Food insecurity continues to be a chronic problem in Southern Africa and particularly in the Southern African Development Community (SADC) region. Across the region, food shortages are reported every year, especially in countries including Malawi, Swaziland and Zimbabwe. In this research, Malawi was selected as a case study to understand how a sustainable and improved cereal production under rainfed conditions can be achieved by means of water productivity enhancement. The research focused on two of the major cereals grown in the country: maize, which is the staple food for the country, and sorghum, which is regarded as a drought-tolerant crop that survives under adverse climatic conditions. Poor weather conditions coupled with high population growth and low soil fertility are the major contributory factors to deteriorating food security in Malawi. Although irrigation has the potential to increase agricultural productivity, the technique is underutilised. Rainfed agriculture still dominates in Malawi making food production only possible in the rainy season from November to April. In addition to that, most of the soils are highly weathered and require regularly enhancement of their fertility. Although potential crop yields are high, the average yield in smallholder farming of the last 10 seasons is only 1.6 ton ha-1 for maize and 0.7 ton ha-1 for sorghum. To make rainfed agriculture the main source of food and livelihood security for rural communities, the yield gap must be reduced. The research started with an analysis of the socio-economic characteristics of the smallholder farmers in the region. A field survey of 60 farmer households in the Lilongwe and Shire Valley Agricultural Development Divisions (ADD), whose main crops are maize and sorghum, was conducted. These ADDs with different climates and agronomical management practices were selected as pilot areas. Results indicate that the smallholder farmers are poor and face considerable limitations in their farming practices. Nevertheless, they strive to improve their livelihoods despite the myriad constraints they face. They have small sized plots for their household food security. The farmers tend to favour staple foods for their survival rather than cash crops as they focus mainly on subsistence farming. A major constraint is the fertility status of the soils which are cultivated every year with little or no replenishment in terms of fertilizer enrichment. Even though most of the farmers use inorganic fertilizers, the quantities applied are usually too small to have a pronounced effect on their yields. The effect of rainfall variability on the length of the crop growing period (LGP) over the past three decades was analysed. Data from five meteorological stations in the central region of Malawi, where 90% of the economic activities are agro-based, were analysed. The analysis showed significant changes in the onset, cessation and length of the growing period. There is a clear delayed onset and advanced cessation, and thus a shorter LGP, in most locations with time within the period considered (1980 to 2009). Since farmers are willing to learn and adapt to the effects of climate change, so that their livelihoods can improve and not greatly impacted, the results of this analysis and the consequent recommendation for introducing crop cultivars with a shorter growing cycle, may be useful to farmers. Next, a calibrated and validated crop growth model, AquaCrop, was used in this research to analyse crop yield gaps and to generate crop management strategies to improve and stabilize crop yields. The multi-crop water productivity model AquaCrop was developed by the Food and Agriculture Organization of the United Nations (FAO) to address food security and assess crop production influenced by environment and management. Two, field experiments with maize and sorghum were set up during three successive growing seasons (2010/11, 2011/12 and 2012/13), both at Bunda (Lilongwe ADD) and Kasinthula (Shire Valley ADD) to (i) evaluate the effect of fertility levels on rainfed crop yield and (ii) obtain field data for fine tuning and validation of AquaCrop for Malawian conditions. The field experiments had two levels of fertiliser application: full dose (F1) and half dose (F0) according to the recommendations of the government extension service. Also the effect of different crop varieties (early, medium and late maturing) were studied in some of the years. As expected, there was a significant increase in yield of maize and sorghum with higher fertilizer application. The experimental data of the F1 treatments from 2010/11 were used for fine-tuning the AquaCrop model to the environmental conditions in Malawi. The F0 treatments were used for calibrating the soil fertility stress module of the model. For model validation, data from 2011/12 and 2012/13 were used. Different statistical indicators (correlation coefficient r², relative root mean square error RRMSE and Nash-Sutcliffe model efficiency EF) showed that the model performed excellent in simulating biomass, soil water content, canopy cover and grain yield of maize and sorghum. It was concluded that AquaCrop was successfully calibrated and validated for maize and sorghum for Malawi and that the model can be used for formulating and evaluating different strategies and their effects on crop production. AquaCrop was subsequently used to assess the yield stability for maize and sorghum for the current weather conditions in the region. The simulations were run for the two study sites and for three fertility levels (i.e., F1 and F0 as considered in the field experiments, FM as applied in farmers’ fields, which implicates a much lower fertility level). Crop yields under FM are between 1.9 to 3.0 ton ha-1 for maize and 2.0 to 2.3 ton ha-1 for sorghum, which is higher than what is reported by the government studies (1-2 ton ha-1). This is due to the absence of the effect of pests, diseases and weed infestation in the simulations with AquaCrop. With full soil fertility (F1), the production can be doubled. Yet, while very good yields can be expected in good rainy years, the crop yield will be lower than under FM strategies in the drier years. Under all management strategies, the occurrence of failure years is relatively high, i.e. almost 1 year out of 10 years for Bunda and 2 years out of 10 years for Kasinthula. To study the effect of climate change on cereal production, local-scale climate projections for the future were generated for central Malawi. Climatic change factors from 15 global climate models (GCMs) from the Coupled Model Intercomparison Project phase 3 (CMIP3) were used. The GCM output was downscaled to local-scale future data following two distinct methodologies, i.e. the self-organising maps (SOM) approach by the University of Cape Town-Climate Systems Analysis Group (UCT-CSAG) versus the stochastic weather generator LARS-WG. Finally, the future climatic data generated by LARS-WG were used to assess the effect of climate change on cereal production. The effects of future climate change on maize and sorghum yields at Bunda were assessed by use of the AquaCrop model the mid-21st century (SRES scenarios A1B). Significant differences in mean yield as compared to the baseline were found. Maize will be impacted negatively while sorghum will benefit from climate change. The projected rather small yield decline of about 5% for maize and the yield increase of 2% to 10% for sorghum contradict the often projected sharp decline of cereals in Southern Africa. Given inconsistencies found between the simulated yields under observed versus generated baseline weather data by LARS-WG, the yield decline of maize might be somewhat larger and the yield increase of sorghum might be slightly smaller. Finally, it has to be noted that despite the small increase or decrease of yield, the occurrence of failure years will almost double from 0.7 year out of 10, to 1.2 years out of 10 with climate change for both maize and sorghum. Acknowledgements iii Abstract v Nederlandstalige samenvatting vii List of Figures xi List of Tables xv List of acronyms xix Table of Contents xxiii Part I Introduction 1 Chapter 1 Problem statement and research questions 3 Part II Environment 9 Chapter 2 Study area 11 Chapter 3 Social-economic status of farmers in the study area 21 Chapter 4 Effect of rainfall variability on the length of the crop growing period over the past three decades 31 Chapter 5 Future Climate Downscaling 51 Part III Model 65 Chapter 6 Crop yield response to soil water and soil fertility stress 67 Chapter 7 Fine-tuning and validation of AquaCrop 79 Part IV Simulation and Assessment 95 Chapter 8 Simulation of historical yields 97 Chapter 9 Assessing the effect of climate change on cereal production 105 Part V Conclusion and Perspectives 119 Chapter 10 General conclusions and recommendations 121 References 125 Annex I: Farmer survey questionnaire 141 Annex II: Experimental plot layouts 155 Annex III: Statistical analysis of future climate data for LARS-WG approach 157 Annex IV: Publication list 159 nrpages: 185 status: published