Lawrence Berkeley National Laboratory

Releases of acutely toxic airborne contaminants in or near a building can lead to significant human exposures unless prompt response measures are identified and implemented. Possible responses include conflicting options, such as shutting the ventilation system off versus running it in a purge (100 percent outside air) mode, or having occupants evacuate versus sheltering in place. The right choice depends in part on quickly identifying the source location, the amount released, and the likely future dispersion of the pollutant. This paper summarizes recent developments to provide such estimates in real time using an approach called Bayesian Monte Carlo updating. This approach rapidly interprets measurements of airborne pollutant concentrations from multiple sensors placed in the building, and computes best estimates and uncertainties of the release conditions. The algorithm is fast, and can continuously update the estimates as measurements stream in from sensors. As an illustration, two specific applications of the approach are described.