Help with the JULES FLUXNET suite, and for Urban modelling

[yuansun]

Dear Patrick,

Thanks for providing useful resources about the JULES.

I tried a JULES run once and not sure if it was successful or not. But I found there already are outputs.

For the future direction, I am interested in the urban module. As the JULES takes a tile approach, its urban representation has 1 tile and 2 tiles for the global scale but a more complex scheme only works within the UK domain. I am glad to contribute to the modelling by Fortran or data analysis on simulation outputs.

As Dr. Zheng mentioned, CLMU, the urban module in the Community Land Model, takes a canyon model with five facets, and its outputs are different from JULES. I am still thinking about the scientific questions.

Best,
Yuan

[pmcguire]

Hi Yuan

I am glad you could get it running, and that it can produce output. Did you successfully get the suite to make plots, as well, comparing the model to the observations?

The suite u-al752 is not specialized to be an urban modeling suite. But it can be adapted to do urban modeling. Running this u-al752 suite is mainly to get you up to speed on running JULES suites on JASMIN, getting all the right accounts, etc.

Furthermore, there are other JULES suites that are more specialized for urban modeling, and that use the MORUSES urban modeling scheme (with canyons, etc.) from the Met Office and the University of Reading. You can find such suites by searching with the rosie program on the cylc1 node of JASMIN. If you want me to search for such a suite for you, just let me know. Or you could search for it yourself.

Patrick

[yuansun]

Hi Patrick,
Thanks for reminding me to check the output plots. I found the previous run only output several files (maybe the spinup) while interrupted later. I re-ran the u-al752 suite.

Below are several questions if you could take an insight:

• I use the GUI window to monitor the job running. Due to the computing limitations, not all jobs run at once. Then, after some jobs running succeeded, I manually triggered the rest jobs that previously failed in submission.
  

  

• Does the job running sequence matter? In this run, I am not sure which job should run first so I manually trigger them blindly. I guess the cylc helps me automatically select jobs it wants to run first since some were submitted while other submissions failed.

• After making plots, I compared them to the sample, it seemed to be similar. ?
  

  

• How should I restart a branch running if one run failed in the middle? I met this issue before in CESM for long-time projections. For the JULES, I found the u-al752 starts with dump spin-up files. That is, if I want to start with other files, should I create a suite by myself? Or I should keep with the old suite and start all over again?

• For the urban, I ran ‘rosie go’ to search urban-related suites. I found some and they could be loaded in my directory.

  • I am not sure if I could use these suites from others.
  • Are these already published or passed model validation in Rose and available for all users?
  • Or are the suites seen in the rose just users’ records?

• For future runs, I am not sure whether I need to configure JULES on the school HPC or other machines. Could I keep using the JASMIN node and JULES workspace?

Best,
Yuan

[pmcguire]

Hi Yuan:

Yes, usually we only run maybe 10 sites at a time (or maybe it’s more than 10? I don’t remember now), and then it goes on to the next 10 sites, if enough cores are available to run on SLURM.
If some of the rose/cylc jobs for some of the sites fail, then you can retrigger the failed jobs in the GUI on the cylc1 node. You can even change the suite, and reload the suite and retrigger the failed jobs. Or you can retrigger successful jobs after changing and reloading the suite.

For the u-al752 suite, the running sequence for the jules subset of jobs doesn’t matter.
But you need to run the build/make job before the jules subset of jobs, and you need to run the make_plots after the jules subset of jobs is successfully finished.

As far as the plots go, it looks like the sample plot that we have there has a different set of sites than in your plot. So, the visual comparison only works for the first few sites.
Also, it looks like both the sample plot and your plot only have the daily JULES model curves in the plots (together with the monthly averages of the daily model curves). You might try to change the suite so that it will overlay the daily (and possibly monthly) FLUXNET observations on the plots. I don’t remember for sure right now if the following is possible, but you might only need to rerun the make_plots part of the suite after making this change.

You can create a copy of the u-al752 suite with rosie go, which will give you a new suite with a new ID number. And if you want to start the spinup with other dump files (or starting up with idealized parameters instead of dump files), then you can modify the new suite, and then use fcm to check in your changes to the suite to MOSRS, and then rerun the suite.

It’s normally ok to use other people’s suites, but I think you might want to contact the owner of the suite, to get their agreement before you start doing serious work with the suite.
Some of the suites have been vetted to some degree, but often the suites are not checked over very much.

You can keep using the JASMIN supercomputer to run JULES. I can best help you if you do your work with JULES on JASMIN. If you need extra space for managing your output, maybe the jules group workspace disk is not the best place on there to put lots of output. Your home directory is a good place there, but that has a 100GB limit per user, I think. But the home directory is backed up often. There’s a lot of space on the scratch disks, but that isn’t backed up, and files are deleted after 30 days anyways. If you need a lot of space on JASMIN, then maybe there is some other group workspace that we can get access for you to? Maybe U. Manchester has a group workspace there? Or maybe there is some other group workspace on JASMIN available?

Patrick

[yuansun]

Hi Patrick,

In Manchester, we do have access to a group workspace on JASMIN. I use it.

Best,
Yuan