Some Key Assumptions
the training data are representative
the predictor variables are appropriate
the effects of missing variables are implicit in the available variables
the neural net architecture is reasonable
that there is a systematic relationship between environmental variables and land use that is modellable