Extended Abstract (604K)J1.5
Combined Methods from Entity and Field Frameworks to Determine the Source Characteristics of a Contaminant
Andrew J. Annunzio, Penn State Univ., University Park, PA; and S. E. Haupt, G. Young, and L. M. Rodriguez
In the event of a contaminant release, it is crucial to ascertain the source characteristics of the contaminant for mitigation purposes and to predict subsequent atmospheric transport and dispersion (AT&D). Here we apply a mixed Lagrangian/entity and Eulerian/field framework to determine source characteristics of both continuous and instantaneous releases. This is done by finding a Lagrangian quantity that determines the contaminant distribution; the plume/puff spread. In previous work, we showed that an entity framework is suitable to ascertain source characteristics of a contaminant. For a puff, no assumption was made about the contaminant distribution, and we inverted a simple set of equations to find the source. For a plume, we had already implemented a mixed Lagrangian/Eulerian framework because a steady flow of contaminants allowed us to average concentration data and infer plume spread from the concentration field. An advantage of these algorithms is that no meteorological input is required.
While these methods are valuable, it becomes risky to implement them when spatial observations of a contaminant are sparse or when more than one entity is present in the sensor domain. For these situations, we must make use of all meteorological data available, and also may need to assume a plume/puff shape. This assumption implies a mixed Entity/Field approach because the field is matched to determine an entity state. To overcome limitations of a sparse data set, we can fit the data to a dispersion model at each time step to use contaminant observations for a greater period than one time instant. This will enable accurate determination of plume/puff spread, a quantity that is crucial to determine the contaminant source. Fitting the data will also allow us to ascertain the number of entities present in the sensor domain; however, an iterative procedure is used here to fit the contaminant data to ensure that the field is approximated by the correct number of entities. With these advances to the models, it will be shown that this combined Entity/Field approach will allow accurate determination of source characteristics of multiple plumes/puffs.
Joint Session 1, Applications of Artificial Intelligence Techniques to Air Pollution Problems
Tuesday, 19 January 2010, 3:30 PM-5:30 PM, B308
Previous paper Next paper