In this paper we propose event-driven power management techniques for wireless sensor networks. To accomplish this we model a sensor network application as a set of application-specific events that the application may contain. Events are first classified into scheduled and non-scheduled events. These events are further classified according to the size and the locality of the data, and the real-time characteristics of the event. For scheduled events we propose schedule-driven power control and global coordination. For nonscheduled events we propose source-driven and sink-driven power control for both lower energy consumption and higher performance. Experimentation results confirm that the event driven power management can substantially save energy compared to existing low energy sensor network protocols while it can meet the performance required by the application.