Wireless sensor networks (WSNs) are developed in an accelerated pace during the last few years. A lot of projects have been launched for various applications, however, none of those projects realize long-term operations and large-scale deployments. Motivated by the need of revealing the difficulties of large-scale long-term WSNs, we launched two long-term large-scale WSNs: “GreenOrbs” in 2009 and “CitySee” in 2011, which work for closely monitoring the forest and urban environments and accurately measuring carbon sequestration and emission, respectively. Through long-term measurements and statistical analysis we have a deep understanding on data collection behaviors of large-scale WSNS, especially for routing performance. This thesis mainly answers questions which cover root cau...[ Read more ]

Wireless sensor networks (WSNs) are developed in an accelerated pace during the last few years. A lot of projects have been launched for various applications, however, none of those projects realize long-term operations and large-scale deployments. Motivated by the need of revealing the difficulties of large-scale long-term WSNs, we launched two long-term large-scale WSNs: “GreenOrbs” in 2009 and “CitySee” in 2011, which work for closely monitoring the forest and urban environments and accurately measuring carbon sequestration and emission, respectively. Through long-term measurements and statistical analysis we have a deep understanding on data collection behaviors of large-scale WSNS, especially for routing performance. This thesis mainly answers questions which cover root causes and impact of network behaviors on routing performances, e.g., What causes routing dynamics? What is the impact of routing dynamics? When, where, and under what kind of circumstances packet losses occur? Why packets are lost? It also shows findings and insights on routing dynamics and packet delivery performances and presents a novel prediction method for traffic modeling and energy saving. We evaluate the performances of proposed methods by experiments with extensive data-trace and verifications in real deployed systems. The results show that our findings are insightful and the proposed method has been applied in our system.