Windborne Systems, a startup founded by Stanford students in 2019, has released its sixth version of AI weather forecasting tool WeatherMesh. The new model claims to be more accurate than traditional and AI forecasts produced by the European Centre for Medium-Range Weather Forecasting (ECMWF), which is currently considered the leading provider of accurate weather prediction.
WeatherMesh 6 produces a forecast every hour, with a resolution as low as 3 km in Europe and the continental US. This is compared to traditional models that generate forecasts every six hours. The tool has been shown to be particularly accurate on surface temperature measurements, outperforming ECMWF's traditional and AI systems five days out.
Windborne's unique combination of model-building and data collection has given it a competitive edge in the field of weather forecasting. The company has launched over 400 balloons around the world, gathering sensor readings that are fed into its deep learning models. This direct ingestion of data is key to the improvement in WeatherMesh 6, according to Windborne's head of AI, Joan Creus-Costa.
Windborne has raised $25 million in venture funding and sells its balloon data to NOAA and military organizations. The company also sells its forecasts to investors and commodity traders, but remains focused on building out its model and data infrastructure over commercial products.
