South Africa is classified by the FAO as an arid region. The western interior's annual rainfall is about 400 mm. The region has an aridity index of 0.17 according to the UNESCO conference of 1977. Farmers rely largely on rainfall for crop production. The region is blessed with deep sandy soils, however, which can store water above an impermeable clay layer at a depth of about 3 m. The purpose of this paper is to highlight the practices that have been adopted to harvest rainfall and use it effectively for sustaining long-term crop production. An experiment was carried out over four years to study different components of the water balance during the fallow period, between consecutive crops, as well as during the growing season. Two sites were selected, both with sandy soils, with clay plus silt content between 5 and 8% in the top 0.3 m and increasing to about 20% at 3m depth. One site at Hoopstad has an impermeable layer at 3 m while the other at Petrusburg has no impermeable layer within a 6 m depth. Crops are planted only when the water table rises to a depth of 0.6 m. Wheat was grown in monoculture with three different cultivation practices: a conventional treatment with a mouldboard ploughing and mechanical weed control and two conservation tillage treatments – one stubble mulch and the other no tillage. The conservation treatments were aimed at leaving as much crop residue on the soil surface as possible to prevent wind erosion, inhibit evaporation and increase rainwater infiltration. Owing to the soil texture, the soil tends to compact easily; leading to severe problems associated with shallow root development and reduced effectiveness of stored water exploitation. A deep cultivation action was consequently included in the conventional and stubble mulch treatments to alleviate this problem. Weed control through the fallow period was done mechanically with different implements in control and stubble mulch treatments while chemical control was used for the no till treatment. Grain and total dry matter yields were determined over all four years of the experiment. From these data and water balance data, water use efficiency (WUE) and rainfall use efficiency (RUE) were calculated for each season. The rainfall storage efficiency (RSE) was determined for the different fallow periods. The results showed that the RSE is mainly determined by the amount and distribution of rainfall. The rainfall both during each growing season and during intervening fallow periods at both sites was below the long-term average although wheat yields exceeded the long-term average for the area. This was mainly due to the water harvesting practices that were introduced. Wheat yields at the site with a perched water table were three times higher than those at the site without a water table. The contribution of stored water from the water table was estimated to be about 70% on average. The RUE on the soil without a water table was calculated to be 5.2 kg ha-1 mm-1 while that of the soil with a perched water table was 12.9 kg ha-1 mm-1. The conventional cultivation practice at the soil with a perched water table has a RUE of 15.1 kg ha-1 mm-1 which was significantly higher than that of the stubble mulch and no till treatments which were 12.2 and 11.4 respectively. No significant differences occurred between treatments at the site without a water table.