Problem with converting timeseries data using mule-convpp

[njharvey]

Hi there,

I have modified the regional nesting suite (vn13.5) to output a number of variables at the timestep level but when the fields files get converted to pp (and then on to .nc) they only contain one of the variables.

The error I get is: skipping field validation due to irregular lbcode

My suite-id is u-dt760 and an example file is here: ~/cylc-run/u-dt760/run8/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um/umnsaa_chil_0_000

I had this previously on Monsoon2 and it was solved by pointing to an alternative (older - I think) version of um-convpp. Is this possible on Monsoon3?

More info on previous similar issue here: https://cms-helpdesk.ncas.ac.uk/t/problem-with-conversion-to-pp-for-um-timeseries-output/1486/2

Thanks in advance,
Natalie

[r-sharp]

I can't comment with any authority on the likelihood of availability of um-convpp on Monsoon3, however, as part of the team responsible for maintaining the "um tools" I can tell you that um-convpp was deprecated and then retired some time ago so IF there is a version on monsoon3, it will be old and unmaintained.

Your time, and that of others who follow in your footsteps may well be better spent investigating the cause of the issue when using the updated pp conversion tool.

Given the error message is complaining about an invalid lbcode - have you considered looking into why it's invalid and what it should actually be ?

[njharvey]

Hi,

Many thanks for getting back to me about this. When I had the problem on Monsoon2 I did a little bit of digging but the NCAS CMS team suggested the use of um-convpp and that solved the issue so I didn't look any further.

The full error message is below.

Natalie

• echo '[INFO] Doing convpp for /home/users/natalie.harvey.ext/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSE
  S/um//umnsaa_chil_0_000'
• mule-convpp /home/users/natalie.harvey.ext/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um//umnsaa_chil_
  0_000 /home/users/natalie.harvey.ext/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um//umnsaa_chil_0_000.tm
  p.pp
  /projects/metoff/umdir/mule/mule-2025.10.1/python-3.7.0_numpy-1.20.2/openmp/lib/mule/validators.py:198: UserWarning:
  File: /home/users/natalie.harvey.ext/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um//umnsaa_chil_0_000
  Field validation failures:
  Fields (0,1,2,3,4, ... 11 total fields)
  Skipping Field validation due to irregular lbcode:
  Field lbcode: 31320
  warnings.warn(msg)
• [[ 0 -eq 0 ]]
• mv /home/users/natalie.harvey.ext/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um//umnsaa_chil_0_000.tmp
  .pp /home/users/natalie.harvey.ext/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um//umnsaa_chil_0_000.pp
• i=1
• for filename in $datapath/a_
  ++ rev
  ++ cut -c1-3
  ++ rev

[r-sharp]

Hi Natalie, I'm having a few issues digging into this.
Firstly, I'm not much of a suite user, so was wondering if you had attempted to perform the conversion at the command line ? and whether that produced the same errors.
Secondly, I could really do with knowing which version of mule-convpp you're using (although I'll mostly be trying to debug at the most current (2024.11.1 I believe))
Thirdly, I asked someone to copy the file that you're seeing errors with to somewhere I could see it, but they said your HOME directory is 'locked down' so they don't have access. The request is that you run chmod 0755 $HOME to make your HOME directory readable and executable for 'all users' (a directory need to be executable in order to see it's contents).
Thanks
Roddy

[njharvey]

Hi Roddy,

Thanks for looking at this for me. My HOME directory should be accessible now.

Below is the error when I run mule-convpp on the fields file. It suggests that it using mule-2025.10.1 (I had to load the um_tools module to get this to run on the command line).

Thanks,
Natalie

natalie.harvey.ext@exd-login11:~/cylc-run/u-dt760/run7/share/cycle/20230612T0000Z/Wessex/ukv_CCI-WC/RAL3p3_MURK_DSSOIL_MORUSES/um $ mule-convpp umnsaa_chil_0_000 test
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

• (CONVPP-II) Module Information *
  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

mule : /projects/metoff/umdir/mule/mule-2025.10.1/python-3.12.10_numpy-1.26.4/openmp/lib/mule/init.py (version 2025.10.1)
um_utils : /projects/metoff/umdir/mule/mule-2025.10.1/python-3.12.10_numpy-1.26.4/openmp/lib/um_utils/init.py (version 2025.10.1)
um_packing : /projects/metoff/umdir/mule/mule-2025.10.1/python-3.12.10_numpy-1.26.4/openmp/lib/um_packing/init.py (version 2025.10.1) (packing lib from SHUMlib: 2025101)

/projects/metoff/umdir/mule/mule-2025.10.1/python-3.12.10_numpy-1.26.4/openmp/lib/mule/validators.py:198: UserWarning:
File: umnsaa_chil_0_000
Field validation failures:
Fields (0,1,2,3,4, ... 11 total fields)
Skipping Field validation due to irregular lbcode:
Field lbcode: 31320
warnings.warn(msg)

[r-sharp]

Hi Natalie,
I've got a copy of the file, thanks.
I'm seeing the same warning message with the latest version of mule-convpp (2025.10.1)
I've got as far as discovering mule won't validate fields with an lbcode setting of anything other than 1, or 101 (which are regular lat-lon grids on regular and rotated poles)
31320 is a gregorian calendar timeseries cross section.
I'm rather presuming that gregorian calendar, and particularly timeseries of those, were a bit scarcer when mule was written.
The actual comment in the code is :
# If the field has a particularly exotic lbcode value, we can't be
# sure about the validation
which would seem to rule out the ability to validate any cross-section diagnostic output, where the lbcode would be codes around 10000 and 30000. Which, to me, implies that none of the users for the first few years, when the mule tools were in development alongside the um ones, no users requested the functionality to create pp files from anything other than full fields.
I don't know how to create a 'robust' solution to this (and the original mule author left some years ago and it's barely been touched since). However, some things that might help me are if you have any examples of input files, and the corresponding output files that you successfully converted with um-convpp (which presumably did a lot less 'validating' of the fields) but fail with mule-convpp.
I'm going to have to experiment with rebuilt versions of the code, and I'd like some suitable files where I can compare the input of the modified code to that of the code which produced 'acceptable' results.
Thanks
Roddy

[r-sharp]

I've produced a file, which appears to have a single point, for numerous timesteps, for 11 fields.
One of our other team members has put it somewhere you shold be able to see it. /home/users/sam.clarke.mon/convpp_test_output.pp
I'd be curious to know if that has worked. Thanks

[njharvey]

This is all really helpful and confirms what I thought about what was happening.

Unfortunately I don't have any of the monsoon2 fields files that were used as input for um-convpp as they were not archived and not monsoon doesn't exist anymore. I do have some of the previously converted pp and nc files if they might be useful?

Thanks for all your help with this,
Natalie

[r-sharp]

A couple of the pp files might be useful please. So I can see what they look like and what's in them.

[njharvey]

Hi Roddy,

That sounds really promising but the file that you shared only seems to have one variable in? Is this the right file?

Best wishes,
Natalie

[r-sharp]

Rats, that's what I get if I load it in xconv too.
Can you tell me if it has at least got that one field correct and with data available at the various times.

[njharvey]

That is one of the variables and it has the right number of timestep when compared to the ff file.

[r-sharp]

Well, that's some good news 😃
I just need to work out where the others went...
is it by any chance the first, or last field in the FF ?

[r-sharp]

One other thing I'm curious about, is how 'desirable' is the 'pp' file ?
I notice from your investigations on the cms-helpdesk that the next stage is to convert to nc.
mule-convpp is only having an issue validating that the file is valid (or at least that appeared to be the case until you pointed out there was only one variable), but mule-convpp is a relatively 'simple' python file that uses the mule library for reading/writing. So the first half of the script deals with readnig a FF and storing the headers and data in a bunch of dictionaries and numpy data arrays - which (whilst I'm not the man for the job) might be easy enough to write out directly as nc and skip the pp file completely.

[njharvey]

I am interested in the nc file really so no great problem if there isn't a pp one too, especially if that makes life easier your end!
Are the pp files converted using um-convpp still useful if you go down this route?

Thanks again for your help!

[r-sharp]

Actually, digging further in, the validators that are complaining are on the 'read' of the FF. Not the write.
So it might not be an easy alternative route.

[r-sharp]

Another days digging, and it's not looking good.
I believe mule-convpp IS writing all the data to the output file, but in a way that xconv can't read beyond the first field.
If I manually "mess" with the loop over the fields, the output, according to xconv, only has one field, the name of which changes as I change which fields I skip and which I process.
I'm pretty stumped.

[r-sharp]

-Quick update-
I threw in a quick hack to see if it helped. xconv now tells me there are 11 fields in that file.
I've asked the team to copy it somewhere you can see and let me know. Hopefully I'll have an update for you on Monday.
Hope you have/had a good weekend.

[njharvey]

Exciting!

[r-sharp]

Hi Natalie,
I've got another file for you to try. It's been copied to /home/users/james.bruten.mon/convpp_test_output_mkII.pp
Please let me know if that's useful.
Roddy

[r-sharp]

No need to apologise. My bad if I put gave the impression a response was expected. My boss keeps leaning on me to keep folk updated on progress when I dissappear down the rabbit hole 😃

[njharvey]

No problem - I am very grateful to you for going down this rabbit hole for me!

File looks good to me :)

What did you have to hack? How can I access the version of mule-convpp you used to create this?

Thanks

[r-sharp]

Have you managed to check the data is what you want ? and/or that convsh can convert to nc ?
xconv reports an error if I try to plot or 'show' the data.

A comment you made in the original/previous indicates this error is holding up your work. How quickly do you need to be able to move forward ? I'd like to try and work up a way of getting the script within mule (pp.py) to logically turn off the operations I've cut out, either by passing an arg in from mule-convpp, or from a test on the headers within the data passed.

At the moment, while what I've got will work for you - it will break for others.
Also I've had to create my own mule installation and then change both PATH and PYTHONPATH to get it to run. These are changes which override where every mule based operation will pick up code/libraries from - so if you were using mule elsewhere in the same environment, they'd pick up the altered version, not the original one.

[njharvey]

Hi Roddy,

I get the same xconv error when I tried to plot or show the data in the pp file. However, the script I use to do the conversion to nc runs fine and I can open the nc in xconv and inspect the data (see below).

It would to good to get this resolved ASAP as we have been delayed quite a bit with the move to Monsoon3. We are planning to run a whole years worth of simulations which will take some time!

How long do you think that it will take for you to develop the changes to mule-convpp that would be the best case? I am not an expert in the nesting suite that I am using but am guessing that mule is used in many places.

Thanks,
Natalie

[r-sharp]

I've emailed you directly for the next steps