The state of the US season ahead, kept current by machine rather than by memory. NOAA’s Oceanic Niño Index is re-read from the source file on every deploy, the next Climate Prediction Center release date is computed from CPC’s own published rule rather than copied off a list that runs out, and the tropical-cyclone panel reports the feed as it stood at the timestamp beside it.
Nothing on this page is a Drawpie forecast. Every number is either fetched from NOAA or worked out from a rule NOAA states — and the page says which, for each one, and how long ago it last confirmed it. The explainers underneath are where the reading and the judgement live.
Every figure in this section is computed from NOAA PSL's Oceanic Niño Index file, last read at the time stamped above. No value here was typed by a person.
| Season | ONI (°C) |
|---|---|
| DJF 2026 | -0.37 |
| JFM 2026 | -0.14 |
| FMA 2026 | +0.13 |
| MAM 2026 | +0.51 |
| AMJ 2026 | +0.98 |
| MJJ 2026 | +1.40 |
Why the series stops where it does. The index is a three-month running mean, so the season after MJJ 2026 cannot be computed until its last month is complete. That is arithmetic, not a lag in our checking — the last completed season is recomputed from the file itself, so it advances without anyone editing a sentence.
What the run count means. +1.40 °C is beyond the +0.5 °C threshold on the El Niño side, and it is the 3rd consecutive season past it. NOAA's own definition of an episode needs five. We print the count rather than a verdict, because the verdict is CPC's to make and it makes it in prose.
Where it sits in the record. +1.40 is the warmest MJJ value in 77 years, and ranks 56 of the 918 completed seasons in the whole 77-year file, counted from the warm end.
When this flips. Every sentence in this section is derived from the sign of that number, so the page needs no rewriting to follow it. Past −0.5 °C the phase reads La Niña, the run counts consecutive seasons below the threshold instead of above it, the rank is taken from the cool end of the record and the table's cells shade cool. Between −0.5 and +0.5 it reads neutral, and neutral is a state in its own right here rather than a gap between two events: the episode counter switches off, because NOAA's five-season rule has nothing to count. None of that is a forecast of a flip — it is whatever the file says on the day the page was last built.
Source file: NOAA PSL, Oceanic Niño Index — the file's own trailer reads “As of 08/4/2026, uses NOAA ERSST V6”.
The probabilities NOAA attaches to this — how likely a very strong event is, and what it means for temperature and precipitation by region — are read out of CPC's written discussion by a person, so they live in one article rather than being restated here: NOAA winter forecast 2026–2027.
Two of these three dates are computed from a rule NOAA states. The third is a pattern, and is labelled as one. None of them is a list that can run out.
Thursday 17 September 2026
CPC states the rule itself: released on the third Thursday of the month, around 8:30 AM Eastern. This page computes that date rather than reading it off a list, so it is still right in 2028. After this one: 15 Oct, 19 Nov, 17 Dec.
CPC release schedule — the rule, stated by CPC's own long-lead forecast release schedule page.
Thursday 10 September 2026
Each discussion names its successor in its own closing line, and this is the date the 13 August 2026 one named. That gives it a one-month horizon by construction — when this date passes, this card stops asserting it and falls back to precedent, because we will not have read the next discussion yet.
CPC ENSO diagnostic discussion — last one read: 13 August 2026.
Thursday 15 October 2026 on precedent
NOAA has not announced this date and may never announce it in advance. The flagship has accompanied the October long-lead release for the last four years, so the third Thursday of October is the day to watch — computed here, not copied. The research behind that precedent, year by year with its caveat, is in the article below rather than restated here.
Working: why that date, and how firm it is.
From the National Hurricane Center's active-storm feed. Advisories are issued on a six-hourly cycle, so this is the one block on the page that goes out of date in hours rather than weeks — and it says so rather than assuming a build is recent.
70 kt (80 mph) · 979 mb
30.5°N, 172.4°W · moving at 10 mph · advisory 22 Aug 15:00 UTC
Public advisory →45 kt (50 mph) · 994 mb
14.8°N, 150.2°W · moving at 12 mph · advisory 22 Aug 15:00 UTC
Public advisory →Basin matters for a US reader and the feed does not spell it out — the storm ID prefix does, and the fetch splits on it. Central and eastern Pacific systems rarely reach the mainland; the Atlantic list is empty — every system above is in a Pacific basin.
6 US weather articles. Everything on this page that required a person to read a NOAA document and form a judgement lives in one of them, and in exactly one of them.
NOAA's December–February outlook is already published. Above 50% odds of a milder north, above 60% odds of a wetter Southeast, and one dry signal the forecaster …
NOAA has an El Niño Advisory and calls it the great snowfall suppressor. Where the record says snow gains and loses — and why it is not a forecast.
2027 is not the super El Niño year — it is the year it falls apart. What NOAA has said, what the five strongest events did next, and what is not forecast.
NOAA's flagship Winter Outlook is not out — on four years of precedent it lands 15 October. What the agency has actually published for winter 2026/27, on El Niño, snow …
A potentially historic heat wave will scorch much of the central and eastern U.S. through the July 4 weekend, with 180+ million at risk and heat indices up to 115°F. …
An Extreme Heat Watch covers the Mid-South from Monday June 29 through Thursday July 2, 2026, with heat index values up to 110°F. Here's exactly which areas are affected, …
The numbers here are fetched, and everything derived from them is arithmetic. scripts/fetch_weather.sh pulls two keyless NOAA endpoints on each deploy, before Hugo runs: the Oceanic Niño Index from NOAA PSL, which mirrors the Climate Prediction Center's own oni.ascii.txt, and the National Hurricane Center's active-storms feed. The last completed season, the +0.5 °C run length, both record ranks and the storm list are computed from those two payloads by scripts/weather_parse.py. None of them is typed into a template, and none can be edited into agreement with anything.
This is a static page, so "fetched" means "as of the timestamp shown". The two stamps at the top are the moments each source was last read successfully — not the moment you loaded the page. Between deploys nothing refreshes, which is why each block carries its own expiry: seven days for the index, six hours for the storm list, both taken from how often NOAA itself republishes them. Past that the page stops asserting and tells you when it last knew.
Two of the dates are computed from a rule NOAA states, not copied from a list. CPC publishes that its long-lead outlooks go out on the third Thursday of the month, around 8:30 AM Eastern, so this page computes the third Thursday rather than storing dates. A stored list has a horizon and a horizon is how pages rot; a rule does not. The 4 dates a person did read off CPC's schedule page — 20 August 2026, 17 September 2026, 15 October 2026 and 19 November 2026 — are kept in scripts/data/weather_us_stated.json for one purpose: an automated check asserts the arithmetic reproduces all four, so if CPC ever changes the rule the check fails instead of the page inventing a date.
One date is stated by a human and carries its own expiry. Each ENSO diagnostic discussion names only its successor, so the next-discussion date has a one-month horizon by construction. It is stored with the date it was read and the date it expires, both taken from NOAA's own naming rather than estimated. Once that date passes, the card stops asserting it and says plainly that we have not yet read the new one — falling back to the second-Thursday cadence the last three discussions followed, labelled as precedent.
The flagship Winter Outlook date is a pattern and is never presented as a schedule. NOAA has not announced when the 2026–27 Winter Outlook lands. It has accompanied the October release for four years, which is precedent. The year-by-year working, including the one year we could not confirm from a noaa.gov timestamp, is in the winter forecast article and is deliberately not duplicated here.
Nothing here is a Drawpie forecast, and the probabilities are deliberately absent. This page prints no chance-of, no regional tilt and no snowfall figure. Those all come from a person reading CPC prose and map products, so each one lives in exactly one article, linked above. The hub prints only what it fetches or computes — which is also why a number cannot drift between this page and an article: they are not both allowed to say it.
What happens when the fetch fails. The script exits 0 unconditionally and the Netlify command fences it again, so a NOAA outage can never fail a deploy; the last-good committed data/weather_us.json ships unchanged. If one source fails, only that block is flagged — it keeps the timestamp of the last run that actually confirmed it, so a repeated outage cannot quietly re-date old data. A failed fetch and a genuinely quiet tropics are different states and are rendered differently, because an empty storm list from NOAA is a real answer and a timeout is not.
Seasonal outlooks are probabilistic statements about three-month averages over large regions. They are not forecasts of individual storms, snowfall totals or conditions on any given date. For decisions that depend on actual weather — and for anything involving a tropical cyclone — use your local National Weather Service forecast office and the National Hurricane Center, not this page.
NOAA, the National Weather Service, the Climate Prediction Center and the National Hurricane Center are US government bodies and are not affiliated with Drawpie. Their data is public domain; the analysis and the mistakes here are ours. No affiliate links and no payment was received for any link on this page.