Long-term swell forecasts are always a gamble. Sometimes they can be spot-on, other times they can be way off. How accurate a long-term forecast is likely to be is therefore a useful thing to know. Luckily, the MSW forecasts have a ‘probability’ factor that tells us just that. Here’s how it works.
In a previous article (HERE) I explained that one of the main problems with forecasting has to do with the initial measurements that are fed into the model. The more accurately we can measure these, the more accurately we can predict the surf.
The ocean-atmosphere system is non-linear and chaotic, so any inaccuracies in those inputs will be amplified in the model. Any tiny errors at the start will increase exponentially as we try to forecast further ahead into the future.
Now, we can simulate those initial measurement errors by seeing what happens if we run the model twice, each time with slightly different inputs.
For example, first run the model with a particular windspeed input, then change that input by, say, 1 km/h and run it again. The predicted wave heights for each run might differ by, say, 0.5 foot for a one-day forecast and two feet for a two-day forecast, but ten feet for a seven-day forecast. The tiny input difference has been amplified, with the forecasts radically diverging after about five days.
The amount of divergence in the forecasts when we simulate input errors like this tells us how much confidence we can put in that forecast. In the simple example above, the seven-day forecast is pretty unreliable – the wave height can’t be predicted within more than ten feet for wind inputs accurate to plus or minus 1 km/h.
This is called ensemble forecasting, and is common practice with weather and wave predictions. Each model run is called a member and the set of forecasts is called the ensemble. The variations in the input values are designed to simulate known measurement errors. To get a definitive value for the forecast, the average of all the ensemble members is typically taken. The spread of the members around the average is quantified to reflect the uncertainty in the forecast.
The MSW forecasts use a 20-member ensemble to come up with a ‘probability’ parameter for each forecast. You’ll notice that these probabilities generally decrease as the length of the forecast increases. But they don’t always decrease smoothly, and not always in the same way for every forecast.
That’s because the confidence, just like the forecast itself, depends on the current state of the atmosphere and ocean. The more stable the situation, the more reliable the forecast.
The most reliable forecasts are usually when a swell is generated a long way away and local conditions are stable. Five days ahead, for example, the swell might already be on its way. If local areas are dominated by high pressure, the swell will probably be clean when it arrives.
The least reliable forecasts are when storms develop very close to the coastline. Five days ahead the low might not have even developed yet, so there could be a huge range of possibilities. Local conditions are less stable, so any slight change in the trajectory of the storm can make a big difference to local wind conditions or swell direction.
Cover shot of Nazare by Helio Antonio.