A critical problem in homeland defense is correctly characterizing the source of hazardous material. Field monitors are expected to measure concentrations of toxic material. Algorithms are then required that back-calculate the parameters of the source and the local meteorology so that subsequent predictive modeling can inform decision-makers. Here, a genetic algorithm is used together with transport and dispersion models to assimilate sensor data to characterize emission sources. The parameters computed include location, time, and amount of the release and meteorological conditions relevant to the transport and dispersion. This methodology is demonstrated for a basic Gaussian plume dispersion model and verified in the context of both synthetic data and actual monitored data from field tests with known release amounts. Its error bounds are set using Monte Carlo techniques and robustness assessed through the addition of white noise. Algorithm speed is tuned through optimizing the parameters of the genetic algorithm.