||A wireless sensor network (WSN) is a self-organized wireless network that consists of a number of sensor nodes. The deployment of a number of sensor nodes enables people to monitor and interact with the physical world. Forest is one of those environments where WSNs are applied. Under the circumstances of global climate changes and environmental pollution, WSNs for forestry applications attract increasing attention in recent years. This dissertation research focuses on developing techniques that make the sensor network systems more applicable and efficient for practical forestry applications. I have been working on GreenOrbs, a long-term large-scale WSN system deployed in the forest. In order to overcome the uncertainty emerging from the forestry deployments, I address several key issues, including error-resilient localization for networked sensor nodes against irregular wireless signals, forwarding-quality-aware data collection against unstable link and node behavior, and quantitative canopy closure estimates against error-prone sensor readings. I develop theoretical principles, design practical protocols, and implement my ideas with GreenOrbs. I evaluate those approaches through real-world large-scale experiments and trace-driven simulations. The results validate their effectiveness and efficiency in overcoming the uncertainty. The proposed approaches can be further applied to other fields.