U-Net Kalman Filter (UNetKF): An Example of Machine Learning-Assisted Data Assimilation

Machine learning techniques have seen a tremendous rise in popularity in weather and climate sciences. Data assimilation (DA), which combines observations and numerical models, has great potential to incorporate machine learning and artificial intelligence (ML/AI) techniques. In this paper, we use U-Net, a type of convolutional neutral network (CNN), to improve the localized error covariances for the Ensemble Kalman Filter (EnKF) algorithm. Using a 2-layer quasi-geostrophic model, U-Nets are trained using data from EnKF DA experiments. The trained U-Nets are then successfully implemented in U-Net Kalman Filter (UNetKF) experiments to predict localized error covariances that possess adaptive localization and some state-dependent features of the model error covariances. UNetKF is compared to traditional 3-dimensional variational (3DVar), ensemble 3DVar (En3DVar) and EnKF methods. The performance of UNetKF can match or exceed that of 3DVar, or En3DVar and EnKF for small to moderate ensemble sizes. We also demonstrate that trained U-Nets can be transferred to a higher-resolution model for UNetKF implementation, which again performs competitively to 3DVar and EnKF, particularly for small ensemble sizes.