Microsoft’s Start weather development team has been recognized as the “World’s most accurate global forecast provider” by ForecastWatch in March 2023. Building on this achievement, Microsoft has now introduced a method to enhance weather predictions up to 30 days in advance. The team’s groundbreaking research, detailed in a paper on the arXiv site at Cornell University, showcases a medium-range forecast model that significantly improves upon the current system utilized by the European Centre for Medium-range Weather Forecasts (ECMWF).
Innovative Approach to Weather Prediction
The new model developed by Microsoft employs a combination of five artificial intelligence (AI) models and three deep learning architectures to generate its forecasts. This approach allows the system to analyze decades of weather data, identifying patterns and predicting future atmospheric conditions. Similar to Numerical Weather Prediction (NWP) models, Microsoft’s AI-based system forecasts future atmospheric states by analyzing the current state across three dimensions: latitude, longitude, and altitude. By predicting the atmosphere’s state at future intervals and iteratively feeding this information back into the model, the system can forecast weather with increased accuracy and speed, thanks in part to the utilization of GPUs for faster computation.
Enhanced Accuracy and Future Plans
Microsoft’s new forecasting model has demonstrated superior accuracy in temperature prediction compared to the ECMWF’s methods. The company reports a 17 percent improvement in one-week weather forecasts and a 4 percent enhancement for four-week forecasts. These advancements are significant, as they promise to deliver more reliable weather predictions to the public and industries reliant on accurate forecasts. Microsoft plans to integrate this model into its Microsoft Start news service for 30-day forecasts, aiming to set a new standard in weather prediction accuracy and reliability.
By leveraging AI and deep learning, Microsoft’s Start team is not only enhancing the accuracy of weather forecasts but also contributing to the broader field of meteorology. The development of such advanced forecasting models is crucial for planning and preparedness in various sectors, including agriculture, transportation, and emergency management.
Last Updated on November 7, 2024 8:41 pm CET
