Agricultural drought is characterized by shortages in precipitation, large differences between actual and potential evapotranspiration, and soil water deficits that impact crop growth and pasture productivity. Rainfall and other agrometeorological gauge networks in Sub-Saharan Africa are inadequate for drought early warning systems and hence, satellite-based estimates of rainfall and vegetation greenness provide the main sources of information. While a number of studies have described the empirical relationship between rainfall and vegetation greenness, these studies lack a process based approach that includes soil moisture storage.

In Chapters I and II, I modeled soil moisture using satellite rainfall inputs and developed a new method for estimating soil moisture with NDVI calibrated to in situ and microwave soil moisture observations. By transforming both NDVI and rainfall into estimates of soil moisture I was able to easily compare these two datasets in a physically meaningful way. In Chapter II, I also show how the new NDVI derived soil moisture can be assimilated into a water balance model that calculates an index of crop water stress. Compared to the analogous rainfall derived estimates of soil moisture and crop stress the NDVI derived estimates were better correlated with millet yields.

In Chapter III, I developed a metric for defining growing season drought events that negatively impact millet yields. This metric is based on the data and models used in the Chapters I and II. I then use this metric to evaluate the ability of a sophisticated land surface model to detect drought events. The analysis showed that this particular land surface model's soil moisture estimates do have the potential to benefit the food security and drought early warning communities.

With a focus on soil moisture, this dissertation introduced new methods that utilized a variety of data and models for agricultural drought monitoring applications. These new methods facilitate a more quantitative, transparent `convergence of evidence' approach to identifying agricultural drought events that lead to food insecurity. Ideally, these new methods will contribute to better famine early warning and the timely delivery of food aid to reduce the human suffering caused by drought.