How Swell Intel Makes Surf Forecasts

How are surf forecasts made?

Swell Intel makes surf forecasts by running SWAN nearshore spectral wave physics over high-resolution bathymetry, so open-ocean swell is translated into the breaking wave height at each specific break rather than reported as a raw offshore number. Swell Intel then corrects that physics with live satellite altimeter data and a correction layer trained on years of NDBC buoy history, and assimilates fresh buoy observations hourly.

Under the hood, Swell Intel drives a SWAN (Simulating WAves Nearshore) spectral model with global swell input, then runs it down through progressively finer grids to high-resolution bathymetry at each break. That is where the model resolves refraction, shoaling, and breaking — the physics that turns an open-ocean swell into the wave that actually stands up at your spot. A live satellite-altimeter correction and a correction layer trained on years of NDBC buoy history keep the swell input honest, and hourly buoy assimilation folds the latest real observations back into the forecast.

Why do surf forecasts say 3ft when it's overhead?

Most forecasts quote the offshore significant wave height a buoy measures miles out, which is a statistical average of the larger swell and reads far smaller than the face you paddle into. Swell Intel instead models breaking wave height at the break, so the number Swell Intel shows is the rideable size after the swell shoals and stands up over the local bathymetry — much closer to what surfers call overhead.

A buoy's significant wave height is a statistical average of the offshore sea state, measured in deep water before the swell has shoaled. By the time that swell reaches a shallow reef or sandbar it can more than double in face height. Because Swell Intel models the breaking wave height at the break instead of quoting the buoy, its numbers track the size surfers experience — and its breaking wave height explainer walks through exactly how that transformation works.

How accurate are Swell Intel's forecasts?

Swell Intel publishes its accuracy: across 362 spots and 7,863 forecast-vs-buoy pairs, Swell Intel's swell calls land within one foot of the NDBC buoy reading 83% of the time, at an average error of 0.59 ft.

Swell Intel scores every forecast cycle against co-located NDBC buoy observations and reports two headline numbers: mean absolute error in feet, and the share of swell calls that land within one foot of the buoy. Buoys measure open-water swell rather than breaking surf, so these metrics verify the swell input that Swell Intel's breaking-height calculations are built on. The live figures, region breakdowns, and trend over time are on the accuracy dashboard, updated as new observations arrive.

What makes a nearshore model different from an offshore model?

An offshore model (or a buoy reading with a spot lookup) describes the open-ocean sea state and leaves the surfer to guess how it breaks. Swell Intel's nearshore model runs SWAN spectral physics on high-resolution bathymetry at each break, resolving how swell refracts, shoals, and breaks — which is why Swell Intel can give a breaking wave height at the specific spot instead of an open-ocean average.

Offshore readers — a raw buoy value or a global model with a spot lookup — stop at the open-ocean sea state. Swell Intel's nearshore SWAN model keeps going: it propagates the swell spectrum across high-resolution bathymetry so refraction around headlands, shoaling over reefs, and the exposure of each break all shape the final number. That is the difference between a forecast that describes the ocean and one that describes your wave.

What Swell Intel doesn't publish

Swell Intel is transparent about outcomes but not about the recipe. We publish verified accuracy, the physics we use, and how we measure error — but not the proprietary calibration coefficients, blend weights, or grid configurations behind the model. Extended outlooks run to 16 days for planning incoming swells; near-term forecasts stay the most reliable.