The VisAD data model is designed to express any numerical scientific data, so that different applications based on the VisAD library can share data or be merged. The key is the integration of abstract mathematical metadata, including data organization, units, coordinate systems, sampling topology and geometry, missing data indicators and error estimates. These abstract forms of metadata can be used to express higher-level scientific metadata, such as map projections, navigation and calibration for data from satellites, radars and other sensors, regular and irregular grid organizations for model data, dates and times, and physical data units.

Because of their level of abstraction, VisAD metadata also fit nicely with programming language syntax. We have done this in our Python programming interface for VisAD. Like IDL and Matlab, operations on entire arrays are supported so for example one imge may be subtracted from another. However, in the VisAD Python interface, metadata are implicit in such opperations. So for example when arrays are subtracted, if the images have different earth navigations or units, then unit conversions and resampling are done automatically.