The impact of class resolution in land use change models

A recent focus of land use/land cover research is the design and validation of spatially explicit predictive models. An often overlooked aspect of model development is the role of class resolution. The objective of this paper is to evaluate the impact of changing the number and breadth of land use classes on a model's calibration and predictions. To address the objective, three model specifications were developed representing urban development in the Barnegat Bay watershed (New Jersey, USA). The models range from one based on a single broad resolution class to one using six fine resolution classes. The results of the analysis indicated that changing the level of class resolution impacts the model's calibration parameters and predicted outcomes, but more finely defined urban conversion classes did not uniformly improve the accuracy of the model. If fine resolution classes are identified, however, the specific types of conversions that are not well captured by the model are revealed. The paper ends with a discussion of the broader implications of class resolution decisions in land use/land cover models. (C) 2009 Elsevier Ltd. All rights reserved.