Final Report Summary - SAFERVIS (Uncertainty Visualization for Reliable Data Discovery)
In the ERC Advanced Grant “SaferVis” we have considerably improved the status quo in visual data analysis by innovative ideas for next-generation technologies that provide uncertainty visualization as a core methodology. We have developed visualization techniques for the communication of the uncertainty in scientific data and features derived from such data. To achieve this, we have derived probability distributions of the occurrence of features using stochastic uncertainty models and ensembles of data representations. For a graphical depiction, we have developed visual mappings for statistical measures and feature distributions, and we have developed new approaches to visualize these quantities in combination with the initial data in a spatio-temporal context.
We have also shed light on the question whether uncertainty visualization can help to discover quantitatively the data generation process. To this end, we have pursued research on the in-situ connection of uncertainty visualization techniques to a running simulation. We have developed high-performance numerical simulation techniques for fluids and deformable solids, and we have used these techniques to interactively change input parametrizations, and to instantly show the sensitivity of the simulation results to these changes and the uncertainty that is represented by the resulting ensemble simulations. In collaboration with researchers from meteorology we have successfully used the developed ensemble visualization techniques to investigate the uncertainty that is represented by a set of simulated samples of the possible states of the atmosphere.
We have also shed light on the question whether uncertainty visualization can help to discover quantitatively the data generation process. To this end, we have pursued research on the in-situ connection of uncertainty visualization techniques to a running simulation. We have developed high-performance numerical simulation techniques for fluids and deformable solids, and we have used these techniques to interactively change input parametrizations, and to instantly show the sensitivity of the simulation results to these changes and the uncertainty that is represented by the resulting ensemble simulations. In collaboration with researchers from meteorology we have successfully used the developed ensemble visualization techniques to investigate the uncertainty that is represented by a set of simulated samples of the possible states of the atmosphere.