Categorical maps are preferred because fragmentation studies are based on landscape metrics. This are calculated with algorithms that operate on categorical maps composed by fragments of different cover type, by programs or extensions like Fragstats and Patch Analyst. More recently, there is a preference for working with surface metrics, which are obtained from continuous maps; this should allow to work with topographic maps for purposes other that human impact on natural vegetation cover. I recommend reading the paper by McGarigal et al. 2009. Surface metrics: an alternative to patch metrics for the quantification of landscape structure, Landscape Ecology 24:433–450.
Because topography does not get fragmented, it is vegetation cover that may get fragmented. Certainly,,land forms do change with time, but usually at a much slower rate than land cover, and in a continuous manner, while the concept of fragmentation is applied to discrete classes, such as land cover types.
As I mentioned 12 days ago, vegetation cover maps are also analyzed with metrics developed for understanding continuous landcover maps. Topography changes continuously (unless a barrier such as a terrace or a river cuts the landscape) and this metrics could be used to understand associations between topography and other operating factors, and not necessarily changes in time.