ATMOSMARINE has great expertise in the calibration of atmospheric and ocean dynamic models and developed a new approach for the forecasting of environmental parameters through the combination of dynamic models calibrated in with statistical models. This hybrid approach (numerical-statistical) manages to use the knowledge of present-time physics, through the source terms of the models, to isolate the error of the numerical model (residue) and characterize these residues on simulations using Neural Networks (RNs).

 

This system was named SOPHIA (Hybrid Forecasting Operational System with Artificial Intelligence) and the results obtained can be applied in the prediction of extreme events or specialty products. This methodology has been proved and published in a few international scientific journals.