Metadata: a case study from the environmental sciences

Environmental data are increasingly being processed automatically into derived producers which are then used for a variety of scientific purposes. Lack of adequate documentation of the quality control and transformation algorithms can seriously diminish product credibility and utility, particularly for studies of global change which require consistent information over many decades. Formal modeling of the concepts, algorithms, and data structures provides an approach for increasing the quality and reducing the burden of structuring and providing appropriate metadata. Pilot studies of seemingly simple examples have revealed significant conceptual issues such as the need for a science-based theory of approximate equivalence among different finite representations of the same physical object, and for a careful division of roles between scientists and software designers when implementing support for physical units as metadata.