![Geostrom](https://mobicheckin-assets.s3.amazonaws.com/uploads/events/647747350824f13e47fd4158/guests/65773f74bab0f615ba198ee4/guest_products/65c465fead3d9a79a37c5984/illustration/medium_geostorm.gif)
Geostrom
AI Application AI Technology AI for Society AI for society DECISION MARKETING DEEP LEARNING FORECAST'Geostorm' is a deep learning-based model for predicting typhoon tracks, demonstrating accuracy in 72-hour forecasts comparable to existing models. It utilizes real-time Unified Model data, setting it apart from most data-driven models that depend on reanalysis data. This approach represents a significant advancement in real-time typhoon trajectory prediction, overcoming the operational limitations of reanalysis data. The combination of data-driven path prediction models and a bias correction model for UM to ERA5 further enhances its effectiveness.