This topic will record our notes for the day. Feel free to include any of your questions here as well.
Current Status + Work Plan
Chair: Claire
Previous year workshop
Four streams:
- Ancils
- Met Forcing
- JULES vs CABLE comparison
- BIOS and code merges
Goal 1: Ancillary workflows
Create a workflow that aligns JULES and CABLE ancillary generation processes that can be used for adaptable domains and resolutions
Mat: Lachlan and I have a technically working workflow for creating CABLE ancillaries. I will go over this in more detail at 3:30pm. We are able to generate ancillaries for CABLE at any resolution (regional and global) using a simple text input defining the grid. It is technically working, but need science input from the community, e.g. input source data, conversion algorathms etc. See more at 3:30.
Gab: Will code diverge from BoM/Met Office?
Mat: It is based on what the BoM use for JULES, but it is forked, so currently static in time and not updated. Ideally we would be able to get it back into the Met Office trunk so it is aligned going forward
Ian: This may be difficult, e.g. Siyuan is having trouble getting the BoM work into the Met Office trunk, we need to jump through their gates, which could be a problem.
Mat: There is also the divergence of this new workflow with what is done for AM3 by NRI (Martin Dix) which uses e.g. NCAR 0.5°.
Ian: An issue with AM3 ancil suite is that we have lost provenance, we can’t recreate the sources yet. So a new workflow that is able to recreate
Paul: Where is the documentation, for example for ESM1.6?
Ian: We need to work on multiple documentation papers.
Ancillary goals for next 12 months
- Get Land WG agreement on workflow science questions, e.g. input datasources, conversion transfer functions etc.
- Align the workflows that AM3 (NCAR 0.5° - Martin Dix) and the “new” CABLE workflow (CCI @300m ).
1 Like
Goal 2: Met Forcing
Abhaas Goyal has been working on the met forcing workflow. Has a working version based on ANU Climate variables, so it does unit conversions, variable name conversions etc. Next will focus on ERA5 met forcing.
Ian: Will it be daily or hourly?
Claire: Initially based on the Weather Generator, so e.g. daily to hourly. Need to decide which Weather Generator to use because the one in CABLE could be improved.
Datasets that are daily and need Weather Generator: AGCD, ANU climate, some BARRA.
Initial Met Forcing goals for next 12 months
- Discussing methods functions/methods for weather generators
- Released version of the software that works with more than one input dataset
- Simplifying the offline driver
- Publish datasets of hourly/ 1/2 hourly forcing using the weather generator
Goal 3: JULES vs CABLE experiments
Alison: Do like-for-like comparisons of JULES and CABLE at 5km over Australia for, say 30 years.
JULES: Siyuan is not here, but we believe is still interested in running JULES. Could be supported by Mat
CABLE: Anna? Supported by a PhD student?
Gab: Should be compared within BenchCab
Gab: Ulrike is standing up a team at BoM that will be running JULES for
Comparison Goals next 12 months
- TBC
Goal 4: BIOS updates
Claire: It is done! but not yet merged, can’t close the branch yet because of BLAZE fire development needs finalising and merging.
Ian: Timeline for BLAZE 6-12 months.
Claire: Groundwater merging is close to merging.
Claire: Matthias Cunz has a postdoc working on further developments.
Gab: Timeline for ESM3?
Ian: To be discussed on Thursday at lunchtime
Seminar Sessions 15 mins each:
Ian Harman:
Early steps towards a fire module for CABLE.
Gab Abramowitz:
Recent analyses of PLUMBER2 output, new metrics for benchcab.
Jiachen Lu:
Urban model implementation in ACCESS-AM3: early steps.
Abhaas Goyal:
Pre-processor for meteorological forcing variables for Land Surface Models.
Claire Carouge:
Reinventing CABLE’s infrastructure for long-term sustainability.
Land ancillary training
Presenting: Mat Lipson, Alex Norton, Lachlan Whyborn
LandAncillaryTraining.pptx (12.0 MB)
Use cases to create/update ancillaries:
- Fix something
- Ancillary perturbations
- Sensitivity to the source of ancillary
- Change of resolution: relation between sets of ancillaries between resolutions.
Bad practices with ancillaries:
- Proprietary format
- Fixed file name in the model code
- Lack of provenance
- Units: problems with identifying the unit
- Testing → UKMO tools have unit tests and KGOs.
- Code to create an ensemble of ancillaries, based on single source of data
- Using non-self-describing formats → importance of adding metadata to ancillary files to describe source and processing
Other discussion
A cookbook for ancillary creation scripts is good. But sharing information on existing ancillaries and sharing ancillaries is important as well.
NGMS program should solve some of the bad practices (proprietary format, lack of provenance)
CAP might not be transferred to be used with LFRic. MULE, likely not (unsure).
Question around contributors which are not on Gadi.-> Library of ancillary datasets comes into play here but it’s not enough. Provide access to Gadi to external CABLE users?
Links to training material:
CABLE restart mapping scripts: Land-Workshop-2025/Restart-remapping at main · ACCESS-Community-Hub/Land-Workshop-2025 · GitHub
Lachlan’s “how to use mule” tutorial: Land-Workshop-2025/How-to-use-mule/How-to-use-mule.md at main · ACCESS-Community-Hub/Land-Workshop-2025 · GitHub
The Land Ancillary Suite (work in progress): Land-Workshop-2025/Ancillary-suite at main · ACCESS-Community-Hub/Land-Workshop-2025 · GitHub
Some example ancillary modification scripts: Land-Workshop-2025/cookbooks at main · ACCESS-Community-Hub/Land-Workshop-2025 · GitHub
ACCESS Land Working Group Workshop Day
Friday Morning Session
Topic 1: Integration with PLUMBER2 Benchmarking Suite
(led by Gab Abramowitz)
Gab has been working to streamline modelevaluation.org and Benchcab integration. In summary:
- Ability to drill down between multiple site / single site, and multiple variable / single variable analysis.
- New Independent Normalised Metric Value approach: only benchmarks are normalized between 0–1 (models can be outside this range if better/worse than all benchmarks).
- Automatic integration with OneFlux for flux QC processing/gap filling.
- All code written by Gab in R, but contributors welcome.
Still needed for contributors:
- Documentation
- Minimal viable plotting template to build from
Plans (with afternoon discussion in bold):
- New analysis engines / modification of existing
- Automated feedback to CABLE developers
- Shared CABLE/JULES experiment and testing strategy
- Configurations for automated testing
Topic 2: Calibration / Evaluation / Data Assimilation Using Earth Observation
(led by Alex Norton)
Around the room: What observations are people using with CABLE?
- Everything by default in iLAMB
- IGBP
- Flux tower data
Further discussion – how can we use observations better?
- Many land products (e.g., iLAMB) are monthly → problematic for analysis.
- Bureau has focused on diurnal performance.
- Soil hydraulic properties not well observed.
- Renato’s presentation showed many observationally derived products that can give insight into different processes.
- Challenge: satellite-derived “observations” often rely on models (e.g., Latent Heat uses the same algorithm as LSM).
- Vegetation height = low hanging fruit.
- Siyuan integrated GRACE for JULES and subdaily LST.
- Simulate SCIF then use as a constraint.
Plans (with afternoon discussion in bold):
- Review observational datasets for calibration/evaluation (beyond flux tower). Gather feedback.
- Data fusion studies (e.g., parameter optimisation).
- Non-categorical parameters/inputs.
- Identify additional outputs comparable with observations.
- Longer-term: generalisable parameter optimisation package (document existing work, limitations).
12-month plan: points 3, 4, 5, 6
Topic 3: Future Science Goals and Technical Challenges
(led by Ian Harman)
- Ian showed a slide with various CABLE developments (some complete) not yet included in trunk.
- Goal: rank projects by technical/scientific maturity, then prioritize.
- Need cross-cell communication for BLAZE, Lateral Flow.
- Cross-cell communication won’t be a problem (per Mikhail).
- Sean: load balancing problem when parallelising on cells → might consider parallelising across tiles.
Plans (with afternoon discussion in bold):
- Create dependency graph of future science and technical challenges. Rank technical/scientific maturity.
- Identify technological impediments holding back science (e.g., categorical inputs, cross-tile communication).
Work Plan Revision
(led by Claire)
Remit of the Land WG:
- Collaborate on WG projects (systems development and model evaluation).
- Expand regular membership (e.g., through broad interest and invited seminars).
- Encourage interaction with other WGs.
Goals 1 and 2 may not align fully.
Proposed agenda:
-
Every second meeting:
- Science seminar (internal or invited)
- WG project update (remind, encourage involvement)
-
Longer interval (≈3 months):
- Joint meeting with ESM / Atmosphere / Cryosphere WGs
Current projects:
- MetForcing
- Shared ancils
- LSM comparison
- BIOS/src
Discussion: Do we want Land cookbooks (like Cosima)?
- Claire: Cosima started with model outputs, ancils, datasets. Script repository may become useful after that.
Friday Afternoon Session
PLUMBER2 / Benchcab
Q: What configurations are we going to test automatically?
- Please review and comment at Gab’s notes here: Google Doc link
Future Science and Technical Challenges
- Dependency graph on whiteboard
Data Assimilation
- Primary notes by Alex
WG Comparison Experiment
- Circulate previous notes
- Meeting with Siyuan needed to finalise metforcing/evaluation using:
- Flux towers / benchcab
- GRACE?
- Runoff?
- iLAMB?
- Oz e FLUX?
- ET?
Final Wrap-Up
Impediments for New Users to CABLE
-
Lack of documentation:
- Standard setups
- Build instructions
- Run instructions
- Parameter sequencing
-
From NRI’s perspective: documentation is a community responsibility.
-
No NRI-supported offline CABLE version → no official documentation.