Robust optimization evaluation of reliance on locally produced foods

Abstract

While local food production may be beneficial in terms of developing the local economy and reducing greenhouse gases from transportation, sustainability strategies focused on local food production may generate their own risks due to yield variability. We have developed a robust optimization (RO) model to determine the minimum amount of land (cropland and pasture) required to grow food items that would satisfy a local population’s (accounting for gender and age) calorie and nutrient needs. This model has been applied to Boone County, Missouri, which has a population of approximately 170,000. Boone County is 1790 km², with 16% of the land defined as cropland and 30% defined as pasture. The model includes 27 nutrients from 17 potential foods that could be produced: six fruits and vegetables, five grains and six animal-sourced foods. Yield estimates are based on the predominate methods of agriculture in the USA. We first run our model assuming no variability, using the midpoint yield estimates. Then, to quantify uncertainty in yield for different food types, we use historical yield data over 10 years to estimate this variability and run our RO model under these variability estimates. We compare the two model results to illustrate the impact of data uncertainty on meeting sustainable local food for communities. Solutions suggest that nutrition needs can be met for the Boone County population within the land area defined.