Surf Forecast Accuracy

Forecasting the weather is always difficult and with any method it's worth bearing in mind the enormous complexity of predicting globally what will happen for the week ahead. We want to make sure we're as clear as possible with everyone using magicseaweed about what you can expect from our forecasts.

The bulk of our forecasts start life as the output of the NWW3 Model. This computer model is a third generation swell forecast model. That's to say it's the third in a series of attempts by meteorologists to forecast swell mathematically from observed conditions using a forecast of surface winds around the globe. It's very sophisticated in the way it does this and is fed every six hours by thousands of actual observations from satellites, buoys, ships, weather stations and aeroplanes as the starting point for it's predictions. It's regarded as one of the most sophisticated models of its kind and is routinely used for shipping forecasts for both commercial and military planning. The full 'validation' giving a wealth of technical reports on the percentage errors etc. can be found here:

It's very heavy going but the key is that it's one of the most accurate systems yet devised for predicting the way swells are generated and travel to our coastlines.

Nearshore Data

Recently we've started improving many of our forecasts with the addition of a complex nearshore modelling tool that allows us to take this open ocean data and predict how it will be transformed right up to the beach, giving an accurate idea of the size of the swell coming right into the surf zone. This allows us to take the highly accurate (see link above) NWW3 model and make the results immediately relevant to a surf session. We're currently building our own accuracy statistics from this compared to certain key wave buoys located just outside the surf zone and so far the results are impressive.

Open ocean data

First (as we make very clear) this is open ocean data. From this it's difficult to give a 100% reliable idea of the actual surf conditions on a given day on a given beach. What we can give is a good indication of the amount, quality and direction of swell heading towards that beach. This data is a powerful tool to allow you to create your own forecast for your local break, but a number of factors can conspire to make your prediction more or less accurate. The first is that you need to read carefully the tutorials and articles on this website to fully understand what swell is and how significant height and period can best be interpreted with the swell direction to estimate the conditions at your local break. Secondly an awareness that these simple descriptions (height and period) of a complex mish mash of waves offshore from our beaches are inadequate to fully describe the surf. Two '4ft@11second' days can be different on the beach even if other factors (swell direction) are the same because they over simplify a complex local situation in which more than one swell, with more than one different size and period wave may be moving in different directions at the same time.

Range and accuracy

The second issue is range and detail of the forecasts. For about the first three days the forecast is at it's most accurate. During this period you can with some confidence use the data as part of the information required to plan trips and travel and the three hour details are quite relevant. As you move into 3-5 days the predictions will start to change more as time progresses and to expect the three hour details to be accurate is less and less reasonable. Between 5-8 days the forecast enters its 'experimental' stage. This means that small errors creeping into the model will mean it's likely that the actual conditions vary more widely from the forecast - it becomes more inaccurate. It would be very unrealistic to imagine that the data for exactly 3pm in seven days time is more than a rough outline of what *might* be happening at about that time. Of course this trend is the same with any forecast and the 1-3, 3-5 and 5-8 day periods are pretty arbitrary. The rule is simple - the longer the period we're forecasting for the less reliable it becomes.

So why show three hour intervals over the full eight days? We thought long and hard and came to the conclusion that this data, this information is the result of hundreds of days of academic work and millions of pounds of investment. It exists. It's there available - why wouldn't we give you it? It's better, we believe, to give the data warts and all with clear explanations than to dumb it down and hype it up.