In drought-prone arid and semi-arid areas, limited plant available water exerts a tremendous negative effect on crop production, leading to undesirable low crop productivity, untold food insecurity, and never-ending poverty. In-field rainwater harvesting (IRWH or In-field RWH) is specifically designed to trap rainfall within the field and optimise its use to benefit crop yield and quality, and improve water use efficiency (WUE) in these regions. Two RWH-crop field experiments were established in the semi-arid area of the Hatfield Experimental Farm, University of Pretoria, South Africa. The first RWH-potato experiment was conducted during the 2009/2010 growing season while the second RWH-Swiss chard experiment was carried out during the 2010/2011 growing season. Three cropping systems were involved: (1) conventional tillage (CT), (2) tied-ridges (TR), and (3) IRWH with three different design ratios of runoff area to cropping area (1:1, 2:1 & 3:1). The runoff area of each design ratio was either bare (B) or plastic-covered (P) and this resulted in six IRWH treatments. Therefore, there were a total of eight treatments: CT, TR, 1:1B, 1:1P, 2:1B, 2:1P, 3:1B and 3:1P. For both growing seasons, the total plot area yields and WUEs of TR and CT were in general higher than those of the IRWH treatments. This is because TR and CT had more plants per plot than the IRWH treatments and the rainfall recorded for the specific seasons were sufficient, so there was little advantage in collecting/harvesting additional water. In terms of yields and WUEs expressed on the net cropped area, the IRWH treatments had higher yields and WUE than CT and TR because they captured more runoff than the latter treatments. Field trials are expensive, laborious and time consuming, therefore models were developed to predict potential runoff and crop growth and yield of different RWH techniques or design ratios. During the current investigation, runoff models such as the linear regression, curve number (CN) and Morin and Cluff (1980) models were used to describe and simulate runoff generation from this ecotope. The empirical rainfall-runoff linear regression model indicated that runoff efficiency declined as runoff length increased. The statistics revealed that the CN and Morin and Cluff (1980) models simulated runoff very well. Moreover, the use of a generic crop growth Soil Water Balance model (SWB) showed potential to simulate crop growth and yield for different RWH techniques and design ratios. During the present study, the SWB model was modified by incorporating linear runoff simulation models in order to predict the soil water balance and crop yield under different RWH design scenarios. Field data collected on the study ecotope contributed to the parameterization and calibration of the SWB model for the crops involved. The SWB model was in general, successfully calibrated for the potato crop, while the calibration for the Swiss chard crop was generally not as successful, most probably because of the continuous growing and harvesting system followed (approach for pastures). The scenario simulation results for potato suggested that for the study ecotope, if land is limiting, CT, TR and smaller design ratios (1:1) are the best options in terms of yield per total plot area. However, if land is not limiting, larger design ratios (2:1 and 3:1) are better options, according to the yields per net cropped area outcomes. The SWB model shows promise as a useful tool to assist in the selection of the best RWH strategy and the ideal planting date under specific conditions with minimal input requirements. However, there is a need to upgrade it to a 2D SWB model for better accuracy under a range of conditions. / Dissertation (MScAgric)--University of Pretoria, 2014. / lk2014 / Plant Production and Soil Science / MScAgric / Unrestricted