Advanced / Special Topics / Verification Statistics

See the DATEM (Data Archive of Tracer Experiments and Meteorology) web page for more detailed information on each of these tracer experiments. The model verification statistics are presented for the un-averaged data, where each measured and calculated concentration is paired in space and time. Zero-zero pairs are always excluded. Results are also presented after temporally averaging the measured and calculated values over the duration of each experiment. These latter results are paired in space but not time and hence represent the calculation's spatial performance. Statistics presented are the correlation coefficient (R), the fractional bias (FB), the figure-of-merit in space (FMS), the KS parameter representing the departure of the measured and calculated cumulative distributions from each other, and the normalized rank (0 to 4) which is computed from the normalized sum of the previous four parameters. The URL includes links to download the measurement data, meteorological data, statistical programs, and HYSPLIT configuration files.

Unless otherwise noted, all HYSPLIT calculations used the NCEP North American Regional Reanalysis (3 hr; 32 km) meteorological data for the dispersion calculations. The main characteristics of each experimental data set are summarized as follows:

Verification Statistics Summary

The HYSPLIT model performance by rank for the time mean (spatial correlation) and individual sample paired concentrations for the most recent version of HYSPLIT is given. The rank is computed from the sum of the correlation, fractional bias, figure-of-merit in space, and the Kolmogorov-Smirnov parameter. In addition, all required data are provided with the HYSPLIT distribution to run CAPTEX release #2 and compute the resulting performance statistics through the GUI.

Table of Contents