Another multiple trajectory approach, an extension of a previous section on sources of trajectory error, is a variation of an old manual technique, where trajectories are started about the point of interest, and the divergence of those trajectories would provide a more quantitative estimate of the uncertainty associated with the center point trajectory. In HYSPLIT, this method has been incorporated directly into the code so that trajectories are automatically computed about a
3-dimensional cube centered about the starting point. However, the initial positions are not offset, just the meteorological data point associated with each particular trajectory, so that all trajectories start from the same point.
- The first step would be to setup the trajectory simulation for the 600 m AGL mid-boundary layer forward trajectory by retrieving the traj_fwrd_control.txt file. Then save and exit the setup menu.
- Instead of using the simple run model tab, go to the Trajectory / Special Runs / Ensemble menu tab to run the calculation. This tab calls a special trajectory executable (..\exec\hyts_ens). If the SETUP.CFG message appears, select to delete the file and then run.
- A single endpoints output file is created which shows the forward trajectory results for the 27 trajectories that make up the points of the data cube about the starting location.
- The same calculation can be performed for the backward trajectory point that was done in an earlier section by retrieving the traj_back_control.txt file into the trajectory setup menu and rerunning the ensemble which should show the equivalent 27 backward trajectory results. Considering just those trajectories not caught in the anti-cyclonic circulation, there is a considerable variation in the upwind origin.
- The computational details of the meteorological ensemble can be configured in the Advanced / Configuration Setup / Trajectory meteorological grid offset ensemble configuration menu #7, which shows the default grid offset of 1 grid point in the horizontal and 0.01 sigma units in the vertical. The 27 members are created from 3 planes of 9 trajectories, each plane at +/- 0.1 sigma (about 250 m). The 9 trajectories on each plane represent all combinations of +/- horizontal combinations in x and y.
The difference between the forward and backward results show a dramatic difference in the "trajectory dispersion", the forward trajectory results are quite constrained while the backward pattern is quite disperse. This is consistent with the weather pattern associated with release #2, where the tracer flowed along the northern side of a high pressure system. The backward equivalent starts in a divergent region where small changes in position resulted in very different trajectories.