What we said. What the ocean did.
A forecast nobody scores is just an opinion. This page is where ours gets scored, including the times it was wrong, because those are the ones that change the model.
Status: the log is just starting
We’re not going to fill this page with entries we haven’t actually run. Below is exactly how the checking works and the one result confirmed so far. Cases get published here as they accumulate. The wrong ones included.
Three ways of being checked
No single method covers a forecast. Each of these grounds a different part of it, and each has something it structurally cannot see.
Satellite: grounds the geometry
A cloud-free Sentinel-2 pass over a spot, with the real swell crest direction pulled out by a 2-D FFT. The crest lines are usually invisible to the eye; the transform separates the periodic signal from the noise. It reports its own confidence and returns nothing rather than a guess when the signal is too weak, which happens often, since it needs both good glint and enough swell.
- Grounds
- Approach angle, refraction around a point, where it breaks.
- Cannot see
- Height, and whether it was any good.
Buoys: ground the height
An offshore buoy measures what is arriving, not what the break does with it. Useful as a check on the inputs, not on the transformation, which is the part that’s hard.
- Grounds
- Offshore wave height and period, continuously.
- Cannot see
- Everything that happens between the buoy and the beach.
Surfers: ground the only thing that matters
No instrument can tell you a wave was fun. A report from someone who was in the water is the only source of truth for quality, and quality is the thing the forecast is ultimately claiming. This is the slowest signal to collect and the most valuable.
- Grounds
- Whether it was actually worth paddling out.
- Cannot see
- Nothing, and that's the point.
Confirmed so far
Santa Cruz · satellite vs offshore buoy
Swell direction to within 3°
The FFT extraction recovered an offshore swell direction from Sentinel-2 imagery that agreed with NDBC buoy 46026 to within three degrees. That validates the measuring instrument, the thing we check the model with, which had to come first. Checking the model itself against it is what fills the rest of this page.
Everything else is still in progress. When a run produces a verdict, good or bad, it appears here with the forecast timestamp, what was predicted, what was observed, and what changed as a result.
Tell us what it actually did
If you surfed one of our spots, this is the most useful thing you can send us, especially if we got it wrong. It takes a sentence.
How the forecast is produced in the first place is written up in the methodology, including the parts of it that are weakly known.