Modelling urban growth over time using grid-digitized method with variance inflation factors applied to spatial correlation

Abstract

© 2016, Saudi Society for Geosciences. Analysis of land use change over time is useful information to support urban planning and management policies. Most land use modelling studies have used polygonal data structure. The main limitation of polygonal data structure is that it is difficult to measure changes in land use. This study proposes another method for predicting land use change. This method is based on an analog-to-digital conversion which replaces polygonal shapes by coded grid points. The method is applied to data from a survey of Phuket province from 1967 to 2009 where land use was classified broadly as forest, agriculture, urban, water bodies and miscellaneous land. Logistic regression was used to predict a binary land use outcome (urban/other), and location combined with land use at a previous survey was a determinant. To account for correlation in land use amongst nearby plots of land, variance inflation factors were used to compute standard errors of proportions of urban growth. The result of the present study discloses that greater urbanization was observed in the southern parts of Phuket during the period of study, and surprisingly, that reforestation occurred in 1985–2000. This study shows that analog-to-digital conversion methods are useful approaches to develop appropriate statistical models for land use change.