Identifying the Impact of Natural Hazards on Food Security in Africa: Crop Monitoring Using MODIS NDVI Time-Series

Most developing countries rely primarily on the successful cultivation of staple crops to ensure food security. Climatic hazards like drought and flooding often negatively impact economically vulnerable economies such as those in Eastern Africa. Effective tracking of food production is required in this area. Production is typically quantified as the simple product of a planted area and its corresponding crop yield. To date, crop yields have been estimated with reasonable accuracy using grid-cell techniques and a Water Requirement Satisfaction Index (WRSI), which draw from remotely sensed data. However, planted area and hence production estimation remains an arduous manual technique fraught with inevitable inaccuracies. In this study we present ongoing efforts to use MODIS NDVI time-series data as a surrogate for greenness, exploiting phenological contrast between cropland and other land cover types. In regions with small field sizes, variations in land cover can impose uncertainty in food production figures, resulting in a lack of consensus in the donor community as to the amount and type of food aid required during an emergency. To concentrate on this issue, statistical methods were employed to produce sub-pixel estimation, addressing the challenges in a monitoring system for use in subsistence-farmed areas. We will discuss two key results. Firstly, we established an inter-annual evaluation of crop health in primary agricultural areas in Kenya. These estimates will greatly improve our ability to anticipate and prevent famine in risk-prone regions through the FEWS NET early warning system. A primary goal is to build capacity in high-risk areas through the transfer of these results to local entities in the form of an operational tool. The low cost and accessibility of MODIS data lends itself well to this objective. Monitoring of crop health will be instituted for use on a yearly basis, and will draw on MODIS data analysis, ground sampling and valuable local expertise. Secondly, a baseline map of cropped areas was established, utilizing MODIS time-series data, Landsat ETM+ data and a custom dot-grid sampling method. This product aids in disaggregating crop location and density, and establishes a nominal quantitative assessment of farming practices. The techniques used to generate these results for Kenya can be expanded for use throughout developing Africa and beyond.