Abstract Simplifications are described that will substantially reduce the computation time and memory requirements of a Gaussian puff diffusion model without impairing its ability to simulate temporally and spatially changing conditions. Among the more important simplifications are: a variable puff generation rate, merging of closely spaced puffs, nonuniform spatial resolution of the wind field and the use of integers to represent some of the variables. A model using the simplifications was able to simulate the 24-h behavior of emissions from a power plant; concentrations were calculated at 25 points and at 48 different times. The time required on a Data General Nova 1200 minicomputer is about 6 min, but varies with different sequences of meteorological conditions. The memory required is about 24,000 words.