A codimension multifractal methodology was used to analyze and model concentrations in SF6 tracer gas plumes from a line source in atmospheric surface layer flows. The universal multifractal parameters (alpha, C1, H) of the codimension methodology were determined from high frequency SF6 time series measurements. Changes in the statistical qualities of the SF6 time series such as peak-to-mean ratios, intermittency, and concentration fluctuation intensities were reflected in changes in these parameters. Time series can be artificially generated using an extremal Levy, stochastic model and the (alpha, C1, H) parameters. The peak-to-mean ratios, intermittency, and concentration fluctuation intensities predicted by the model are within a factor of two of the comparable parameter of the measurements and lie within the uncertainty of the experimental measurments. Application of the method to other situations where there is a need to predict extremes in the behavior of scalar (e.g. CO2, H2O) signals and large departures of peak values from the mean is suggested.