Abstract This paper addresses water uptake by roots as calculated in summary crop growth models for use at the land use systems level. Simple comparative exercises were carried out using three models. The results show considerable differences in daily and cumulative water uptake. One model calculates moisture changes for subdivisions of the root zone, even though the procedure on which it is based was intended for use over the rooted zone as a whole. This is combined with an algorithm to allow for compensatory effects in soils with non-uniform water distribution, resulting in artefacts. The results of another model based on the electrical analogue, confirm that theoretical models help to improve understanding of complex processes and guide research, but using them for predictions may give worse results than conventional empirical methods. The third model uses a ‘soil water depletion fraction’ to indicate the limit of readily available water in the rooted soil as a whole. The procedure was originally developed for irrigated agriculture and seems inappropriate for rain fed conditions characterised by irregular wetting and drying cycles that result in non-uniform soil water distribution. Simple comparative exercises as carried out, help to reveal weak points in the models studied and indicate ranges of uncertainty. Error propagation analyses using local data must be added to indicate whether improvements at local level are necessary and to identify priorities for further research, feeding an upward spiral to enhance model accuracy.