Realistic Tropical Cyclone Wind And Pressure Fields Can Be Reconstructed From Sparse Data Using Deep Learning

Tropical cyclones are a major source of destruction and loss of life, motivating the need for accurate risk assessment and forecasting. Here, we show that a physics-informed neural network can be a promising and computationally efficient algorithm for tropical cyclone data assimilation. Using synthetic training data and real-time observations, the neural network is able to accurately reconstruct full realistic 2- and 3-dimensional wind and pressure fields which capture key features of the cyclone. Our results demonstrate how recent advances in deep learning can augment data assimilation schemes and be applied to other flow problems.