ERROR ASSESSMENT IN DECISION-TREE MODELS APPLIED TO VEGETATION ANALYSIS

Methods were developed to evaluate the performance of a decision-tree model used to predict landscape-level patterns of potential forest veg etation in central New York State. The model integrated environmental databases and knowledge on distribution of vegetation. Soil and terrai n decision-tree variables were derived by processing state-wide soil g eographic databases and digital terrain data. Variables used as model inputs were soil parent material, soil drainage, soil acidity, slope p osition, slope gradient, and slope azimuth. Landscape-scale maps of po tential vegetation were derived through sequential map overlay operati ons using a geographic information system (GIS). A verification sample of 276 field plots was analyzed to determine: (1) agreement between G IS-derived estimates of decision-tree variables and direct field measu rements, (2) agreement between vegetation distributions predicted usin g GIS-derived estimates and using field observations, (3) effect of mi sclassification costs on prediction agreement, (4) influence of partic ular environmental variables on model predictions, and (5) misclassifi cation rates of the decision-tree model. Results indicate that the pre diction model was most sensitive to drainage and slope gradient, and t hat the imprecision of the input data led to a high frequency of incor rect predictions of vegetation. However, in many cases of misclassific ation the predicted vegetation was similar to that of the field plots so that the cost of errors was less than expected from the misclassifi cation rate alone. Moreover, since common vegetation types were more a ccurately predicted than rare types, the model appears to be reasonabl y good at predicting vegetation for a randomly selected plot in the la ndscape. The error assessment methodology developed for this study pro vides a useful approach for determining the accuracy and sensitivity o f landscape-scale environmental models, and indicates the need to deve lop appropriate field sampling procedures for verifying the prediction s of such models.