Hi,
1. Can we start back trajectories below planetary boundary layer(PBL) ? Is that rational?
There were some pappers (JGR) which started back trajectories PBL.
2. If try to use back trajectories calculated by Trajectory Model (online), should I avoid starting below PBL?
Question about back trajectories starting blow PBL
Re: Question about back trajectories starting blow PBL
Hi Ren2014,
You can absolutely start back trajectories below the PBL, and in fact, this is probably done in the majority of cases in which back-trajectories are run. It basically depends on the problem you are investigating with the back-trajectory.
If, for example, you are wondering where the air came from that you sampled at a ground-level sampler, then you would definitely want to start the back-trajectory within the PBL, as the sampler is "seeing" air that is in the PBL. In this ground-level sampler case, one might think to start the back-trajectory at the height of the sampler (say, 2 meters), but this is generally not recommended as the trajectory would often then hit the ground and lose accuracy. In that case, if you are only going to use one starting height, I usually recommend that one start the back-trajectory at a height of 0.5 * PBL. There is a way to do this in HYSPLIT, where you tell it (through the advanced menu) to interpret starting heights as fractions of the PBL. After doing this, you then give the starting height as 0.5 (or whatever fraction you want). Another thing to consider trying is to start at different heights, and see what sort of difference there is. When looking at ground level sampling results, I sometimes try trajectories starting at 0.1, 0.3, 0.5, 0.7, and 0.9 * PBL. If there is a "huge" difference, then this is useful to know, as then one wouldn't be unjustifiably confident in results starting from a single height. If the results are relatively similar, then you can be even more confident that you are getting a reasonable answer for your case.
Note that the HYSPLIT model estimates the PBL height at any location and time from the meteorological data being used the drive the particular simulation being run. And of course, the PBL height changes within any day, and from day to day. So, if you choose a starting height of say, 0.5 * PBL, and then run numerous trajectories starting at different times or days, you will notice that the actual starting height (in meters, above ground level, say) will generally be different with each trajectory. This is what would be expected.
I have put together a short batch of slides that talk about the starting height a little more, with some conceptual graphics, at the following URL (one is a PDF version, one is a PowerPoint version, with identical content):
http://www.arl.noaa.gov/documents/repor ... er_01.pptx
http://www.arl.noaa.gov/documents/repor ... ver_01.pdf
An example of the case where you might want to start the trajectory above the PBL would be if one is attempting to investigate samples taken from an aircraft flying above the PBL. In that case, one would want to start the back-trajectory at the height the aircraft was flying at the time any given sample was taken.
Hope this helps!
Mark
You can absolutely start back trajectories below the PBL, and in fact, this is probably done in the majority of cases in which back-trajectories are run. It basically depends on the problem you are investigating with the back-trajectory.
If, for example, you are wondering where the air came from that you sampled at a ground-level sampler, then you would definitely want to start the back-trajectory within the PBL, as the sampler is "seeing" air that is in the PBL. In this ground-level sampler case, one might think to start the back-trajectory at the height of the sampler (say, 2 meters), but this is generally not recommended as the trajectory would often then hit the ground and lose accuracy. In that case, if you are only going to use one starting height, I usually recommend that one start the back-trajectory at a height of 0.5 * PBL. There is a way to do this in HYSPLIT, where you tell it (through the advanced menu) to interpret starting heights as fractions of the PBL. After doing this, you then give the starting height as 0.5 (or whatever fraction you want). Another thing to consider trying is to start at different heights, and see what sort of difference there is. When looking at ground level sampling results, I sometimes try trajectories starting at 0.1, 0.3, 0.5, 0.7, and 0.9 * PBL. If there is a "huge" difference, then this is useful to know, as then one wouldn't be unjustifiably confident in results starting from a single height. If the results are relatively similar, then you can be even more confident that you are getting a reasonable answer for your case.
Note that the HYSPLIT model estimates the PBL height at any location and time from the meteorological data being used the drive the particular simulation being run. And of course, the PBL height changes within any day, and from day to day. So, if you choose a starting height of say, 0.5 * PBL, and then run numerous trajectories starting at different times or days, you will notice that the actual starting height (in meters, above ground level, say) will generally be different with each trajectory. This is what would be expected.
I have put together a short batch of slides that talk about the starting height a little more, with some conceptual graphics, at the following URL (one is a PDF version, one is a PowerPoint version, with identical content):
http://www.arl.noaa.gov/documents/repor ... er_01.pptx
http://www.arl.noaa.gov/documents/repor ... ver_01.pdf
An example of the case where you might want to start the trajectory above the PBL would be if one is attempting to investigate samples taken from an aircraft flying above the PBL. In that case, one would want to start the back-trajectory at the height the aircraft was flying at the time any given sample was taken.
Hope this helps!
Mark
Re: Question about back trajectories starting blow PBL
thank you! Mark~ I am very appreciate your help.
Re: Question about back trajectories starting blow PBL
Thanks a lot!
Re: Question about back trajectories starting blow PBL
Hi Mark,
I have a new quistion. Isn't ture that the higher resolution meteorological data are the more accurate back trajectories, starting at below the PBL, we would get ? Do you have some suggsestion about the meteorological data resolution ?
Thanks a lot~ : )
I have a new quistion. Isn't ture that the higher resolution meteorological data are the more accurate back trajectories, starting at below the PBL, we would get ? Do you have some suggsestion about the meteorological data resolution ?
Thanks a lot~ : )
Re: Question about back trajectories starting blow PBL
Hi Ren2014,
In general, yes, higher-resolution meteorology should usually give more "accurate" trajectories. All things being equal, one could always pick the highest resolution meteorological data available for a simulation. Note that trajectories will stop once they leave the domain of the met data available for the simulation. So, if you have a high resolution, but limited domain, dataset, the trajectory will stop once it gets to the edge of the domain. However, if you also provide the simulation with a larger domain dataset (generally coarser), then HYSPLIT will continue on the trajectory on this coarser, larger domain. When I do trajectories, I almost always include a global meteorological dataset (e.g., the 2.5 degree global reanalysis) along with any finer scale data that I want to use. By doing this, I ensure that all trajectories I run will be able to "finish", even if they end up using relatively coarse resolution met data towards the end.
Also, it can be very useful and instructive to do an analysis using different meteorological data, to see what differences might arise. This is a good strategy with this and other related questions -- try it and see what happens! With any of this, there is no perfect answer.
Also, your choice of met data will also depend on the nature of the area you are trying to model. For example, if you are in the middle of relatively flat, non-coastal terrain, and the extent of your trajectories generally stay on this terrain, then relatively coarse data (e.g., 40km EDAS, or 32 km NARR, or 0.5 deg GDAS, or even 1 deg GDAS) will all probably be fine, and should all give relatively similar results. However, if you are in more complex terrain (e.g., mountains) or in a coastal area where sea-breeze or lake-breeze effects may be present, then you would want to use the finest met data available (12 km NAM, or finer, if available), and even then, the met data may not be able to really capture the local wind fields very well.
One other thing to consider is this. If you have meteorological measurements at your site (or somewhere nearby), then you can try to compare the met measurements with what the met model data thinks is happening at that site, using for example the Meteorology > Display Data > Text Profile functionality of the HYSPLIT Graphical User Interface. If you can demonstrate that the met model that you are using for the trajectory is giving results relatively consistent with the measurements (e.g., particularly for wind speed and wind direction), then you can have more confidence in the accuracy of the trajectory you are running. If not that consistent, then you know there are more uncertainties involved.
Hope this helps,
Mark
In general, yes, higher-resolution meteorology should usually give more "accurate" trajectories. All things being equal, one could always pick the highest resolution meteorological data available for a simulation. Note that trajectories will stop once they leave the domain of the met data available for the simulation. So, if you have a high resolution, but limited domain, dataset, the trajectory will stop once it gets to the edge of the domain. However, if you also provide the simulation with a larger domain dataset (generally coarser), then HYSPLIT will continue on the trajectory on this coarser, larger domain. When I do trajectories, I almost always include a global meteorological dataset (e.g., the 2.5 degree global reanalysis) along with any finer scale data that I want to use. By doing this, I ensure that all trajectories I run will be able to "finish", even if they end up using relatively coarse resolution met data towards the end.
Also, it can be very useful and instructive to do an analysis using different meteorological data, to see what differences might arise. This is a good strategy with this and other related questions -- try it and see what happens! With any of this, there is no perfect answer.
Also, your choice of met data will also depend on the nature of the area you are trying to model. For example, if you are in the middle of relatively flat, non-coastal terrain, and the extent of your trajectories generally stay on this terrain, then relatively coarse data (e.g., 40km EDAS, or 32 km NARR, or 0.5 deg GDAS, or even 1 deg GDAS) will all probably be fine, and should all give relatively similar results. However, if you are in more complex terrain (e.g., mountains) or in a coastal area where sea-breeze or lake-breeze effects may be present, then you would want to use the finest met data available (12 km NAM, or finer, if available), and even then, the met data may not be able to really capture the local wind fields very well.
One other thing to consider is this. If you have meteorological measurements at your site (or somewhere nearby), then you can try to compare the met measurements with what the met model data thinks is happening at that site, using for example the Meteorology > Display Data > Text Profile functionality of the HYSPLIT Graphical User Interface. If you can demonstrate that the met model that you are using for the trajectory is giving results relatively consistent with the measurements (e.g., particularly for wind speed and wind direction), then you can have more confidence in the accuracy of the trajectory you are running. If not that consistent, then you know there are more uncertainties involved.
Hope this helps,
Mark