SLEUTH Case Study

6.1
In this section four discussion points are elaborated. Which are: the inclusion of policy as a factor in the MCE; the calibration of the self-modification module; the spatial resolution; and land use modeling. The latter was not explicitly mentioned throughout the report, but there still one important remark ought to be made regarding this component of SLEUTH.
Policy
It becomes evident that none of the excluded layers were able capture the actual urban dynamics that have taken place in the study area. As none of the excluded layers not seem to be able to capture the dispersed patterns of the study area. The core of the issue is that the growth coefficient values are uniformly applied over the entire study area, even though these values
This last optional phase is often recognized as an important component of SLEUTH-3r, as it would allow SLEUTH-3r to simulate dynamic growth rates over time, so that growth rates replicate the typical S-curve of urbanization and population growth (Silva & Clarke, 2002). From literature, there is the original application of SLEUTH when self-modification was calibrated to the San Francisco Bay Area by Clarke et al. (1997). The default self-modification values are inherited from this application, of an earlier SLEUTH version applied to a different study area (Clarke, et al. 1997). Additionally, it may be likely that due to changes made in the transition rules in the newer versions of SLEUTH and SLEUTH-3r, the default self-modification values may not always be fitting. Candau (2002) reports extensively on the self-modification module, all its different parameters and what they do. She even suggest that the self-modification module ought to have its own calibration procedure, as at that point it did not. Therefrom, there are only brief mentions of self-modification and calibration, e.g. Jantz et al. (2014) and Liu et al. (2012) who both altered the parameter settings in their attempt to improve modeling results. Liu et al. (2012) state: “These additional parameters also contribute to the model accuracy, and the significance of their influence, determine that a method to calibrate them is
The key argument to be made here is that, this only occurred under specific conditions, i.e. the application of fuzzy set theory whilst coarsening the resolution. This led to more agreement between comparing clusters of cells as opposed to cell-by-cell comparison. Therefore, in this study, this was only possible due to the availability of finer data, whence the coarser data came from. Also, due to the stochastic mechanisms in SLEUTH, random pixels are randomly selected during the various growth cycles. This is the case for all of the four growth types. Imagine an existing village observed urban growth shaped like a chessboard pattern, and SLEUTH would have predicted that same chessboard pattern during simulation, just one row or one column off. Validating the simulation results would produce zero percent agreement, whilst a human visually checking the data would be inclined to think SLEUTH did a great job. This reason this happens could partly be attributed to SLEUTH’s stochastic mechanisms that is randomly selecting pixels to urbanize in the correct more area, albeit on the wrong specific location. One way to deal with this is coarsening the spatial resolution via fuzzy set theory. Then, the agreement would be perfect if the resolution would be two times the cell side length. However, if input data would have been coarser to begin with, additional coarsening would be required to resolve nearby allocation disagreement