Modeling Land Use Change: Using Knowledge Discovery and Cellular Automata in a GIS Environment

Abstract

We develop a parcel-based spatial land use change prediction model by coupling machine learning and interpretation algorithms such as cellular automata and decision tree in a Geographic Information System environment. We collect and process historical land use data and various driving factors that affect land use changes in Hunterdon County of New Jersey using decision tree J48 Classifier to develop a set of transition rules that illustrate the land use change processes during the period 1986-1995. Then we apply the derived transition rules to the 1995 land use data in a cellular automata model Agent Analyst to predict spatial land use pattern in 2004. We validate these by the actual land use in 2002. The developed decision tree-based cellular automata model has a reasonable overall accuracy of 84.46 percent in predicting land use changes. It shows a much higher capability in predicting quantitative changes (92.5%) that location changes (74.8%) in land use. With such an encouraging measure of validity, we use the model to simulate the 2011 land use patterns in Hunterdon County based on the actual land uses in 2002. We build two scenarios: the “business as usual” scenario and the “policy” scenario (with imposed government policy). The simulation results show that successfully implementing current land use policies such as down-zoning, open space, and farmland preservation could prevent 973 agricultural and 870 forest parcels (a total of 2,856 hectares) from future urban encroachment in Hunterdon County during the period 2002-2011. It becomes a significant policy instrument for government to reckon with. 

Key words: land use change, cellular automata, decision tree, parcel, geographic information system, J48 Classifier, Agent Analyst, Hunterdon County