Visualization of multi-scale dynamics of hydrous cold plumes at subduction zones

Recent increases in the computational power of high-performance computing systems have led to a large gap between the high-resolution runs of numerical simulations—typically approaching 50–100 million tracers and 1–5 million grid points in two dimensions—and the modest resolution of 1–2 million pixels for conventional display devices. This technical problem is further compounded by the variety of fields produced by numerical simulations and the limited bandwidth available through the Internet in the course of collaborative ventures. We have developed a visualization system using the paradigm of web-based inquiry to address these mounting problems. We have employed, as a case study, a problem involving two-dimensional multi-scale dynamics of hydrous cold plumes at subduction zones. A Lagrangian marker method, in which the number of markers varies dynamically, is used to delineate the many different fields, such as temperature, viscosity, strain, and chemical composition. We found commercially available software to be insufficient for our visualization needs and so we were driven to develop a new set of tools tailored to high-resolution, multi-aspect, multi-scale simulations, and adaptable to many other applications in which large datasets involving tens of millions of tracers with many different fields are prevalent. In order to address this gap in visualization techniques, we have developed solutions for remote visualization and for local visualization. Our remote visualization solution is a web-based, zoomable image service (WEB-IS) that requires minimal bandwidth while allowing the user to explore our data through time, across many thermo–physical properties, and through different spatial scales. For local visualization, we found it optimal to use bandwidth-intensive, high-resolution display walls for performing parallel visualization in order to best comprehend the causal and temporal relationships between the multiple physical and chemical properties in a simulation.