Using spatial autologistic regression for predicting urban growth

ABSTRACT

This paper aims to contribute to the discussion around the statistical performance of spatial autologistic regressions. We provide empirical evidence using spatially explicit land use change data. The data is fitted using both the traditional logistic regression and the spatial logistic approach. Results show that the spatial autologistic regression outperforms the traditional logistic approach. We conclude that the results from the spatial autologistic regression run on our land use change data are preferable to those from the traditional logistic regression. Various estimates from the logistic regression are non-significant in the autologistic approach. This may provoke misleading actions among urban planners.

KEYWORDS:

• Spatial autologistic regression
• land use change model
• urbanisation

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. B: Binary|

C: Continuous.

2. Calculated within a distance covered by 3 neighbouring cells from the regression cell.

3. Reference case, therefore not included in the model calibration.

4. Calculated within a distance covered by 4 neighbouring cells from the regression cell.

5. Calculated within a distance covered by 6 neighbouring cells from the regression cell.