Neural network models for spatio-temporal geophysical fields reconstruction : application to sea surface temperature

In this work, neural network representation is investigated for the representation and assimilation of dynamical systems and the development of data-driven strategies to learn computationallyefficient representation of hidden dynamics.