This paper proposes Efficient Sleep Scheduling based on Application Timing (ESSAT), a novel power management scheme that aggressively exploits the timing semantics of wireless sensor network applications. We present three ESSAT protocols each of which integrates (1) a lightweight traffic shaper that actively shapes the workload inside the network to achieve predictable timing properties over multiple hops, and (2) a local scheduling algorithm that wakes up nodes just-in-time based on the timing properties of shaped workloads. Our ESSAT protocols have several distinguishing features. First, they can save significant energy with minimal delay penalties. Second, they do not maintain TDMA schedules or communication backbones; as such, they are highly efficient and suitable for resource constrained sensor platforms. Moreover, the protocols are robust in highly dynamic network environments, i.e., they can handle variable multi-hop communication delays and aggregate workloads involving multiple queries, and can adapt to varying workload and network topologies. Our simulations showed that DTS-SS, an ESSAT protocol, achieved an average node duty cycle 38-87% lower than SPAN, and query latencies 36-98% lower than PSM and SYNC.