Cluster-Based Visualization for Merger Tree Data: The Challenge of Missing Expectations

Scientific simulations are yielding increasing amounts of data; to visualize the full output from a simulation, one must first reduce clutter and obstruction. Clustering algorithms are common tools for condensing information and decreasing clutter when analyzing and visualizing simulation output. Often, simulation data have intuitive groupings. In some cases, though, such as merger trees from N-body dark matter simulations, there are limited expectations for clustering results. We investigate cluster-based visualization design for merger tree data, testing whether multidimensional encodings and opening the "black box" can allow for meaningful representation and exploration of these data.