Peel Watch alerts send you one email, one to three days ahead, when a spot you chose reaches your score threshold for your skill level, and stay silent otherwise. That's the whole idea behind our surf alerts Portugal-wide: you stop checking the forecast every morning and let it check itself. The score is our 0–5 rating per spot, per level, and it's multiplicative, so a single dealbreaker (strong onshore wind, the wrong tide, swell too big for your level) pulls the hour toward zero. An alert can't fire on a day that's a 4 for an intermediate and a 0 for a beginner if you're the beginner, because it's computed for your level only. Sign-in is by magic link, no password and no card, and every email carries a one-tap unsubscribe. We don't send alerts for anything further out than day three, because forecast error grows with lead time, and we leave out provisional spots that no human has checked yet.
Setting one up, in order:
- Sign in with a magic link sent to your email. No password, no card.
- Pick a spot from the 115+ we cover between Viana do Castelo and Tavira.
- Pick your level, L1 Novice to L5 Expert.
- Choose the score on the 0–5 scale that's worth a drive to you.
- Leave it alone. You hear from us 1–3 days before the spot lines up, and you can unsubscribe in one tap.
Surf alerts Portugal: what actually triggers an email?
The trigger is the score. Nothing else.
Every spot gets an hourly forecast 192 hours (8 days) out, refreshed four times a day at 03:30, 09:30, 15:30 and 21:30 UTC. For each hour, the model computes:
score = 10 × swellQuality × windFactor × tideFactor × skillFit × daylight, halved and printed to one decimal, 0.0–5.0.
Because the terms multiply, they can't average each other out. A clean, long-period ondulação (groundswell) with a howling onshore wind still scores low, because a windFactor near zero drags the whole product down with it. That's the point. An alert built on an averaged score would happily email you about a day that has one excellent ingredient and one ruinous one.
The bands, as printed on the site:
- 0–1 Not so great
- 1–2 Marginal
- 2–3 Decent
- 3–4 Great
- 4–5 Fantastic
When the score for your spot, at your level, crosses your number inside the 1–3 day window, the email goes out.
Why does the alert depend on your skill level?
Because the same water is a different day for different people. The skillFit term compares the modeled breaking height against an ideal band and a hard cap for each level. Above the cap, skillFit sinks the hour regardless of how good the other terms are.
| Level | Ideal breaking height | Hard cap |
|---|---|---|
| L1 Novice | 0.3–0.9 m | 1.3 m |
| L2 | 0.5–1.2 m | 1.8 m |
| L3 | 0.8–1.8 m | 2.6 m |
| L4 | 1–2.6 m | 3.8 m |
| L5 Expert | 1.3 m+ | No cap |
For scale, 1.3 m is a little over 4 feet and 3.8 m is about 12.5 feet.
This is the part most alert systems skip. A generic swell alert tells everyone that 2 m is coming. For an L4 that's comfortably inside the ideal band. For an L1 it's past the cap. Our alerts are set per spot and per level, so the L1 alert on that day simply doesn't send. Silence is a forecast too.
Why don't alerts look further ahead than three days?
Because by day four or five the error bars are wide enough to flip the answer, and an email that's wrong half the time trains you to ignore it.
What our own accuracy panel shows
Every 72 hours or so we lock in the forecast and compare it to the day-of call, and we publish the result on the honesty panel at /en/features. Average swell-height error is about ±0.1 m, growing to about ±0.2 m at five days out. Average wind error is ±1–3 kn depending on lead time.
That sounds small. Now apply it to the table above. A five-day call of 1.2 m for an L1 means anything from about 1.0 m to 1.4 m, and 1.4 m is past the 1.3 m cap. The same forecast could be a good novice day or a non-starter. Add a couple of knots of wind error on a day when the nortada (the summer north wind on the west coast) is building, and the windFactor moves too. Multiply those together and a five-day alert is a coin with opinions.
What the wave modelers say
This isn't a quirk of our setup. ECMWF, whose models sit under much of Europe's wave forecasting, states that the accuracy of weather and wave forecasts worsens with increasing forecast period (ECMWF, Ocean waves at ECMWF). A study using Iberian Peninsula coastal buoy data found the same for ECMWF and GEFS wave-height forecasts: accuracy decreases as forecast days increase (ScienceDirect, 2025). And ECMWF's 2025 note on its AI-based wave modeling says forecasts stay sharp for the first few days, with smoothing noticeable at longer lead times, and that the improvements shift skill by roughly a day (ECMWF AIFS blog, 2025).
What a storm does to a long-range call
Winter makes the point loudly. During depression Bram, IPMA warnings for the west coast showed waves of 4 to 5.5 meters, with direction shifting temporarily near the mouth of the Douro (SAPO, citing IPMA). A direction change of that kind moves swell in or out of a spot's window, which is exactly the term that decides whether a sheltered corner works. You can still look 8 days out on the spot page. We just won't email you about it.
Why are provisional spots left out of alerts?
Because an alert is a recommendation, and we won't make one on a profile no person has checked.
Each spot's model is part measurement, part judgment. The coastline facing is measured from OpenStreetMap. The swell window, its shadow factor, the sheltered wind sectors and the tide preference are hand-authored. When we add a spot, the measured parts arrive first and the judgment parts need a human to look at them against the real beach.
Until that happens, the spot carries a provisional badge on the dashboard, its size and skill-fit terms are capped for safety, and it's excluded from alerts. You can still read its forecast. We'd just rather not send an email saying a spot lines up when we haven't confirmed which way it faces to swell.
The principle isn't unusual on this coast. Portugal's Bandeira Azul (Blue Flag) listing is reviewed every year; the 2026 round recognized 396 beaches, 350 coastal and 46 inland (Turismo de Portugal, Bandeira Azul 2026). Different question, same habit: check first, then label.
What the coast and its buoys tell us about the scale of the job
Our forecast is modeled from Open-Meteo data, not read from buoys. The national buoy network is still the measured record we keep in mind when a model number looks too tidy.
| Fact | Figure | Source |
|---|---|---|
| Length of Portugal's coastline | 1,794 km (1,115 mi) | Wikipedia, Geography of Portugal |
| MONIZEE multiparameter buoy sites | Leixões, Nazaré, Sines, Faro | Instituto Hidrográfico |
| Wave-buoy sampling interval | 30–60 minutes | SNIG/DGT record |
| Max wave height, Nazaré buoy, depression Hermínia (Jan 2025) | 17 m | Região Sul, citing Instituto Hidrográfico |
Four buoys over nearly 1,800 km of coast is why neighboring spots share one offshore forecast in our model, and why orientation, seabed and maré (tide) do the work of separating them.
Which threshold is worth setting?
This part is opinion, not model output, so read it as our judgment.
- 3.0 (Great) if you have to drive more than an hour. It filters out days that are fine on paper but compromised on one term.
- 2.0 (Decent) if the spot is close and you'd paddle out on an ordinary day anyway.
- 4.0 (Fantastic) if you only want to hear about the rare ones. Expect long silences, which is the intended behavior.
Given the error figures above, treat a 0.2 difference in score as noise. Setting 3.0 versus 2.8 won't change much, and nudging it daily defeats the purpose. Pick a number, leave it for a few weeks, then adjust once.
If you swim or fish rather than surf, the alert isn't the tool. The swim board ranks beaches 0–10 for calmness, a separate scale, with a day picker for the calmest hours. It doesn't include rips, flags or lifeguards, and neither score means a beach is safe.
Setting one up takes a spot, a level and a number
The Peel Watch team built alerts because we got tired of opening five tabs before coffee and still driving to a blown-out beach. There's nothing to buy and no card to enter. When you have a minute, open the dashboard, pick a spot you actually surf, set your level and your number, and let the forecast do the checking. If it never emails you, that tells you something too.