1. General Model Information

Name: Classification and Regression Tree model

Acronym: CART

Contact:

Author(s):

Abstract:

The CART (Classification and Regression Tree) model is a predictive, statistical model used to classify remote sensing information. It employs statistical regression techniques (e.g. stepwise multiple regression) to construct dichotomous decision trees. Training set data are used as input to produce the decision trees. The typical types of training set data used are geology, soils, and topography.

The model produces a classification decision tree on its first pass. For example, it would split all pixels that are classified as vegetation into forest and non-forested pixels. Then, using the training set data, it can split the pixels classified as forest into forest types, such as coniferous and deciduous. On its second pass the model assesses which of the decision tree branches are statistically significant. Those determined to be not significant are "pruned."