Mozambique is a semi-arid area with unreliable rainfall distribution; therefore optimal planting dates are critical to ensure that maize is not stressed during critical stages. The objective of this research was to study the effect of sowing date and cultivar on maize (Zea mays L.) yields in Mozambique. A further objective was to establish whether the SWB model could be utilized to help select the optimum planting window for different maize cultivars and localities. An experiment was conducted during the 2007/08 season at the Chókwè Agricultural Research Station, Mozambique, in which a short (or early cultivar, Changalane) and long (or late) season maize cultivar (Tsangano) were sown on three different dates: 5 December 2007 (PD1), 25 December 2007 (PD2) and 15 January 2008 (PD3). Sowing date had a significant effect (p<0.05) on yield and yield components. The 25 December planting (PD2) out yielded (4.3 t ha-1) the 5 December (PD1) (2.5 t ha-1,) and 15 January (PD3) (1.5 t ha-1) plantings for cv. Changalane. However, for cv. Tsangano, PD1 (3.2 t ha-1) out yielded PD2 (2.3 t ha-1) and PD3 (0.7 t ha-1). Cultivars varied significantly in yield potential. The most responsive cultivar to water supply was Changalane, which when planted late in December (PD2), gave a water productivity (WP) of 17 kg ha-1 mm-1, while Tsangano, the late cultivar, performed better when planted early in December (PD1), with a WP of 8.5 kg ha-1 mm-1 The Soil Water Balance (SWB) model was calibrated on the data from one planting date per cultivar and successfully validated on independent data sets from the other two planting dates. Long-term historical weather data sets were obtained for Chókwè and Umbeluzi, two important dry land maize production areas in Mozambique. The calibrated SWB model was used to simulate maize yields for different planting dates to establish the best planting date for different cultivar x plant date x soil combinations. Simulation results for the two cultivars across three planting dates showed that the simulated grain yields per planting date varied substantially from year to year and between the two sites. The SWB scenario simulation results showed that for both Umbeluzii and Chókwè sites, in four out of five years, best yields can be achieved by planting Changalane late in December and Tsangano early in December. It can be concluded that the SWB model can be a very useful tool to help select the most suitable maize cultivars and planting dates for different localities, based on differences in plant water availability during the growing season. Copyright / Dissertation (MSc(Agric))--University of Pretoria, 2011. / Plant Production and Soil Science / unrestricted