The impact of variation in scale on the behavior of a cellular automata used for land use change modeling

The cellular automata (CA) model is an important tool in land use studies. Scale is a central concept in both land use research and CA models. To better understand the connection between different scales and processes of land use change, more studies on the behavior of a CA model at different scales are necessary and urgent. The objective of this study was to explore the impact of variation in scale on the behavior of a CA employed for land use change modeling. The scale variation was represented by the variation in the spatial extent of the images used, in the cell size and in the neighborhood configuration, including the size and shape. Two aspects of the behavior of the model were examined: the algorithms expressing the land use transitions, and the precision of the model’s output. The average enrichment factor, which generates the transition rules, was used to quantify the interpretation of the land use transitions. The kappa index was used to quantify the precision of the model’s output. The results showed that the variation of the spatial extent, cell size, and neighborhood size and shape influenced the model’s behavior in different ways. A combination of small cell size and small neighborhood size generated improper expressions of the land use transitions. Here, we believe that an increase in the neighborhood size might be the solution. However, the precision of the simulation decreased when the neighborhood size increased with ring shape. In addition, the precision first increased and then decreased once the neighborhood size reached a certain value for a scope-shaped neighborhood. Based on these results, we recommend the finest resolution and a neighborhood size of 9 × 9 cells arranged in a scope shape. This set of parameters reflects the most accurate land information. The model interprets well the land use transitions that are identical to those found in actual “real” land transitions and also provides high-precision outputs. For different model applications at different spatial scales, a preliminary discussion of the scale parameter setting is recommended.