| Land use change has caused traumatic environmental damages in many destination areas and thus threatens sustainable tourism development. Reliable prediction of land use change is essential to a successful planning and management of destination land use. However, due to the complex and chaotic nature of land use systems, our ability to predict land use change in the conventional way is decreasing as the rate and scale land use change continues to increase. There is a need for alternative approaches and models.; This research proposed a parcel GIS-based multinomial logistic model for destination land use prediction. Based on the random utility theory and a logistic framework, the model explicitly took tourism related factors into account to better represent destination land use systems; it utilized parcels as the units of analysis to enhance spatial processions and prediction accuracy; and it was integrated with GIS to make prediction practice more efficient. The model was tested in the Murrells Inlet area with a five-category land use system using 24 predictor variables representing tourism factors, parcel characteristics, service accessibility, market factors, government regulations, and initial conditions.; The results indicate that the model has generated a fairly good prediction for parcel land use, with 80.76% of parcels and 94.29% of the area correctly predicted. Prediction success rates are better for major land uses (vacant land, residential, accommodation, and commercial) than for the minor uses (institutional). Tourism related factors, as a whole, were found closely related to spatial differentiation of land use. Beachfront location, major roads, residential neighborhood, housing unit value, and vacant land are significant predictors of land use in the coastal destination area. Parcels as the spatial units were found suitable for the local-level land use modeling because they are the real world objects, the basis of land transaction, and the smallest legal land units. GIS proved to be useful in integrating different data sources, deriving variable layers, and visualizing the predicted results during the modeling process. Although there are many issues yet to be resolved, the research suggests that the parcel GIS-based multinomial logistic model is applicable for destination land use prediction. |
