A new AI weather forecasting tool released today by startup WindBorne Systems provides more frequent and accurate forecasts of key variables than a world-leading system developed by European governments, thanks to advances in how sensor readings are incorporated into deep learning models.
Founded by a group of Stanford students in 2019, WindBorne began by building a better weather balloon, with the idea of selling weather data. But with the advent of deep learning models to predict the weather in 2022, the team realized that they could capture a lot of value by building their own model.
WindBorne claims that the new version of its model provides a more accurate forecast than the traditional ECMWF and AI systems across several variables. Another easy way to understand it, says WindBorne chief product officer Kai Marshland, is that WeatherMesh 6 is “about five days as accurate as the traditional forecast was yesterday,” especially for surface temperatures.
Today marks the release of the sixth version of that model, WeatherMesh, which the company says is more accurate than traditional and AI forecasts produced by the European Center for Medium-Range Weather Forecasting (ECMWF), a European intergovernmental organization recognized by meteorologists as the leading provider of accurate weather forecasting today.
WeatherMesh 6 produces a forecast every hour, as opposed to every six hours, as traditional models do. Its resolution is now down to 3 km in Europe and the continental US, where the data quality is much higher.
Traditional weather forecasts are generated by complex physics models that require expensive computers to run, and take a long time to do. AI models – developed by startups and large labs like Google DeepMind – tend to move faster than physics models, but currently do not have high resolution, many variables or accurate predictions over long time horizons.
Nevertheless, the AI climate is developing rapidly and is already being used in major government agencies around the world. Researchers are working to integrate it into systems used to synthesize weather data and generate social forecasts.
WindBorne benefits from its unique integration of modeling and data collection. The company now has about 400 balloons in flight collecting sensors at any given time, deployed at 15 sites around the world. Improvements to its current model come from improvements in the way the data collected by the balloons are incorporated into the models.
“I personally do not understand how to do business [an] An AI-based weather company without the advantage of a data set,” WindBorne CEO John Dean told TechCrunch.
ECMWF’s superiority stems from the organization’s capabilities in “data assimilation,” the task of turning disparate sensor readings into a comprehensive, machine-readable picture of the world. Currently, AI climate models rely on datasets produced by ECMWF and the US National Oceanic and Atmospheric Administration.
But WindBorne and other organizations are working to feed data directly into the models, and the company’s head of AI, Joan Creus-Costa, says that the direct inclusion of data from their balloons and other sources is the main reason for the development of the new version of WeatherMesh. It took a year to modify and redesign the transformer-based model so that the model could deliver these predictions without losing stability.
“When we started doing it [data assimilation] we were still very dependent on ECMWF,” said Dean.
The company suffered a scare last year when a United Airlines plane crashed into its balloons. While the plane suffered a bit of damage, no one was hurt, in part because WindBorne followed US rules about how big its sensor package could be. However, the company has now added transponders to its balloons that report their location via the global aviation surveillance system, ADS-B, in an effort to reduce the chances of another accident.
WindBorne, which has raised $25 million in funding for a reported $85 million by 2024, sells the balloon data to NOAA, where it is used in the U.S. weather forecasting business, as well as the U.S. Air Force and Navy. The company also sells its forecasts to investors and commodity traders, but Dean says the company is still focused on building its model and data infrastructure on top of commercial products, in part because of the changing nature of information.
“I’m not trying to invest a big team in building a SaaS product, if the way people are looking for consumer data two years from now is using an agent, right?” Dean said.
If you shop through links in our articles, we may earn a small commission. This does not affect our editorial independence.
