Catching the breeze may bring you gold but sail too close to the wind and you’ll miss the boat. Combined with a sailor’s intuition meteorologist Sukanta Basu’s highly detailed  wind forecast may well help the Dutch sailing team secure a win at the Olympics this year.   

Imagine a warm August morning in Japan. The Olympic Games are in full flow and the Dutch team is about to take to the waters of Sagami Bay. Just before the race kicks off the coaches receive the wind forecasts  for the day on their iOS or Android phones, with wind speed and direction specifications for every ten minutes of the race. The forecasts help the team make the right tactical moves on the water and can make the difference between going home with a medal or going home emptyhanded.


Such detailed data cannot be gleaned from online weather forecasts and that is why the sailing team enlisted the help of meteorologist Sukanta Basu who created a sophisticated wind prediction model  especially for competitive sailing. 

The model generates many terabytes of data on a daily basis but Basu only relays the information the athletes really need. ‘Of course they want to know how hard the wind blows but the exact magnitude of the wind is less important,’ Basu explains. ‘What is important, however, is how wind conditions rapidly change during the course of the day.’


Basu’s approach has been put to the test during earlier races at Sagami Bay and the results have been encouraging. ‘We are right in the middle of cracking the Tokyo code,’ head coach of the Dutch sailing team Jaap Zielhuis says.   

The Sagami Bay code is not an easy one to crack. ‘The weather there is not easy to predict,’ Basu says. To the Northwest of the sailing area lies mount Fuji, which is almost 3,800 metres tall, and in close proximity of the bay sits the island of Izu Oshima. ‘The island modulates the airflow,’ Basu explains, ‘and that will cause a problem for the team, particularly if the wind is south-westerly.’ At that moment in the race they will be sailing downwind, i.e. on the side of the island where wind conditions will be erratic.

The wind in images

The  disruptive effect on airflow caused by the presence of Izu Oshima is clearly illustrated by the charts produced by the predictive tool, showing dozens of lines visualising the path and speed of the wind. Former MSc student Kars Trommel helped Basu by applying deep learning, a subset of AI, to discover the range of wind patterns which may occur.

Lack of data

'Another challenge we have to contend with is that few meteorologists are doing research into wind conditions and the available data is limited,’ Basu says.  To help him collect the necessary data Basu relies on the help of the Sailing Innovation centre which also analyses the data.

Basu’s starting point is the Global Forecast System (GFS), an open source metereological dataset developed in the US.  ‘But we didn’t have sufficient spatial resolution for our project. That is why Basu turned to dynamic downscaling using the Weather Research and Forecasting model (WRF), and machine learning approaches.  The code of WRF model is millions of lines long,’ Basu explains. ‘You can compare the GFS data to a photograph of, say, two megapixels. Dynamical downscaling turns that into 200 megapixels which allows us to zoom in and extract more detail.’

Dynamic downscaling is not just about pixels but about grid points, which are spread across the globe. ‘The GFS model will solve equations for every grid point and that in turn will produce a weather forecast,’ Basu says. ‘But those grid points are dozens of kilometres apart. Sagami Bay has only two or three grid points and we need more information than these can give us. So we need a grid with a higher resolution.’

Zoom power

Thanks to the WRF model, Basu can achieve a resolution of a kilometre. ‘And it looks as if, during the coming Olympics, we’ll be able to use a grid whose points will only be 500 metres apart,’ Basu says. ‘But I don’t think we’ll narrow it down to 250 metres. The higher the resolution, the more processing power you need.’

Basu will have to find a balance between zooming in as closely as possible and not overloading his equipment. ‘We will be running our predictive model at the supercomputing centre SURF in Amsterdam,’ Basu say. ‘Thanks to the supercomputer we will be able to provide the sailing coaches with the most up to date wind forecasts every morning for every race.’

Intuition trumps knowledge

If the racing team are able to anticipate the changing winds it can make the difference between a medal and disappointment. But ultimately the team will have to assess the reliability of the day’s forecast. The more changeable the weather, the harder it is to predict. ‘In the end it’s about sailing’, coach Zielhuis says. ‘In the deciding moment the intuition of the sailor is what counts. That person sees a  little wind ripple on the waves and says: this is what I’m going to do, let’s make it work.’