What would be the most accurate way to determine back trajectories (BT) for airmass sources at a particular field site. The sampling time at the field site was ~24 hours. Should I run a 24 hr BT for every hour during sampling? Would I then combine all of these BT’s to get an accurate BT for the airmass collected over the 24 hr sampling period? Also what height would you suggest to end the BT at, 10 m or 950 h Pa?


You have stumbled across a difficult question. It’s not always a straightforward thing to determine the best way to get at the question of source regions using trajectory models, particularly in the absence of additional information.

The best scenario is where the 24hr period is mostly unchanging — like under a high pressure system, or perhaps with easterly offshore winds. If the airmass isn’t moving much and if you are interested in source regions that are not terribly distant (within several hundred km or so) chances are that trajectories calculated over the period are not going to change much so it doesn’t matter if you calculate only one or many trajectories over the period, and it probably doesn’t matter much about the height variable either.

The worst case occurs when the opposite is true. Frankly, if you are collecting samples across a frontal boundary, I’m not sure if you can say much about the source region. In the absence of additional information, I would probably discard the sample. Unfortunately, these cases are frequently among the most interesting.

When we collect precipitation samples for chemical analysis, we also keep a record of hourly rainfall. For a 24hr case we would run trajectories corresponding to the heaviest periods of precipitation, plus, perhaps, the onset and cessation of precipitation. If the trajectories are similar over the period, we can be fairly certain about source regions. If there is a lot of sheer, you really need additional information to support further analysis.

The bottom line: If 24hr sampling results in insufficient information to pursue additional analysis, you should move to a shorter sampling period. If you cannot move to a shorter sampling period, you need to collect additional information to support your analysis. Sometimes a compromise is possible. An shorter sampling period may be useful to establish some baseline information about your sources — and then you can go to 24hr or longer periods once you establish a notion of what is happening out there. You should also have some apriori knowledge about the sources you expect to track — such as an emissions inventory.

With regard to 10m or 950 h Pa, I’m not sure it matters a whole lot. Do a sensitivity test on the model and see what difference it makes. My guess is not much. For precipitation trajectories, we tend to shoot for just above the top of the boundary layer — typically ~850mb.

So, sort out your assumptions up front and make sure you have stated them when you present your results. There is no “correct” way to do this. You may also wish to look into cluster analysis for a semi-objective way of looking at your trajectory results.

Rick Artz