Abstract

To examine the effects of historical land-use change and static land-use state on the modeling, we established three cellular automata (CA) models using the spatial autoregressive model (SAR). The models are CA_SAR-Cha based on the change data, CA_SAR-Sta based on the start-state data, and CA_SAR-End based on the end-state data. The models that considered five different neighborhood sizes (from 3 × 3 to 11 × 11) were applied to simulate the urban growth of Jiaxing, China from 2008 to 2018, and predict the urban scenario to the year 2048. All three models can accurately reproduce the urban growth from 2008 to 2018, and the CA_SAR-End model performed best in calibration and validation. The differences in historical land data did affect the spatial distribution of the simulated urban patterns. The neighborhood size has a significant impact on the model's allocation ability, yet the appropriate size depends on the unique landscape context being studied.