Three-dimensional ocean observation is the foundation for accurately predicting ocean information. Although ocean observation sensor arrays can obtain internal data, their deployment is difficult, costly, and prone to component failures and environmental noise, resulting in discontinuous data. To address the severe missing data problem in three-dimensional ocean flow fields, this paper proposes an unsupervised model: Three-dim...