Scaling of classification systems—effects of class precision on detection accuracy from medium resolution multispectral data

Abstract Context Land-cover class definitions are scale-dependent. Up-scaling categorical data must account for that dependence, but most decision rules aggregating do not produce scale-specific definitions. However, non-hierarchical, empirically derived classification systems common in phytosociology define classes using species co-occurrence patterns. Objectives Evaluate tradeoffs precision and representativeness when up-scaling across natural landscapes the multi-dimensional grid-point (MDGP)-scaling algorithm, which generates definitions; compare spectral detection accuracy of MDGP-scaled to ‘majority-rule’ aggregated classes. Methods Vegetation maps created from 2-m resolution WorldView-2 two Everglades wetland areas were scaled 30-m Landsat grid with MDGP-scaling algorithm. A full-factorial analysis evaluated effects class-label on compositional information loss multispectral data. Results MDGP?scaling retained between 3.8 27.9% more than majority rule as increased. Increasing retention also increased 1 8.6%. Rare removal increase similarity controlled by threshold, leading higher Conclusions When landscapes, negotiating trade-offs thematic precision, landscape-scale map results greater class-detection lower-resolution, multispectral, remotely sensed provides a framework weigh make informed decisions parameter selection.