A Knowledge-Based Approach to Urban Growth Modeling in Gorgan City Using Logistic Regression



The logistic regression (LR) method, as an empirical estimation method, was used to model urban growth in Gorgan city between the years 1987-2001. Three groups of variables including Social-economic, land use and biophysical variables were used. Similarity among variables is probable. To remove the correlated variables, covariance was calculated and distance to administrative, sporting centers and cities, with covariance greater than 0.9, were deleted from the group of independent variables. Relative operating characteristic (ROC), was used to assess success of modeling approaches and to assess the model sensitivity for independent variables remove. ROC value for the LR with full data was 0.87. Using probability image of urban growth predicted by the LR model, which shows likelihood for land use change to urban use in future; urban distribution patterns for the years 2010, 2020, 2030 and 2040 were created. New urban location is selected from the cells with the most likelihood for land use change. Also, the relative effect of the 10 predictor variables were evaluated through ROC using 10 reduced-variable models and the full model. Cultivated areas and pasture were the most influential variables in comparison with the others in driving urban growth. This shows that present land use has the most important rule in urban growth in Gorgan city.