The next generation of wireless sensor networks will be dynamic systems with the potential to revolutionize understanding of environmental change, provided they can assimilate large amounts of heterogeneous data in real time, rapidly assess (optimize) the relative value and costs of new data collection, and schedule subsequent measurements accordingly. Thus, they are Dynamic Data Driven Application Systems that integrate sensing with modeling in an adaptive framework. Keen interest in broad application of wireless sensing of the environment, as in NEON and CLEANER, awaits DDDAS technology that can estimate the value of future data in terms of its contribution to understanding against the costs of deployment, acquisition, transmission, and storage. This balance is especially important for environmental data, because networks will typically be deployed in remote locations without access to infrastructure (e.g., power), and sampling intervals will range from meters and seconds to landscapes and years, depending on the process, the current state of the system, the uncertainty about that state, and the perceived potential for rapid change. Network control must be dynamic and driven by models capable of learning about both the environment and the network. The focus of this project is the dynamic sensor network application involving understanding how biodiversity and carbon storage are influenced by global change. Specifically, this project is designed to learn how the growth, survival, and reproduction of forest trees are influenced by changes in climate, CO2 and disturbance, in the context of these and other variables that can fluctuate rapidly. This goal involves models of how tree growth and resource allocation are influenced by variables that can be understood through adaptive sampling across diverse scales in both time and space. The project will enable a general framework for dynamic data-driven wireless network control that combines environmental modeling and sensor network modeling both in and out of the network. Out of the network, environmental modeling entails full assimilation of all information, with exploitation of computing resources available there. Environmental modeling in the network is based on simplified representations that provide real-time, approximate answers. The in-network control model provides rapid scheduling for new measurements, and it communicates network information to the server, for diagnostics, supervisory control, and data assimilation. Periodically, the in-network model is updated based on this most complete understanding of the environmental variables, parameters, and battery life. Specific goals are (i) to construct a wireless sensing and networking infrastructure that supports a new paradigm of joint in-network and supervisory measurement, modeling, and prediction, (ii) to develop the modeling strategy needed to combine system understanding with costs for efficient wireless sensing of the environment, (iii) to make significant progress in understanding the maintenance of biodiversity and in measuring ecosystem properties, and (iv) to improve collaboration between computer sciences, engineering, statisticians and environmental scientists.
