Marine Weather Forecasts Are Getting an AI Upgrade

By Vanessa Minke-Martin

Machine learning systems—powered by new data—are taking some of the guesswork out of maritime safety.

Jake Spink fished British Columbia’s craggy coast for four decades. Now, as the president of the British Columbia Coast Pilots—an association of highly trained captains that guide thousands of tankers, cruise ships, and other large vessels into the province’s ports each year—Spink can draw detailed maps of the shoreline from memory. Even so, navigating the region’s fickle and sometimes perilous weather can be a challenge. There have been many times when Spink has motored up a flat calm channel, coffee in hand, and turned between two islands to suddenly face blinding spray and waves breaking over the bow. Twice, while piloting through storms near Haida Gwaii, off the province’s north coast, he fought 100-kilometer-per-hour wind—strong enough to uproot a tree—blowing from the southeast only to have it flip within ten minutes to howling just as hard from the northwest.

“For wind to change direction that fast, it’s shocking,” says Spink. “That’s what catches people unaware.”

Checking the weather forecast doesn’t always help. In British Columbia, as in many coastal regions, predicting the weather is notoriously difficult. High mountains and narrow channels funnel wind, trap fog, and cause moist air to dump torrents of rain, while extreme tides and fast currents build imposing waves. The result is a patchwork of hazardous meteorological microclimates, where conditions can differ wildly between neighboring fjords and across small islands.

The problem, explains Melissa Westland, a graduate student with the Weather Forecast Research Team at the University of British Columbia, is that weather models are based on regional weather patterns and don’t account for the dramatic influence of such diverse terrain.

Image Credits: The Australian Institute of Marine Science.