Skip to main content

El Niño methodology · Method fixed August 2026

How SnowSure builds its El Niño forecast

Most El Niño content asks you to trust a vibe. Ours asks you to check our math. This page is the complete chain from raw data to the ranking on your resort's page — including the parts we're least certain about.

See the rankingsPrediction ledgerSite-wide methodologyOpen dataset API

Step 1 — Thirty years of snowfall, resort by resort

The foundation is ERA5, the European Centre for Medium-Range Weather Forecasts' global reanalysis — the scientific community's standard reconstruction of past weather. For every resort SnowSure tracks, we extract daily snowfall at the resort's coordinates for 30 complete seasons, 1996-97 through 2025-26, and roll them into season totals. Reanalysis is not a marketing snow report: it won't match the number on a resort's brochure, because brochures come from summit snow stakes and reanalysis comes from a physical model of the atmosphere on a grid. That's why we never compare resorts by raw centimeters — every figure we publish compares a resort to its own 30-year baseline. The bias cancels out.

Before anything is published, each station passes a quality screen. Series with era discontinuities or grid artifacts are marked "under review" and excluded from rankings rather than quietly included. Four of 143 stations are currently held out on those grounds. When neighboring resorts share a model grid cell in early years, we say so.

Step 2 — Classifying the winters

We classify each of the 30 winters by NOAA's Oceanic Niño Index (ONI). That gives ten El Niño winters in the window — and three strong-to-super events on ONI: 1997-98, 2015-16, and 2023-24. Those three stay the published analog set (n = 3, frozen with the August 19 ledger). We now also ingest NOAA's operational Relative Oceanic Niño Index (RONI), which subtracts the tropical warming trend so 2023-24 reads weaker than on ONI, and 1982-83 is the RONI-era record. RONI is event-strength context. It does not reshuffle resort ranks. Southern Hemisphere resorts use the same analog winters in their own season (June–November).

Step 3 — The signal

For each resort we compute one number: average snowfall in the three strong El Niño winters, versus that resort's own 30-year average. Alta: +12%. Portillo: +81%. Mt. Baker: −15%. We publish the all-ten-El Niño figure alongside it, and the three individual analog winters, so you can see the spread — including the years that broke the pattern. The signal becomes a rank (out of 152 ranked resorts) and a tier:

TierSignalMeaning
A≥ +8%Prime El Niño terrain — book with confidence
B+3 to +8%Favored
C−3 to +3%ENSO-proof — the pattern barely touches it
D−8 to −3%Late bloomer — aim February–March
E≤ −8%Deep timing play — the base builds late; ski it at peak depth, without the crowds

Step 4 — The event forecast (the part that isn't ours)

Whether a strong El Niño actually happens is not a SnowSure call. We now read three independent event-strength sources and publish them next to the ranks:

  • NOAA CPC — official outlook. Operational index is RONI as of February 2026. August 2026: >90% chance of a very strong event. The 69% "strongest since 1950" figure is on raw ONI / Niño 3.4.
  • CCSR / IRI multi-model plume — physics ensemble. Mid-August 2026: weekly Niño 3.4 +2.7°C; 25 of 26 models put OND in the very-strong bin; no La Niña before spring 2027.
  • SNU ACE Lab CNN (Ham et al. 2019) — a deep-learning model trained on ocean patterns, not the same equations as the plume. July issuance peaked near +2.4°C RONI and still has a La Niña flip in early 2028. Long-lead, low verified skill.

Agreement between a physics plume and an independent neural net is the reason to take a very strong event seriously. The exact peak, and the date of any following La Niña, are not. We restate these numbers when the issuances update and date every revision. They never edit the prediction ledger.

What we're confident about, and what we're not

Confident: the historical record (ERA5, quality-screened), the event forecast (NOAA, high skill at this range), and the direction of the big signals — the southern storm track in North America, the Andes' outsized El Niño winters, the weakened Sea-of-Japan snow machine. These reproduce across events and match published teleconnection science.

Honest limits: three strong analogs is a small sample, and we will not pretend otherwise — every signal we publish carries "n = 3, a lean, not a law." A record-strength event may behave differently from past strong events. Snowfall also depends on temperature and storm timing that no seasonal method predicts: a warm pattern can turn above-normal precipitation into rain at low elevations, which is why our copy repeats one rule — ski high. And in every strong El Niño year, at least one region has broken its own pattern (Niseko finished +23% in 2023-24 against a −15% signal). A tilt in the odds is exactly that.

The scorecard — how you'll know if we were right

A forecast you never grade is a horoscope. The 2026-27 prediction ledger is the public record: twelve headline calls and 139 predicted tiers, frozen August 19, 2026, with scoring rules locked before the Northern Hemisphere season starts. That document does not get edited. It gets graded.

  • Monthly (year-round): we re-check NOAA's ENSO diagnostic and re-issue the outlook with a dated changelog.
  • Weekly (December–March): every resort page's El Niño module shows predicted signal vs. measured season-to-date snowfall — the same live totals SnowSure verifies daily. Southern Hemisphere pages use the matching winter window (June–September).
  • Monthly (December–March): a public scorecard grades every regional call — hits, misses, and what the record event is doing that the analogs didn't.
  • April: a full season post-mortem. The misses stay published.

The Southern Hemisphere is already the first test: the Andes — our #1-ranked El Niño terrain — are running one of the deepest winters in our 30-year archive right now, in real time, during this El Niño's onset.

The maps

The regional maps are the same numbers as the ranked table, drawn in SnowSure's snowfall-map language: cyan → indigo → magenta over a dark grayscale basemap, not a pin for every resort. Magenta is well above that resort's own 30-year ERA5 average; navy is below; near-zero stays close to the basemap (the ENSO-proof band). We interpolate only where SnowSure has a station — we do not paint the Great Plains or the open ocean. The analog-year tabs swap the underlying winter (1997-98, 2015-16, 2023-24) so you can see which event is doing the work. They are not a live snowfall forecast, and they are not NOAA's official CPC outlook.

Changelog

Method fixed August 2026. Changes to the signal definition, analog set, or quality screen will be versioned here rather than silently rewritten.

  • 2026-08-19 — v1. Published the 143-station fleet (139 ranked, 4 under review) using the three strong analogs 1997-98, 2015-16, and 2023-24.
  • 2026-08-20 — maps v2. Replaced resort-pin maps with shaded overlays and working analog-year layers. Signal definition unchanged.
  • 2026-08-20 — v1.1. Expanded the public ranking to 150 SnowSure resorts (154 stations, 4 still under review). Same signal and tiers. The August 19 prediction ledger stays the grade sheet for the original 139.
  • 2026-08-21 — v1.2. Added Monarch Mountain (Colorado) at #54 / Tier B (+6%). Fleet is 151 ranked, 155 stations, 4 still under review. Same signal and tiers.
  • 2026-08-21 — v1.3. Added Purgatory Resort (Colorado) at #79 / Tier C (ENSO-proof). Fleet is 152 ranked, 156 stations, 4 still under review.
  • 2026-08-21 — v1.4. Ingest NOAA RONI next to ONI. Publish IRI plume and SNU CNN as event-strength context. Analog winters and the August 19 ledger unchanged.

Data: ERA5 (ECMWF) via SnowSure's resort archive · Event strength: NOAA CPC ONI + RONI, CCSR/IRI plume, SNU ACE Lab CNN · 156 stations, 152 ranked, 4 under review · Method fixed August 2026; changes are versioned on this page.

Full El Niño rankings and map →