How the forecast is built
Every part of the forecast, what it runs on, and how far that part has been checked.
1 · What goes in
2 · What the model does
- Swell to the surf zone3.4–4.5% high against a full-physics model at Ocean BeachChecked
- Wrap around points and into bays33 of 47 satellite passes, none failedChecked
- Breaking at 0.60 of depthTwo spots; 15.5% low on a reefPartly
- Fort Point tidal currentOne satellite pass, Jan 8, 2025Partly
- Breaker typeAn indication; weakly known by anyoneNot yet
3 · What you see
How far ahead
Hourly for the first 72 hours, 3-hourly to day 10, 6-hourly to day 16. Every hour carries a confidence word: high for about 3 days, medium to day 5, low after. California spots also show what the week normally does, from 25 years of modelled waves (2000–2025), labelled as an average.
| Row | From day | To day | Value |
|---|---|---|---|
| Time step | 0 | 3 | Hourly |
| Time step | 3 | 10 | 3-hourly |
| Time step | 10 | 16 | 6-hourly |
| Confidence | 0 | 3 | High |
| Confidence | 3 | 5 | Medium |
| Confidence | 5 | 16 | Low |
| Swell · California | 0 | 6.5 | CDIP MOP |
| Swell · California | 6.5 | 16 | NOAA GFS-Wave |
| Swell · East Coast, Australia | 0 | 16 | NOAA GFS-Wave |
Sources
| Input | Source | Breaks |
|---|---|---|
| Swell, CA days 0–7 | CDIP MOP (Scripps) | 110 |
| Swell, elsewhere and days 7–16 | NOAA GFS-Wave | 59 + all |
| Seafloor, US | NOAA CUDEM 1/9″; USGS CoNED in parts of SoCal | 137 |
| Seafloor, Australia | NSW Marine LiDAR 2018, 5 m | 32 |
| Tide | NOAA CO-OPS · TICON-4 (AU) | 169 |
| Wind, weather | NOAA NWS to ~day 7 (US) · NOAA GFS | 169 |
| Water temperature | Nearest NOAA buoy · JPL MUR | 169 |
| Seasonal average | CDIP MOP hindcast 2000–2025 | 110 |
What it does not know
Breaker type is weakly predicted, by anyone~14% of breaking intensity, in field data
We say whether a wave should spill or plunge, from the Iribarren number. In a field study of 187 breaking waves at three sites (Robertson and others, 2014), it explained about 14% of the variation in measured breaking intensity, and nothing else they measured did much better. We tested four published alternatives against our own output and all four failed. Treat breaker type as an indication.
Sandbars move; the survey does notUS surveys 2015–2023 · Australia 2018
On a reef the seabed is fixed. On a beach the bars migrate through the season, and the seafloor we use is a snapshot from the survey year.
Currents and rips, except at one spotOnly Fort Point's tidal current is modelled
The Golden Gate ebb is in the forecast. Rips, longshore drift and other tidal flows are not.
The rating is judgementStructure from the literature, weights are ours
There is no published standard for how much a given wind matters against a given size. Reasonable people would weight it differently.
It is a model, not a measurementIts inputs are forecasts too
It cannot see a passing squall or the sandbar that shifted last week.
What the model does
March the whole spectrum shoreward. Every swell train, with its own height, period and angle, crosses the real seabed in a parabolic mild-slope solver: refraction, shoaling, diffraction, breaking. Deterministic physics, so the same inputs give the same waves.
Break at 0.60 of the depth, on the combined sea. Until August 2026 this was 0.78, a single-wave limit that put the break too far inshore against a full-physics model. The change moved the break line seaward and lowered the published height by about 5%.
Let the biggest breaker decide. A 2 ft 17-second groundswell out-breaks a 3 ft 6-second windswell, because breaking follows energy.
Bend and block what the grid cannot. A traced wrap turns swell around points and into bays; headlands block the directions they shadow. At Fort Point the Golden Gate ebb steepens the swell.
Corrected on this page
This page and Forecast vs reality used to say things the product no longer does. Each line below was wrong on the previous version.
- Horizon: 16 days, hourly to 72 hours. It said 7 days was “a hard limit”, with an outlook to 10, and 3-hourly to 72 hours.
- Swell input: CDIP MOP covers California only. 59 breaks on the East Coast and in Australia run on NOAA GFS-Wave.
- Breaking: 0.60 of depth on significant height. It said 0.78.
- Seafloor: US surveys run 2015–2023, and Australia uses NSW LiDAR from 2018. It said every survey was flown in 2022 or 2023.
- Tide, wind, water temperature: Australia’s tide comes from TICON-4; wind from NOAA NWS and GFS; water temperature from NOAA buoys and JPL MUR. It said NOAA CO-OPS and Open-Meteo for all of them.
- Seasonal average: California only. It read as if every spot had one.
- Currents: one tidal current is modelled (Fort Point). It said none.
- Breaker type: the 14% is against measured breaking intensity, and the study’s “do not pursue further” was about a different measure. It said 14% of breaker type and pinned that conclusion on the Iribarren number.
- Forecast vs reality: the 3° result checks the satellite tool, not the forecast, and satellite does not score where a wave breaks. It said the log was just starting with that as its one result.