CA-based urban growth model considering the temporal dynamic adjustment of local spatial driving factors: An application in Wuhan City

Abstract

Cellular Automaton (CA) is widely used because of its ability to simulate complex spatiotemporal dynamic processes through applying simple rules. The basis of the CA model is the definition of transformation rules. During a simulation process, the rules determine the change of the cell state. However, existing processing methods calculate the driving factors based on single-point time (start time or end time), making it difficult to reflect the fact that numerous driving factors affecting the cell conversion dynamically change with time. Based on the time dynamics perspective and the data set of multiple time series, this paper designs a method of dynamic adjustment of driving factors of urban expansion on the local cell-scale. It uses linear, exponential, logarithmic, and polynomial fitting to develop a CA model of dynamic adjustment that conforms to the characteristics of local spatial evolution. The main conclusions of the paper are as follows: (1) The polynomial fitting has the highest average R2, indicating that the driving factors experiences large fluctuations over time; (2) Secondly, the simulation result kappa obtained by the four fitting methods is between 0.781–0.810, which is higher than the simulation accuracy obtained by using only a single time point. In other words, the factor does not dynamically fit with time and (3) The fitting accuracy of road density is a key indicator of correct and incorrect simulation parts of construction land. Our results demonstrate that the precision of the CA model may be significantly improved by capturing the time development law of environmental variables affecting urban development at the micro-scale.