2. Temporal Predictive Modelling

It seems clear that it would be possible to account for temporal differences only by doing modelling as a series of time slices. However, this approach is only a partial solution to creating a true temporal model, as it still essentially treats time as a static, rather than a dynamic, process. There are both conceptual and software limitations to handling time dynamically. Because of these limitations, true temporal GIS is still something of a future desire rather than a current reality. Therefore, even though this solution is at best partial, it is better than treating large masses of time homogeneously.

The time slice approach has been adopted by White (2002), who attempted to deal with the temporal aspect of predictive modelling. However, White takes a significantly different approach to modelling time, in that he uses a series of palaeoenvironmental reconstructions with each of the time slices used. While these could provide for more robust models, in that they would be more likely to explain settlement patterns rather than simply correlate with aspects of the modern environment, there are significant difficulties in taking a palaeoenvironmental approach. First, palaeoenvironmental data are fairly rare and often not extensively collected over a region. In the case of the study region, much of this sort of reconstruction comes from lake cores (as discussed in Petch et al. 2000), which can mean a limited number of samples are extrapolated over large regions. Second, it is questionable whether pollen counts accurately reflect plant communities, as some pollens travel further or are released in greater amounts. Since palaeoenvironmental data suffer from scale and sampling problems, a correlative approach is adopted here.

While the addition of a temporal aspect of modelling is possible, it creates a number of logistic difficulties. These must be considered before the creation of a temporal model.