Prospective monitoring for land use change prediction: Applications of spatiotemporal pattern analyses to land transaction data
MetadataShow full item record
Effective urban growth management requires fundamental information on how the landscape has changed, as well as how it will change in the future. Past change is useful for the study of drivers and environmental consequences of land use changes, while prediction of future change is necessary to develop proactive strategies for urban growth management or against undesirable land use changes. Given that current monitoring methods have only focused on describing what change has already occurred, the primary goal of this study is to propose a new approach to statistically predict future land use changes. Prospective monitoring methods were developed by combining three space-time pattern analyses and a cumulative sum method. The analytical procedure of the local Knox was customized by introducing a likelihood method and the space-time scan statistic was applied only to a new individual event instead of scanning the whole study area. A new method, the space-time chain statistic, was devised to delineate the irregular shape of a cluster. In the application of the methods, a set of spatial/chain and temporal thresholds were determined for the three methods. The maximum likelihood ratio method was used to choose the most likely cluster candidate among statistics from all combinations of spatial and temporal thresholds. The simulation was repeatedly performed 1,000 times to obtain the expectation for each observation. The methodology was then applied to monitoring land transactions for the years of 2002 through 2004 in Deokjin-gu of the city of Jeonju in Korea. The results of the prospective monitoring methods showed similar patterns on cumulative sum charts. A close examination of the behavior of the cumulative sum charts with z -scores revealed that the cumulative sum method signaled a little late, because it needed to accumulate deviations until the sum exceeded a critical value. A test for the assumption of normality revealed a positive skewness. Thus, significant clusters were determined in a complimentary way using both z -scores and p -values. The investigation of land use changes on the detected clusters revealed that the spatiotemporal clustering of land transaction was associated with future land use changes even though each monitoring method illustrated strengths in a different year with different predictability. This implies that the prospective monitoring can be useful for proactive urban growth management or effective land use planning because it provides information about future land use changes.