Abstract
This study examines how land-use change simulation outcomes can vary based on the way the simulation model is applied, attempting to support informed model choices and model applications. This is accomplished through a series of experiments using a hypothetical model that represents the basic logic of various cell-based dynamic land-use change modeling environments. In the experiments, consideration is given to the sensitivity of the simulation results with respect to the following four application specifications: (1) the spatial resolution, (2) the temporal resolution, (3) the probability distribution, and (4) the degree of the influence of stochastic factors, under multiple growth scenarios. The experiments show that all four factors, particularly the spatiotemporal resolution and the degree to which stochastic factors are involved, can generate substantial variation in the simulation model outcomes. It is also found that the magnitude of the variation can be affected by changes in regional growth rates and the level of fluctuation, which determine the demand for new development to be allocated over the simulation time horizon.
Acknowledgment
The author thanks the editors and anonymous referees for their helpful comments.
Notes
1. 1. To be consistent, the number of cells is adjusted depending on the spatial scale used in the experiments. For instance, in the case of the 72 × 72 scale (i.e., Setting 2, high resolution), is set to 16/year (i.e., even distribution of the total 320 cells), which is equivalent to four 36 × 36 cells per year.
2. 2. In fact, map comparisons and the identification of differences have long been placed at the center of the spatial modeling and GIS literature (see e.g., Pontius Citation2000, Power et al. Citation2001, Hagen Citation2003, Hagen-Zanker Citation2006, Pontius and Cheuk Citation2006, White Citation2006, Remmel Citation2009, Ruiz et al. Citation2012).