Meeting notes archive 2021

23 Dec 2021 climate data weekly meeting 33 2021

Paola, Claire

oi10 quota
- Currently 545 of 550TB quota used, not much remaining
- What to do about it? Community fund additional storage? Petition for more RDSI-type?
- MAS harvest probably still weekly, check if new data appears after Christmas?
- Grafana alerts for quota usage
- What is update frequency of nci_account/files report? Seems slow to pick up file deletions etc. lquota is correct but files report seems very slow to update.
ACS reference data
- repo naming
  - Having “NCI” in the name may cause contention with NCI?
  - Can we change the name of the repo without breaking things?
  - Would it be more community inclusive if it was called ACS reference data?
  - Indicate NCI as location – e.g. change repo name to “ACS-ref-data-NCI”, leaves open possibility of “ACS-ref-data-S3” etc.
- Paola has pushed 3 repos
  - Discussion with Damien about how to store update logs, etc.
  - How to be open and manage user expectations
- Put reusable code in a separate repo
Intake-thredds
- Adding to hh5 conda unstable
- Open a remote thredds catalogue as an intake catalogue
- More limited than Siphon but could be useful
DAP API
- How to manage passing authentication information?
- Maybe talk to Katie in the new year about more documentation needed – researchdatasupport@csiro.au
- Surprisingly few hits returned for ‘climate’ keyword (~25)
- Use of FOR codes – many datasets still using the old codes?
Log4j2
- NCI (Syazwan) confirm THREDDS and Geonetwork have been patched.
Merry Chrsitmas! We’ll restart these meetings late January.

//16/12/2021 Climate data weekly 32 2021

Claire, Tim, Paola. Apologies: Chloe

Alicia now with data 61. Hopefully she’ll still be able to join these meetings in the future
Log4j2 CVE
- THREDDS affected – CCiA patched, Sharon T has a list of known services, IM&T should see that and maintain TDS as a service.
- Not hard to patch, check any java based things.
- InvenioRDM is okay as it’s a full python stack.
- NCI status unknown.
- Nectar have notified users.
Paola met up with Damien re ACS data. Put together a survey of what datasets people use.
- Will need to follow up to work out which versions and products.
- VM on NCI cloud to manage downloads,
- Might use Jenkins to manage scheduling of downloads – Scott manages accessdev.nci.org.au/Jenkins accessible to anyone with NCI credentials. Let’s people check when last update occurred. Separate to PBS queue ie running in Jenkins worker setup locally, or submitting copyq jobs?
- Introduced Damien to CREATE-IP
- Could link downloads managed in Jenkins to catalogue entries in Invenio?
- Decided on one repo per dataset, standards around readmes and doco TBD
InvenioRDM
- Test instance updated to latest version, a few bug fixes. URLs not producing links still an outstanding problem, Paola will report.
- Paola has been working on ingesting example from Zenodo
- Tim to provide Paola with info about harvesting from CSIRO DAP using the API
- DMPonline and GeoNetwork tested, Paola has written a script to harvest geonetwork to an intermediate standard, and from there to Invenio.
- Do a curated crawl – filter geonetwork based on categories. For DAP, use keywords, FOR codes, other options? It’ll be hard to ensure we don’t go too wide or too narrow.
- Allows bulk ingest but it’ll be difficult to automate

02 Dec 2021 climate data weekly meeting 31 2021

Paola, Alicia, Chloe, Tim E, Claire

Big data working group
- Alicia wasn’t able to review Claire’s changes before commit went in but working with Paola on tools page
- Tim will review list of R tools and make sure we’re covering everything
- Tim to see if the can provide some sample R code demonstrations too (e.g. Tropical cyclones)
Alicia moving to Data61
- working to Climate Services for Agriculture
- Can still contribute to working groups
- Big problem for ACS though
- “Research data scientist for climate risk”, still not a CSRIO contract though
- CSA currently using CMIP5, need to move to CMIP6 – work with ACS for downscaled stuff?
ACS reference data replication
- Paola needs to clean up her download scripts anyway
  - Needs better documentation
  - Would be ideal time to hand over to Alicia but can still do it for handover to someone else later
- Alicia won’t be able to help now and Chloe is short-staffed for help
- Need written agreement between CSIRO and Clex – Paola to escalate between Damien and Claire C
- Not a big hurry to get downloads going given holiday period coming up, probably get things going about February anyway
- Storage quota adjustment for ia39 still hasn’t been applied – it’s been a few months now, and Steve has escalated it.
- Could give Paola write permission to xv83 if needed but not urgent yet.
- Code management – BoM can’t use github, but NCI have the no-project (user only URL) limitation. Damien to take up with Steve & Roger whether we can get an git.nci.org.au/ACS repo group established for us.
Single access WG
- Paola has pulled together code to pull from GeoNetwork and prepare for upload to InvenioRDM
- Want to go via an intermediate format so can harvest from other sources too like CSIRO DAP
- Design a two-step format to harvest from input data catalogue into a standard format, and from that format to the Invenio input
- Tim had an example scrape from DAP
- Tim to review Paola’s code/approach
Federated climate data
- Tim setting up the CSIRO Eratos node
- He’ll copy Marcus’ data from THREDDS – ESCI, VicDELWP, CCiA
- Other needs in Ag?
- What happens with the BoM ESCI stuff?
- Tom has been trying to sell Eratos stuff to ACS too.
- Eratos rep will be contracted to CSIRO to give them access to set up their software on our infrastructure

25 Nov 2021 climate data weekly meeting 30 2021

Paola, Alicia, Claire. Apologies: Chloe

FAIR data
- Claire published a wave dataset yesterday, already had a query! Yay Findable data!
Alicia’s contract is not being extended beyond December ☹
- May not be able to help out as much with the ACS data management / replica management
Big data guidance WG book: Paola and Alicia working on tools page.
- Claire made edits.
- Damien’s Pangeo tools survey should inform what we include (make sure we at least capture all of that)
- Paola rearranging, put each major tool description in as an expandable so that it doesn’t take up heaps of space.
- Table of considerations in choosing each tool where there are similarities (e.g. pandas/numpy/xarray)
- How to arrange and present tool groupings
- R content – check with TimE, check Mick Sumner’s blog and Simon Wotherspoon

11 Nov 2021 climate data weekly meeting 29 2021

Chloe, Alicia, Claire, Paola

Reanalysis/reference data
- ia39 storage quota has not been actioned still
  - Chloe to follow up with Steve to escalate
  - Ticket resolved without action
- Start with sharing precipitation datasets
  - VMs can run cron but there are risks with management, and with comms failures between components (cloud, storage teams)
  - Currently managing downloads through Scott’s Jenkins server
  - Get Alicia access to Jenkins server? or find alternative?
    - Sometimes auth fails but otherwise it’s mostly reliable
    - Effectively use Jenkins as a cron-like workflow manager
    - Hosted on access-dev
  - Some things can be done on VDI/OOD
  - Need to consider how to manage things that require authentication vs freely available data
- ACTIONS
  - Chloe Mackallah follow up with Steve to escalate quota request
  - Paola Petrelli review precipitation download scripts and share locations with Alicia
  - Alicia Takbash review Paola’s scripts and also consider where to host (ACDguide github, NCI gitlab, other?) – discuss with Damien

04 Nov 2021 climate data weekly meeting 28 2021

Claire, Paola, Chloe

Single access climate data WG
- Paola has been working on InvenioRDM
- Can now automatically ingest relevant records from NCI geonetwork into test instance
- Record deletion is now easier.
- Records come in okay but missing things and will need improvement – but that is the point of using Invenio!
- Issue warning where things are missing?
- Temporal coverage now available in dates field – from, to, range
- Search guide, should be able to search on many things including spatial and temporal extents even if they’re not displayed in the record if they are present on the back end
- working only on the test node for now.
- Harvest from NCI based on atmosphere climate processes tag – can also search ocean, environment etc.
- Can show additional information that doesn’t display in GeoNetwork, appears in “description” field in invenio but may need a lot of editing
- Discuss invenio front page view
- Invenio roles are based on datacite but GeoNetwork is based on ISO19115-2
ACCESS data paper (Mackallah et al)
- With authors for review
ACS
- Claire C talking to Richard, likely to have agreements with both Clex and NESP that are very similar
- With Andy for review
- Partners request universal access to reference data
- Discuss where to put code – could use ACDguide github org, although everything else there is currently doco

28 Oct 2021 climate data weekly meeting 27 2021

Paola, Claire, Chloe, Tim. Apologies: Alicia

ARDC Skills impact workshop
- Met Sally Lowenstein casually in gather networking session afterward
  - Will invite her to next climate guidelines WG meeting
- Paola had an interview with ARDC about Clex DMP process
NCI reference collection
- Use NCI gitlab to host codes
- Formal agreement for Paola to work with Alicia (under Chloe and Damien) to manage datasets on behalf of the community under ACS storage
- Replaces rr7 properly
- Start with a couple of small datasets, put code to manage data in a git repo
- We still need to address `ua6` clean out. Most can go to ua6_4.
  - What about CESM LME etc?
  - Demonstrate importance of datasets
- Don’t want to have to justify the importance of many small datasets to NCI
- Good opportunity to tidy everything up
- Note: Raktima’s re-generated CCiA data will be CSIRO-hosted.
InvenioRDM for single access WG
- playing with vocabularies – cross listing between vocabs on search
- Developers have seen similar things with custom keywords
- Assign ID to each vocab item, but each of our vocabs included “other” with same ID, which might cause issues. Need unique IDs
- Still need to address variables, need a strategy!
  - Created google doc – please review (and associated spreadsheet of potential candidates).
  - Create list of geonetwork entries at NCI and harvest them
  - Could use keywords in DAP that’s searchable to harvest
  - How to keep maintained – check for new versions, how to know about new datasets to add?
Post-processing for ACCESS
- Chloe working with Scott to broaden APP, okay but not talking to CMOR properly yet. Moving along.
- Similar to CCAM DRSing code (e.g. axiom project)
- Handling CMIP6 output variables is okay but very painful to merge custom variables to have similar metadata
  - Can use JSON files to do what you want but difficult to trick CMOR in just the right way.
ACS
- Chloe started a data standards document for ACS
  - Basically CORDEX, but that is not yet set in stone.
  - Also incorporate non-CORDEX
  - things
- Python access across systems
  - gadi_jupyter and hh5
  - Paul Branson and DCFP pangeo approaches
  - singularity
  - Is it worth trying to bring things together?
  - Damien and Chloe will be distributing a survey to project leaders that may help capture some of these needs
Python instability on NCI/OOD
- Very slow and unreliable at times, dropping out, falling over, very slow response times
- Turns out to have been a load problem – NCI identified
  - Unclear whether resolved
  - CLEX researchers asked to monitor for unexpected python slowdowns and note times so NCI can investigate
- OOD seems to be maybe less reliable than VDI was
  - Sometimes need to go to gadi sooner
  - Maybe often user problems but not always, sometimes external processes seem to be interfering with jupyterlab
    - kernel death leads to lots of unmanaged memory, hard to track down and clean up.
Enabling access WG
- Claire interviewing IM&T (Mik?) for DAP contribution.

14 Oct 2021 climate data weekly meeting 26 2021

Paola, Chloe, Claire

ARDC Skills summit
- Paola met Sally Lowenstein from BoM who is a data steward
- BoM looking to establish a mentorship programme in data stewardship
- She’s at a governance level rather than technical
- Make contact for data governance group? via BoM members? (e.g. Avijeet)
AMOS working groups
- Data governance meets next week
- This week is big data working group
- Paola added content on intake catalogues
- Including a couple of more detailed examples where DRS along isn’t sufficient to resolve data of interest
- Governance document for big data working group is a good point to start (there’s also governance/goal re website for single access WG)
Marine heatwaves code
- code and dataset published
- DAP record created pointing to Zenodo publication of data
- Paola getting lots of questions about the code
- dask is difficult to debug weirdness in!
  - It really is hard to understand, Scott also gets confused by it
  - Separation of comprehensible parallelisation from magic, and what level of efficiency is really required for this task.
  - Comes a time to just run jobs in the queue
- optimum chunk size round O(100MB) but still get weird memory growth.

30 Sep 2021 climate data weekly meeting 25 2021

Claire, Chloe, Paola

ACS
- Chloe on a committee including EMA and NRRRA
  - Help set stakeholder expectations around data
  - Need recognition that Chloe has the authority to determine these governance structures
- Damien leading a work package on reanalysis so work with us
- Tom R organising a meeting with NCI (Sean and Ben), CSIRO (Marcus, Richard and Damien) and Paola(?) on the Federated Climate Data Portal
  - Tom looking to include ACS in the FCDP
  - What are ACS data delivery expectations on the stakeholder side?
  - Clarity around copyright and IP ownership?
  - Clarity around data expectations (format, structure)
- Need to clarify separation of original data from post-processed data
- Discuss project mgmt – `ia39` and `xv83`
- Reanalysis clean up
  - Some replicas are only a few variables, shouldn’t count them
  - Precip replica – FROGS is a regridded product kind of like CREATE-IP
- Chloe’s next steps – set expectations around readmes, data structure, governance
  - Which standards to use and why
- Learn from mistakes of the past!
- Are we going to be producing climate indices? Crossover with Clex work? (climdex etc)
- Ben Schroeter’s Axiom code for DRS-ing CCAM output – could he help wtih BARRA/BARPA as well? (apparently that’s CRE not ACS but related???)
Marine heatwaves code
- Paola continuing clean up and development

23 Sep 2021 climate data weekly meeting 24 2021

Claire, Paola

Reanalysis data
- Alicia has worked through the spreadsheet ot identify datasets of relevance to ACS
- Data in qv56 is managed by Synda but may not be actively maintained?
  - Are CREATE-IP not being maintained by NCI or is the updated data not going into CREATE-IP in the first place?
  - Sometimes may need to get data directly from the provider after the CREATE-IP timeseries ends, but then lose the DRS.
- Could use Synda to replicate in xv83 if that’s more efficient
ACS
- Paola may be able to provide in-kind support for ACS, helps CLEX to recover storage
- Formal agreement with CSIRO to inherid data and data mgmt scripts
- Formal handover of data mgmt responsibility to CSIRO/succession plan → role clarity
  - Management of long term updates
Intake
- Can it be used with ESMValTool?
  - GPCC would be a good test case, it’s the same data just minor renames.
  - Does CMORising on the fly write the intermediates that ultimately get removed again?
  - Why the NCL?!!?
- FROGS seemed like it’d straightforward but actually wasn’t.
- ERA-Int
- Contribution and comments on Intake in the Big Data
  - Include local examples (Scott’s work!)
Pangeo
- Ocean Community hack week – should Paola contribute or too late?

09 Sep 2021 climate data weekly meeting 23 2021

Chloe, Paola, Claire, Tim, Francois

Data retractions
- More ACCESS data to be retracted and republished
  - metadata error found by BARPA
p73/p66
- p73 for ACCESS data is now available
- p66 membership will be cleaned up
  - Quasi-active users like Claire and Paola will be low on the list of people to be culled
CMIP6 regridded data
- Francois still getting requests so that’s good
- Not being updated automatically, could get out of control
- regridded some DAMIP data
Data working groups
- InvenioRDM
  - Claire had a play, had a bunch of questions so general success adding a metadata record
  - Server has a certificate now – https://oneclimate.dmponline.cloud.edu.au/
  - May well be taken up by NESP2, demoed to Brad Murphy last week.
  - Tim and Scott helping on the technical side
- ESMValTool
  - Need to play with the initial set up and see what we can and can’t do.
  - Create intake catalogues for other data (other than CMIP data which is already done in hh5 as well as by Francois)
  - Can ESMValTool read from intake catalogues? Be great if it does!
ACCESS-NRI/ACCESS Day
- Gather a bunch of ideas and expectations, hand to Richard Coleman – “this is what the community expects you to do”
- Set expectations from the community instead of the scoping document
- Understand what’s in and out of scope
- Community building
Reanalysis data clean up
- CSIRO supporting with actual allocation for someone to do stuff!
- Meeting this afternoon, Alicia Takbash will do the work with our support
- Tim also has aclocation the climate Mission, working with data organisation with Raktima etc which will help
- No one clearly appropriate from BoM side (was Aurel) – Francois has permissions
  - Justin is a new section head from but only 3 people in his section!
  - Reorg-triggered hiring may help here, Francois will keep us in the loop
- Also reliant on Kelsey/NCI to help with storage availability
- Paola, Claire, Chloe and Alicia meeting this afternoon.

02 Sep 2021 climate data weekly weekly 22 2021

Thomas, Claire, Paola. Apologies: Chloe

Vaccines!
- They kick your butt but it’s better than covid!
Reanlaysis data
- Discuss when Chloe can join us
ESMValTool
- Scott has installed a conda env in hh5
- Many mentions of it at ACCESS Day last Friday (and in ACCESS-NRI context)
  - e.g. Gab for land surface evaluation
- Thomas had looked at it and decided it wasn’t right at the time
- We’re evaluating it again now it’s a bit more mature
- Standard tools to produce particular metrics, really good for quick standardised analysis
- How much effort to CMORise datasets like AGCD/AWAP?
- How extensible is it to new/different methods/metrics?
- No longer need two copies of datasets, can CMORise on the fly with ESMValTool v2.3
- How well does it work with dask?
- User config vs central admin config (where to find files, write intermediates etc).
  - Value of shared intermediate files for reuse?
- Share new code/analyses
- Validate model runs as they come off the production line
- Apparent necessity to explicitly list input files is a worry? Is there no way to script the inputs/use intake/stacc etc?
- ~24hr build time makes it really unappealing to try.
- Surprisingly many NCL dependencies
  - What if any is the connection with GeoCAT? (scalable ncl) – surely necessary to make the recipes scale
- How does Damien Irving see the landscape of GeoCAT, ESMValTool, others?
  - https://geocat.ucar.edu/
    https://pyaos.github.io/packages/
- CMIP6-preprocessing could be used and then build recipe on top of that instead?
- Interest to, say, R users?
ACCESS-S2 meeting (ACCESS day)
- Things need to be actively maintained – e.g. if an error is found in data it does need to be corrected, not having everyone work around problems!
- Everyone needed in the same room – Griff, Paola, Kelsey, Thomas, .?
- Effort to fix metadata in files is surely worth the time, shouldn’t put all the effort on the user.
  - One year has different metadata/inconsistency.
  - 3 weeks of Thomas’ effort to work around it, could use nctools or jupyter to inject/fix the problems.
- Learn from CMIP5 mistakes!!
- Correct shared files when necessary so everyone benefits. (e.g. replicated datasets that come with errors: document and fix).

26 Aug 2021 climate data weekly meeting 21 2021

Paola, Claire.

InvenioRDM metadata catalogue
- Paola has installed the latest version of invenioRDM
- Record type vs rescource type
- Manual changes to JSON schema to enable additional fields (e.g. spatial extent)
- Invenio package management is a pain, need to start over
- Met with Tim & Scott, interested in helping
  - Tim immediately picked up the relationship between JSON and postgresDB – use json type in db so every record is in one table with a JSON for each record
  - Changing JSON schema therefore doesn’t change the database, new records just validate against new schema
- GCMD keywords, can get vocabulary as YAML file. https://gcmd.earthdata.nasa.gov/static/kms/
- Can use multiple vocabularies for each field if available (e.g. GCMD measurements + CF standard names)
- NESP2 – Brad sent Claire a proposed metadata template to review
ACCESS Day
- Tomorrow, whole day workshop
- ACCESS-NRI status update?

12 Aug 2021 climate data weekly meeting 20 2021

Claire, Chloe, Thomas. Apologies: Paola

Reanalysis data
- What is needed for ACS in the ACCESS space??
- Data in rr7 is not being updated/maintained other than MERRA2 which is now split between rr7 and ua8
- Community effort required
- NESP2 – Brad Murphy supportive of an IP register of input data
  - Identify what input dataset each research group is likely to use
  - Nominate which data will be required for projects
  - Who will maintain each dataset?
ACS/NESP/ACCESS-NRI
- Projects all seem to be in scoping phases at the moment
  - NESP co-design
  - ACS project scoping
  - NRI??
- NRI: secondments vs hires. No idea what’s happening in this space
Ocean Hack Week
- Last week, based in Maine
- India-Oceania local area group
- project on CMIP6 with cloud hub connected to Pangeo CMIP6 archive
- Hopefully will end up turning into some guides or something.
  - https://github.com/oceanhackweek
  - https://github.com/oceanhackweek/ohw21-proj-cmip-ard
IPCC report
- Very high impact
- Final preparation done online this round – we should have more messaging about emissions savings from not having all the representatives/lead authors travelling for AR6!

05 Aug 2021 climate data weekly meeting 19 2021

Paola, Claire

CMS training
- Paola delivered data training
- Attendance fluctuates hugely
- Utilise external materials for common topics and focus on things that Clex specialises in
- about 16 attended data training on managing and publishing data
- 2 part series – importance of data standards etc.
- lots of interest in data publishing
NESP2
- Co-design phase
- Knowledge broker role is currently advertised
- Sonia Bluhm and Mandy Hopkins both leaving
- Tentative science plans are in place but codesign needs to work through
AMOS working groups: Single Access Point
- InvenioRDM is close to Long-Term-Release version
- Review and comment on or add to template spreadsheet
NCI OOD
- OnDemand service in testing phase (see last week’s notes)
- VDI being phased out rapidly
- Works on CSIRO VPN
- Not working for Paola on home network or UTas VPN – is this happening to other people?
Reanlaysis datasets
- C20C needs updating and maintaining but not by clex – move to CSIRO share?
- MERRA2 is split between ua8 and rr7, as there’s not enough storage in rr7
- ua6 still has some datasets in (e.g. LME data), other LE and LME are in ua8
- JRA55 is partially in synda (qv56) but the rr7 is not maintained and not the same, retire rr7 copy?
- ORAS5 ocean reanalysis has some demand
  - Claire has a copy of just sea surface height to 2018 plus extensions directly from data producer
  - Paola has key vars in ua8/CMEMS_reanalysis/..
- Paola downloading GECCO2, SODA and CHOR all at monthly frequency and possibly other ocean reanalysis for ENSO research
- We should reserve one meeting to talk about how to re-organise all these datasets

29 Jul 2021 climate data weekly meeting 18 2021

Claire, Chloe. Apologies: Paola

Paola is in a future of CMS meeting
- CMS are invaluable!!! Happy to help promote in any way we can
- Permission to work on cmip6-preprocessing
projects
- ESCI
  - Complete but Chloe had to do a lot of reprocessing after some significant problems in the data
- ACS
  - xv83 is funded 1/3 Decadal and 2/3 CSC – not whole of CSIRO
  - Talk to Mark H about how to approach data management
  - Chloe ought to be a CI on xv83
  - CIs (or at least leadCI) should ideally have sudo-like powers to manage data in their projects
NCI OOD
- VDI replacement service
  - guacamole based remote desktop
  - Gadi terminal in browser
  - Jupyterlab – can specify modules and conda envs to use (so can leverage hh5)
  - VDI migrates old VDI including home dir
    - still not integrated with gadi filesystems
- Seems to be basically functional
- Removes need for Strudel
- A few problems noticed, some reported
- VDI switch off in a month?

22 Jul 2021 climate data weekly meeting 17 2021

Thomas, Claire, Paola, Chloe, Damien, Michael G, Tim, Julius Busecke

CMIP6 preprocessing module
- Guest meeting with Julius Busecke, developer of cmip6-preprocessing, a tool to align metadata and perform ocean de-drifting on cloud-based CMIP6 data
- ACCESS-ESM has different metadata than IPSL, e.g. (x,y vs i,j)
- Demonstration of aligning ACCESS-ESM and IPSL data, performing a multi-member ensemble analysis of the two models and producing a nice plot of hist+ssp585
- Extensive discussion of errata processes and how to handle (and whether modelling groups will action various things).
- From the chat
  - From Damien Irving to Everyone: 10:13 AM
    In terms of “pint coming soon” for unit conversion, the xclim library has a nice implementation of pint for unit conversion that I’ve been using recently (I found pint-array wasn’t really ready?) https://github.com/AusClimateService/unseen/issues/1
    From Claire T to Everyone: 10:20 AM
    I think Paola was working on a library that was meant to interact with the ESGF Errata service but not sure it worked as well as we’d hoped – due to lack of/inconsistent reporting? But that’d help you know when/what fixes to apply on the fly, I think – Paola?
    From Michael Grose to Everyone: 10:30 AM
    I like the visualisations on the right sides as it runs – I haven’t seen that before!
    From Chloe Mackallah to Everyone: 10:31 AM
    I was watching that too, so pretty 🙂
    From Claire T to Everyone: 10:31 AM
    The dask dashboard plugins for jupyterlab are very cool! 🙂
    From Damien Irving to Everyone: 10:35 AM
    With respect to grid information, many of the modelling groups have bogus volcello data (i.e. the volume of the global ocean is non-sensical), so it would definitely be an advantage to be able to calculate areas and volumes from your independent grid information.
    (Similar to the way region mask provides you with a basin mask rather than using the basin files provided by the CMIP6 modelling groups, as they are also inconsistent with their basin boundaries)
    From Claire T to Everyone: 10:47 AM
    Cool, so it sounds like you’re somewhat hooked in to ESGF, that’s good
    One nice thing is we have NCI (the ESGF node) handling downloads for us so we at least don’t have that particular headache anymore! but yes, sympathise hugely from CMIP5 days
    (so we get to use pangeo on the clef tool / intake catalogue on the HPC system)
    From Damien Irving to Everyone: 10:51 AM
    Hierarchy of problems:
    - 1. Inconsistent metadata (e.g. x,y vs i,j)
    - 2. Incorrect metadata (e.g. branch times)
    - 3. Missing metadata (grid information)
    - 4. Incorrect supporting data files (e.g. bogus volcello, basin boundary inconsistency)
    - 5. Incorrect data files with easy fix (flux wrong sign)
    - 6. Incorrect data files with no easy fix (wrong definition for masso)
  - Crowd sourcing for fixes to 1-5 sound like a good goal.

08 Jul 2021 climate data weekly meeting 16 2021

Claire, Thomas

Ocean Hack Week 2021
- Thomas leading a CMIP-based project
- Paige helping out
- Lots of ecologists, biologists etc, R users who want to use CMIP
- Convert specific datasets to CSVs or whatever as necessary for use in ecology fields
- Github contributions
- Julius code to be used
- Setting up a hub, cloud machine and platform with R, python etc
  - Connect to publicly available CMIP dataset in the cloud (pangeo data or ESGF nodes?)
- Who will join this project? People good with data or people who want to work with the data?
- Us to contribute here and there? Maybe August?
  - Global – so timezones but we can help asynchronously.
EASIhub
- When to work on HPC and when is cloud appropriate?
- If we work in EASI are there costs associated with accessing cloud-based climate data?
- Do we spin up EASI instances where the data is??
- Sometimes HPC remains the best option – e.g. NCI’s CMIP data holding does what we need and has no surprise costs
HPC
- filesystem instability with /scratch1 on Pearcey (also /datastore failovers)
- Petrichor migration coming soon
- Advice to move back to NCI but is that the best option??
- Australian Climate Service will fund a fairly significant mount of stroage at NCI for climate (-related) data
Python
- Scott’s intake example is excellent
- Thomas has demos of how to use the hh5 intake catalogue
- Claire is using the catalogue to compute Nino3.4 indices for a couple of CMIP6 models
- Francois’ notebooks also demonstrate use of intake_esm from his own catalogues but they are much more advanced than we are at and we struggle to reproduce.
- xarray – dataArray name vs variable name?! .rename is a very useful command!
Lots of internal commitments getting in the way of doing fun python things!
School holidays so a few people on leave at the moment.

24 Jun 2021 climate data weekly meeting 15 2021

Thomas, Paola, Chloe, Claire

CMIP6-preprocessing
- Chloe has made contact with Julius about getting ACCESS into the tool
- Thomas finds these tools useful for doing advanced cell maths
- What’s missing? is it not the fx files?
- Michael/Zeb – is this the tool to rule them all? probably not! cloud based, but just another thing to know about along with ESMValTool, PCMDI metrics, there are/will be many preprocessing and analysis tools for CMIP6
- Invite Jules to join this meeting one time?
- What does XGCM need?
CORDEX CCAM issue
- Scott found Marcus’ CORDEX-Antarctica runs from late last year have inconsistent calendars within a dataset

The CORDEX run cordex.output.AUS-44i.CSIRO.CCCma-CanESM2.historical.r1i1p1.CCAM-2008.v1.day.tasmin.v20201120 has some wonky metadata, the 1960 file has a calendar marked as ‘standard’ (i.e. gregorian), however the 2005 file has a 365 day calendar.

- This data was not produced by Tony & my post-processor, there may well be other fun issues in that dataset!
AMOS data working groups
- Big data group setting up a jupyter book for documenting issues and approaches Working Group Goals — Working with Big/Challenging Data Collections (acdguide.github.io)
- Could use this sort of set up for other working groups too, particularly the dataset guidelines group
- Similarly CMS wiki data pages which Paola has been revising in preparing training
- Pull new data training pages together into a jupyter book too
- Research Vocabulary Australia (similar to RDA, part of ARDC) have CF which is some start toward a climate vocab… also GCMD but that’s really machine-readable rather than human-readable. Lack of a climate repository limits availability of searchable climate vocabularies.
Data storage grants
- ALCG previously came with no storage grant/funding
- NCI will now be running a storage grant scheme again for nationally significant collections (similar to RDSI?)
- Not clear how to apply for this storage
- Important for ongoing support for datasets from groups with not-indefinite funding (clex)
- Important for supporting data that’s critical to a wide community – replicated reanalysis data e.g.
- This scheme is internally administered by NCI
- Data governance committee with NCI communications aren’t open/transparent
- No direct line of comms between technical staff in partner orgs and NCI
Observational dataset coordination
- NESP suggestion of collating individually held obs data into a centrally managed place
- But that sounds like rr7
- In ~August let’s revisit this coherently and discuss funding, what should be included, how to structure etc.
- splitting of data across projects makes it very hard to find and use from a user perspective

03 Jun 2021 climate data weekly meeting 14 2021

Thomas, Claire, Paola

ACS (Australian Climate Service)
- Data management strategy is still unclear.
- We seem to have opportunity to speak up and contribute through this group, AMOS data working groups, NESP2 Data Group
- Want to ensure alignment of requirements across projects where data may deliver to multiple things – NESP2 plus ACS may have different aspects on the same data
- Damien is open and driven by best practise, open science, documentation
- Damien and Mitch Black seem like the people to work with in the ACS space to progress this
- Richard is CSIRO’s rep so follow up with him too
Marine Heatwaves code
- Paola looked at Nick’s code (a dask version of Eric O’s code)
- Paola’s version reviewed by ???
  - Matches thresholds except at boundaries?
- Another code from clex students
- xarray efficiency issues with `groupby` operations limit when it’s appropriate to use xarray vs pandas.
CMIP6 (pre-)processing repo (Lamont/pangeo)
- Making cmip6 data analysis-ready
- becoming quite mature
- Repo is here: https://github.com/jbusecke/cmip6_preprocessing
- Questions about ACCESS model grids – direct them to Chloe Mackallah
- Funny that even with all the learnings from CMIP5 there are still so many exceptions in CMIP6!
- Julius happy to talk to us if we want to know more
Data working groups
- Claire C will try to mention to NCI again and see if they want to participate
ERA5
- Data is natively 0:360 longitude but we request -180:180
- A student found an issue where the longitudes had been amended but not the data so S.America appeared where Australia should be!
- Trust issues with the authoritative data now
- data governance committee m’ship?

27 May 2021 climate data weekly meeting 13 2021

Claire, Paola, Francois

Last week this meeting was replaced by climate dataset guidelines WG
- Alicia, Katie, Chloe, Paola only attendees
- Agree on guidelines, but need to agree on levels of specificity, generic things.
- Don’t want to invent things, use existing conventions etc
Paola has been busy preparing training materials on data publishing
- Journals often have their own policies around data but increasingly they talk about FAIR principles and the Coalition of Earth & Space Science (framework for data sharing in that field). No publication without data DOIs (can create problems with people trying to put DOIs on data they don’t own).
- Sometimes publish code where it’s inappropriate to publish data (e.g. Francois’ regridded CMIP data)
  - NarCLiM have been using Francois’ code for ENSO, teleconnections work, yay!
- Journals recommend discipline repositories – e.g. AAD, AODN, maybe TERN? but otherwise can use Figshare but this is less preferred. Avoid using just FTP servers!
- There’s still issues around regularly updating data. Mint DOIs periodically?
  - Can mint a DOI on the metadata instead of the data to avoid this issue – e.g. upload the metadata record and sensor metadata files as a file in Zenodo.
CARSA/ACS code sharing
- How to make jupyter nbviewer work with NCI git, is it possible?
- Have to submit all code to the CARSA git org, but that’s problematic where code is better hosted on NCI where it’s being developed/used, or inappropriate like CCAM
- Good in principle to share data but problematic to have it all together in a single git org
- Clex don’t do that, researchers share code where is appropriate for them and mint DOIs through Zenodo to the clex org
- Hard to get buy in from researchers too! They want to share code but not in this central way. COSIMA cookbook example
- Damien has good reach in research community and people want to do what he says but need to take the right approach.
InvenioRDM
- Redeploy on new larger VM (100GB persistent on UTas Nectar)
- Long term release of the tool soon, but we can provide feedback to the developers now
- Test uploading records, what’s missing?
  - URLs don’t render??
- How to manage who can register for accounts to ensure only the local community can contribute? AAF? specified email domains?

13 May 2021 climate data weekly meeting 12 2021

Paola, Thomas, Claire

Marine Heatwaves code (Eric Oliver)
- Used by a heap of people in the ocean community (Thomas knows some eg Nick Mortimer, Hillary ?, Claire knows others eg Vanessa Hernaman, Paola works with many in CLEX)
- Paola has been working on parallelising it, changing structures but keeping variables and everything intact
- Jules Kajtar’s work was a minimal parallelism of Eric’s code
- MHW will be covered at Nick M’s workshop at upcoming Dask Downunder summit
- Dask/xarray can empower so much that makes this work much more scalable.
- Opportunity to combine CSIRO and CLEX efforts here? Repos currently private during validation but could be shared.
Intake catalogues/CMIP
- Thomas has solved his CMIP5 workflow problem!
- Select CCiA historical & RCP model data quickly and start working.
- Thomas’ code: https://github.com/Thomas-Moore-Creative/NCI-CCiA-ARD
  - https://github.com/Thomas-Moore-Creative/NCI-CCiA-ARD/blob/main/CCiA_CMIP5_intake_sandbox.ipynb
  - CMIP5 catalogue is implicitly “latest” (cmip5_all contains versions) so version is unique in this catalogue for each dataset.
  - Necessary because rr3 contains ‘latest’ symlink but al33 does not (same for fs38 vs oi10).
- Scott has made a dummy version of Clef based on Intake instead of MAS, can compare to ESGF, will present at our AMOS working group meeting later this morning.
- Intake catalogues still have to be routinely updated to keep accurate
Ocean hack week – link climate and ecology, e.g., ‘how do we get a timeseries out of CMIP to look at biology changes?’
- Training/upskilling topics, e.g. How to do initial large-scale processing in python then transition reduced dataset to R for traditional analysis.

22 Apr 2021 climate data weekly meeting 10 2021 – finding CMIP data

Paola, Thomas Moore, Scott Wales, Claire, Raktima, Francois

Thomas’ use case is how do we find data matching particular criteria when getting started with CMIP5/6 to get going quickly
Working with python
Ideally want to avoid duplication of effort
- Clef indexes data based on NCI’s MAS database and also searches ESGF for remote files – close to “authoritative”
- Francois’ intake catalogues are good and intake-esm is much more mature now, but not “authoritative” – scanned to update ad hoc (~ every few weeks)
NCI manage the metadata database that Clef is based on, but issues occur now and then (e.g. checksums, ‘project’ issues in CORDEX)
- Clef could use multiple or different backends, e.g. could use something lighter weight like sqlite or intake-esm catalogue
Clef has additional functionality like confirming checksums, but maybe people don’t rely on this too much?
Intake that Francois uses selected the latest version of each file available but does not check against databases (e.g. ESGF) for checksum – best approach for lightest weight, avoid database management hassles.
Intake is great for easy access of data. Clef meets needs of researchers who need to know version/tracking_id information when publishing papers.
Intake can catalogue all file metadata but trusts that the replication was correct.
Scott has been developing a hybrid system, create CSV from intake plus postgres from MAS (I think?)
Intake is heaps better from jupyter notebooks – clef is a bit clunky in python for filelists
- Francois’ demo of whole catalogue was really impressive!!
As a community, move toward centralised intake catalogues?
- Catalogue scans are taking longer, takes a few hours.
- Scott is scanning nightly ~26M files via Jenkins server.
- CMIP data is relatively static, weekly scans should be sufficient.
- No complaints from NCI about Scott’s scans so filesystem impacts are probably fine.
- Scans vastly more costly if opening files to access metadata.
- Is it viable to only add changed files to catalogue? Does that reduce impact?
Jenkins server for automation of regular jobs – what is it, how does it run?
Francois had some issues with intake-esm and needed modification to work so unwilling to update. Scott hasn’t seen these problems so may be resolved now.
Intake-esm in Francois’ conda env in `eg3`, Scott’s is in conda in `hh5`.
- NCI also maintain `dk92` as a “centralised conda” to accompany their pangeo module which is broadly a copy of hh5 structure, not sure what it contains though.
- Thomas uses a Decadal team conda env
Centralised catalogues
- Provide demonstrator working envs and catalogues
- Propose NCI host the catalogue for each dataset in its relevant project (NCI manage these projects but it’d be ideal for users, so it’d be good to try to have this conversation).
- Need to make it mature first
Scott and Francois to work together to perfect their catalogues!

15 Apr 2021 climate data weekly meeting 9 2021

Raktima, Claire, Paola, Francois.

Data indexing
- Next week meeting focus on data finding – clef, intake etc options. Invitations extended to Thomas Moore (CSIRO) and Scott Wales (Clex)
- intake not currently a viable back-end for Clef (still file based not db?)
- Clef having problems with MAS for CORDEX data (new database, permissions issue)
- BoM SDM data is published but causes many problems due to incomplete/incorrect attributes
- Francois’s intake catalogue covers whole of CMIP6 (not only monthly), ASCII file,
  - query time appears to be fast.
- intake catalogue creation (data crawl) takes ~20min (but NCI systems may have stability issues with crawls)
- Francois CARSA example notebook using intake-esm: https://git.nci.org.au/pfd548/carsa-work-package-3/-/blob/master/src/notebook_exemple.ipynb
- Clef could potentially support multiple backends (ie MAS and intake-esm)
- Intake-esm used by many climate facilities internationally
- Clef’s power is it searches ESGF and local fs simultaneously, Intake is you get what you get.
Data working groups
- Meeting today on working with large data collections
Dataset cataloguing and usage comments
- CRE Mission data catalogue will include a lot of metadata (everything including quality assessments and “market readiness”)
- Will be able to be made public on the platform, so can be used by single catalogue database too potentially
- Single access working group catalogue to use invenioRDM (similar to zenodo).
  - Include entries for datasets but also data portals so people can find the portals (even if they don’t have access).
  - suggested use field (e.g. ‘forcing’, ‘validation’, ‘analysis’ etc)
- CERN funded long term, with developer support for at least 20 years – so zenodo, invenio etc tools are likely to be very stable in terms of long term support.
- Demonstrate the invenioRDM product to NESP2 to meet their “data wrangling” objectives

01 Apr 2021 climate data weekly meeting 8 2021

Paola, Claire

Single access point catalogue group
- RDM (research data management) tool installation hassles, trying to set up a test install locally
  - DB with ElasticSearch front end
- Rebecca Cowley circulated link to Programme – IMDIS 2021 – Online, 12-14 April (seadatanet.org)
- group meeting this afternoon 1pm
- Paola looking at less info than what Raktima is capturing in her pages, but maybe we can leverage overlapping data set entries (but R’s pages probably won’t be publicly visible?)
- Going for breadth of datasets rather than depth – want to find everything even if info is limited
- Goals and scope are relatively clearer for this group compared to others
- Be good to check in soon with Tim re RDM tools.
Other working groups
- Access across institutions group is a bit more nebulous – we can have agreements but is it working?
- Group 2 on large datasets may be more of an ongoing CoP
- Funding model group requires engagement from NCI to understand path. ARDC trying to work this out but need NCI engagement/leadership in _collaborative_ approach.
CORDEX
- New table set up in MAS, but can’t see the view properly yet.
- test/dev/ops is fine but need more control at the test level to reduce risk of mistakes being introduced in doing the stages in different ways
ERA5
- ub4 version being retired at end of month
- era2grib tools are used for models (e.g. WRF, UM) and that causes some hassles. (netcdf → grib → netcdf)
CCiA data
- Raktima and others to recreate the data underpinning CCiA – data reproducibility, “unencumbered IP”.

25 Mar 2021 climate data weekly meeting 7 2021

Claire, Paola, Francois, Raktima

CORDEX
- Syazwan is downloading Marcus’ CCAM data from the Cantabria ESGF node
- Clef searching CORDEX now but:
  - BoM-SDA is published at NCI but doesn’t match ESGF conventions
  - Jack’s CCAM data has wrong project name “CORDEX-Australia” not “CORDEX”
  - CORDEX data is a mess!
  - Need NCI to refresh tables to fix some issues, before the Clef udpate can be released to users.
- Users can request CORDEX data now
- Might be interesting to get CORDEX-SEA data to enhance collaboration with Singapore group
- Paola has some CORDEX data also in `ua8` including some from Jack – is this a copy of what’s in `rq8` downloaded from the CSIRO data server?
xclim
- Model evaluation code to calculate climate metrics
- Paola has come across it before but no one seems to particularly be using it
- Paola looked at in context of events eg marine heatwaves but they have alternate code for that – e.g. Eric Oliver’s code being updated from numpy to xarray and dask
ARCDAP
- Francois and Claire remotely attended a workshop in Singapore working with ASEAN countries on regional climate projections
- Constructive meetings, will hold follow ups with Singapore group
CMIP6
- There doesn’t seem to have been much data published lately?
- Not that many models in our selections? (55 for historical), but it’s not obvious that we’re missing data… Paola Petrelli to investigate
- Copernicus hosting a subset of CMIP6 data for selected variables and experiments with additional QA/QC.
  - Might be good for sub-selection for quick look things?
  - ACCESS is not in the model list wonder why? Local ESGF node? More data to come in May
- There is data on NCI that isn’t showing up in Clef, maybe not in MAS or maybe checksums?
ERA5
- Only access through rt52, ub4 data to be removed soon.
- CSIRO looking to reproduce CCiA data using ERA5

11 Mar 2021 climate data weekly meeting 6 2021

Claire, Paola. Apologies: Raktima

Many data requests recently!
Clex ERA5 download tool
- Being used by US polar agency
  - Need to update documentation so it’s appropriate for external use!
  - Gets data at variable level rather than table
- NCI don’t yet have a clear solution for CDS model data
- vertically integrated surface data is being downloaded, who are the users?
Data publishing
- Clex Himawari product in the pipeline, hold ups with citation information in files (incorrect name/s).
- Licence and readme – Readme no longer recommended by NCI? Cover in data description in geonetwork.
- BARPA data may need to be published soon-ish?
Data working groups
- WG1 single access catalogue is off to a good start, good engagement. TomR & PeterL to join group.
- WG2 has first meeting today. Seems like a lot of pangeo focus, need to not lose sight of lower level issues like data structure, chunking, usability for researchers, end users?
  - Aim for guidelines for users producing data and consuming data
  - Would be nice to have intake with a db back-end so we could get datasets ‘intake-ready’
Clef
- Add CORDEX support for data in rr3 and al33
- Can request additional downloads (via VDI, if on gadi give warning that the request needs to be manual now)

04 Mar 2021 climate data weekly meeting 5 2021

Paola, Claire, Raktima, Francois

Data working groups
- doodle polls for initial meetings sent out for 3 groups staggered by a week each
- Only a few responses but going ahead and organising meetings
- First group (“Creation of a single catalogue and/or access point for climate data”) meets tomorrow at 1pm
NCI ESGF unpublishing
- FGOALS data erroneously published in both CMIP and ScenarioMIP has now been removed
NCI data publishing
- Keywords can be corrected via confluence process for new/updated publications (e.g. BRAN2020)
- Another field for at least one clex publication contained an incorrect flag (magnetism)
- Data collection contact is recorded but not displayed in geonetwork
- Data management portal exists but not functioning (datamgt.nci.org.au has a nice interface but can’t get into the DMP Online tool or anything)
- Previously used form in datamgt.nci to create/submit form for data management plan and then mint DOI, could copy previous plan and update.
- Mapping to geonetwork database issue
- Current process is to create data management plan on a confluence page and NCI presumably manually transfer to geonetwork
- Paola provides XML from her own DMPOnline tool implementation to NCI
- A template might be useful?
Xenodo collection for datasets
- Many (CLEX) researchers are publishing data files ad hoc to support papers
- This is not ideal but if it’s happening maybe should at least pull together in an centre-level xenodo collection
- Can add records from other data repositories (like Figshare) as well
- might be useful for datasets too small for NCI to worry about or that aren’t in the clex collection
- Decision not made yet but likely to happen
CMIP6
- Fewer requests from researchers than in CMIP5 it seems – less usage or better support and doco this time?
- CREATE-IP is partially up to date in qv56 but still not being routinely maintained yet, still use ua8 for now.
ARDC phase 2 of Data Retention Project announced
- Links to working group 4/5, provide community aspect to NCI?
- https://ardc.edu.au/news/ardc-launches-phase-2-of-the-data-retention-project-significant-national-data-collections/

25 Feb 2021 climate data weekly meeting 4 2021

Claire, Raktima, Paola, Francois

CSIRO dataset documentation
- Raktima working on confluence pages for our datasets, Claire (and others) to review
AMOS workshop
- No word from NCI re working groups, Kelsey is active in other data publishing things though
- List of emails from AMOS to contact the non-attendees to check interest in working groups
- Aim to start meetings in 2 weeks time.
- Paola to prepare general agenda/guidelines to keep WGs moving, communicating, goal setting etc.
NESP2 data working group met
- Claire and Francois are in the group so should have 2 way communication. IanM represents CLEX. Ben from NCI. Reps from other unis and the Bureau.
- Presentation from AMOS workshop seen as a good basis to build up list of NESP-related datasets
- There’s a lot of uncertainty for people about how to actually get data from where it is to delivery. Need people who are willing to just dive in. It’s hard to know where to start!
Data publishing/datamgt NCI
- Is datamgt still under redevelopment?
- Current publishing is through Confluence pages describing datasets/collection
- Paola still providing XMLs for geonetwork
- Big projects haven’t been through the same DMP process – e.g. ESGF publishing is different; BoM use big projects with generic DMPs (may be insufficient content)
- How to edit/update an existing publication?
- Publication guidelines?
- Data curation is a pain on disparate datasets – coordinate systems, standard name variables, time conventions etc. Data wrangling role is vital!!
Pangeo data management
- Thomas Moore’s team use zarr
- They like intake-esm catalog on Francois’ regridded data
Other data management
- TRMM comes in a weird binary format. Or HDF. Can convert to netCDF (Francois) and update regularly but then need to publish/register.
  - Paola has netCDFs in ua8
  - Looks like they’re being downloaded from NASA via opendap so it is available in multiple formats from source (netCDF is new)
  - Can’t find it at NCI via catalog so people don’t know it’s there.
- ERA5 at NCI is managed by ‘table’ rather than ‘variable’
  - not getting accumulation variables
Code publishing
- Clex gathering code, to publish centrally via zenodo?
  - Other options being prepared from pre-prepared offerings for flexibility (e.g. ruby on rails, db+web front end)
  - Install on a VM and adapt as needed, fill in.
- Could use the same database to capture a dataset list
- Is Birdhouse still relevant?? (Tim Erwin)

18 Feb 2021 climate data weekly meeting 3 2021

Paola, Tim, Francois, Raktima, Claire

AMOS collaborative data workshop
- Review self-nominated working groups
- shared docs in Creating a collaborative approach to climate data – Google Drive
- Proposed 6 (or 7?) working groups
- Objective of WGs is to develop guidelines or policies which we can then try to adopt at an institutional level
  - trying to implement new tools or services is likely to be too far.
- Contact each group in the next week to start discussions.
  - Update NESP, CSIRO etc mgmt.

28 Jan 2021 climate data weekly meeting 2 2021

Claire, Paola.

AMOS
- Paola preparing docs for collaborative data workshop
- email potential attendees in the next couple of days
CMIP6
- Some incorrect attributes Scott found doing climpact indices
- data withdrawal inconsistent (e.g. ssp data in CMIP for one model)
- metadata issues are a problem, tend not to retract for them but makes it near-impossible to do programmatic work with data like that
- Errata service is down – certificate expired and no page loads.
- Review NCI synda replication files

14 Jan 2021 climate data weekly meeting 1 2021

Paola, Francois, Claire.

AMOS
- Claire registered, others still need to.
- Paola working on Code of Conduct for the workshop
- Platform for conference is OnAir
- Google docs for collaborative doc editing
BoM R&D workshop
- Best ever
- Better keynotes could be attracted when no travel involved
- Easier to pick and choose talks of interest
- 250 ppl attended
- Gather worked well for social events
- Just standard Teams offerring
Singapore ARCDAP meeting
- Rescheduled to this year
- Need to check if it clashes with AMOS workshop
ESMValTool
- Hard to justify when we already have working code
- Hard to combine our codes centrally without significant support
- Researchers need to invest in that process
NESP2, CARSA
- NESP2 is underpinning science, but delivers to stakeholders
- CARSA is delivering operational products
- Alignment?