Search

Abstract Numerous machine learning-based land change models have been presented by researchers over the last two decades. To date, however, far less have simulated multiple land use classes and specific land use subclasses at the same time. In some areas of the world, it is important to simulate both of these dynamics to understand fully the drivers and consequences of land change. One important example is the process of urbanization in China, as urban policies have been developed that guide urban expansion, rural protections, and urban subclass development. This paper presents a new model integrating geographic information systems (GIS) with artificial neural networks (ANNs) to predict multiple transitions among land use types and urban subclasses in the Zhanggong District of Ganzhou city in China. We show that the model produces satisfactory goodness of fit values-based on location, quantity and spatial configuration-between simulated and observed land use maps for 2015. Our simulated future maps produced by the model for 2020 and 2025 demonstrate that transitions from farmland and forest to urban will remain the main pathway of urbanization although we predict that the rate will slow after 2025. The goals of urban planning should be aligned with land use planning according to "Combining multiple laws and regulations" in China. Taking into account the current and future land use transitions will enhance the accuracy and timeliness of land use policy making and urban land planning. For the sustainable land use in Zhanggong District, we argue for a strengthened regulation of the land market by government and believe that planning officials should guide the spatial distribution of land supply actively. Furthermore, improving the production, living and ecological functions of land resources are the key points to optimize urban land use functions and the allocation of land resources. We discuss how our model can be adapted to other areas to benefit land use management and urban planning in China. Highlights • GIS and ANNs were used to predict multiple land use transitions and urban subclasses. • A case study demonstrates that the model performs relatively well validated with several measures for goodness of fit. • The simulated maps suggest that farmland and forest are the most common candidates for urban expansion. • The methodology can be adapted to other study areas to benefit land use management and urban planning in China. [ABSTRACT FROM AUTHOR]