Skip to main content
Humans+Robots

AI can predict the perfect surfing day

‘That wave is still going to crash on your head’

By Matt Crossman

I paddled out to catch my third wave of the day and found my surfing instructor, Rocky Canon, sitting on his board, staring off into the great blue infinity. He rose and fell with each wave, lost in an oceanic reverie, and he didn’t realize I had returned until I asked what he was doing.

Canon, a former pro and long-time instructor, has spent much of his life on Hawaii’s beaches. He is well-known as a surfing commentator and personality, a surfing expert among experts. And yet here he was, studying the waves in a spare few minutes on a Tuesday afternoon in February. These were baby waves, suitable for beginners, on a beach he has been to hundreds of times. But like Bill Belichick unable to take his eyes off of a pee-wee football game, Canon focused on the waves nonetheless.

They rolled under him and lifted his board, his legs, his butt, his back, his shoulders. He stored that feeling alongside innumerable other waves that have lifted him in his lifetime. This wave told him about the next one and the one after that. No two waves are exactly alike, but collectively they follow patterns. They break in roughly the same place, at roughly the same speed, with roughly the same ferocity. The difference between roughly and exactly is the difference between predicting and knowing.

The more Canon knows how waves will break, the better an instructor (and surfer) he’ll be. It helps his clients, too. I’d been worried about surfing on that beach on Oahu’s North Shore. The day before, to my untrained eye, the waves looked way too big for a beginner. But on the day Canon taught me, they seemed fine. The next day, they raged again.

The day-to-day variations prompted questions as old as surfing: How will the waves be today? How will they be tomorrow?

Thanks to artificial intelligence, our ability to answer those questions is better than ever.


The biggest day in the history of wave predicting was also the biggest day in the history of weather forecasting and arguably the biggest day of the 20th century: D-Day. Or, actually, the day before D-Day. The Allied invasion of Normandy, France, on June 6, 1944, was originally scheduled for June 5. But the weather forecast that day called for high winds and choppy waves, so it was pushed back a day, when the forecast was far better.

That bad forecast for June 5 (and the better one for June 6) turned out to be correct. If the invasion had gone forward on June 5, it likely would have failed, because the size of the waves would have made it nearly impossible for the Allied soldiers to get to shore.

The stakes couldn’t have been higher, and the technology used to make that prediction was rudimentary by today’s standards. Today, many companies use far more advanced technology to predict the waves, and the only stakes are a fun day at the beach.

For much of surf-predicting history, the “tech” involved calculations a person could figure out on a piece of paper (a mathematically skilled person, at least). It was this simple: If there are big waves in Place X at a certain time, they will arrive in Place Y at a predictable time and be of a predictable size when they get there.

Today, the idea is, broadly speaking, the same. But the information involved — billions of data points from satellites, buoys, and more across the world — has increased exponentially. That massive data is fed into machine-learning algorithms, and the result has been far greater accuracy.

The system predicts many factors — including shape, slope, and how quickly or slowly the wave breaks.

While some wave-predicting companies are more sophisticated than others, they all accumulate huge amounts of data and distill it into a prediction of what the conditions will be like. Wind patterns are scrutinized, ocean depths measured. Real-world eyes double-check the accuracy of virtual predictions. That’s how they do it. What they do is far more romantic: They turn data into joy.

The wave-riding, surf-loving, number-crunching scientists who run these predictive companies pore over nitty-gritty data, suss out root-mean-square error and identify and eliminate bias in mass amounts of numbers, all to say, “surf’s up.”

Mike Forte, a scientist at North Carolina-based AcuSea, which predicts conditions on the southeast coast of the United States, was a surfer long before he was a scientist; he’s surfed beaches there since he was in elementary school. In pursuit of waves, Forte was always looking for information about how the weather would be, because the weather would tell him about the waves. “When you’re a surfer, you automatically become an amateur meteorologistslashwave forecaster,” Forte says.

A bad day on the water always beats a good day at the office. But those bad days on the water could be good days if you go to a beach that you know in advance will have good waves. AcuSea’s use of AI and machine learning, combined with improvements in understanding the ocean’s depth, allows users to pick where they will surf based on what will happen there.

Nearing 50 now, Forte uses AcuSea’s predictive model to sketch out his week, choosing which beach to go to based on what the models tell him. “I try to plan my work around the surf,” he says. AcuSea created a ranking system for waves called “quality,” which takes into account factors such as height, angle, and speed. The higher the number, the better the surfing.

The predictive technology can help athletes in just about every water sport. Windfinder, a Germany-based company founded by sailors and windsurfers, presents the data and lets users apply it to their own passions, whatever those passions might be. Co-founder Jonas Kaufmann says the definition of a good weather day depends on the activity a person wants to participate in. Good conditions for surfing might be terrible for boomerang throwing.

Wait, what? Boomerang throwing? Yep. The company received a note over Facebook from James Hoy, a champion boomerang thrower, reporting that winds of 8 to 15 mph are ideal for his sport.


One big change in the surfing-prediction business in the last 10 years is in the data available. Wave buoys and satellites generate more data, and it has become much more affordable. Satellite data used to show wave heights only in narrow strips of ocean. Now, that information is available for huge swaths of water. The more data companies accumulate, the better they understand how that data affects waves around the world.

Surfline, headquartered in California, has the most robust data set in the industry. In addition to the satellite and buoy information that virtually all companies have, Surfline has 1.5 million handwritten reports of surf conditions dating back 30 years and HD cameras pointed from poles and roofs at 700 beaches around the world. (The camera at my beach offered an eye-level shot of someone standing on the shore; cameras at other beaches offer overhead perspectives, and some spots have multi-point views.) Thanks to the increased data, as well as improvements in machine learning, Surfline has cut its error rate in half in recent years.

Years ago, predictive models were based on wave-size data sought by coastal engineers, who didn’t know or care what a good wave was. Surfline predicts many other factors — including shape, slope, and how quickly or slowly the wave breaks.

The cameras, in particular, are transforming the business because they provide real-time, hard data to validate or invalidate predictions. Computers digitize the waves, allowing for more precise measurements than mere human observation. That gives instantaneous proof that the predictive models accurately forecasted what happened — or that they need to be modified.

Four million surfers across the world rely on that information — including me, if I may call myself a surfer after two days on the waves.

After I surfed with Canon, I spent an hour in my hotel room poring over Surfline’s camera feed from the time we spent on the water. I watched the waves, much like Canon had. They were hypnotizing. I eventually found video of myself surfing (or what passed for surfing, the way I did it).

If I had worn an Apple Watch, Surfline could have synced the video to it, clipped out my rides and eliminated all the waves I didn’t surf. Not only can Surfline help me find waves before I get to them, they also can find only the ones I ride.

That watch also would have fed into a leap Surfline has made in the predictive business. Ben Freeston, Surfline’s vice president of innovation, says Surfline is moving from just predicting waves to curating them.

Freeston created a smart surfing camera and collected data from various beaches, including rides per hour, how long those rides were, and the shape, size, and speed of the waves that provided those rides. The key step came next: Surfline analyzes surfers’ preferences, based on their Apple Watch data and their membership profiles on Surfline’s app and website, where they can list their favorite beaches. Just like Netflix can predict which movies its members will like based on their past viewing patterns, Surfline predicts which waves will be right for individual surfers. “There’s lots of different versions of perfect,” Freeston says.

For example, Freeston lives in California, where he takes every chance he can to surf the punchy beach break waves of Salt Creek. If he goes on vacation, Surfline can tell him which beaches will have waves like the ones at Salt Creek and when they will have them. “We can say to you, ‘These waves are like the waves where you went out and had the most fun,’” Freeston says.


Like Forte, Freeston was a surfer before he was a data scientist. He lived in the United Kingdom and worked in the computer industry. He collected publicly available data about surfing as a side project because he wanted to know when the waves would be good. The side project turned into a growing business that he called Magic Seaweed, which he sold to Surfline three years ago.

The title — Magic Seaweed — is a winking acknowledgement that prediction is an imperfect science. As advanced as Freeston’s models are, as accurately as they predict a day’s waves, they will never predict each wave.

A few minutes after I found Canon staring into the deep, he held the front of my board as I lay on it. My arms draped in the water. I faced the shore, and he watched the waves, looking for one suitable for me.

He saw one he liked and maneuvered to the back of my board. Like a dad teaching his child to ride a bike, he gave my board a gentle shove and hollered at me to paddle. I did, and then I pulled my feet underneath me and stood. I was upright for a fraction of a second when — CRASH! — the wave swallowed me and I pitched face first into the ocean.

The crash surprised Canon almost as much as it surprised me. I didn’t see that wave coming, and neither did he.

In the two days I spent surfing, I crashed several times for no reason, other than lack of balance. So to fall for a legit reason — the wave landed on me — was a thrill, the kind of wipeout I craved so I could tell my friends about it. The wave gets bigger every time I tell the story.

I rode a handful of waves that day and watched dozens, and that was the only one that broke that way. The rest of them unfolded much more slowly; it was the difference between a lazy fly ball and a screaming line drive. “What the hell happened?” I asked Canon after I paddled back out to him. “Why did that one crash so early?”

He laughed and shrugged, as if delighted to be unable to answer my questions beyond, “I don’t know.”  

I understood then why he studies waves intently, even after all these years. He went back to watching them and soon found another one for me. That one behaved like the others, and I rode it all the way in. The distance gets longer every time I tell the story.

A few months later, I told Freeston about my crash in the fast-breaking wave. Over the phone, he had the same reaction as Canon had in the water. My wipeout brought all three of us joy.

Surfline, AcuSea, Windfinder and other predictive companies aim to save surfers the time and energy required to find good waves. Gone are the days where you show up at the beach to see an ocean too timid or too volatile for your tastes. But as much as artificial intelligence can lead you to the waves, it’s still up to you to ride them — or wipe out trying.

“We’re not trying to pull the magic out of surfing. We’re just trying to get you to the place where you can enjoy that journey as best as possible for you. That wave is still going to crash on your head,” Freeston said. “It’s still magical.”

Published on

Matt Crossman is a writer based in St. Louis.

 

Illustration by Errata Carmona

Humans+Robots

Venture into the linguistic unknown

A glossary of terms, technologies, and innovators you'll find reading Experience stories about tech and the outdoors

By Schuyler Velasco