Over the past few years, the use of dispersion model ensembles has been an increasingly attractive
approach to predict atmospheric transport. The ensembles are generally constructed by combining multiple
numerical weather prediction simulations or output from different dispersion models, by introducing
variations in a particular model's physics parameterizations, or by different combinations of these
variations. In addition, the source term and height distribution can be varied to create dispersion
For this application, we create a HYSPLIT ensemble simulation using the 26 meteorological model
members from the National Weather Service (NWS) operational Sort Range Ensemble Forecast (SREF).
The HYSPLIT model runs are generated by using the Transfer Coefficient
Matrix procedure that is updated with the most recent meteorological data at 03, 09, 15 and 21Z.
The figures below show each of the simulation ensemble members as well as a series of statistical
measures that describe the ensemble properties such as:
The spatial distribution of the number of members producing concentrations greater than zero.
The mean concentration of all ensemble members.
The variance of all ensemble members (the mean square difference between individual members and the mean).
The coefficient of variation of all ensemble members (the square root of the variance*100 divided by the mean).
The probability of concentration produces contours that give the probability of exceeding a fixed concentration
value at one of three levels: 1% of the maximum concentration, 10% of the maximum, and the maximum concentration.
The maximum is determined to be the first concentration to a power of 10 that is less than the actual maximum
value (exponent of 10). The concentration level for the probability display is shown on the graphic with
the pollutant identification field set to something like C14, where 14 represents the concentration to the
power of 10-14.
The concentration at percentile levels shows the concentration contours of areas where concentrations will
be exceeded only at the given probability level. The probability level choices through the menu are
limited to 50, 90, and 95th percentiles.
Refresh/reload your browser for the most recent images. Click on a thumbnail image to expand.
The charts are updated four times per day (~00, 06, 12, 18 UTC).
Number of members
Coefficient of Variation
Probability that concentration exceeds
1% of maximum value
10% of maximum value
100% of maximum value
Individual member plots
This web page is under development and is experimental.
Products may change without notice.
Concentrations are relative to the source of 1 unit.
Contour values may vary from day to day and chart to chart.