Technological advances have enabled the deployment of large-scale sensor networks for environmental monitoring and surveillance purposes. The large volume of data generated by sensors needs to be processed to respond to the users queries. However, efficient processing of queries in sensor networks poses great challenges due to the unique characteristics imposed on sensor networks including slow processing capability, limited storage, and energy-limited batteries, etc. Among various queries, top-k query is one of the fundamental operators in many applications of wireless sensor networks for phenomenon monitoring. In this paper we focus on evaluating top-k queries in an energy-efficient manner such that the network lifetime is maximized. To achieve that, we devise a scalable, filter-based localized evaluation algorithm for top-k query evaluation, which is able to filter out as many unlikely top-k results as possible within the network from transmission. We also conduct extensive experiments by simulations to evaluate the performance of the proposed algorithm on real datasets. The experimental results show that the proposed algorithm outperforms existing algorithms significantly in network lifetime prolongation.