We compared three different soil water models to evaluate the extent to which variation in plant growth form and cover and soil texture along a topographic gradient interact to affect relative rates of evaporation and transpiration under semiarid conditions. The models all incorporated one-dimensional distribution of water in the soil and had separate functions for loss of water through transpiration and soil evaporation but differed in the degree of mechanism and emphasis. PALS-SW (Patch Arid Lands Simulator-Soil Water) is a mechanistic model that includes soil water fluxes and emphasizes the physiological control of water loss by different plant life forms along the gradient. 2DSOIL is a mechanistic model that emphasizes the physical aspects of soil water fluxes. SWB (Soil Water Budget) is a simple water budget model that has no soil water redistribution and includes simplified schemes for soil evaporation and transpiration by different life forms. The model predictions were compared to observed soil water distributions at five positions along the gradient. All models predicted soil water distributions reasonably well and, for the most part, predicted similar trends along the transect in the fractions of water lost as soil evaporation versus transpiration. Transpiration was lowest (about 40% of total evapotranspiration (ET)) for the creosote bush community, which had the lowest plant cover (30% peak cover). The fraction of ET as transpiration increased with increasing plant cover, with 2DSOIL predicting the highest transpiration (60% of total ET) for the mixed vegetation community (60% peak cover) on relatively fine textured soil and PALS-SW predicting highest transpiration (69% of total ET) for the mixed vegetation community (70% peak cover) on relatively coarse textured soil. The community type had an effect on the amount of water lost as transpiration primarily via depth and distribution of roots. In this respect, PALS-SW predicted greatest differences among stations as related to differences in plant community types. However, since PALS-SW did not provide as good of fit with the soil moisture data as did 2DSOIL, the differences in the morphology and physiology of the life-forms may be secondary to the overall control of water loss by the primary factors accounted for in 2DSOIL: vertical distribution of soil moisture, degree of canopy cover, and evaporative energy budget of the canopy. Soil texture interacted with the amount and type of plant cover to affect evaporation and transpiration, but the effect was relatively minor.¿ 1997 American Geophysical Union