This thesis compares the in situ water balance of common annual production systems (wheat, lupin, subclover and serradella) with a grazed perennial system (lucerne) at a site in the Avon Catchment, Western Australia. Using a physically-based water balance approach the value of a plant based solution in redressing the hydrological imbalance that has become a feature of much of the dryland agricultural region of Western Australia is investigated. The effectiveness of lucerne in providing greater available storage for buffering large rainfall events, as compared to the annual systems, is illustrated. Continued transpiration following out-of-season rainfall events maintains a larger available storage capacity. In contrast, the annual systems that are fallow over summer only withdrew a small fraction of water by soil evaporation between rainfall events. Under annual systems, the profile moisture store was sequentially increased to the extent that additional increments of rainfall could potentially contribute to deep drainage. A particular focus of this study has been to investigate the presence of soil constraints to root growth, and to assess how these constraints affect the water balance. A site survey indicated the soil penetration resistance was sufficient to impose a physical constraint to root growth. Published literature on the site shows soil acidity is also at a level imposing chemical constraints to root growth. A root growth model “Rootmodel”, for predicting root growth with and without soil constraints is examined in detail as a method for providing root growth parameters for inclusion into the numerical water balance model, SWIM based on Richard’s equation. Functions developed from “Rootmodel” adequately describe the effect of profile limitations to root growth, such as soil strength, moisture availability and temperature. Recommendations are made for inclusion of a growth suppressing function in “rootmodel” based on the chemical limitation of low pH. The effects of soil acidity on the root growth of several species is investigated experimentally and the resultant root data provided a reference point by which the simplified prediction of root growth built into SWIM could be adjusted using a linear reduction function. A similar linear reduction function is also employed to impose a physical constraint in the form of high penetration resistance.