Welcome, Guest
Username: Password: Remember me
Forum on HARMONIE Climate Branch

TOPIC: Prep_ini_surfex failed

Prep_ini_surfex failed 9 months 1 day ago #2294

Hello,

I am trying to run a simulation with the NorCP set-up and it crashed in the preparation of initial field for Surfex. So basically, the model try to copy this file

/lus/snx11062/HCLIM_NORCP_LBC/HCLIM38h1_ALADIN_ECEARTH_FPSCP_HIST_1995_2005/1999/12/ICMSHFULL+00000.sfx

And in this repertory, what we have is only

/lus/snx11062/HCLIM_NORCP_LBC/HCLIM38h1_ALADIN_ECEARTH_FPSCP_HIST_1995_2005/1999/12/ICMSHHARM+00000.sfx

However, in the model, this line is called as follow

/lus/snx11062/HCLIM_NORCP_LBC/HCLIM38h1_ALADIN_ECEARTH_FPSCP_HIST_1995_2005/1999/12/ICMSH${CNMEXP}+00000.sfx

Where in my sms/config_exp.h
CNMEXP=HARM

But somehow, this variable get recall FULL later...
Can somebody have a hint for me?

Thank you in advance

Prep_ini_surfex failed 9 months 1 day ago #2295

  • Bert van Ulft
  • Bert van Ulft's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 98
  • Thank you received: 22
Hi Dominic,

in the configuration file sms/config_exp.h there is an option SURFEX_LSELECT. If this variable is set to "no" all possible SURFEX fields are written to output files named ICMSHHARM+$FFFFF.sfx. To reduce the size of the files SURFEX_LSELECT can be set to yes (the default nowadays). In this case only the fields listed in nam/surfex_selected_output.pm are written to the ICMSHHARM+$FFFFF.sfx files. However, this is not enough to restart a run or continue into the next month. For this purpose a ICMSHFULL+$FFFFF.sfx file is written at the end of a month, so ICMSHFULL+00744.sfx for a 31-day month. This file contains all output fields again, same as when SURFEX_LSELECT is set to "no".
It looks like somehow this option doesn't behave properly in your case. Have a look at the bdstrategy file in your archive directory and the script scr/Boundary_strategy.pl, perhaps this can give a clue.

best wishes,

Bert

Prep_ini_surfex failed 9 months 1 day ago #2296

Hi Bert,

thank you for your very quick answer, I really appreciate.I will look it up.

Dom

Prep_ini_surfex failed 9 months 1 day ago #2298

Hi Dominic,

Bert explained very well the "theory" behind this, although the issue here is more of a practical nature where links have been lost.
The model expects/creates an initial surfex file for each month, ICMSHHARM+00000.sfx if SURFEX_LSELECT is false and ICMSHFULL+00000.sfx if true which is the case in the NorCP setup.
This initial file is linked to the surfex file of the last time step in the previous month, ICMSHFULL+00744.sfx for example. The links for these initial surfex files are created automatically during the run but when data (for some reason) is copied to some other location, these links might become lost. This is the case here.

I stand as the owner of the data under HCLIM_NORCP_LBC, so I can help with manually relinking the necessary file(s). I don't have a script that can do this efficiently for many months, but if you point me to a specific year and month you want to start the simulation I can do this.


Petter

Prep_ini_surfex failed 8 months 4 weeks ago #2303

Hi Peter,

Thank you for our answer and sorry not to have been quickly reactive to it, I was out of my office. So, let me resume to be sure that I understand correctly. I am not sure that I fully understand what’s going on. So for the NorCP set-up we should use SURFEX_LSELECT=yes meaning the model will look for ICMSHFULL+00000.sfx to initialize the surface scheme.

However, when I look into the in the repertory where those file is for the hindcast simulation for example (here: /lus/snx11062/HCLIM_NORCP_LBC/HCLIM38h1_ALADIN_ERAI_NorCP_1997_2017/2011/01) I do not see any ICMSHFULL+00000.sfx but only ICMSHHARM+00000.sfx (same time for any other month). So in that repertory we are suppose to have but type?

Furthermore, in my sms/config_exp.h
CNMEXP=HARM

which is called later by the model with something like ICMSH${CNMEXP}+00000.sfx, which make sense with what we find in /lus/snx11062/HCLIM_NORCP_LBC/HCLIM38h1_ALADIN_ERAI_NorCP_1997_2017/2011/01

I am sorry to bother you, but I am still struggling a bit…

Thank you again

Dom

Prep_ini_surfex failed 8 months 4 weeks ago #2304

Hi again,
FYI,
here I am using another set of LBC as used in my initial question. I wanted to test, but I have the same problem.


Dom

Prep_ini_surfex failed 8 months 4 weeks ago #2305

Hi Dom,

It is correct that we use SURFEX_LSELECT=yes, and therefore surfex initialization of the nested domain uses ICMSHFULL+00744.sfx instead of ICMSHHARM+00744.sfx (for a 31 day month) from the parent domain. These files are from the month before the one you are starting your simulation in.
However, the nested model actually searches for the parent's model ICMSHFULL+00000.sfx from the same month that you are starting your simulation in. To make this file generally available, the parent model every month automatically creates a symbolic link called ICMSHFULL+00000.sfx that points to ICMSHFULL+00744.sfx from the previous month.
When the parent model data is copied, these symbolic links are usually broken. Or, sometimes the ICMSHFULL+00744.sfx file is copied only for the known start date of the nested domain, since it is used only for initialization.
For example, you can see here that ICMSHFULL+00000.sfx exists only for January 1995, which is the start year of the historical run:
ls -l HCLIM38h1_ALADIN_ECEARTH_FPSCP_HIST_1995_2005/1995/01/ICMSHFULL+00*
and it is missing in all other months.
Someone created the link manually in this case, and this is what Petter offered to do for some other specific starting date that you would be interested in.

Does this help?

Danijel

Prep_ini_surfex failed 8 months 4 weeks ago #2306

Hi Danijel,

Ok, now I got it, thank you for your help.

Then, yes Peter, I would appreciate if you could created the link for my starting date.

I will write you an email.

thank you everyone

Dom
Time to create page: 0.080 seconds