A risk assessment procedure is described for use in managing a system of pumped-storage reservoirs in the Thames basin during a drought. Historical daily rainfall sequences are used as equi-probable scenarios of future rainfall. These are transformed to flow, reservoir level, and demand restriction sequences through the use of rainfall-runoff and water resource system models. The risk assessment information required is then obtained through a statistical analysis of these sequences. A novel technique is presented for incorporating monthly rainfall forecasts, presented as probabilities of rainfall being above average, average, or below average, into the risk assessment scheme. Information on current hydrological conditions is incorporated in the procedure through the use of recently observed natural flows to adjust the internal state variables of a conceptual rainfall-runoff model to achieve agreement between observed and model flow. The overall procedure is accommodated within a decision support system for drought management which is implemented on a microcomputer and makes extensive use of interactive menus, forms and colour graphic displays. A key feature of the system is the maintenance of an up-to-date archive of hydrometric data which is achieved through a real-time communication link with a second computer dedicated to real-time data acquisition via telemetry. Monitoring the reliability of the water resource system during droughts is made a quick and easy task, and the effect of a change in the operating policy on system reliability can be readily assessed. The information obtained provides valuable support for tactical decision-making within the overall long-term operating strategy.