Statistical Landscape Change Modelling
Statistical (empirical) models use observed the relationship between independent variables and a dependent variable (for example land cover type) to predict the future state of that dependent variable. The primary limitation of such an approach is the inability to represent systems that are non-stationary, i.e. systems in which the relationships between variables are changing through time. This assumption of stationarity rarely holds in landscape studies. Two primary empirical models are available for studying lands cover and use change; transition matrix models and regression models. My research has particularly focused on the latter, particularly the logistic regression model.

The figure above shows observed land cover for 3 years (1984 - 1999) for SPA 56, with a fourth map (2014) predicted from this data. Models run for observed periods of change for SPA 56 were found to have a pixel-by-pixel accuracy of up to 57%. As Gil Pontius has found, other models for observed change in SPA 56 were unable to perform as well as the null model of no change (i.e. assuming the landscape does not change). Thus, this map of future land cover should be understood as a projection of future land cover given that the observed trends continue unchanged into the future (i.e. the stationarity condition is maintained).

I have also worked with the hierarchical partitioning method to use statistical modelling in an explanatory capacity rather than for predictive purposes. Work on this has been published in the journal Ecosystems.