Wireless Sensor Networks (WSNs) are collection of tiny sensor nodes capable of sensing, processing and broadcasting data correlated to some occurrence in the network area. The sensor nodes have severe limitation, such as: bandwidth, short communication range, limited CPU processing facility, memory and energy. Enhancing the lifetime of wireless sensors network and efficient utilizations of bandwidth are essential for the proliferation of wireless sensor network in different applications. We provide an in-depth study of applying wireless sensor networks (WSNs) to real-world habitat monitoring. A set of system design requirements were developed that cover the hardware design of the nodes, the sensor network software, protective enclosures, and system architecture to meet the requirements of biologists. Although researchers anticipate some challenges arising in real-world deployments of WSNs, many problems can only be discovered through experience. We present a set of experiences from a four month long deployment on a remote island. We analyze the environmental and node health data to evaluate system performance. The close integration of WSNs with their environment provides environmental data at densities previously impossible. We show that the sensor data is also useful for predicting system operation and network failures. Based on over one million data readings, we analyze the node and network design and develop network reliability profiles and failure models.