The Columbia River (CoRie) Observation Network includes an extensive array of fixed stations monitoring the Columbia River estuary and nearby coastal ocean. At each station, variable combinations of in-situ sensors measure one or more physical properties of water or atmosphere. Using a multi-scale data assimilation model, the CORIE modeling system integrates models and field controls to produce a simulation of 3D circulation, in a region centered in the estuary and plume. The CORIE data assimilation framework [1] combines observational data with numerical data models to produce an estimated system state for the physical process. To augment the fixed observational network, additional data is collected during periodical cruises of a mobile sensor station. Because these cruises are expensive and rare, an important goal for scientists is to sample data at points that most reduce the uncertainty of the data assimilation model. This is challenging, since the estuary environment is very dynamic, and therefore the optimal cruise path cannot be determined in advance. The goal of our system is to move the mobile station as the data assimilation proceeds in order to maximally reduce the uncertainty in the data assimilation process.

Localization of small wireless sensor devices, with the deployment of the minimal infrastructure or hardware, has been the topic of significant research over the past few years. We have developed the Robust Extended Kalman Filter (REKF) localization system [1], which enables a mobile, data gathering robot to localize static sensor devices, by combining the RSSI data received from the motes, with estimates of its trajectory. The REKF localization system is particularly well suited to delay-tolerant sensor networks, where node positions need not be known in real time. We have observed accuracies ranging from approximately 30cm to 1m in practice.