6.4 Source Geolocation




One aspect of running multiple backward trajectories from various locations is if these trajectory locations and times correspond with high measured values then by superimposing these trajectories it may be possible to determine the source location of the measured pollutant. The GUI has an option to execute a script to run multiple iterations of the backward trajectory calculation which can then be summarized using the frequency plot menu option.

  1. Before getting started it is necessary to define a forward trajectory using the trajectory setup menu that encompasses the time period of the sampling. We have already used this trajectory configuration for several previous examples: the 750 m trajectory saved to file name traj_fwrd_control.txt, which should be retrieved into the menu. Rename the endpoint output file from tdump_fwrd to just tdump. Then just save. It is not necessary to run the trajectory again. Retrieving and saving just loads the run information into the GUI variables.

  2. Now click the Trajectory / Special Runs / Geolocation menu tab which will open a relatively simple menu composed of just three steps. Step 1 defines the measured data input file used in this series of calculations. The trajectory model will be run backward, from each of the sampling locations. The sampling data for CAPTEX in the DATEM format can be found the file captex2_meas.txt.

  3. Step 2 creates three CONTROL.{xxx} files at the beginning, middle, and end of the sampling period for each sample with a non-zero measurement. The CONTROL files are numbered sequentially from 001 to a maximum of 999. The trajectory output files are labeled using the base name from the CONTROL file followed by the sampler ID number and the starting date and time of the trajectory. Execute Step 2 and a completion message is shown after all 597 CONTROL files have been created in the working directory.

  4. As an example of this process, examine the first non-zero sample in captex2_meas.txt which shows 1983 09 25 1800 0300 41.30 -82.22 15.6 316. Three CONTROL files were created in the working directory for this sample: CONTROL.001, CONTROL.002, and CONTROL.003. In file 001, the trajectory starts at the end of the sample 83 09 25 21 and has a duration of -4 hour to take it back to the start of the release at 17 UTC. In contrast, file 003 corresponds to the start of the 3-h sampling period at 83 09 25 18 and has a duration of only -1 hours back to the release start time. The trajectory output file name consists of the sampling station number (316) and the trajectory start time. CONTROL files are created for all samples with some of the later ones having trajectory durations of as much as -64 hours.

  5. Next Execute Step 3 which invokes a Tcl script to sequentially run all the trajectory simulations starting with CONTROL.001 through the last available control file. Each simulation uses the same namelist configuration. Once the calculation has completed, the individual trajectories can be gridded and displayed.

  6. Once all calculations have completed, go to the Trajectory / Display / Frequency menu to grid and display the results. Make sure no unwanted (without the number suffix) tdump files are listed in INFILE. Using a 0.5 degree grid (no residence time) shows a pattern with the maximum probabilities centered over Lake Erie rather than the actual location in western Ohio. This calculation was done assuming trajectories started at 750 m AGL for air samples actually taken near ground level (about 2 m AGL). As an exercise, reconfigure the simulation for a starting trajectory height of 10 m AGL, recreate the CONTROL files with the new starting height, run the trajectories, and the display the frequency plot graphics. With this calculation, the results are even more ambiguous, with a maximum probability region in eastern Pennsylvania. This result is consistent with previous sections that showed the low-level trajectories to be not very representative of boundary-layer flow.

The trajectory geo-location results shown here only present a summary of the methodology rather than a definitive analysis for this particular event. Part of the issue is that the geography of the orientation of the tracer release location with the sampling network always required the same flow conditions for a successful experiment. For a more accurate geo-location calculation using only trajectories, it is necessary to sample a variety of different flow orientations. For instance, a geometric triangulation using trajectory analysis would be easier if the tracer release had been located within the center of the sampling network. The problem presented here with this single event analysis will be studied further in relation to the concentration calculation, where we can also account for the effects of sample timing and zero measured values.