Initial heights and mixed layer depth

[fvasallol]

Hello!
We're running simulations from polar regions. We have chosen KMIXD=0 and KMSL=2 in the SETPUT. We have selected as initial heights 25%, 50% and 75% fractions of the boundary layer depth a we are using as atmospheric data input the GDAS1 model.
We have found several outputs that show an inconsistency between the initial height of one, two or the three simultaneous trajectories and the initial mixed layer depth. All these "anomalous" files show that the three initial heights are coherent with each other, since they keep a proportion 1:2:3.
We have distinguished three anomalous cases:
a) Two different initial values of mixed layer depth: the one of trajectories 1 and 2 is consistent with the initial heights but the one of the trajectory 3 does not.

trayectory_a.gif (7.46 KiB) Viewed 3915 times

b) Two different initial values of mixed layer depth: the one of trajectory 1 is consistent with the initial heights but the one of the trajectories 2 and 3 does not.

trayectory_b.gif (7.49 KiB) Viewed 3915 times

c) Only a single initial value of mixed layer depth, but it's inconsistent with the initial heights of the three trajectories. We think that the second value of the mixed layer depth is "hidden" and can be inferred from the values of the initial heights.

trayectory_c.gif (6.41 KiB) Viewed 3915 times

We usually calculate the trajectories with a Hysplit version for linux installed in a computer center, but we have repeated several "anomalous" trajectories in a laptop with a Hysplit for Windows, and the results are identical.
We estimate that this issue impacts - to a greater or lesser extent - less than 0.5% of the total number of trajectories that we have launched, and we are very interested in knowing the reason.
Can someone help us?
Thank you!

[MarkCohen]

Hello! Thank you for alerting us to this issue, and our apologies for taking so long to respond.

First, thanks for providing the details that you did, and for very clear explanations (with color coding!). With the information you provided, I was able to attempt the same runs here (i.e., same starting locations and heights, same met data, etc.). I was able to reproduce the results you showed, including the inconsistencies in starting height.

For the first case you mentioned, I ran the "profile" program at the starting location & time and found that the "correct" mixed layer depth (PBLH), according to the gdas1p0 met data was ~142 m above ground level (AGL). But, as we both found, the fractional PBLH starting heights were based instead, erroneously, on a PBLH of ~79 m AGL. It turns out that there was a small bug in the HYSPLIT code in how the model dealt with locations at the horizontal edge of the grid, i.e., at longitude = 0. This sometimes caused a nearby grid point to be used instead; in this case, a nearby grid point with PBLH ~ 79 m AGL was used. The issue would not be apparent if the PBLH was relatively uniform near longitude 0.0, but could be apparent when there are non-trivial gradients in PBLH around longitude 0.0. In any event, we have attempted to fix this issue in revision #941 to the code repository.

Testing was carried out on the three cases you provided, and this code change appears to resolve the problems that you found.

For case 1:

For case 2:

For case 3:

Additional note: In carrying out the testing for this case, I noticed that the default value of KMIX0 (the minimum mixed layer depth) in the trajectory model has been set to 0 meters. With this setting, if one encounters very low PBLH values in the met data (e.g., it seemed that there were a few values of 0 PBLH in the gdas1p0 data), then an array bounds violation can occur in the program when it tries to locate meteorological levels relative to the PBL. So, I changed the default value of KMIX0 to 25 (meters) to get around this problem. The user can always set their own value of KMIX0 in the "advanced" SETUP.CFG namelist file, but now, if nothing is specified regarding this in SETUP.CFG, then a KMIX0 value of 25 (meters) is used rather than 0 (meters). This small change was implemented in revision #942 to the code repository. I do not believe this change will have any significant impact in the vast majority of situations.

[fvasallol]

Hello!
Thank you for your attention and for answering my question. I'm glad that my information was useful to correct this small issue in Hysplit.
I realised that most of incoherent trajectories - and the ones with biggest deviation - started at longitude 0, but I didn't include this detail to not lengthen the post. Now, I think it's a significant fact and I think I should have include it.
Regarding the PBLH, we have found some problems because we are studying polar regions and PBLH is critical in them. At first, we found that the GUI include an option "Default minimum" for the Mixed Layer Computation in the Concentration menu, but this option was not included in the Trajectory menu. So we tried to include a line KMIX0=valueselected in the SETUP.CFG, and it worked.
So that's what we did when we needed to set a minimum value of PBLH. This matter has been addressed in this forum by my colleague sgonzalezh in the topic "HYSPLIT Boundary Layer Issue".
All your information is very useful for us.
Thank you so much!