Voting-based ensemble-averaging visualization for water mass distribution

The distribution of water masses has become one of the most important topics in recent oceanic research. Water masses flow dynamically and have interannual variability, and hence the distribution can change dramatically even over short time periods and may differ from year to year. In this paper, we use a high-resolution ocean dataset, which contains multiple ocean variables, to visualize the details of the water mass. Because water mass can be defined by multiple ocean variables (e.g., temperature and salinity), we develop a multi-variate visualization system, which allows us to extract the time-varying distributions of water masses from multiple variables. The visualization is then adjusted by multiple ocean specialists because directly applying the existing definition to extract the water mass would result in incorrect rendering results. This leads to another problem that different ocean specialists may have different perspectives on the distribution of water masses, so that the adjustments would also be different. To solve this problem, an ensemble average process is performed for the adjusted rendering results from multiple ocean specialists. To increase the authenticity, we also add a voting scheme to the system, so that a majority rule can be applied to the ensemble-averaging result. As the application of the proposed voting-based ensemble-averaging visualization system, we first show the interannual variability of the significant water mass and then visualize the dynamic behavior for the period of interest in different years. We also highlight a mixing phenomenon that has a strong influence on the distribution of the water mass. As a result, we can obtain a clear and accurate visualization of the water mass distribution.