SciVL: A Descriptive Language for 2D Multivariate Scientific Visualization Synthesis

We present SciVL, the Scientific Visualization Modeling Language, a new way to describe scientific visualizations of multivariate, twodimensional datasets. Our goal was to create a language with which a user could rapidly prototype complex and precise data-driven visualizations as well as facilitate their modification during the iterative design process. Our language can combine three types of basic visual elements: discrete icons, color planes, and streamlines in layers. A text file fully describes the resulting image and controls the different elements’ layering, appearance, relative locations, and spacing, including the mapping assignments of any of these characteristics to data variables. In addition the language supports extensions to include other basic visual elements and to accommodate 3D data. An interdisciplinary scientific visualization course that synthesized the efforts of computer scientists as well as graphic design and illustration students inspired us to choose the three basic types of visual elements. A review of graphic design literature and feedback from expert visual designers provided us with an anecdotal evaluation of our design’s adequacy for the visual language. We have applied our language to fluid flow and polarimetric radar data visualizations. These visualizations elucidated some issues in our framework, such as limited handling of different types of 2D data and the perceptual efficacy of our iconand streamline-placing algorithm. In sum, the visual language we present, in conjunction with the simple text interface used for generating the visualizations, accomplishes our primary goal of providing a way to quickly describe and modify complex data-driven visualizations for twodimensional, multivariate datasets.