AbstractA key component of cellular automata (CA) models is the transition rules that determine the transformation of cells at each iteration. However, most previous studies use a single set of transition rules across the entire study region, and therefore do not fully account for spatial heterogeneity. In this research, a vector CA model has been implemented that calibrates transition rules by taking the entire study region and partitioned sub‐regions into consideration. The changes in residential areas were modelled for the city of Ipswich, Queensland, Australia, from 1999 to 2016. The results confirm that the spatially partitioned rules can generate more accurate and stable results compared to calibrated rules using the whole study area, with an increase of mean producer's spatial accuracy of 72.07 and 75.59% in two sub‐regions (2‐2 and 2‐3). The implementation of CA models with partitioned transition rules enables a better understanding of spatial heterogeneity in land use change.