A new AI tool introduced by the startup WindBorne Systems is delivering more accurate weather forecasts than the world's leading intergovernmental organizations. Founded in 2019 by Stanford University students, the company has developed a revolutionary method for feeding sensor data into deep learning models. This is reported by Techcrunch.com reports.
The sixth version of the company's model, called WeatherMesh, is outperforming the systems of the European Centre for Medium-Range Weather Forecasts (ECMWF). Kai Marshland, WindBorne's Director of Product, notes that the five-day forecast of the WeatherMesh 6 model is as accurate as the one-day forecast of traditional systems. This is particularly evident in surface temperature measurements.
Traditional weather forecasting relies on complex physical models that require expensive supercomputers and are time-consuming to process. AI models created by large laboratories like Google DeepMind and startups run much faster. WindBorne distinguishes itself from competitors through an exclusive dataset collected via its fleet of nearly 400 weather balloons.
Currently, WeatherMesh 6 provides new forecasts every hour, whereas traditional models do so every six hours. Forecast accuracy in Europe and the US has been refined to 3 kilometers. WindBorne CEO John Dean emphasized the importance of owning proprietary data, stating that being an AI-based weather company without a data advantage makes no sense.
Read “Zamin” on Telegram!
