In March 2023, weather forecast monitoring company ForecastWatch awarded the Microsoft Start Weather Development Team the title of “World’s Most Accurate Global Forecast Provider”. It seems the Microsoft Start team isn’t resting on its laurels. This week Microsoft announced that the team has invented a more accurate way to predict the weather up to 30 days in advance.

B Bing blog postMicrosoft says its Start team has published a paper on a new medium-term forecasting model on the subject arXiv Cornell University website. It shows how this new model compares to the current one used by the European Center for Medium-Range Weather Forecasts (ECMWF). In summary, Microsoft’s new system combines the use of five different artificial intelligence models and three deep learning architectures to arrive at their predictions.

Microsoft says these AI models can take decades of weather data so they can recognize patterns that are familiar today and then predict what they will do in the future. It added:

However, they work in the same way as a NWP model: given the current state of the atmosphere on a 3D sphere (latitude, longitude and altitude), predict the state of the atmosphere for some future time, say one hour later. They then feed that forecast back into the model to forecast two hours later, and so on.

Microsoft’s AI models can run predictions faster with GPU usage and, as a result, can be run more frequently. In theory, this should lead to better predictions.

So far, Microsoft claims that its methods for measuring temperature errors are 17% more accurate than those used by the ECMWF for weekly weather forecasts and 4% for four-week forecasts. The company will add this model for 30-day weather forecasts into its Microsoft Start model.