Planning the virtual ESMValTool workshop, 24–26 November 2026

[rbeucher]

ESMValTool virtual workshop · 24–26 November 2026

Three days of short talks, discussions, coding, reviews and documentation work. Below is the proposed programme, with actions to help us finalise it.

Organisers: Romain Beucher (@rbeucher) and Felicity Chun (@flicj191).

Talk offers are identified below. Time slots, session hosts and task owners still need agreement. The linked PRs are starting points for discussion and work; their authors are not automatically expected to attend or present.

Time zones and joining

Most participants are in Europe, so we’ll have morning sessions with Australia and afternoon sessions with Canada.

• Berlin: CET, UTC+1.
• British Columbia: UTC−7, including Vancouver and Victoria.
• Eastern Australia: AEDT, UTC+11, covering Sydney, Canberra and Melbourne.

Dates are shown when a session falls on another local day. Join the sessions that work for you—overnight attendance is not expected.

Talks will be recorded with the speaker’s agreement. The session host handles recording, and the organisers share recordings and slides. Meetings and coding sessions use shared notes: results, questions, blockers and next steps.

Tuesday 24 November — priorities, ocean diagnostics and evaluation pipelines

We’ll start by identifying useful work, then use the afternoon to compare evaluation pipelines and begin coding.

Morning: Europe and Australia

Berlin	British Columbia	Eastern Australia	Session, people and format
09:00–09:10	01:00–01:10	19:00–19:10	Welcome and how the workshop works. Organisers: Romain (@rbeucher) and Felicity (@flicj191). Introduce the programme, shared notes and ways to contribute.
09:10–09:30	01:10–01:30	19:10–19:30	Technical update: CORDEX and preparation for v2.16. Talk offered by Valeriu Predoi (@valeriupredoi). Proposed coverage: cmip7_repack, PyActiveStorage, CORDEX/CORDEX-CMIP6 and configuration changes. 15-minute talk + 5-minute Q&A. Scope to agree.
09:30–09:50	01:30–01:50	19:30–19:50	CMIP7 and REF: what should we work on? Recipe priorities, derived variables and remaining CMIP5 needs. Topic input from Ranjini (@rswamina), Axel (@axel-lauer) and Katja (@katjaweigel); speaker to confirm. 15-minute overview + 5-minute discussion.
09:50–10:00	01:50–02:00	19:50–20:00	Break.
10:00–10:45	02:00–02:45	20:00–20:45	Improving ocean diagnostics. Bring the related ocean work into one session: hydrography, COSIMA, regional metrics, ocean transport, GLORYS, sea ice and Python migration. 15 minutes on progress, 25 minutes on priorities and testing, 5 minutes to capture tasks. Presenter and facilitator needed.
10:45–11:00	02:45–03:00	20:45–21:00	Task overview and choosing what to work on. Review tasks across all topics using a shared project board or task list. Choose a task, find collaborators and agree a first step. Facilitator to confirm.

Ocean work to use as starting points:

• Hydrography — #4610 and COSIMA — #4633.
• Indian Ocean metrics — #4606 and Southern Ocean — #4265.
• Ocean transport — #4549 and GLORYS CMORizer — #4525.
• Antarctic sea-ice seasonality — #4611 and Russell diagnostics migration — #4612.

Other references: CORDEX project board, configuration migration guide, REF ozone — #4604, and derived variables — Core issue #2984.

Afternoon: Europe and North America

Australian times below are Wednesday 25 November.

Berlin	British Columbia	Eastern Australia · Wed 25	Session, people and format
16:00–16:10	08:00–08:10	02:00–02:10	The ECCC / CCCma evaluation pipeline. Talk offered by Kristi Webb (@k-a-webb). 7-minute overview + 3-minute Q&A.
16:10–16:20	08:10–08:20	02:10–02:20	ICONEval. Short overview and questions. Manuel Schlund (@schlunma) was suggested by Birgit; presenter to confirm.
16:20–16:30	08:20–08:30	02:20–02:30	ACCESS evaluation workflows. Proposed 10-minute prepared recording, with questions collected in the notes. Presenter and recording to confirm.
16:30–16:40	08:30–08:40	02:30–02:40	CMEW and reliable recipe runs. Topic suggested by Emma Hogan (@ehogan). Short overview or group review; presenter to confirm.
16:40–16:50	08:40–08:50	02:40–02:50	What can our pipelines share? Compare common needs, testing and possible collaboration. Discussion proposed by Birgit Hassler (@hb326); chair to confirm.
16:50–17:00	08:50–09:00	02:50–03:00	Break.
17:00–18:00	09:00–10:00	03:00–04:00	Coding: data access, pipeline tests and CMIP7 recipes. Choose a small task and work together. Optional final hour for Europe. Task owners needed.

Possible coding tasks:

• Intake integration — Core #2690.
• Catalogue configuration — Core #3234.
• Recipe Test Workflow retries — #4127.
• Optional datasets — Core issue #3278: test recipes with and without optional datasets, including different combinations.

Tuesday’s aim: everyone working on a task has a clear first step, a place to leave updates and someone to ask for help.

Wednesday 25 November — shared coding, AI and recipe tools

Optional coding hour for all three regions

Berlin · Wed 25	British Columbia · Tue 24 → Wed 25	Eastern Australia · Wed 25	Session
07:00–08:00	23:00–00:00	17:00–18:00	Shared coding hour. Five minutes to choose tasks, 50 minutes working together and five minutes to leave updates. Host, helpers and task owners to agree.

This is late in British Columbia and early in Europe, so it is optional. Please say whether you would use this session.

Morning: Europe and Australia

Berlin	British Columbia	Eastern Australia	Session, people and format
09:00–09:15	01:00–01:15	19:00–19:15	NCL to Python: lessons from the migration. Talk offered by Romain (@rbeucher). 10-minute overview + 5-minute Q&A.
09:15–09:30	01:15–01:30	19:15–19:30	Using AI in contributions: what is expected? Bouwe Andela (@bouweandela) agreed to discuss the policy. 10-minute talk + 5-minute Q&A.
09:30–09:40	01:30–01:40	19:30–19:40	Break.
09:40–09:50	01:40–01:50	19:40–19:50	Recipe language server and editor support. Demo offered by Bouwe (@bouweandela). 7-minute demo + 3-minute Q&A.
09:50–10:10	01:50–02:10	19:50–20:10	Finding and building recipes: GUI demo. Presentation offered by Romain (@rbeucher). 15-minute demo + 5-minute Q&A.
10:10–11:00	02:10–03:00	20:10–21:00	Coding and optional Pixi/configuration help. Choose recipe tools, ocean testing or setup help. Bring a failing command or configuration example if you need support. Helpers and task owners to confirm.

References: ESMValTool AI policy — #4631, ESMValCore AI policy — #3259, validation and editor support — Core #3261, and recipe editor support — #4638.

Afternoon: Europe and North America

Australian times below are Thursday 26 November.

Berlin	British Columbia	Eastern Australia · Thu 26	Session, people and format
16:00–16:20	08:00–08:20	02:00–02:20	New recipes: how do we prioritise and review them? Discussion proposed by Liza Malinina (@malininae). Agree practical priorities and review needs. Chair to confirm.
16:20–16:35	08:20–08:35	02:20–02:35	AI contributions and review: questions from North America. Discuss examples and questions from the earlier talk. Additional slot to agree with Bouwe (@bouweandela).
16:35–16:50	08:35–08:50	02:35–02:50	Recipe discovery: filters, tags and useful output. Topic input from Kristi (@k-a-webb), Bettina (@bettina-gier) and Bouwe (@bouweandela). Choose two or three useful improvements and volunteers. Chair to confirm.
16:50–17:00	08:50–09:00	02:50–03:00	Break.
17:00–18:00	09:00–10:00	03:00–04:00	Parallel coding and review. Work on recipes and tools, or continue ocean diagnostics. Optional final hour for Europe. Reviewers and task owners needed.

Possible tasks and examples:

• Review MJO / BSISO — #4613 or cold-point tropopause — #4565.
• Show CSV output as HTML tables — Core issue #3277.
• Link recipes to documentation and example output — language-server issue #25.
• Build on the existing recipe/debug table and recipe tags proposal.
• Continue testing and scientific review of the ocean work listed above.

Wednesday’s aim: make progress on a few manageable tasks and leave enough notes for another person or region to continue.

Thursday 26 November — community paper, tutorials and next steps

The morning starts one hour earlier. The paper and User Engagement sessions run one after another so people can attend both.

Morning: Europe and Australia

Berlin	British Columbia	Eastern Australia	Session, people and format
08:00–09:00	00:00–01:00	18:00–19:00	Community paper working session. Proposed by Birgit (@hb326) and Bettina (@bettina-gier). Review the draft, settle remaining points and agree final edits. Facilitator to confirm; actions captured in shared notes.
09:00–10:00	01:00–02:00	19:00–20:00	User Engagement Team meeting and tutorial planning. Lisa Bock (@LisaBock) requested the meeting; Ranjini (@rswamina) raised tutorial work. Agree priorities and practical tasks. Facilitators to confirm.
10:00–10:15	02:00–02:15	20:00–20:15	Supporting IPCC authors: training and webinars. Update offered by Lisa (@LisaBock), with Felicity (@flicj191) suggested as co-presenter. 10-minute update + 5-minute Q&A. Joint presentation to confirm.
10:15–10:25	02:15–02:25	20:15–20:25	Break.
10:25–10:30	02:25–02:30	20:25–20:30	Results: CMIP7 and pipelines. Five-minute update from a volunteer chosen by the group.
10:30–10:35	02:30–02:35	20:30–20:35	Results: ocean evaluation. Five-minute update from a volunteer chosen by the group.
10:35–10:40	02:35–02:40	20:35–20:40	Results: AI and recipe tools. Five-minute update from a volunteer chosen by the group.
10:40–10:45	02:40–02:45	20:40–20:45	Results: tutorials. Five-minute update from a volunteer chosen by the group.
10:45–11:00	02:45–03:00	20:45–21:00	Wrap-up: what happens next? Review outcomes, remaining work and next steps. Organisers: Romain (@rbeucher) and Felicity (@flicj191); presenters to agree.

Tutorial starting point: Basic tutorials — #4592.

The paper session will capture its own decisions, so there is no separate paper report-back later in the morning.

Afternoon: Europe and North America

Berlin	British Columbia	Eastern Australia · Fri 27	Session
15:00–16:00	07:00–08:00	01:00–02:00	Finalise outcomes and agree next steps. Bring in North American contributions, review outstanding questions and identify who will continue each task. Facilitator and available task representatives to confirm. Shared notes for everyone.

Thursday’s aim: each task has a short result, any remaining blockers and a named next step.

Time to work in your own daytime

These are suggested independent work blocks, in each location’s local time on the date shown. Use them for coding, reviews, tutorials, paper edits or catching up with recordings.

Day	Berlin	British Columbia	Eastern Australia
Tuesday 24 November	11:15–12:30 and 13:30–15:45	10:15–12:00	10:00–12:00 and 13:00–15:00
Wednesday 25 November	11:15–12:30 and 13:30–15:45	10:15–12:00	10:00–12:00 and 13:00–15:00
Thursday 26 November	11:15–12:30 and 13:30–15:00	09:30–12:00	10:00–12:00 and 13:00–15:00

Who does what?

• Organisers: confirm speakers and hosts, provide meeting links and the task board/list, and share recordings and slides.
• Session hosts: open the call, introduce speakers, keep time, ask someone to take notes and record agreed talks.
• Speakers: prepare the short talk or demo, leave time for questions and share slides.
• Task owners: add a small task with an issue/PR link, explain the expected result and setup, and help contributors get started.
• Contributors: choose a task, say what you’re working on, ask for help and leave results or blockers before signing off. Testing, scientific review and documentation are all welcome.

For prepared recordings, the speaker makes the recording and sends its link and slides to the organisers. The suggested deadline is Friday 20 November. Prepared demos and the ACCESS overview still need agreement.

Actions to finalise the programme

Please reply with your availability and what you can take on. These are requests to confirm the arrangements.

• @rbeucher @flicj191 — confirm session hosts, meeting links, shared notes and the project board or task list.
• @valeriupredoi — confirm the technical-update slot and proposed coverage.
• @k-a-webb — confirm the Canadian pipeline talk and time.
• @bouweandela — confirm the AI and language-server slots, including whether the afternoon Q&A works.
• @rbeucher — confirm the NCL-to-Python and GUI demos, and whether prepared recordings would help.
• @rswamina @axel-lauer @katjaweigel — suggest a few concrete CMIP7/recipe tasks and who could introduce them.
• @schlunma @ehogan — say whether you could introduce ICONEval and CMEW respectively, or suggest another presenter.
• @hb326 — help shape the pipeline discussion and suggest a chair.
• @hb326 @bettina-gier — outline the paper session’s goals and any reading needed beforehand.
• @LisaBock @rswamina — outline the User Engagement and tutorial session.
• @LisaBock @flicj191 — confirm the IPCC training update and presenters.
• @malininae — suggest the main questions for the recipe-priorities discussion and whether you could facilitate it.
• @k-a-webb @bettina-gier @bouweandela — suggest two or three priorities for recipe discovery.
• Volunteers: offer to introduce the ocean work, help with Pixi/configuration, host a session or support the shared coding hour.
• Everyone: add a task you would like to work on, with a relevant link and a small, useful outcome.

[rbeucher]

I'd be happy to give a presentation on using AI to convert NCL diagnostics to Python, including lessons learned and how to review the converted code. I'd also be glad to discuss the use of AI more generally in ESMValTool development—where it helps, where it falls short, and how to check its output.

[rbeucher]

I'd also like to propose a discussion about the discoverability of ESMValTool recipes. In meetings and workshops, I find it difficult to answer questions about which recipes are available for a particular use case, because it is hard to get a clear overview of what exists. How could we make recipes easier to find and understand?

I've been working on a user interface to help people explore available recipes and design new ones, and I'd be happy to show its current state as a starting point for the discussion.

[k-a-webb]

I would be very interested in this topic, as while I've found this, https://esmvaltool.dkrz.de/shared/esmvaltool/stable_release/, generally helpful, I do think there is room for improvement.

The recipe overview:

• It would be nice if the search bar was more comprehensive; could filter on variables, datasets, "new" or "updated", etc.
• It would be nice to have an alternative layout that minimizes scrolling, perhaps just a table?

Within the summary of a recipe's output:

• Within an output summary the sections are labelled according to the <diagnostic alias>:<diagnostic script alias> which is only helpful if the recipe uses sufficiently descriptive aliases (an example of where this could be improved).
• Only figures are shown. I would be nice to display tables (I'm not a fan of tables being shown as images) or at least indicate they exist..
• There is excellent documentation of the "description of the recipes" in the ESMValTool documentation, however this is separate from both the recipe, and the nice HTML output summary. Would there be a better way to pair recipes with their documentation?

[bouweandela]

It would be nice to have an alternative layout that minimizes scrolling, perhaps just a table?

Does the debug page cover that? https://esmvaltool.dkrz.de/shared/esmvaltool/stable_release/debug.html

Only figures are shown. I would be nice to display tables (I'm not a fan of tables being shown as images) or at least indicate they exist..

This should be easy, I opened ESMValGroup/ESMValCore#3277

There is excellent documentation of the "description of the recipes" in the ESMValTool documentation, however this is separate from both the recipe, and the nice HTML output summary. Would there be a better way to pair recipes with their documentation?

Opened ESMValGroup/esmvaltool-language-server#25

[bouweandela]

The easiest way to discover recipes for a particular purpose today may be to point an LLM at the esmvaltool documentation and ask. But I agree that it would be nice to improve on this.

[bettina-gier]

I've done some thinking on this before too and agree on the searchability for filtering. My idea was for recipes/diagnostics to have tags for this, but I'm not sure about the technical possibilities to achieve this within the documentation. Quick google search brings me to this: https://github.com/sharm294/sphinx-datatables, with a last update 9 months ago.

[rbeucher]

It looks like I have not been very successful in gathering suggestions so far, so apologies if I am being a bit pushy! 😄

Here are a few more ideas inspired by the minutes from the last ESMValTool Monthly Meeting.

I think CMIP7 and REF should be high on our agenda, and it might be worth dedicating a full session to them. Some possible topics:

• Updating recipes for CMIP7: Could this be a task for one of the coding sessions? @rswamina?
• ESGF downloads: Testing data discovery and download workflows for CMIP7.
• Recipe compatibility: Adapting existing CMIP5 and CMIP6 recipes to work with CMIP7 data.

@bouweandela @valeriupredoi – Is there anything from the @ESMValGroup/technical-lead-development-team that we should include?

@hb326 @alistairsellar – Would it be useful to have someone from the Steering Committee provide an update on the longer-term strategy and direction of ESMValTool?

It would also be great to hear from the @ESMValGroup/userengagementteam, particularly about recent activities and plans for the IPCC author training.

These are just a few ideas to get the discussion going. Please feel free to suggest other topics, priorities, or activities you would like to see included.

I would really appreciate your input so I can start putting together a draft programme.

[rswamina]

If we have any IPCC chapter scientists, perhaps we can get input on how to prioritize recipes for CMIP7/REF/IPCC. I am not sure how exactly to do this. It will involve at the minimum changing variable names to branded names and testing on CMIP7 data?

[axel-lauer]

Would it be possible to include a facet in any given dataset (e.g., CMIP5) like skip_if_not_found which permits the recipe to run if the dataset is not available, but will use it if it is available? This could perhaps avoid the loss of backwards compatibility.

🚀 I like this idea!

[bouweandela]

Would it be possible to include a facet in any given dataset (e.g., CMIP5) like skip_if_not_found which permits the recipe to run if the dataset is not available, but will use it if it is available?

There is a command-line/configuration option for that, but it may not work in various cases, e.g. when the recipe authors have chosen to regrid all datasets to the grid of the one that is missing. Example usage: esmvaltool run --skip_nonexistent=True recipe_example.yml, from the docs

[bettina-gier]

I think the point is to have it as a dataset facet, so you can add CMIP3/CMIP5 data that's more rare, while still keeping some vital datasets to the analysis as mandatory

[bouweandela]

Opened ESMValGroup/ESMValCore#3278

[rbeucher]

@bouweandela Will you be happy to discuss the work on the AI policy?

[bouweandela]

Sure!

[rbeucher]

DEADLINE FOR PROPOSALS: Friday 6th of November 23:59UTC!

Thank you!

[axel-lauer]

One more item for the CMIP7 topic could be on updating the scripts for derived variables: ESMValGroup/ESMValCore#2984 (maybe for a coding session).

[k-a-webb]

I would appreciate this discussion, especially a coding session

[rswamina]

The User Engagement Team usually have a one hour meeting and some time to work on tutorials. @LisaBock - what do you think?

[LisaBock]

Thanks, @rswamina ! Sorry for not bringing it up earlier...
Yes, we would need one hour (separate) for the UET meeting. Ideally with an Europe–Australia overlap. If these time slots are already needed for other topics, I am thinking about organising a UET meeting shortly (the days) before the ESMValTool workshop.

While we have coding time, we could definitely work on the tutorial (porting it to the rst-format).

[malininae]

With the giant efforts from literally everyone to submit papers pre March 2027, I was wondering if some sort of triage system/policy should be created for the recipes influx? And is it worth to discuss it all together.

[LisaBock]

If it is of general interest, then I can briefly report on our planned support/webinars for IPCC authors and chapter scientists, together with @flicj191.

[hb326]

It would be great to have a discussion about the different available pipelines we have developed now to run ESMValTool in the production of model simulations. We have developed ICONEval here at DLR, and I heard that there are potentially similar frameworks available from Australia and Canada. Maybe @k-a-webb could present something for Canada, and @schlunma something for ICONEval?

[k-a-webb]

I'm happy to briefly present how we (ECCC, CCCma) have developed a wrapper around ESMValTool to support model development. It's similar to REF in being a framework to pair datasets with relevant recipes/diagnostics.

[ehogan]

Would CMEW fit into this? 🤔

[bettina-gier]

Should we rediscuss the paper that was started since the last workshop? Maybe have a last push to finalize a draft?

[hb326]

Absolutely! The plan is to have a way more mature draft at the time of the workshop so that we hopefully only have to add the final touches to it.

[valeriupredoi]

pls pen me down to give an update on technical matters, including eg https://github.com/NCAS-CMS/cmip7_repack and https://github.com/NCAS-CMS/PyActiveStorage Danke muchly 🍻

[bouweandela]

It would be good to repeat here that support for the config-developer.yml file will be removed in the upcoming v2.16 release and answer questions on how to upgrade to the new configuration file format.

[bouweandela]

And I believe you wanted a session where everyone can bring their pixi problems and we help them with those?

[bouweandela]

It would be nice to mention the much improved support we have for CORDEX(-CMIP6) data too: Cordex support (ESO4clima WP2)

[malininae]

Not a topic, but just an FYI: BC is not changing time anymore, which screws up a lot of timezone stuff.

[bouweandela]

I can present the brand new esmvaltool language server that I've been developing, it gives you a lot of information in your editor when writing recipes: https://github.com/ESMValGroup/esmvaltool-language-server

[rbeucher]

I can present a new GUI I have been developing to navigate and build recipes.

[rbeucher]

We will release a draft agenda tomorrow.

[rbeucher]

Thanks everyone for your suggestions! We’ve worked through them and put together a draft programme, which is now in the updated description of this discussion.

Please have a look and share any comments. Tagging those mentioned in the programme: @valeriupredoi @k-a-webb @bouweandela @rswamina @axel-lauer @katjaweigel @schlunma @ehogan @hb326 @bettina-gier @LisaBock @malininae @flicj191.

Could you please confirm your availability for the proposed sessions and whether you’re happy to contribute as outlined? The roles are still provisional. Your replies will help us finalise the programme and share it more broadly.

Organising a virtual workshop across three widely separated time zones is challenging. We’ve tried to balance everyone’s needs, but there’s unfortunately limited flexibility to move sessions around. Thanks for your understanding!