Opened 2 years ago

Closed 23 months ago

#2201 closed help (fixed)

Using custom ancilliaries

Reported by: shakka Owned by: willie
Component: UM Model Keywords: ancilliaries, orography, lsm
Cc: Platform: Monsoon2
UM Version: 10.4

Description

Hello,

I'm currently trying to run Stuart Webster's nested suite for the UM with new custom orography and land/sea mask ancilliaries for the Antarctic Peninsula. I'd greatly appreciate some guidance on the steps I need to take to do this.

  1. I understand that I need to use xancil to first convert the existing files that I have (in netCDF format) to UM ancils. Having tried this, I get the following error:

Writing Orography ancillary file /home/d04/elgil/ancils/orog

Warning: variable ht has different variable type to its attribute missing_value
ERROR: in procedure get_ncdata_r 1 : NetCDF error number -49 :
NetCDF: Variable not found :

— is this likely an issue with the files I have, or the way I am using xancil?

  1. I now have the problem of getting the UM to read these ancils, rather than the ones that it would otherwise automatically generate. Where do I need to specify the input folder to make sure that it uses the new lsm and orog files rather than the old ones, and doesn't overwrite these? Does it matter that the new ancils are only for the regional model (I need to run the global model to generate the LBCs to drive the first nest)?
  1. Will the UM be able to calculate SSTs and sea ice fields using the new lsm? Do I need to perform a 'create new ancilliaries' run first to get the model to do this?

If there are any steps that I've missed with this, I'd be grateful if you could let me know.

Thanks,

Ella

Change History (15)

comment:1 Changed 2 years ago by shakka

Hello,

Is there anyone available who could help me with this problem?

Thanks,

Ella

comment:2 Changed 2 years ago by jeff

Hi Ella

Regarding 1. You only have the height field in your netcdf file but xancil needs more fields than that to create the ancillary file. At a minimum your will need the standard deviation of orography variable as well, but other fields are also probably going to be needed this will depend on how your run has been set up.

Do you have a suite id? This will help answer your query.

Jeff.

comment:3 Changed 2 years ago by shakka

Hi Jeff,

Thanks for your response. I do have access to the SD of orography as a separate .pp file for each resolution - is there a way to combine the SD and height into one file?

I was previously using u-ah710 but have another copy of the same suite on Monsoon2 named u-an505.

Thanks

Ella

comment:4 Changed 2 years ago by jeff

Hi Ella

You could use xconv http://cms.ncas.ac.uk/documents/xconv to read both files and then write out a single netcdf file. Make sure Netcdf output precision is set to 64 bit, this can be found under "Setup Xconv Defaults".

Even if you do this I'm not sure if its going to work, xancil creates files for direct use by the UM but the nesting suite uses the central ancillary program to create the ancillaries and I'm not sure how (if?) this will work with a xancil generated file. Hopefully someone who knows more about the nesting suite will comment on this.

Jeff.

comment:5 Changed 2 years ago by shakka

Hi Jeff,

I know that it's possible in theory as several people using the nesting suite have updated the SST and sea ice ancillaries previously. However, as you say, I'm not sure about how they generated them in the first place. I'll have a go at combining them in xconv now.

Thanks for your help

Ella

comment:6 Changed 2 years ago by shakka

Hello - I have generated ancils that I think should work. I am now having difficulties with getting the model to run using them.

Following the advice of colleagues, I have taken the following steps:

  1. Added two namelist entries at the bottom of my rose-suite.conf file pointing to the orography and land/sea mask ancillaries
  1. Changed the ancil_filenames and ancil_versions files to reflect this
  1. Created a directory containing these two files, alongside copies of ancils from a control run (the ones that aren't orography or lsm) and my orography and lsm files.

The model is now failing on the reconfiguration (I have encountered various errors*, which all seem to suggest that they are not handling the ancils correctly). Is there a way of getting the UM to overwrite the ancils generated by the CAP with my own before the SSTs and sea ice ancils are generated? Unfortunately my Fortran coding skills are essentially non-existent so I'm not too sure where to start with this.

I'd be grateful for any advice.

Thanks,

Ella

  • Such as 'calendar year doesn't match calendar of model setup' - this one is strange, as orography and lsm shouldn't have a time dimension!

and 'Incorrect E-W resolution for Land/Sea? Mask Ancillary'- again strange, as they should match.

comment:7 Changed 2 years ago by willie

Hi Ella,

Have you checked that the Peninsula 4.4km mask produced by the Nesting Suite aligns with the one you have created in Xancil? I could only locate share/dataancils/Peninsula/km4p0/qrparm.mask which I'm guesssing is what the Nesting Suite produced.

Regards
Willie

comment:8 Changed 2 years ago by shakka

Hi Willie,

In my rose-suite.conf file I have set the ancil build mode to custom and then pointed the UM to read ancils from /projects/cascade/elgil/ancils. This is the folder that I have put my own data in. They should align, as they were produced from the original UM ancils generated by a previous run. However, the new ancils don't contain all of the information that the default ancil does, e.g. XX gradient and XY gradient. The files I have produced only include the orography itself and the standard deviation of orography. Is this likely to be the source of the problem?

Ella

comment:9 Changed 2 years ago by willie

Hi Ella,

The files

e3afe409d3f1f16be78549dd45a0174b  /projects/cascade/elgil/ancils/qrparm.mask
e3afe409d3f1f16be78549dd45a0174b  /home/d04/elgil/cylc-run//u-an505/share/data/ancils/Peninsula/km4p0/qrparm.mask

are identical. Which suite are we talking about?

Regards
Willie

comment:10 Changed 2 years ago by shakka

Hi Willie,

The suite I am referring to is u-ai781.

Best,

Ella

comment:11 Changed 23 months ago by willie

  • Owner changed from um_support to willie
  • Status changed from new to accepted

Hi Ella,

Sorry this is taking so long. I've been on vacation for two weeks.

As I understand it, the two files of interest are

  1. /projects/cascade/elgil/ancils/orog
  2. /share/data/ancils/Peninsula/km4p0/qrparm.orog

The first is your custom orography and the second is the CAP/Nesting suite generated version.

I've looked at these and they are correctly registered in terms of lat/long, but your custom orography is missing the orographic gradient, silhouette, half peak and unfiltered orography that the CAP version has. You should regenerate the custom orography to match.

I've run a copy of your u-a9781 plus job plus local modifications and it is running out of memory in Peninsula_4kmp0_ctrl_um_recon when it attempts to process the ancillaries.

I've also noticed that your custom orography is not ending up in the share/data/ancils/Peninsula directory where it needs to be. So the modifications you've made to app/um/rose-app.conf to achieve this have not worked. I think your custom orography should have the same name as the one you're trying to replace and there should be some sort of app that inserts it in the right place after the CAP/Nesting suite has created the originals. Then there is no more to do.

I don't think you need to modify ancil_filenames and ancil_versions as these relate to the global driving model.

I hope that helps.

Ticket #2210 is the same problem.

Regards
Willie

comment:12 Changed 23 months ago by shakka

Hi Willie,

Thanks for your reply. I've since come to the same conclusions about the ancil_versions and _filenames myself.

On the advice of Stu Webster I updated the symlinks to qrparm.orog.mn so that this points to my own custom orography. It seems to work.

I've managed to overcome the memory issue by changing the endianness of the ancil file after reading another ticket about a similar issue. I didn't generate the custom orography myself, so if it's possible, it would be easier to omit the gradient, roughness etc.

Rather, I now get an error in the replanca_rcf_replanca routine:

Error message: replanca_rcf_replanca: PP HEADERS ON ANCILLARY FILE DO NOT MATCH[76]
Error code: 231?  Error from routine: Rcf_Ancil_Atmos

I've deduced from the advice given in ticket #807 that this refers to the sea ice fraction (STASH code 31), which suggests to me that the model has not interpolated the sea ice fraction to match the new land sea mask correctly.

Do you think this could be the case? If so, how do I go about altering this so that this doesn't happen?

Thanks,

Ella

comment:13 Changed 23 months ago by willie

Hi Ella,

Is this still an issue?

Regards
Willie

comment:14 Changed 23 months ago by shakka

Hi Willie,

Sorry - I forgot to update this. I've had a wide array of issues since this but have managed to sort them all out and get the news ancils to run. Of course now it fails due to numerical instability, but that's a whole different kettle of fish!

I realised that for this particular problem I had an incorrect namelist entry in rose-app.conf, and deleting it solved the above error.

Thanks,

Ella

comment:15 Changed 23 months ago by shakka

  • Resolution set to fixed
  • Status changed from accepted to closed
Note: See TracTickets for help on using tickets.