DEEP LEARNING FOR LAGRANGIAN DRIFT SIMULATION AT THE SEA SURFACE

We address Lagrangian drift simulation in geophysical dynamics and explore deep learning approaches to overcome known limitations of state-of-the-art model-based and Markovian approaches in terms of computational complexity and error propagation. We introduce a novel fully-convolutional architecture, referred to as DriftNet, inspired from the Eulerian Fokker-Planck representation of Lagrangian dynamics. Numerical experiments for Lagrangian drift simulation at the sea surface demonstrates the relevance of DriftNet w.r.t. state-of-the-art schemes. Benefiting from the fullyconvolutional nature of DriftNet, we explore through a neural inversion based on a trained DriftNet how to diagnose model-derived velocities w.r.t. real drifter trajectories.