I am working on estimating the degree that aerosols encountered clouds along their back trajectories. Because it is near ground sampling, I am setting the initial height to be 0.1 * PBL height (~60-120 m). 27 ensemble trajectories were used to account for uncertainties. The default offset in z dimension in HYSPLIT is 0.01 sigma units (~250 m). In this case, is it better to increase the initial height to be 250 m, or to use a smaller offset factor (e.g. 0.005 or 0.0025) so that the vertical deviation is smaller than the starting height? Not sure what offset should be used for uncertainty analysis.
Additionally, I think the ensemble analysis only accounts for the error due to limited resolution of the meteorological data. I wonder how does this error compared with the error associated with trajectory integration - which error is more significant? Are there any other sources of error that one should consider in error estimation? Thank you!
Ensemble trajectories and uncertainty assessment
Re: Ensemble trajectories and uncertainty assessment
The error associated with trajectory integration is much smaller than error due to uncertainty in the meteorological wind field.
However, depending on the terrain, you may also run into uncertainty in trajectory path due to influence of terrain.
You can try running a few forward trajectories from your backward trajectory end points to see if terrain has an impact.
e.g. https://www.ready.noaa.gov/documents/Tu ... _isob.html
Even for near ground sampling, it is acceptable to use something like half the mixing layer height as a starting point unless you are interested in very short range transport. This is because the material is expected to mix fairly quickly throughout the boundary layer. That is another source of error with trajectories since they do not take into account transport due to turbulent mixing. One way to account for that is to start trajectories at a few heights throughout the boundary layer. Or one could consider backward dispersion runs.
However, depending on the terrain, you may also run into uncertainty in trajectory path due to influence of terrain.
You can try running a few forward trajectories from your backward trajectory end points to see if terrain has an impact.
e.g. https://www.ready.noaa.gov/documents/Tu ... _isob.html
Even for near ground sampling, it is acceptable to use something like half the mixing layer height as a starting point unless you are interested in very short range transport. This is because the material is expected to mix fairly quickly throughout the boundary layer. That is another source of error with trajectories since they do not take into account transport due to turbulent mixing. One way to account for that is to start trajectories at a few heights throughout the boundary layer. Or one could consider backward dispersion runs.
Re: Ensemble trajectories and uncertainty assessment
Thank you for your answer! This is super helpful!