Machine-Learning-Assisted Hybrid Earth System Modelling
We propose a novel hybrid modeling approach that combines a physics-based, numerical model of the Earth system with a machine learning model. We also envision to integrate the hybrid model training with a data assimilation cycle, so that new observations update both the state estimate used as initial conditions and the machine learning model parameters.
58 GEOSCIENCES↗