Error in HYSPLIT reading model data from 360-day calendar

[prerita]

Hi,

I am using meteo data from a GCM model with 360 day calendar and converting the NC files to arl format.

The HYSPLIT model runs fine and without obvious errors. However, upon checking individual output trajectories some months show error reading the previous month's data and stop when it needs the previous month data.
It is happening for only in particular months and for other months it completes fine with reading all the 10-day back data.
I suspect it is because the Global model data follows a 360-day calendar format and that is causing this issue.

If anyone can confirm if this is what might be why the HYSPLIT is not reading previous months, it would be very helpful.
Any help with solving this issue and converting my files to gregorian format for HYSPLIT would be greatly appreciated.

Thanks
Prerita.

[alicec]

Yes - that is almost certainly the issue. HYSPLIT will not recognize a modified calendar. For instance the CAM model
had a related issue in that it doesn't follow the normal calendar for leap years. the work-around was to just change the year
to one which was not a leap year.
viewtopic.php?f=25&t=2147&p=6188&hilit= ... +day#p6188

You can probably follow the approach outlined in that post to use the edit_head utility to modify the time of the ARL file.
https://www.ready.noaa.gov/hysplitusers ... #edit_head

[prerita]

Dear Alice,

Many thanks for your reply.
What I did was to use an xaaray utility that convert 360day to 365day calender and modify my nc files to make the datetime standard. Then
converted the nc files to arl and ran the hysplit.

But now the trajectories are not being calculated and I see the error -

Code: Select all

ERROR metset: meteorological data time interval varies. Check input meteo file.
  Changed from          180  min to         1620  min
  At day/hr           20           0

My monthly files are now in standard format but there is this strange error.
I see this issue has been discussed but my headers of the arl files are in the correct date format.

Only thing different which I can think of is I had cf time in the original nc files whereas the modified nc files have datetime64[ns] time format and then they are converted to arl files.

Could this be causing the issue?

I would appreciate any help.

Thanks, Prerita

[alicec]

It sounds like something went wrong in the conversion so the times in the ARL files are not consistent.
Try using some of the utilities listed here under 'Meteorological Data Examination' to test the ARL files.
For instance chk_times or profile or chk_file.
https://www.ready.noaa.gov/hysplitusersguide/util.htm

[prerita]

Dear Alice,

Thanks for your answer.

After trying a lot of different things I think leap year is not the issue.

For modifying the time in arl files approach using edit_head how do I modify the time for a monthly 30 day ARL file to 31 31-day ARL file ?
I need an additional day with 31st date in the monthly ARL files which I do not have. All my monthly files have up to 30 days which is the issue.

Secondly, The add_miss utility seems to do this by interpolating the data. But doesn't work because it requires 'missing data records must already exist in the file to use this program'. I do not have missing records corresponding to the 31st dates in my files.
So if I can create empty records for the dates maybe add_miss can work.
But I do not know how to do that for arl files.
Could you help?

Alternatively, if you can advice on any other workaround, it would be very helpful as i have not been able to solve it.

[alicec]

Yes, similar to the leap year problem, HYPSLIT will terminate if it cannot find meteorological data covering the next time period that it expects (e.g. if a month that is supposed to have 31 days only has 30).
I think the best course of action would be the following. It's a bit laborious, but with some scripting, it shouldn't be too bad.

1. use xtrct_time to create daily files rather than monthly.
2. shift the times of the daily files so they are contiguous. (e.g. shift Feb 1 to Jan 31, Feb 2 to Feb 1, and so on).
3. concat the files to get monthly files again (or just use them as daily files).