A Bayesian statistical algorithm is developed and demonstrated for real-time interpretation of pollutant monitoring sensor data in buildings. The approach is rapid and capable of reducing uncertainty in (i) locating and characterizing a source, (ii) assessing the current building operating conditions, and (iii) predicting the time-dependant pollutant transport as sensor data are obtained. The approach is particularly useful for situations where sensors may detect hazardous airborne concentrations that require immediate interpretation and response. In an illustrative example, a hypothetical pollutant release is modeled in a five-room house. Two data collection scenarios are considered: (1) concurrent sampling: sensor measurements are obtained simultaneously in each of the five rooms at five-minute intervals, and (2) sequential sampling: sensor measurements are obtained sequentially, one room at a time, at five-minute intervals. Uncertainties in the source location, duration, and release rate, as well as in the building operating conditions, are updated in real-time, as the synthetic data becomes available. In the concurrent sampling scenario, the location of the source is correctly identified, and uncertainties greatly reduced, at the first measurement sequence (t=5 minutes). The uncertainties continue to be reduced as the hypothetical sensors send subsequent measurements. In the sequential sampling scenario, the location of the source and the model parameters are not characterized with high certainty until all of the rooms have been measured once (t=25 minutes). Implications for sensor placement and operation in buildings, and future research needs, are discussed.