Hazardous weather events affecting populated coastal are among the most devastating natural disasters in terms of mortality and economical losses due to their low predictability. Currently, the generation of useful predictions, reliable and anticipated of hazardous weather events affecting populated coastal regions remains an ambitious challenge for the scientific community. Deficiencies in the accurate prediction of such events are tightly related with the initial value problem, which states that better the state of the atmosphere is estimated, the more accurate the forecasts. This problem is addressed by using advanced Data Assimilation (DA) techniques, which play an important role in current numerical weather prediction and is currently at the forefront of atmospheric and oceanic sciences research. However, although using the most current sophisticated DA algorithms, the estimation of the atmosphere is not accurate enough to improve the predictability of hazardous weather events, mainly because their linear and Gaussian underlying assumptions. The main aim of the present project is to go beyond the state of the art in DA by developing and implementing a novel and advanced DA technique that takes nonlinearities and non-Gaussianities into account, enabling us to to improve high-impact weather forecasts. The new DA will be tested in real cases in combination with a high-resolution atmospheric model to improve the predictability of several poorly forecasted Mediterranean Hurricanes. This novel technique will significantly improve global, regional, and climate forecasts. The applicant’s strong mathematical and theoretical skills in DA together with his broad experience running numerical weather models using HPC facilities will facilitate the achievement of the key goals of this proposal. This project will also expand the applicant’s experience, research competencies and professional networks, enhancing the development of his career as an independent researcher.