Almost all data is affected by uncertainty — either of epistemic or aleatoric character, or both. While graphs of R → R functions mostly show error bars, visualization methods for high-dimensional data typically do not consider uncertainty. An important subfield of data visualization is feature-based visualization, i.e. the extraction and visualization of, e.g. geometrical or topological structures. In this project we aim at finding probabilistic equivalents to crisp features for scalar, vector and tensor fields fraught with uncertainty, quantifying these probabilistic features and devising methods for their interactive visualization.