1. Any scheme for structuring data that is used to group individuals. In ecological and taxonomic studies numerical classification schemes have been devised, but various hierarchical or non-hierarchical classificatory strategies have also been used. In taxonomy, the fundamental unit is the species. Among living forms species are groups of individuals that look alike and can interbreed, but cannot interbreed with other species. In palaeontology, where breeding capability cannot be determined, species are defined according to morphological similarities. In formal nomenclature, taxonomists follow the binomial system developed by the Swedish naturalist Carolus Linnaeus (1707–78). In this system each species is defined by two names: the generic (referring to the genus) and the specific (referring to the species). Thus various related species may share a common generic name. Genera (sing. genus) may be combined with others to form families, and related families combined into an order. Orders may be combined into classes, and classes into phyla (sing. phylum) or divisions in the case of Plantae (although this level is omitted in modern angiosperm classification). For example, the brachiopods comprise some eleven orders split between two classes and these two classes are the major subdivisions of the phylum Brachiopoda. The basic groupings, the phyla, are combined together into kingdoms, e.g. Plantae (the plants), Fungi, and Animalia (animals), and the kingdoms into domains. Some workers have tackled the uncertainties arising from subjectivity in classification by using numerical methods. In their view, if enough characters were measured and represented by cluster statistics, the distances between clusters could be used as a measure of difference. Even so, the worker has to decide (subjectively) how best to analyse the measurements, and so objectivity is lost. Other workers emphasize those features shared by organisms that show a hierarchical pattern (see cladistics).
2. In remote sensing, the computer-assisted recognition of surface materials. The process assigns individual pixels of an image to categories (e.g. vegetation, road) based on spectral characteristics compared to spectral characteristics of known parts of an image (training areas). Assignation of pixels is not always possible when the parameter space of different training areas overlaps. In such cases a principal component analysis prior to classification may be used to allow better separation of training areas by increasing the overall parameter space. See also box classification; minimum-distance-to-means classification; maximum-likelihood classification.