The works confirm the opening of new horizons for AI techniques used in operational forecasting. These cutting-edge ideas combine deep learning (from statistics observed in the past) with computational geometry (highly efficient computing structures) applied to the design of their algorithm architectures. La implementación de AI en predicción meteorológica en Europa se ha ido desarrollando en las últimas décadas y, con los recientes resultados de ambos trabajos, compiten directamente con los métodos tradicionales.

However, despite the undoubted advances, these types of techniques could present limitations in the adequate prediction of events that have never happened before. This point is critical especially in the current context of abrupt climate change (with oceanic warming never recorded before). To ensure the greatest success of deep learning, these techniques need to be combined with the dynamic equations of numerical weather prediction models. The combination of statistical and dynamical techniques is key to model uncertainty and reduce it through weightings according to the skill shown in each case.