An approach to early warning of slowly evolving crises with reference to food shortage forecasting

An approach is described for forecasting slowly evolving food crises for which an early warning index is assumed to change linearly with time. The difficulties faced in decision making are outlined. It is seen that the `α-β tracker' can provide optimal estimates of the early warning index and its rate of change. Equations are developed for obtaining the optimal estimates, the corresponding variance-covariance matrix, and estimates of false alarm or failure-to-detect probabilities at each decision point. Decisionmakers are also provided with information relating to the decisions of whether and when to acquire further data for another crisis-no crisis decision.