Nonparametric classifier for GIS data applied to kangaroo distribution mapping

A supervised nonparametric classifier, previously applied to classify remotely sensed data, is used to classify GIS layers. The algorithm is trained using GIS data layers as the independent variables, and predicts the spatial distribution of a dependent variable using a nonparametric technique. A GIS database of kangaroo distribution in Australia tests the algorithm. Results are satisfactory, with the presence of kangaroos being mapped with a producers accuracy of 93 percent for the western grey, and 100 percent for the eastern grey and red kangaroo. The algorithm appears robust to variations in training sample size and a priori probabilities.