Exercise #12




  • Problem #1 - Recompute the multiple meteorological data ensemble and determine whether decreasing the concentration grid size and increasing the particle number will improve the ensemble mean results.

  • Hint - Edit the ens_data.bat batch file to double the resolution (0.125 deg and 20,000 particles).

  • Solution - Although some individual simulations showed substantial improvement, others showed poorer performance, the statistical results for the mean model were quite similar to the original coarse resolution result.

  • Problem #2 - Using the existing output files from the physics ensemble, compute the performance statistics for each member and the ensemble mean.

  • Hint - Using the binary file merge menu, compute the ensemble mean for the 20 members. Then using the ensemble statistics menu, create the file of the summary statistics.

  • Solution - In this example, the better performance of the ensemble mean is not that evident, although more of the metrics show the best results compared with any one ensemble member (except the Gaussian 002). Note that the physics ensemble calculation was only a subset (26th 06Z to 28th 15Z) of the entire experimental period and too few particles were released in this example. Although with fewer points along the 1:1 line, the scatter diagram of the physics ensemble mean is comparable to the scatter diagram of the meteorological ensemble mean calculated in the previous section.