Network coverage is an important Quality of Service (QoS) measurement for many sensor network applications. Many existing studies on network coverage are based on the knowledge of sensor node distributions in sensing fields, which are often represented by given probability distributions. In this paper, we study the impacts of sensor node distributions on network coverage. We first show the impacts on network coverage by adopting different sensor node distributions through both analytical and simulation studies. We observe that assumed different sensor location distributions may lead to significant differences in coverage estimation. Then, we adopt a distribution-free approach to study network coverage, in which no assumption of probability distribution of sensor node locations are needed. The proposed approach has yielded good estimations of network coverage. Though only network coverage is studied in this paper, we believe that this methodology can be generalized and extended to estimation of other sensor network performance metrics. (C) 2011 Elsevier Inc. All rights reserved.