A Computational Topology Approach to Automatic Partition and Visualization of Oceanic Water Masses

This paper addresses the issue of precisely locating boundaries of water masses in the oceanic water mass analysis. A novel computational topology-based solution is presented. First, employing the concept of multidimensional transfer functions, the sampled data of temperature and salinity in investigated area are transformed into a feature function of water masses that represents the variation ratio distribution of temperature and salinity on the domain. Then, a 2-manifold Morse-Smale decomposition over this domain is made to produce boundaries among the different water masses automatically. The results of some comparative experiments verified effectiveness and objectivity of the presented method, and its much better accuracy and efficiency.