Parameter Estimation of Atmospheric Release Incidents Using Maximal Information Collection Reza Madankan 1, * , Puneet Singla 1 , and Tarunraj Singh 1 1 Mechanical and Aerospace Engineering Department, University at Buffalo, Buffalo, NY,14228 {rm93, psingla, tsingh}@buffalo.edu Abstract. The effects of data measurement on source parameter es- timation are studied. The concept of mutual information is applied to find the optimal location for each sensor to improve accuracy of the overall estimation process. For validation purposes, an advection - diffu- sion simulation code, called SCIPUFF, is used as a modeling testbed to study the effects of using dynamic data measurement. Bayesian inference framework is utilized for model-data fusion using stationary and mobile sensor networks, where in mobile sensors, the proposed approach is used to locate data observation sensors. As our numerical simulations show, using the proposed approach leads to a considerably better estimate of parameters comparing with stationary sensors. 1 Introduction With increasing number of instances of toxic material release, there is tremen- dous interest in precise source characterization and generating accurate hazard maps of toxic material dispersion for appropriate disaster management. There is no doubt that proper sensor placement is intimately tied to the performance of the source estimation and model uncertainty characterization. Different strategies have been suggested to determine the optimal path of the mobile sensors for source parameters characterization. Earlier works in this area can be categorized as Chemotaxis [1], Anemotaxis [2], and Fluxotaxis [3]. In chemotaxis approach [1], mobile sensors follow the concentration gradient. Therefore, the motion toward the largest concentration is the goal direction for the chemotaxis. In anemotaxis strategy [2], mobile sensors always move upstream while they locate inside the plume, hence the upstream is the goal direction for mobile sensors. With fluxotaxis approach [3], the mobile sensors compute the amount of dispersal material flux passing through virtual surfaces formed by neighboring sensors. Where, each individual sensor independently calculates the amount of local material flux relative to the current position of its neighbors. Even though, each of the above approaches has its own advantage and appli- cation, but the major drawback of aforementioned approaches lies in the possi- bility of being trapped in local maxima and plateaus of the concentration field. Recently, there have been numerous works focusing on application of infor- mation theoretic concepts in optimal sensor placement [4–10]. For instance, an