Analyzing spatiotemporal land use change using an urban growth model based on multilevel logistic regression and future land demand scenarios

This study establishes an urban growth model using future land demand scenarios based on multilevel logistic regression (MLR) in the Seoul metropolitan area and then explores the effects associated with spatiotemporal land use changes in different scenarios. We divide urban growth factors into spatial allocation and demand variables and use the MLR to identify spatial allocation factors and distribute the probability of urbanization. Then, we estimate the land required for urbanization depending on future population and density scenarios. This study predicts the spatially urbanized land area in 2030 using data collected across a span of 40 years. We obtain mean absolute error values of 6.52%, 6.55%, and 5.15%, Cohen's kappa (k) values of 0.60, 0.66, and 0.79, and mean demand error values of 30.30%, 20.41%, and 12.17 for the late 1990s, late 2000s, and late 2010s, respectively. Our analysis finds that the largest portions of urbanized land were previously agricultural and forest land, and we predict such urbanization patterns will continue until 2030. In 2030, urbanized land is expected to expand around principal cities. Although agricultural land will tend to be urbanized near cities with extensive plains, high ratios of forest lands will be urbanized in cities near Seoul.

Automated summary

This study establishes an urban growth model using future land demand scenarios based on multilevel logistic regression (MLR) in the Seoul metropolitan area and explores the effects associated with spatiotemporal land use changes in different scenarios. The study predicts that urbanized land, previously agricultural and forest land, will continue to expand until 2030, with urbanization patterns being influenced by the proximity of cities to plains and forests.