Although such a partitioning may seem unimportant, or even arbitrary, the two components are identified with reason, as they develop at different rates, while simultaneously affecting the other, either directly through competition for water or through light interception, or indirectly through their effect on, say, fire.
The models proposed by Walker & Noi-Meir (1982) and Scholes & Walker (1993) differ substantially. Walker & Noi-Meir propose that the herbaceous and woody vegetation components obtain their water from different soil layers; the herbaceous layer intercept any infilltration before it becomes available for the woody plants. While this division seems to be appropriate for the savannah ecosystem type, Knoop (1982) was unable to support it for the savannah woodland ecosystem at Neylsvlei, South Africa, where woody plants showed little response to the removal of herbaceous plants.
Scholes & Walker (1993) subsequently proposed a temporal division. Many woodland plants flower and produce leaves before the onset of the rains (Childes 1989). This growth is either supported by root reserves, but probably also by soil moisture from the previous season. Condensation is generally found on soil from even 40cm in depth, that is kept in plastic bags even at the end of the dry season (pers obs). Grass plants sprout later, unless they are stimulated to do so by fire.
The decision of subdividing the vegetation components is only a starting point towards explaining a very complex system. This model facilitates the navigation through the factors that influence the development savannah woodland system, showing the dependencies and interactions between the various components of the system.
As the model is followed through its various links, feedback loops will cause the model to jump from the woody component to the herbaceous component, and back.