Abstract
Automated processing of environmental data is hindered by the wide array of unit representations provided in the metadata of digital datasets. For example, gm/m2, g/m2, gm-2, g/m^2, g.m-2 and gramPerMeterSquared are all representations of a single complex unit that might be human-readable but are not machine-interpretable. Connecting ad hoc units to a single unit concept in an ontology permits the identification of datasets sharing units and provides additional information regarding labels, definitions, dimensions and transformations provided in the ontology. Here we use successive string transformations to link ad hoc unit representations to units in the QUDT ontology (e.g., unit: GM-PER-M2). Although only 896 of 7,110 distinct units in a corpus of ecological metadata from DataONE, the Environmental Data Initiative and the U.S. National Ecological Observatory Network were matched, 324,811 unit uses (instances) out of 355,057 of total unit uses were successfully mapped to QUDT units (91%). The resulting lookup table was used to enable a web service and R functions for adding annotation elements to Ecological Metadata Language documents.
