Problems associated with the application of monthly rainfall-runoff models to arid and semiarid areas for water resource estimation purposes are discussed with respect to the representation of hydrological processes and the model rainfall input. The problems are illustrated using a specific monthly model applied to catchments within the semiarid to arid parts of southern Africa. Some model improvements are suggested and briefly evaluated and the application of a daily model to one of the catchments is compared. The overall conclusion is that while a detailed examination of the available rainfall data can account for many of the inadequate simulation results on an individual month basis, it is difficult to make allowances for the lack of resolution in the normally available information. Other problems related to the structure of individual models can be surmounted by limited model developments, but very little can be done to bypass the rainfall quantification problem.