Refining Landscape Change Models through Outlier Analysis in the Muskegon Watershed of Michigan

Balancing natural resource protection and urban development is of concern to researchers, planners and citizens who are aware of the environmental, social and economic impacts of urban land use. Land-use change models can assist in finding this balance. An objective of this research was to build a better model of land-use change by integrating quantitative and qualitative techniques. A modelling approach is presented that combines statistical logistic regression with field-based outlier analysis. To this end, a collaborative effort between researchers, who are adept at building models, and local experts, who better understand the dynamics of landscape change in their communities, was undertaken. The findings indicate that this modelling approach is successful in improving overall model performance, as measured by pseudo r-squared value, and identifying additional drivers of land-use change, namely zoning, soil suitability and distance to highway interchange. Most importantly, this collaborative modelling process, involving researchers and local planners, has practical utility for land-use decision making.