Date: 20/6/2024
Participants: 14
Chair: @ShayneM
1. Admin
- Speakers: @HIMADRI_SAINI has volunteered to talk on 4th July, but we have no speaker for the meeting after that (18th July). Please consider presenting, even work in progress that you’d like some feedback. Message co-chairs or @Aidan.
- Remember to add any items you want this WG to discuss to ideas for topics
- There is a call for feedback on what subjects should be covered in the training day before the ACCESS Workshop. Please read and respond and share with anyone else who is likely to attend and wants to have some input, particularly ECRs.
There 8 topics. Would like you feedback.
- How to find and access datasets on NCI
- Running model experiments with Payu and Git
- How to debug common errors and tackle ancillaries when running ACCESS-CM
- Running a high-resolution regional model using a two-level nest
- 3d visualisation for earth systems data
- Advanced git & GitHub best practices for software developers
- How to handle large model output
- benchcab: scientific evaluation framework for CABLE
- CMIP7 Evaluation Hackathon
Anything not in training day will be covered as online training or similar.
Please reach out to @paigem if you want to discuss anything training related.
- There is also a call for ideas for breakout sessions. Please do give your feedback.
Aidan make a topic to get ideas for what to cover in ESM WG “day”.
@Harun_Rashid can we talk about biases in models? Utilise climate expertise to identify source of biases and attempt to remedy.
@ShayneM is there a list of metrics from hackathon to target/complete? Are model biases included in hackathon metrics?
@Harun_Rashid Romain has a new team to improve ESMValTool into people’s workflows. Is there a specific bias we could focus on to feedback to ACCESS-NRI to improve these. Not easy, will not solve every problem but would be beneficial for the model and the community.
@ShayneM agree this could be a useful target for ESM WG compute resources.
@nicolamaher this is important, evaluation only tells us good/bad, but no pathway to improve the model.
Evaluation planning is done through this topic:
@Harun_Rashid Met Office has such teams targeting biases in the UM/HadGEMx models. Here, we can do something similar focusing on improving the ACCESS model performance.
- Workshop registration and call for abstract submission will open at the end of June or early July.
2. Shared experiments/ACCESS-NRI resources
- ESM project
lg87
usage unchanged from last meeting. Dietmar can no longer utilise the extra compute resources, so they are being redistributed to other working groups. - Any proposals for shared experiments for next quarter? See guidelines for how to do this.
3. Science talk
@nicolamaher: “Using large ensembles to constrain temperature variability”
SMILE = Single Model Initial-Condition Large Ensemble
Value of SMILES comes quantifying forced response and internal variability. Can compare internal variability and forced response between SMILES.
Want to spruik involved with: MMLEAv2
- New regridded archive of SMILES. Will be released an NCAR soon, already available at NCI.
- 16 models & 11 variables
- 2.5 x 2.5 degree common grid
- Combination of CMIP5/6 forcing
- Allows for easy initial analysis and interannual model comparison of SMILEs
Summary:
- Rank histograms applied to large ensembles are a useful tool to evaluate model’s forced response and variability
- When evaluating over IPCC regions in D]F/|JA find some models generally perform better than others
- DIF is better simulated than JJA - NAO? Summer processes harder to model?
- Regions such as North America / Europe evaluate well for most models, while regions such as
- Africa and South-East Asia are poorly represented in most models
- Constraint lowers the range of proiections
- Constraint lowers present day variability in models, but can either increase or decrease change in variability depending on whether detrended constraint is used
- In general model agreement on the sign of the change even with the constraint on variability proiections is poor - similar to the IPCC recent assessment
Overall: constraint can be used for present day, but does not improve agreement in proiections
Quality and coverage of observations are a limitation for this method.
Archive anything over 10 members. This study limited to ensembles > 20.
4. What’s going on?
@harun there is a membership opportunity for a WCRP Core Project: Earth System Modelling and Observations. See the link below (or contact me if you need more info): OpenCall-SSG-2024.pdf - Google Drive
Gives opportunity to mix and collaborate with many scientists around the world.
This is a wiki post , if you want to update any of the details in these meeting notes, or to add your own recollection of what was discussed then edit me rather than replying.