Volatile climatic conditions, poor soil fertility and low resource endowments are major causes of poor crop productivity on smallholder farms in southern Africa. Despite these barriers, innovative pathways are needed to produce more food per unit area of land, while rebuilding soil fertility. The objective of this study was to use climatic as well as agronomic data from a long-term trial at Monze Farmer Training Centre in Zambia as a case study to illustrate the existence and impact of climate variability on maize productivity under conventional tillage and conservation agriculture (CA) in southern Africa with the aid of the Agricultural Production SIMulator (APSIM) model. Analysis of historical climatic data (1980-2010) showed that the 2001-2010 period had higher rainfall variability (CV = 13.6) compared with the previous decade (CV = 8.2), and the incidences of either too low or too high rainfall were also more frequent in the 2000s compared with 1980s. Both maximum and minimum temperatures increased by about 0.5 °C from 1980 to 2010. The effect of this variability was well simulated by the APSIM model which showed that in severe water limiting environments such as the devastating drought of 1991/1992, no cropping system could achieve any meaningful grain yield. However, in the 1989/1990 season the conventional tillage treatment recorded no grain yield whereas 0.2 t ha-1 were achieved with CA. However in the wettest season in 2003/2004, conventional tillage outyielded CA by about 0.3 t ha-1. These results suggest that CA has potential to mitigate against moisture stress but may depress yields when moisture is abundant. Generally, the rainy seasons are starting late which reduces significantly the planting window thus the option of staggering planting dates to deal with climatic uncertainties may not be viable in the future. Thus, intercropping systems may play an important role in dealing with climate uncertainty. (Texte intégral)