Land working group, meeting minutes 2024

Land Surface Working Group
, ,
1

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: 20/02/2024
Attendees: 10

CABLE-JULES comparison
There is definitive interest in running some comparison of the 2 land models. The following runs could be used in this comparison:

  • ACCESS-AM3 runs with CABLE and JULES. These will be done as part of the development work for CMIP7. Currently, only the biophysics is coupled in AM3. There will be runs done for the March hackathon, but keep in mind the science in these early runs is flaky in the CABLE version. This should be seen as an exercise to identify where improvement is needed.

  • AB is going to run CABLE over Australia with AGCD and ERA5 at 0.25° with the NESP2pt5 branch. This includes POP. AU says it would be good to include in the comparison as CABLE-POP and standard CABLE have shown significant differences before.

  • CR’s team has run JULES with AGCD and ERA5 for 20 years at 5km. Focus on recent years.

  • RECCAP runs? Would these be useful? Maybe but CR points out run focusing on carbon don’t usually output runoff.

CR would like to have runoff outputs for comparison of hydrology. Could plug these through their river routing model.

AU: will the groundwater be available in these comparison runs.
RK: solved 3 bugs, testing to see if any other bugs left and to see if the fixed bugs have changed the behaviour during spin-up. All fluxnet sites run with debug option on with the bugs fixed. Then, we need to review changes to the code outside the groundwater to ensure agreement these are acceptable bug fixes.
CC: Timeline maybe in a month.

GA: will we leverage ilamb for the comparison? Does ilamb have the datasets we need?
A few Australia-specific high-resolution datasets exist that we may want to add to ilamb. Use the working group to create an agreed list of datasets we can then send to MED team.

CC: NRI land team can run ilamb on the runs outputs and make the results available to the community as we need to get more comfortable with ilamb.

Experiments with Working Group resources
CC: Need better reporting of the resource usage. Need to provide a list of proposed experiments to the SAC by mid-March for Q2 of 2024.

AU: can the experiments for the CABLE/JULES comparison use these resources?
CC: yes. It helps with planning the development of CABLE and JULES. And the scheme is not oversubscribed anyway.

Action: CC to provide a way to propose experiments.

Agenda items
CC: it would be good to have input from everyone for the agenda. What is the best way to collate ideas?
Possible solutions:

  • send email one week before to ask for items. Downside: the email will likely be sent to a lot more people than those interested in the working group.
  • could the Announce post be a wiki and people could populate the agenda directly? Possible but if the Announce topic is open, there is a risk people reply via emails and spam everyone.
  • create a forum topic with a list of potential subjects of discussion. People populate the list when they want. Co-chairs pick items from the list and other urgent items to create the agenda for the meeting. A link to the list and a call out for ideas is sent with all meeting announcements. chosen solution

Strategy paper
CR: meeting in a couple of weeks for interested parties.
RL: could we use the WG meeting in 2 weeks for that purpose?
IH: should we instead keep that meeting to discuss the land-focused evaluation we would want out of the Evaluation Hackathon?
RL: there will be time at the Hackathon itself for these discussions.
Decision: next WG meeting will be dedicated to the preparation of the strategy paper on “challenges and future directions of land surface models for hydrology - Australian community perspective”.

2

DATE: 19/03/2024
Attendees: 12

Mengyuan Mu’s presentation: How do land surface models behave as vapour pressure deficit increases?

Hackathon:

  • ESMValTool could be very useful for science evaluation on the longer term but a very very steep learning curve.
  • need timeseries of grid cell at FLUXNET sites from the coupled run.
  • action items from the Hackathon: Issues · ACCESS-NRI/CMIP7-Hackathon · GitHub
  • Lauren Stevens has code from previous efforts here: umplot in branches/Users/lxs599 – cable
  • worth keeping an eye on UKMO’s porting auto-assess to ESMValTool.
  • pre-processors look potentially useful.
3

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: 02/04/2024
Attendees: 12

We discussed what global met forcing datasets are being used across the community to provide information to ILMF (mainly GSWP3, WATCH-WFDEI, ERA5, CRU-JRA). New/updated datasets are being developed that start around 1950, we discussed whether these are suited to carbon cycle studies needing spin-up from pre-industrial conditions.

Gab: Information we want to see for model evaluation and how we want this information to be presented. Showing a mock-up of how the analysis could be presented. With several levels of granularity: PLUMBER plots per tested configuration, window inserts for summative plots and for plots per variables (with a drop-down to choose the variable), plots per site (with timeseries, diurnal cycle etc.) and colour-coded information to quickly show at each site if the test branch is better.

Negin: Should we add urban sites?

  • should we add them from the start? Is the data ready? Globally there are sites that are ready to pull from. Mat can give some sites. We will discuss in a future meeting if different metrics need to be applied to urban models.

We are also not representing agriculture for now in benchcab.
Permanent ice is not represented. Lakes neither. We have sites at wetlands and some points that are frozen half of the year.

Anna: summary tables like in ILAMB. Gab, these already exist in ME.org. Claire asked if these tables can be visually similar to ILAMB’s so it’s faster to interpret when switching between the 2 systems.

Beta testing: The system is a bit too clunky right now for generic use. Once we have something nicer to use, we are happy for people in the community to get involved with testing as they wish.

Ramzi: It might be good to have one overall grading “number” across sites, metrics and configurations.
Gab would like a system that gives the user a statement of what has significantly changed but that would be hard to do.

4

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: 16/04/2024
Attendees: 11

Evaluation follow-up
Bringing Ian and Rachel up to speed. The mix of high-level and detailed analyses is welcome. We need metrics that give a broad view but we also want evaluation that gives some information on processes.

Urban
Jiachen wants to run CABLE with urban coupled to the atmosphere at high resolution.
Running coupled to atmosphere is tricky
Might be best to start working in standalone. Need to introduce urban sites in benchcab for testing. It would be useful to have an urban flag to only run on urban sites for some developments. Might be useful to have other flags to filter sites per type (ice, agriculture…)
High-resolution configurations: 5km configuration from Anna.

Available datasets: BARRA2?

Momentum partnership workshop
Claire to get more aware of what is happening with NGMS - JULES as it has implications for the coupling of CABLE to LFRIC in the future.

Recruitment update at ACCESS-NRI
New training team lead: Paige Martin
Acting ocean team lead: Dougie Squire

ACCESS-NRI workshop program committee representative
ACCESS-NRI is looking for representatives of the working group to be part of the ACCESS-NRI workshop program committee.
Tammas has been proposed to represent the land community on the committee. But if anyone is interested in being on the committee, please let @inh599 @aukkola and @clairecarouge know.

MPI support for CABLE

Do we still want to have an MPI implementation? Or do we want simply the capacity of running a bunch of serial runs and combining afterwards? The TRENDY branch has given up on the MPI implementation so we need to decide what we do for the work around BIOS and CABLE4.

  • What about running on GPU?
  • What about lateral flow? It might come soon. Everything we do should take this into account.
  • Whatever we do should be machine-independent
  • Do we need a full MPI implementation or can we do it with a simple MPI setup and parallel I/O for now? Parallel I/O will not be sufficient when lateral flow is introduced.
  • NGMS is looking at coupling JULES to LFRic via a coupler. In this case, CABLE would also use a coupler and would need an MPI implementation of some sort.
5

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: 07/05/2024
Attendees: 13

Working Group Annual Workshop
Value in having CABLE and JULES people in the same room at the same time. There is too much happening in this fortnight on the JULES side, we need a different day for the WG Annual Workshop.

Momentum workshop: the Bureau would give talks on flood modelling and current JULES development. Gab could give a talk on model evaluation with PLUMBER and current plans.

Supported configurations
Whatever we use for benchmarking should be in there.
Having some simplified test cases might be good.
UKMO has a codified way of choosing their configuration but that takes time and is always behind the development: think about the different types of users:

  • Configuration that is stable and well-tested for new users
  • Someone who will end up developing their own configuration and just need a starting point.

AMIP configurations: these will come with the release of ACCESS-AM3. We will have to decide what configurations we want for CABLE.

Other business
UKMO and ECMWF looking at organising an ancillary workshop next year. Date isn’t clear.

6

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: 21/05/2024
Attendees: 8

Working Group Workshop

Dates: between Tuesday 8/10 - Thursday 10/10.
Proposing 2 days: 1 day for the workshop and an optional “working bees” day to advance WG projects.
Workshop hosted at CSIRO Black Mountain.

Training day at ACCESS-NRI annual workshop

  • Evaluation tools use:
    – ilamb
    – ESMValTool
  • Using GitHub for code development.
  • benchcab (could be done at another time since it is a small cohort. It is ideal as a YouTube-based training).
  • drop-in session for people to come up with their current problems. This might require some preparation with people sending their issues in advance.

Other topics

Christoph leaving for ECMWF. Putting together a strategy for land modelling development for the Bureau.

Ancillaries workshop next year from ECMWF and MetOffice.

Update on ACCESS-NRI work plan: likely coming at the next meeting or the one after.

7

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: 4/06/2024
Attendees: 15

Urban modelling in CABLE
Presentation by Jiachen Lu on his work to implement urban representation in CABLE.
It’s possible to represent urban type without an urban model and simply using the right properties to represent the surface. But there is a problem of finding the right data for doing this.

Discussion on the Bureau’s work on urban representation in NWP

Experiment proposal: TRENDY
The cost of the experiment is relatively low (~100 kSU).
It was requested to perform an additional run with a different branch of CABLE to increase the relevance of the experiment to the community.

No objection was voiced so the experiment was agreed to go ahead.

Workshop date
Considering there was no strong objection due to the school holidays and the chosen dates are the best for known availability of people, we choose to keep the dates. The Land Working Group workshop will happen on 9-10 October 2024.

ILAMB datasets
We need more information on what datasets are acceptable by ILAMB:

  • spatial and temporal resolution
  • point vs grid datasets
  • spatial extent

ACS was using ILAMB, it might be worth checking with them if they have datasets for Australia already curated for ILAMB.

Claire will gather the required information for the next Land Working Group meeting.

8

Date: 18/06/2024
Attendees: 10

ILAMB datasets

ILAMB datasets are resampled at 0.5° which seems a big limitation when looking at regional and high-resolution simulations.

Ideas for additional datasets for ILAMB:

  • Urban flux tower data
  • Reanalyses: BARRA - But is that useful if it does not assimilate land properties?
  • CSIRO ET product on TERN, and other TERN datasets
  • COSMOZ in-situ observations
  • SMAP L4 gridded soil moisture
  • Land surface temperature: new datasets from satellites. Himawari. But these are still very new products and it’s not certain how useful they are yet.

ILAMB is regridding the datasets at 0.5 degree resolution. This is too low resolution for high-resolution simulations. How can we do evaluations for high-resolution simulations? We will need something else for higher time and spatial evaluation.

ESMValTool should be what we use if we want to develop other evaluation.
The issue with ESMValtool is the learning curve.

We should have a test of adding a couple of Australian datasets to ILAMB: CMRSET, AusEFlux. @clairecarouge to do that with the MED team.

9

Date: 2/07/24
Attendees: 8

Feedback on ACCESSS-NRI draft work plan:

  • will new hydrology be included in ESM3? It will be technically but testing will depend on community involvement.
  • beta release for CABLE4 is too ambitious. Alpha release maximum.
  • How plans from W21C in the land space will modify this work plan?
  • How much work from ACCESS-NRI will be needed concerning other developments happening in CABLE: Matthias work, urban development?
  • need to fix the analytic spinup for CASA-CNP for ESM3: does not seem to speed up the spinup when the phosphorus is included.
  • Is there enough time attributing for replying to help requests from users?
  • JULES ancillaries for regional ACCESS: how much involvement from the land team?

Will there be a list of new features for new releases? Yes. The list might not be complete for work that happened in historical branches.

Update on CABLE as ACCESS development.

10

Date: 16/07/24
Attendees: 8

Draft program out for the Bureau’s convection scheme workshop. Some relevant topics to LSM research. http://www.bom.gov.au/research/workshop/2024/program.shtml

Update on FLUXNET sites in benchcab for crop and urban. Nothing new in benchcab but Jiachen ran 17 urban sites with offline CABLE and modified site characteristics for urban.

How to reference the CABLE version used in publication when it isn’t a released version?
Could give the commit hash or add a tag. Need to check what other communities do. Would need to come up with a tag naming scheme that makes obvious the reason for the tag. Could also use annotation in the tags to reference the DOI of the published paper.

Hard-wired vegetation types:
Issue is now opened to catalogue all occurrences. Occurences can refer to a specific vegetation type or assume an ordering of the vegetation types (ice always last, woody types always before grassy etc.). Problem with some inputs having hard-wired values (iveg in FLUXNET input files).

How to deal with the question of consistency between ice tile and ice soil in CABLE? This is being worked on for the offline code but the coupled case works differently.

Is it possible to have tiled soil inputs for our coupled model? It is possibly quite a bit of work and would need to decide if it’s worth the effort.

11

Date: 05/08/24
Attendees: 10

Conferences reminders
ACCESS-NRI workshop: registrations open until 21st August

CABLE configuration work
Claire: Lachlan has worked on a new organisation of configurations. We would like feedback from the community on the design. But it’s tricky as a presentation will not give the feel for using the new design.
Urban initiative
Negin Nazarian has put together a workshop for an Australian Urban Climate Initiative. This happened last week.
The workshop focused on putting together a proposal for NCRIS funding around solving the issues in urban modelling, looking at models, data and impacts.
Ian: will it focus on biophysics only or include greenhouse gases, carbon etc?
Mat: for the moment the focus is on infrastructure. Later discussions will look into this aspect.
Ian: CSIRO does observations and modelling of methane in city. Air mass tracking with Cape Grim observations for regional information on GHG.

12

It was difficult for me to take notes and participate in the discussion. Any update on the notes is welcome. This is a wiki post so please edit.

Date: 20/08/24
Attendees: 12

Put the responsibility on the developer to know what else is going on. Don’t put the responsibility of checking compatibility between developments on the NRI land team. Request for people to update their branch to the latest main state before any testing and review request. If the review takes a long time, it might be necessary to update again before merging.

Developers to provide information on what you are doing and what part of the code it is going to impact → integrate that in PR template.

Original author of PR to merge the code:

  • Have 2 approvals: science and technical reviews. CABLE developers will have to volunteer to spend some time on this science review if we want a scientifically correct code. benchcab could make this review simpler and lighter when it develops enough to provide a comprehensive view of the correctness of the code.

Gab: discussion with Belinda and Jürgen on TERN-NRI project. Running CABLE, LPJ-guess and JULES. Have technical help and might be interested in making benchcab more flexible with other models.

Call for science configurations for benchcab. The current set is not representative of the science in CABLE. We do not want to have too many simulations required for standard but we want a better set than what we are using now.

13

Date: 19/11/24
Attendees: 10

Reporting publications using ACCESS models
Use the form on the ACCESS-NRI website, under “For Users” tab.
Under “For Users”, there is also information on how to acknowledge ACCESS-NRI in publications. Please acknowledge ACCESS-NRI and report publication for CABLE offline use if you received help from the ACCESS-NRI land team (or other team) in your project.

There is also a new logo and a new ACCESS model graphic, please update your communication material. Contact access.nri@anu.edu.au if you need the updated material.

Co-chairs renewal
WG have to put a proposal to the SAC who gives approval.
What is the role of co-chairs: post the role statement on the forum.
People to express interest in the role by emailing the current co-chairs. Co-chairs to give a proposal to the WG before sending to the SAC.

Root distribution
High sensitivity of GPP on the root distribution. Using Jackson et al values give low seasonal variability. ESM1.5 is already producing low variability and we don’t want to reduce it more by changing the root distribution.

AU: Look at the effect of FWSOIL or GW because the soil might experience too much water stress.

IH: Look at adapting the phenology? Prescribed in ESM1.5.
Allocation of NPP into leaf area?

Model Evaluation Feedback
Discussion on Enhancing Model Evaluation Feedback Mechanisms for the ACCESS Community

CSIRO is using p73 at NCI to store and share data. Could use the ACCESS-data category to advertise the datasets to the community.

This would be a great way to archive results found by the community. But it won’t change the development of the model. Need some additional step to connect to the development community. User expectations: need to make clear that it’s not because a problem is found that it will be fixed.

It would be good if specific issues identified are reported in the model repositories in GitHub.

Evaluation that shows good results is as important as evaluation that shows bad results.

Good to document the current strengths and weaknesses of ACCESS. Could be useful for universities to find new research projects.

Workshop outcome follow-up
Mat to give a presentation at the next WG meeting about the work he’s done for using ANTS for CABLE ancillary preparation.
ANTS seem to create some odd ancillaries with very sharp boundaries. There are some science choices in there we will want to review and maybe modify.

Abhaas is working on putting in place a pipeline to prepare met forcings for CABLE. Focussing on ANUCLIMATE to start with.

14

Meeting Summary

Date: December 3, 2024
Attendees: 6

Agenda:

  1. Land Working Group Schedule: The group discussed the meeting schedule for 2025, deciding to keep the current frequency and timing.

  2. Presentation by Matt:

  • Matt presented his work on creating an ancillary creation workflow for the CABLE model, following up from a previous workshop.
  • The goal is to document and share the workflow, making it easier for researchers to generate necessary ancillaries.
  1. Discussion Points:
  • Ancillary Creation Workflow:
    • Matt explained the process of converting the Jules workflow into a bash script for easier manipulation and understanding.
    • The workflow involves creating land cover masks and fractions, with the potential to adapt it for CABLE.
    • Discussion on the use of different sources for land cover data and the need for a crosswalking table to convert CCI classes into PFT fractions.
  • Soil Properties:
    • The group discussed the differences between the Jules and CABLE approaches to soil properties.
    • Jules uses a dominant soil type approach, while CABLE uses spatially varying fields.
    • Concerns were raised about the impact of these differences on model outputs, particularly soil moisture and latent heat flux.
  1. Next Steps:
  • Matt will work with Lachlan to create a variable table for CABLE gridinfo file.
  • The group will continue to refine the workflow, considering the best practices and potential improvements.
  • Further discussions and inputs from other members will be sought to finalize the workflow.
  1. Additional Topics:
  • The potential use of Australian-specific datasets for land cover and soil properties was mentioned.
  • The need for evaluation tools for dynamic vegetation in CABLE was briefly discussed.

Action Items:

  • Matt & Lachlan: Collaborate on the initial mapping table for CABLE.
  • Group: Provide further inputs and feedback on the workflow.