Thesis (M. Sc. (Botany))--Rhodes University, 2000.

Phenomenological models are presented which predicts transpiration rates (E) of individual leaves of Scaevola plumieri, an indigenous dune pioneer, and Ammophila arenaria, an exotic grass species introduced to stabilise mobile sand. In both cases E is predictably related to atmospheric vapour pressure deficit (VPD). VPD is calculated from measurements of ambient temperature and humidity, hence, where these two environmental variables are known, E can be calculated. Possible physiological reasons for the relationships of E to VPD in both species are discussed. Scaling from measurements of E at the leaf level to the canopy level is achieved by summing the leaf area of the canopy in question. E is predicted for the entire canopy leaf area by extrapolation to this larger leaf area. Predicted transpiration rates of individual shoot within the canopy were tested gravimetrically and shown to be accurate in the case of S. plumieri, but less so in the case of A. arenaria. Using this model, the amount of water used by a known area of sand dune is shown to be less than the rainfall input in the case of S. plumieri in wet and dry years. The water use of A. arenaria exceeds rainfall in the low-rainfall year of 1995, while in 1998 rainfall input is slightly higher than water extraction by the plants. Using a geographic information system (GIS), regional maps (surfaces) of transpiration were calculated from surfaces of mean monthly temperature and mean monthly relative humidity. Monthly surfaces of transpiration were subtracted from the monthly median rainfall to produce a surface of mean monthly water deficit. Areas of water surpluses along the coast correspond with the recorded distribution of both species in the seasons that the plants are most actively growing and reproducing. This suggests that unfavourable water availability during these two species growth periods limit their distributions along the coast. In addition to unfavourable water deficits, additional climatic variables that may be important in limiting the distribution of these two species were investigated using a discriminant function analysis.