Land modelling working group, meeting minutes 2023

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important please feel free to add it below or edit this post (it’s a wiki post).

DATE: 14/02/2023
Attendees: 10-15

Introduction on the role of the working group:

  • build a community
  • foster collaboration.
  • advise ACCESS-NRI on ACCESS development

Discussion on how to organise the working group meetings:

Gave a short description of COSIMA meetings. The COSIMA community came together because they had to collaborate on the high-resolution. We need to find our points of collaboration.

Proposed points of collaboration:

  • datasets - reference datasets and model outputs
  • evaluation and benchmarking

The meetings should also allow for discussions of interest to a smaller group only (most likely CABLE-specific discussions).

What people expect from the working group:

For the working group to be effective we need to know what participants expect from the working group. A few participants shared their thoughts, we’ll continue at next meeting:

  • Bring JULES and CABLE closer together.
  • Share datasets with more standardisation and quality control.
  • Create a community that has a voice for funding and advocacy.
  • More discussions on the science: how to best represent key processes in the model
  • Get to work on the same code base

Meeting structure:

We will start with fortnightly meetings. Short presentations of work in progress by volunteers at each meeting.

2 Likes

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important, please add it below or edit this post (it’s a wiki post).

DATE: 14/02/2023
Attendees: 5

The low attendance was noted. We need to make a stronger effort to communicate the meetings. In particular, we need to emphasise the need to watch the forum topic for people who want to receive announcements via emails.

Introduction of Sean Bryan
Sean joined the land surface infrastructure team at ACCESS-NRI last December. He is currently working on the evaluation framework for CABLE.

ACCESS-NRI plans and status
Claire C. gave an overview of our current projects for the land at ACCESS-NRI and our current directions. We currently focus on two aspects:

  • Evaluation framework for CABLE
  • Migrating CABLE to GitHub

Evaluation framework for CABLE
We are working on a release of benchcab, a Python software to manage the execution of standard configurations for CABLE.
Sean has finished a refactoring of the code to improve its navigation.
The plan is:

  • focus first on single site runs
  • release a version with 3 modes of operation. It runs 2 branches of CABLE, for a standard set of configurations for various sets of met. sites. The analysis is then handled by modelevaluation.org. The site options are:
    • 1 site
    • 5 sites
    • 42 sites
  • extend to spatial configurations and use iLAMB for the analysis.
  • extend to CASA

Migration to GitHb
Description of the migration plan is posted on this forum.
Current blocker for the migration: gaining access to the TRAC database. Hopefully, resolved soon.
Discussion on whether the move to GitHub should coincide with a complete overhaul of the development practices or the development practices could come later in stages. Some arguments either way. It is likely the migration to GitHub will happen first and the development practices will change later.

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important, please add it below or edit this post (it’s a wiki post).

DATE: 21/03/2023
Attendees: 8

We continued a roundtable of introductions. Some interests and current projects include model evaluation, JULES development, merging improved CABLE hydrology into the trunk, land management and regional modelling.

Working group suggestions:
The working group should be model agnostic as much as possible. There was a suggestion to have a list of code developments that people are actively working on. This would help technical work and code merging, as well as help scientists find synergies across projects.

Land working group workshop:
A date in late August was suggested and a hybrid meeting is strongly preferred. It was suggested the workshop would run over 2 days and be a mix of science talks and breakout sessions. The breakout sessions could tackle “grand challenges” common to both JULES and CABLE such as model evaluation and drought/climate extreme responses. Other themes could include met office coding and documentation standards, as well as dataset sharing. The wider community will be consulted in due course about timing and themes.

Call for science talks:
We hope to start science updates at the working group meetings. Gab offered to give the first talk on PLUMBER2 results. We welcome other contributions in future meetings.

MODSIM:
Christoph and others are running a land modelling focused session at MODSIM. Abstract submissions are currently open.

Model evaluation:
Noting that we will run demos of modelevaluation.org and iLAMB as soon as possible, hopefully in the next month or so.

1 Like

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important, please add it below or edit this post (it’s a wiki post).

DATE: 5/04/2023
Attendees: 7

Land working group workshop:
Timing of the workshop: the ACCESS-NRI will run a workshop during the week of 5 September with the intent to reserve a large portion of the time for the working groups to meet and tailor their own content. We are now thinking of holding the Land Modelling Working Group in the same week as the ACCESS-NRI workshop, with each workshop lasting for 2 days.

A poll is to be sent to the community to gather topics for the workshop. A few ideas raised were:

  • Introduction to the different versions of CABLE
  • Should we have separate breakout sessions for CABLE and JULES to discuss specific topics for each?

Science talks
Juergen offered to give an overview of TRENDY.

ESM and POP-POPLUC
We discussed the status of the project to get POP-POPLUC working within ACCESS-ESM.
First analyses showed it would have to be built using 2 executables, one for the fast land processes and the atmosphere, the second for the slow land processes called annually. Ian is trying to put together a working plan of what would be needed to achieve this. The need for a good testing framework appears as the major first step. Additional funding for the ACCESS-NRI is being sought to enable this project.
Anna expressed interest in an offline version of CABLE with functionalities from POP-POPLUC and other branches merged together.

CABLE development: Rachel commented it would be good to have better visibility of everyone’s development plans to help better manage concurrent developments.

ilamb: analyses of ACCESS-ESM1.5 with CABLE3 showed less carbon variability in amplitude than ACCESS-ESM1.5. It would be good to have ilamb available soon to diagnose this.

CABLE docathon: the effort will be revived soon with virtual events, one Tuesday per month. Email to be sent soon by Claire

CABLE move to git: it is currently blocked. Claire is awaiting access to the CABLE TRAC database to work on transferring the tickets out of the TRAC system. Once the tickets are sorted, the process will resume with information sent to users.

Actions

  • Claire to contact Gab and Juergen to set dates for presenting at the working group meeting
  • Claire to send calendar invite for the working group meetings
  • Claire to set up CABLE docathon virtual events
1 Like

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important, please add it below or edit this post (it’s a Wiki post).

DATE: 18/04/2023
Attendees: 6

Discussions made it clear people did not receive email notifications about the meetings.
:red_circle: ACTION: Send an email to cable-users and jules-users about the WG with information:

  • how to get notified
  • meeting frequency, time and Zoom link.

Workshop for the land community:
It is clear we would want a 2-day hybrid workshop for the community.

:red_circle: ACTION: send a poll to the community to finalise the dates: either some week in August or during the week of the ACCESS-NRI workshop.

Miscellaneous:
There were some discussions around the difficulties in using CASA because of missing:

  • code modifications in the trunk which are sitting in the TRENDY branch
  • running script to handle the various stages of the spin-up required.

We had a quick look at the figures currently produced in modelevaluation.org for the CABLE evaluation tool.
:green_circle: Recommendation: include more summary tables, add more variables in the comparison (see ilamb), add comparisons between science configurations, not only per science configuration between CABLE versions.

1 Like

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you think is important, please add it below or edit this post (it’s a Wiki post).

DATE: 02/05/2023
Attendees: 11

Gab Abramowitz presented results from the PLUMBER 2 experiment. The entire discussion of the meeting focused on the PLUMBER 2 results.

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you think is important, please add it below or edit this post (it’s a Wiki post).

DATE: 16/05/2023
Attendees: 12

There was a comment it would be good to have a way to collect CABLE’s publications so people are aware of the research that is happening with CABLE.

LWG workshop:

discussed ideas for themes for content:

  • evaluation: benchcab and ilamb
  • science presentations
  • discussion on prioritisation for development of CABLE and JULES to improve coordination
  • datasets
    • input datasets. What are the best datasets.
  • code management update:
    • coding standards
    • what needs to be done for code submissions to new versions.
  • planning experiments for WG resources.

benchcab discussions

benchcab is the name of the tool for the evaluation of CABLE. Sean Bryan (NRI) has given a presentation of the tool. Discussion focused on the next priorities for development on the tool:

  • support for spatial simulations with ilamb:
    • need it to see patterns.
    • additional variables from i lamb.
    • comparison to ACCESS runs.
  • support for CASA, seen as lower priority:
    • pft lookup table: how to test with ≠ tables.
    • Spinup software: POP’s setup tool and Martin’s site spinup tool.
  • Configurations benchcab should run:
    • High-res Australia run in ilamb

DATE: 6/06/2023
Attendees: 14

Workshop
Set for 7-8 September
We presented the proposed structure and ask for feedback.

AMOS
Mengyuan Mu, Jon Page, Lina Teckentrup, Abhirup Dikshit have proposed a session on land processes. Yohanna Villalorbos and Siyuan Tian were also going to propose a session. Both group will coordinate.

TRENDY
Jürgen talked to us about the TRENDY experiment and the version of CABLE, CABLE-POP_TRENDY, that is used to participate in TRENDY.

CABLE performs very well in this experiment. The results from ilamb can viewed here.

CABLE strengths:

  • N cycling
  • gross land-use transitions
  • demography
  • good performance (for carbon and water)

But poor computational performance was discovered: 82% of the runtime is spent in MPI_Recv. We are not sure why or if it only affects the CABLE-POP_TRENDY code.
Jürgen is working now on running small domains in serial runs and combining them at the end.

Ways forward:

  • Improve computational performance so the focus can be more on improving the science
  • Evaluating POP at large scales under different disturbance regimes
  • N cycling
  • Land use change and management

Evaluation
NCAR looking at building an automated tool to build an LSM and run an evaluation with ILAMB. Gab to talk to them at the end of July, and will keep Claire and Sean in the loop.

Date: 4/7/2023
Attendees: 9

ACCESS-NRI workshop
Headcount: it is hard without a program. The poll is the best way to get that number. Claire will advertise the poll more.

Some ideas of topics of interest for the land WG at the ACCESS-NRI workshop:

  • CMIP7
  • ACCESS-NRI presentations on achievements and current work.
  • A discussion on high-resolution models: what would it mean for atmosphere and land? What are the challenges? What is the boundary in resolution for the land models where it becomes too complex?
  • Interactions between land and ocean along the coast: inundation of land by the sea, tidal inundation, coastal weather forecast.
  • Fires: ACCESS-Fire has no feedback to the land surface. Currently, we have not been looking at the short-term consequences of fire, only looking at the impacts on climate and carbon. This is part of a broader topic on dynamic vegetation.

LWG workshop:
Call for volunteers out. We have 2 volunteers so far, with potentially one or more coming in.

Discussion about the program:
We need to cover the Git transition and how the community is to engage with Git.

Would there be interest in getting someone from JULES to give a presentation? Maybe a presentation by Heather Graham, who worked on designing the land-only configuration, could be of interest but it depends on what we would want to get out of it.

Other discussions
Interest in getting more support for the creation of ancillary files to become more independent. Need more training and documentation on that topic.

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important please feel free to add it below or edit this post (it’s a wiki post).

DATE: 18/07/2023
Attendees: 9

ILAMB:
Romain and Rhaegar (ACCESS-NRI) joined to discuss the plans around providing access to ILAMB at NCI.

Romain: The MED team is currently supporting the following frameworks:

  • ILAMB
  • ESMValTool
  • COSIMA-cookbook approach to MED.

ILAMB observational data is now installed at NCI. The plan is to create a data collection at NCI to make it available to everyone.
ILAMB itself is packaged in a conda environment maintained by the MED team.
They have run ILAMB on CMORised data.
There is some documentation and a tutorial written to help new users get started.
The plan is to release ILAMB in phases, release it to small user groups to get feedback and iron out the original issues first.

Rhaegar currently working on using ILAMB on raw model output (i.e. non-CMIP data). It is working for ACCESS-ESM 1.5 output. It is working on the basic 20 variables ILAMB needs. This is on on-the-fly CMORiser, it is reusing some parts of APP4 CMORiser.
Eventually, the MED team plans to submit the ACCESS-specific code to ILAMB to read in ACCESS output directly.

Gab is talking to the US ILAMB team next week. Happy to discuss anything useful to us.

Rachel: On-the-fly CMORiser is good but storing CMORised test data might be useful as more people might then be able to analyse the data. We need more standardised methods to analyse the data.

Romain: the ILAMB observational dataset is very NH centric. Do we need other datasets for the Australian community? Romain thinks the ILAMB developers team would be interested to get the extra data.

Christoph: ILAMB is mostly developed for climate needs since it is focused on monthly data analysis. He is also looking into METplus to analysis the short-term processes.

Ian: Is there a problem with different variable names between CMOR standard and CABLE and JULES offline output for ILAMB?
Gab: He had ILAMB working before on CABLE output so the CABLE output should be closed to be compatible with ILAMB.
Christoph: There are also differences in units between the couple and offline models outputs. That could be a problem.

Rachel: how to visualise the outputs?
Romain: the output is in HTML format, still looking into a solution to get an easy way to visualise the output. ARE or me.org could be solutions?

Romain: looking for test users for a first release of ILAMB to them. Rachel has a list of ACCESS runs that would be good to run through ILAMB.

Gab: giving a talk about the vision behind benchcab: running tests on the code directly from a repository. Will talk with Claire.

Ian: we need a community effort to revise APP4 to ensure the metadata is correct and the definition of all the CMIP variables is correct.
Romain: CMS is working on a new version of APP4.
But CMS will not know all the details. We need to add this to the CMIP7 effort.

Gab: ILAMB used to output tables with absolute score and with relative score. Is it still the case.
Romain: it is possible to get it all out but it depends on how you visualise the plots.

Other business
Claire, Ian and Anna are away overseas so we are cancelling:

  • the next Land Working Group meeting on 1 August.
  • the next docathon on 25 July.

Here are the minutes from the Land modelling working group meeting. If we’ve missed anything you feel is important please feel free to add it below or edit this post (it’s a wiki post).

DATE: 15/08/2023
Attendees: 7

Forcing data for CABLE and JULES.
It would be good to get a central place to put these. Some were found in a user folder. Some of the forcing datasets we have are not recommended by the community anymore. There is already the wd9 project at NCI to collect some datasets for CABLE, but it proved sometimes difficult to add data to that project in the past. We need clear guidelines on how to add data to it. We want to ensure all relevant metadata is included with the data to provide provenance information.
There is also the question of where the storage would come from. ACCESS-NRI will have storage for reference datasets (about 1.5PB in total) but the funding and the guidelines to access this storage are not finalised.

Working Group compute and storage allocation
The Working Group currently has a compute allocation available of 1.25MSU per quarter. We do not have storage allocated yet because of the same funding delays. However we are hoping to get some storage somewhat soon, we will want to discuss how to use this compute and storage at the workshop.

Date: 19/08/2023
Attendees: 12

Workshop feedback:
IH would like to know:

  • Was the balance between science and discussions good?
  • Was it helpful to be split by model for some discussions?

CR: more effective if we stay together during this workshop as we have our own communities we can discuss with at other times

Problem with encouraging broader engagement during discussions?

GA: require participants to nominate what they would like to discuss and upvote or downvote these ideas beforehand.

RL: one of the challenges was to balance several WG workshops at once.
GA: Could we have the WG workshops staggered? Could we have coupled systems on some days and components on other days?

RL: breakout group sizes for the main workshop were a problem. Having breakouts of breakouts might work better in this case.

IH: the training ran out of time. For hands-on training, provide the background material beforehand and only go through the exercises at the training.

AU: online attendance was challenging because the room kept logging out.

CR: problem with no internet access for Bureau staff because of no phone connections in the rooms.

GA: For some Working group workshops, the online attendance was small. Was that worth the cost of supporting a hybrid meeting?

Short-term usage of WG computational resources
IH: For the development of CABLE4 with the integration of POP-POPLUC and eventually BLAZE in ACCESS, there are various configuration options we would like to experiment with:

  • TRENDY runs but with the default soil model instead of SLI (10 kSU)
  • other modules switching could be tested as well.
  • work on whether we can reduce the memory requirements of POP by reducing the number of cohorts and patches. TRENDY runs again.

Other ideas:

  • test the NRI implementation of ILAMB
  • testing for CABLE3-UM coupling
  • UNSW could run CABLE versions and JULES with the same inputs, waiting on groundwater in CABLE.

RL: status of CABLE offline at ACCESS resolution?
IH: we don’t necessarily have the same ancillary inputs in the offline setup as the coupled. We need to think more carefully about the exact setup we have and how to get it to what we want.

JULES workshop reporting
IH: Useful meeting. A lot going on in the UK community. A lot of potential synergies between various projects.
Complete software framework redesign. For the land, the project is exaJULES with an aim to have a prototype by next year. JaC might be best on hold because the target keeps moving fast right now.

Call for modules leader for JULES science module on vegetation, biogenic and model evaluation. If anyone wants to nominate themselves, contact @inh599 .

IH looked for interest in forming a regional user group for JULES.

SciApps committee will have a webpage for each module that is included in JULES. So we need a CABLE module page.

CR: Should there be a higher authority above the JLMP and SciApps committee? They are now recognising there are other countries involved. There are expectations that the Australian community will contribute to the hydrology. They are interested in the ACCESS-NRI structure and trying to get something like it.

CR: went to the JULES calibration and evaluation breakout group. He discussed we want a common data collection and a reference period to compare JULES and CABLE.