Since its inception in the 1990s, Surfline has gone from a popular website that surfers use to forecast to being the only option for surf forecasting in the market. But now they suck at it.
Whether you’re planning on surfing Pipeline on the north shore of O’ahu or Lower Trestles in San Clemente, you’re undoubtedly going to be checking Surfline before you go out. Surfline allows you to check the swell (height and time interval in between waves), tide, wind and water temperature.
Yet, the information that most surfers will focus on is the wave height measurement and conditions, which Surfline assigns to various breaks across the world.
Unfortunately, since Surfline’s recent AI integration with their “LOTUS” model, the wave height measurement has become wildly inaccurate and possibly hazardous.
Instead of getting your forecast from a human forecaster checking the conditions in person or reviewing the live cameras at a break, now that forecast is coming from an AI model. Surfline said on a blog post that LOTUS is “using high-resolution bathymetry, 35 years of surf reports, and 20 years of camera data… and real-time observations from satellites, buoys, and forecasters to continuously refine its accuracy.”
However, the reality of this current forecasting method is that the LOTUS model isn’t properly using live observations to call the surf. Instead, it relies more heavily on data from previous years while averaging out the data to give you an inaccurate estimate of what the surf is doing.
Alexander R. Stine, who holds a doctorate in earth and planetary science and is a professor at San Francisco State University’s School of the Environment, weighed in on Surfline’s switch to an AI model and its switch to relying on empirical data.
“You are throwing away everything we know about physics when you do that,” Stine said. “There’s quite a bit of data cooked into the fact that we know the equations of motion. That’s a pretty firm constraint. Then you’d probably do better with just actually doing the calculation.”
For less popular Northern California waves like Linda Mar in Pacifica or Princeton Jetty outside of Half Moon Bay, you’re rarely getting an accurate reading from LOTUS. For the surfers who aren’t yet able to read swell information to determine wave height and conditions for themselves, they could see LOTUS calling the surf 1-2 feet when there are really 4-foot waves that they might not be ready for. LOTUS even gets the water temperature wrong, with last week’s forecast being off by 4 degrees Fahrenheit.
“At the end of the day, it seems like AI is a cost-cut versus a legitimate enhancement,” said Joey Thompson, a member of SFSU’s Surf and Snow club.
Thompson also teaches surf lessons, often to new surfers who want to know which days they should be going out. Surfline, despite all of its issues, is the only choice with its simple interface that Thompson can recommend.
“I’ll refer them to Surfline, but then you either miss out on so many days because it’ll miscall it or you’ll go and it’ll be not what you expected,” Thompson said. “I feel like there’s such a push for any tech company to be like, ‘Look, we’ve automated like 80% of our company’… and then it just ruins it for everyone.”
Inaccurate forecasts from Surfline are made worse with there being almost no alternative for the large majority of surf spots across the globe. And why is that? Surfline bought all the alternatives. In 2017, Surfline acquired Magicseaweed, its main competitor, and in 2019, it acquired the Australian equivalent, Coastalwatch.
With no real alternative to Surfline, surfers are either stuck trusting inaccurate forecasts or learning how to forecast themselves.
