On further reflection - this may not be quite right
The models is expecting lbcs each hour (rg01_rs01_m01_lbc_freq=3600s), so for the 12 hr cycle there should be 12 times in the lbc file, but there is only one time.
(does the model interpolate to get hrly lbc updating)
Hi Grenville,
I haven’t changed anything in um_createbc from the default settings, and can’t see whether it does any interpolation in time - do you think it is a matter of adjusting the rg01_rs01_m01_lbc_freq=43200 and CRUN_LEN to match the output frequency of the input era data (i.e. 12 hours)?
I’ve given that a whirl and submitted it, but it hasn’t started running yet…
Cheers
Ella
can you supply ec_cb000 ec_cb012 ec_cb024 ec_cb036 ec_cb048 type files, but at 6hrly intervals?
Hi Grenville,
Unfortunately the ERA5 data I have is only every 12 hours, for December 2000 only. If there’s an alternative 6-hrly source on archer that includes all the fields required to force the model, I could try that…? Do you know if there is another place where era5 sits on disc?
Cheers
Ella
Hi Ella
I see that you changed to lbcs each 12 hrs & that produced a more sensible lbc file. We never got to the ancil problem. It might help if you switch off packing - that way there is a chance to see which field is causing problems.
Grenville
Hi Grenville, I did turn the stashpacking off and am now getting some different errors about the recon. It’s complaining about the topographic index ancils:
[FAIL] namelist:items(7ea74479)=ancilfilename: UM_ANCIL_LSH_TOPIND_DIR: unbound variable
I originally didn’t have these set up to reconfigure because I didn’t think they were necessary but I added them in because initially I was getting this same error in the um_rs01_rg01_fcst step. I noticed this namelist entry file in app/um/opt/rose-app-ra3_pack3.conf so assumed it was necessary for the RA3 config I’m using:
> # #84 land consolidation package
> [namelist:items(4786818a)]
> ancilfilename='$UM_ANCIL_LSH_TOPIND_DIR/$UM_ANCIL_LSH_TIND_FILE'
> domain=1
> !!interval=1
> l_ignore_ancil_grid_check=.false.
> !!netcdf_varname='unset'
> !!period=1
> source=2
> stash_req=274,275
> update_anc=.false.
> !!user_prog_ancil_stash_req=
> !!user_prog_rconst=0.0
I re-ran the global recon so that the inner domain would pick up the changes from there but am now getting the error in the ec_recon.
I can’t figure out whether this means I should have it in or out, as it seems to complain either way.
Ella
Ella
I can’t see why switching off packing would cause this – did the science configuration change somehow?
Actually, the logs show that it was STASH 1202 that was causing the problem (net downward sw flux) - that stream us still packed (pp2 output stream)
Perhaps back up a step & we can try to understand the problem with the radiation – could a dodgy ozone ancillary be the source?
Grenville
Hi Grenville,
Good idea. When I first went through the recon I was getting errors in some of the ancils, which I set to warnings so the recon could progress. One of those was definitely the ozone ancillary.
I’m actually trying at the same time to get the suite to use a higher res source for the ancils - n2560e. I’ve got copies of all of these on jasmin and moving them over slowly to /work/n02/n02/shakka/n2560e/.
Is there a way of updating the master ancillary directory to point to that instead? That might overcome the problems with ozone/smc/whatever else was throwing up those errors.
Cheers,
Ella
p.s. I’ve figured out how to update the source via /app/ancil_ozone/rose-app.conf (etc) for all ancil programs except the mask - which is the most important one because it allows all the subsequent ones to run.
I can’t access the default source file $ROSE_DATA/etc/ancil_master/vegetation/cover/igbp/v2/qrdata.igbp
to see what’s in it, so I can’t tell what I should be pointing it to instead. I’ve tried pointing to the /land_sea_mask/qrparm.mask, /vegetation/func_type_modis/qrparm.veg.func and /vegetation/fractions_igbp/qrparm.veg.frac files but they all fail.
Do you know what I need to do to get the CAP to pick up details for producing the mask? (Also well aware that this is more of an answer to this question, not the ERA5 one, but it seems like they’re linked…)
Thanks
Ella
Ella
I’m assuming you are in u-co447, then the install_cold_hpc app sets up links to ancil_master etc in /home/n02/n02/shakka/cylc-run/u-co447/share/data/etc/, so $ROSE_DATA/etc/ancil_master/vegetation/cover/igbp/v2/qrdata.igbp is /work/y07/shared/umshared/ancil/atmos/master/vegetation/cover/igbp/v2/qrdata.igbp
does that answer your question? You can change install_cold_hpc to set up your own ancil files as you like.
I’m not sure I understand what you want the CAP to do?
Grenville
Hi Grenville,
The problem is that I can’t see what’s contained in qrdata.igbp as xconv can’t seem to recognise the file type. I’d like the CAP to generate the nested ancillaries using the higher resolution n2560 global data instead of the default igbp data, so it would be helpful to know what it contains so I can replace all the necessary fields. But if there’s a way to do this with install_cold_hpc I will have a look at that.
Thanks,
Ella
Ella
The igbp data set is in a strange ascii format. Can you not use ANTS with the CCI dataset to create a mask. The CCI dataset is at much higher resolution than n2560.
Grenville
Hi Ella,
We’re downloading ERA5 data from the CDS for our work (or will be, once they sort out their server issues) - you could adapt our scripts for your work if you want: UM_Operational_Scripts/download_scripts at master · UoMResearchIT/UM_Operational_Scripts · GitHub
cheers,
Doug
Hi Grenville,
So it turns out I misunderstood what Stu Webster was suggesting I do with the n2560 ancils - what I actually need to do with them is use the um recon to regrid them onto the nested domain (to overcome the missing pole problem).
His instructions were to:
work through all the ancillaries currently being reconfigured in, and set source=1 (use the data in the start dump), rather than source=2 (use the data in the ancillary file).
If you find after doing this that you hit an “ancillary file not found” error, then you should also unset the filename for that ancillary file, i.e. set ancilfilename='’ .
I have tried to do this, but before I can get to the um recon, the ec_um_recon fails with what looks like a cylc error:
[FAIL] 'CYLC_TASK_ID'
/var/spool/slurmd/job1944684/slurm_script: line 86: FCATE: unbound variable
2022-07-01T16:18:22Z CRITICAL - failed/EXIT
Any ideas what is going on here?
Cheers,
Ella
Thanks Doug, much appreciated. Ella
Hi Ella
The immediate problem seems to be a typo in suite-runtime/dm.rc
FDATE=$(rose date -c -f %Y%m%d%H%M --offset=PT{{lbint}}H)
AINITIAL={{DRV_MOD["ec_ic_lbc_dir"]}}/era5_1deg-model-levs_N48L137_${FCATE}_all.nc
I guess FCATE should be FDATE, but I don’t see where n2560 is coming in here?
Grenville
Hi Grenville, thanks for spotting that. I corrected the typo and things proceeded.
On your point about the n2560 ancils - I think actually Stu meant to let the recon regrid the n2560 ancils onto the global grid, and not generate ancils in the first place. I’ve tried that but getting errors about the LSM not having the right resolution, I assume because the glm_recon expects ancils in N320, not N2560.
Error looks like:
????????????????????????????????????????????????????????????????????????????????
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 41
? Error from routine: RCF_OPEN_LSM_ANCIL
? Error message: Incorrect E-W resolution for Land/Sea Mask Ancillary
? Error from processor: 11
? Error number: 2
????????????????????????????????????????????????????????????????????????????????
Ella
p.s. did I see you and Ros at the ASC meeting yesterday? I was going to say hello but I think you left before I had the chance!
Ella
The reconfiguration won’t regrid ancillary files - it expects ancillary files to be on the target grid.
I may be wrong, but I’m thinking that you’d use a UM n2560 dump as the starting point and allow the reconfiguration to regrid fields from the n2560 dump down to n320 (or whatever the driving model resolution is), so at that point you’d have effectively used n2560 ancillary fields.
I don’t understand Stu’s instructions in any other context.
I was at the meeting and heard your talk, but had to disappear for a zoom meeting then the train.
Perhaps it’d help (me anyway) to discuss this - I’ll email.
Grenville
Hello again,
I’ve sorted the ancil issue now, so I’m back to actually getting the suite to run with ERA data. I’ve followed the instructions in https://code.metoffice.gov.uk/trac/rmed/wiki/suites/nesting/ECdriver so that I have a domain at 0.11 deg, 702 x 620 LAM and a slightly larger domain at the same res to interpolate the ERA data onto which is 722 x 640 points and should be larger than the LAM + LBCs.
I’m getting an error like:
???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 4
? Error from routine: LAM_INCLUSION
? Error message: Eastern boundary at 575.1450000 not within source LAM with 215.0350000
? Error from processor: 0
? Error number: 2
???
…which I also get in a suite that I’m using to produce Arctic ancillaries. That suite is complaining that the LAM+LBC domain (i.e. the EC ancillary domain) is not within the parent domain - which makes no sense because the parent is the glm!
Can you see where this problem is coming from? The suite I’m using to run the Antarctic ERA-forced test run is u-co433, and the one I’m using to produce Arctic ancils is u-co447.
Cheers
Ella
Hi Ella
I can’t find the LAM_INCLUSION error in u-co447.
I’m not sure it’s complaining about the glm domain.
I’m guessing that 575.145 is 215.1435 + 360 - presumably the extra 360 is not relevant, but 215.1435 is East of 215.035, which might be. Do know where 215.1435 and 215.035 come from - are they from the domain and the extra nested region?
I can’t see what you set up for the extra nested region.
Grenville
Grenville