A procedure is developed and tested to provide operational plume forecasts in real-time by continuously updating the previous day's simulations as new meteorological data become available. Simulations are divided into smaller time segments and each segment is continued as an independent calculation using a unit source emission. Multiple computational species are tracked at the same time to represent different classes of radionuclides, each with different dry and wet deposition characteristics. When quantitative air concentration results are required, the unit source calculations are multiplied by the appropriate temporally varying emission rates and decay factors for the radionuclide species involved. Air concentrations for multiple emission scenarios can easily be created in a few minutes and used to optimize model results as more measurement data become available. The procedure was evaluated for the Fukushima accident using publically available emission estimates and some I-131 and Cs-137 monitoring data. The model performance was evaluated at four sampling locations (Dutch Harbor, Alaska; Seattle, Washington; Dublin, Ireland; and Huelva, Spain) at various distances from Japan. The model results showed a very high correlation for the I-131 particulate predictions (0.94) and a moderate correlation for the Cs-137 predictions (0.40). The cesium predictions at Seattle showed five distinct time periods of concentration over-predictions associated with two peak emission periods. Adjusting these emission rates downward to correspond more closely with the time-adjacent rates eliminated the over-prediction but resulted in total emissions of Cs-137 (3 PBq) that were much less than estimated by other researchers (36 PBq).