Alternative representations of in-stream habitat: classification using remote sensing, hydraulic modeling, and fuzzy logic

Improved techniques are needed to characterize complex fluvial systems and monitor ecologically important, yet highly vulnerable riverine environments. This paper explores potential alternatives to traditional mapping of in-stream habitat and presents fuzzy set theory as a means of departing from the rigid, Boolean, object-based framework. We utilize hydrodynamic modeling, remotely sensed data, and fuzzy clustering to obtain classifications that allow for continuous partial membership and gradual transitions among habitat types. Methods of assessing cluster validity are available, but data quality is a crucial consideration. Crisp, vector-based representations can be derived from raster fuzzy classifications by applying a threshold to maximum membership values. This process results in conditional objects separated by ambiguous transition zones, and a compromise must be reached between the proportion of the channel assigned to polygons and the certainty with which this assignment can be made. Spatial patterns of classification uncertainty can also be used to identify areas of confusion, infer boundaries of variable width, and highlight areas of increased habitat diversity. Hydraulic modeling and remote sensing complement one another and, together with field work, could provide a more realistic representation of the fluvial environment.