across a population-weighted service territory, and translate that into expected send-out. This is where a weather map becomes a volume, and where regional detail dominates: the same map carries opposite implications for a New England distribution utility and a Gulf Coast industrial customer.

Step three: the storage buffer decides whether the demand number matters

A demand estimate reprices the curve only if the system cannot absorb it. That is the function of the 3,153 Bcf figure: storage is the shock absorber between a forecast and a price, and its size sets how much forecast error reaches the front month. The August outlook makes the chain explicit when it attributes sub-$3.00 pricing to near record-high storage heading into October rather than to any weather expectation.

The implication runs against the intuitive version. Entering a winter with a thin cushion, a small seasonal revision moves the curve a lot, because it changes the probability the system runs short. Entering with a thick cushion and record production, the same revision moves it very little, because it does not change that probability at all. The forecast has not become less accurate. It has become less informative to price, which is a different thing.

Scenario A: The Cushion Holds and the Forecast Never Gets Paid

Here the seasonal tilt broadly verifies across the northern tier, November-through-March withdrawals run at or below a normal-winter pace, and the front month stays inside a narrow band near the sub-$3.00 zone the August outlook describes. The forecast turns out correct and worth almost nothing, because the information was public in August and the storage position made it non-binding.

What would confirm it: end-of-October working gas near the top of the recent historical range; November and December withdrawals that fail to close the gap to the five-year band; production tracking toward 122.5 Bcf/d rather than short of it. What would break it: a production disappointment, or freeze-off losses in Appalachian and Permian supply concentrated in the same weeks as elevated load.

The uncomfortable part is that this is the branch in which the seasonal narrative is correct. Being right about the weather and wrong about the price is the standard outcome when a well-supplied system meets a published outlook.

Scenario B: A Cold Deviation Inside an Otherwise Mild Season

This is the branch a seasonal map is structurally incapable of ruling out. A three-month outlook expresses a tilt in the mean and says nothing about variance around it, and cold outbreaks are a variance phenomenon. A season can finish above normal on the three-month average and still contain a five-day stretch that sets regional send-out records.

The asymmetry is that this branch shows up in basis and deliverability, not in the national storage balance. The 7 August 2026 regional split is the reason: South Central held 1,093 Bcf and the Midwest 829 Bcf, but the Pacific region held 300 Bcf and the Mountain region 239 Bcf. A national total of 3,153 Bcf does not help a constrained region during a five-day event, because gas in a South Central salt dome cannot reach a New England city gate on peak day at any price.

Triggers to watch: a stratospheric disruption signal in operational guidance during the December-to-February window; a ten-to-fifteen-day forecast shifting colder with no change in the monthly or seasonal outlook, the configuration that marks a variance event rather than a mean shift; regional pipeline maintenance announced ahead of withdrawal season. What would break it: substantial rebuilds in the constrained Pacific and Mountain inventories before November, which would remove the deliverability constraint this branch depends on.

Scenario C: The Weather Stops Being the Marginal Variable

The third branch is the one the winter-forecast frame ignores. Export demand is now large enough to compete with weather for control of the balance: LNG exports averaged 16.5 Bcf/d in the third quarter of 2026 on the August outlook, a volume comparable to swings in heating load, and unlike heating load it does not respond to a warm January.

Here the cushion drains for reasons unrelated to temperature: a terminal ramping faster than assumed, an unplanned outage reversing the effect, a production plateau below the 122.5 Bcf/d path, or a change in pipeline exports to Mexico. The outlook could verify perfectly and still be irrelevant to price, because the marginal molecule was allocated by an export decision rather than a thermostat.

Triggers to watch: revisions to the export figure in successive Short-Term Energy Outlook editions; feedgas deliveries diverging from nameplate assumptions; the production forecast moving in either direction between the August and September editions. What would break it: export demand tracking the published path, which returns the balance to weather control and collapses this branch into A or B.

A Historical Analogue With Actual Numbers

December 2023 in Southern California is a cleaner case than the usual national cold events, because all three mechanisms operate at once and the figures are published. The SoCal Citygate averaged $3.61 per MMBtu that month, the lowest December level since 2015 after adjusting for inflation. California recorded 140 fewer heating degree days than in December 2022, and residential and commercial consumption fell 21% year over year. California storage held 249.3 Bcf entering December 2023, the most since 2020, when it held 276.1 Bcf. Entering December 2022 the same region held 185.5 Bcf, the lowest since 1997, and that winter priced very differently.

The same regional demand shortfall produced a mild price outcome in one year and would have produced a severe one in the other. The difference was the storage position and the pipeline configuration, not the weather forecast.

Two non-weather factors compounded that outcome and illustrate Scenario C in miniature. California hydroelectric generation rose 81% in 2023 against 2022 after the atmospheric river sequence, displacing gas-fired power burn, and the return to service of El Paso Natural Gas Line 2000 in February 2023 improved access to lower-priced supply. Neither appears on a temperature map.

What This Cannot Predict

The argument above says the seasonal outlook carries less marginal price information than usual because the cushion is thick. Several readings would invalidate it.

• If the cushion is smaller than the headline implies. The 3,153 Bcf figure is a national total for one week. If much of that inventory sits in regions or facility types that cannot deliver at high rates on peak day, the effective buffer is smaller than the number suggests and seasonal revisions regain their leverage. A national aggregate is a weak proxy for peak-day adequacy.
• If a record-strength El Niño behaves differently from a merely strong one. The 13 August discussion flags a 69% chance of an event exceeding anything since 1950. The composites used to translate El Niño into a regional temperature tilt come from a small sample, none of it at that strength. Extrapolating a composite beyond its observed range fails without warning.
• If price response is convex rather than proportional. The claim that a thick cushion damps forecast sensitivity assumes a roughly continuous demand-to-price relationship. Below a threshold of perceived adequacy the response can turn convex, producing almost no reaction to successive revisions and then a large one, with no intermediate stage that would have signalled the change.
• If production misses the forecast path. The cushion argument leans on the 122.5 Bcf/d marketed production forecast, and production forecasts get revised. A downward revision of a few percent would change the framing materially, through a channel unrelated to any weather outlook.

One item is simply unresolved. The link between event strength and North American regional outcomes at the top of the ENSO range is not well constrained by the observational record. Whether a historic-strength event produces a proportionally stronger regional signal, or saturates, is not known.

What to Watch Next Week

1. The Climate Prediction Center outlook update, third Thursday of the month. The most recent full set of thirteen three-month outlooks was issued 16 July 2026. Check not the headline tilt but whether the November–January and December–February panels shift in the same direction as the ENSO strength revision. A strengthening event that leaves the regional panels unmoved means the models are not converting strength into a proportional regional response.
2. The weekly EIA storage report, released Thursdays. The 3,153 Bcf print for the week ending 7 August 2026 was released 13 August, with the next on 20 August. Between now and 31 October the number that matters is injection pace, because the end-of-season level sets the starting condition for every branch above.
3. Regional injection detail rather than the national total. Pacific at 300 Bcf and Mountain at 239 Bcf are the constrained positions, and whether they close the gap before November decides whether Scenario B has a price channel at all.
4. The next Short-Term Energy Outlook, scheduled for 9 September 2026. Three lines are worth reading against the August edition: the Henry Hub path relative to $2.87 and $3.00, the LNG export figure relative to 16.5 Bcf/d, and any revision to the 122.5 Bcf/d production forecast.
5. The gap between the ten-to-fifteen-day forecast and the monthly outlook. A short-horizon forecast turning sharply colder while the monthly and seasonal outlooks hold still is the pattern that distinguishes a variance event from a mean shift, and the pattern under which Scenario B pricing appears.

Concrete Framework — Where to Look First

A monitoring checklist for the rest of the injection season and the winter that follows — a way of tracking which branch the data is selecting, not a set of positions.

1. Log the end-of-injection level on 31 October, not before. Weekly prints until then are noise around a trajectory. The number that conditions every later reading is the level at which withdrawal season starts.
2. Track the five regions separately against their own histories. The 7 August 2026 baseline is East 693, Midwest 829, Mountain 239, Pacific 300, South Central 1,093 Bcf. A comfortable national total masking one lagging region is the precondition for a basis event.
3. Separate the two forecast horizons. Monthly and seasonal outlooks in one column, ten-to-fifteen-day guidance in another. Never average them; divergence between the columns is itself the signal.
4. Read the ENSO discussion twice, once for strength and once for regional language. The 13 August 2026 edition gave >90% for a very strong event and 69% for a record one in October–December. Track whether later editions move those numbers, and separately whether the regional impact language changes. The two need not move together.
5. Check three STEO lines each month. Henry Hub against $2.87 and $3.00, LNG exports against 16.5 Bcf/d, production against 122.5 Bcf/d. A revision in the latter two is a Scenario C signal regardless of the weather.
6. Write down the falsifier before the season starts. Record in advance, for each branch, the observation that would rule it out. A scenario framework with no pre-committed falsifier degrades into narrative by January.
7. Date every figure. Storage, price and outlook numbers all revise. A framework carrying undated numbers into December is reasoning from the August world without noticing.

A loud seasonal signal and a large price move are different things, and 2026-27 may be a season in which the first arrives without the second. The forecast is not the story. The gap between it and what the system can absorb is.

Disclaimer: Macro and market-structure analysis, not investment or financial advice. Scenario weights are illustrative framings of uncertainty, not forecasts. Figures cited were published as of mid-August 2026 and are subject to revision.