In sub-Saharan Africa, forecasts of regional or national scale malnutrition are prepared using a variety of ‘Early Warning Systems’, based upon supply-side data such as crop forecasts and satellite images of vegetation growth. Three agencies are developing more localised targeting systems using historical indicators to prepare ‘vulnerability assessments’ and so predict malnutrition at sub-national scales. This work argues that an alternative approach of short-term simulations of local food systems, may offer benefits.
The design of a modelling framework to carry out such a simulation, for a rural community of Zimbabwe, is presented together with the associated data requirements. The thesis reviews the current literature concerning food security, particularly the monitoring of food shortages, the targeting of emergency food aid and the economic and nutritional perspectives of the causes of malnutrition. The extent of spatial and temporal variability amongst households is analysed from primary survey data. The design implications of this variability and of the hierarchical structure o f the rural socio-economy and grain trading are discussed. Two versions of the modelling framework are reported, the first using systems dynamic modelling and the second using expert systems simulation techniques.
The first framework uses the UNICEF diagram of the ‘malnutrition-infection complex’ to develop the central component of the simulation. The second framework combines a ‘rulebase’ of household and com m unity behaviour with rainfall and health statistics to effect changes upon a database of households, data for which are extrapolated from the primary survey and secondary data obtained. The effectiveness of the framework and the direction of future work thereon are discussed.