This topic contains discussions, questions and thoughts for Session 3 - Breakout room 3 of the CMIP7 Workshop
Time and Date: Tuesday 28th February, 2023 from 2:00 - 3:30 pm
CMIP7 and ACCESS
At breakout meeting: Gab Abramowitz, Harun Rashid, Kelsey Druken, Yiling Liu, Ramona Dalla Pozza, Tim Raupach, Christine Chung, Matt England
Question: should we get involved in CMIP7 and what would be the benefits?
Pros:
- Focus to southern hemisphere processes - community interest in this region.
- Maintenance of climate modelling ability and connectivity with international community.
- Leveraging of international community for testing/verification/publication.
- Contributions to educational program and training next-gen climate researchers in Australia.
- Having a world-class tool to address our own problems (ie southern hemisphere).
- Continuation of in-house knowledge and knowledge transfer to new specialists.
Cons:
- Regional downscaling, not global models, is how decisions regarding impacts get made.
- Other models exist so compelling reasons for investing heavily in ACCESS are needed.
Detailed discussion notes:
- Gab & Matt: Southern hemisphere processes are import. The community is interested in this region.
- Ramona: Regional downscaling is how decisions get made, and other models exist. We still need a compelling reason for investing a lot of effort into ACCESS.
- Gab and Harun: International science perspective: all climate models are biased and a simple representation of a complex system. Need multiple independent estimates. We don’t want the “best performing model” but rather lots of independent models that help us understand uncertainty in our representation of the system.
- Christine: It’s important for AU to maintain a climate modelling ability. It would be a shame to have no model developers in the country, who care about local processes. International centres may not care about these processes e.g. teleconnections.
- Ramona: Could we put more effort into the development for say 5-7 years then do a better CMIP8? Matt: Development effort takes a lot of resources and can be never ending. Where is the best use of resources? E.g. NZ example where they use another model off the shelf, but that has not served the community well in terms of engagement.
- Harun: Model development is ever changing and ongoing. If we only did model development we would miss benchmarking. Gab: CMIP allows for proper testing, you get papers published on your model, you have access to lots of users. Harun: The process allows for a lot of feedbacks.
- Gab: Our small team here would not get at everything that happens in the model. By publishing it, you get leveraging of resources and lots of users who get at a lot more.
- Harun: Comparison to other models is helpful for verification of variables and processes.
- Kelsey: Infrastructure perspective - tier 1 nodes for NCI was the main publishing mode. Similar story for shared resources for community working on common challenges. This tech and publishing lessons have been working so well it’s become the model for other communities in terms of building data science capabilities.
- Gab: Validation tools that already exist allow for broader comparison with global metrics, which are useful for problem solving and give a more global picture. We don’t have the resources to build something like that and we should use the resources that others have built.
- Matt: Why CMIP7 is good for the AU community: we don’t realise with CMIP submissions that it trains graduates and next gen climate scientists. We would bring in more climate science activity and help educational programs. At PHD level, when people choose which model to work with, ACCESS being a CMIP model is a drawcard.
- Harun: a competitor is WRF. Having a developer team here allows for in-house experience that allows for education of new users.
- Ramona: There are interactivity issues between WRF and ACCESS. The downscaling community has said it’s difficult to use WRF as a regional model to downscale ACCESS outputs. Harder than with other CMIP models. We need a list of priority areas for work. An example is NARCLIM, who would like it to be easier to use ACCESS with WRF. Need to get these conversations going. Noted as an issue that needs to be discussed and solved.
- Christine: What are the relative roles of NRI vs NESP vs other groups for developing CMIP capacity? NESP doing model improvements and NRI during user tutorials?
- Harun: NRI is there for software and infrastructure, needs to provide a flexible modelling framework and a limited number of configs to cater for different ACCESS users. Will not DO development but will participate.
- Kelsey: Short answer: there are fuzzy boundaries. NRI is not funded to do science.
- Gab: NRI does code management (which is difficult, needs central management), testing (needs management and scientific calls), dataset management for all. Australia’s a small community and needs to play together. NRI has authority to manage the system.
- Harun: NESP does research and funds feedback to developers.
- Gab: NESP is application for user-end feedback, NRI is code/model management.
- Ramona: If they don’t understand properly, government will just reassign FTEs to CMIP7. So important to articulate roles.
- Gab: We should get a lot better at knowing what we have and where we are at.
- Matt: If modelling groups want to be risk averse, safest thing to do is make only incremental changes to CMIP6. Challenge for CMIP7 in Australia is deciding how much we decide to do that is new - and we have to understand the risk around what we choose and resources we use.
- Gab: In-house knowledge and risk of loss of knowledge if highly specialised people in a small team leave. If you skipped CMIP7 then there is a risk that the people who are trained (on previous CMIPs) would be retired by the time we get to CMIP8.
Question: what are the key science questions we can address by participating in CMIP7 with an AU model?
Summary of key science questions:
- Antarctic ice melt inclusion.
- Vegetation representation in terms of hydrological impacts and carbon economy.
- Make sure ENSO is represented properly and how does it translate to rainfall.
- Extremes: temp, rainfall, wind, bushfire, etc.
- Best way to get a list would be survey different communities.
Summary of problems to address:
- Gap between model output and decision making.
- Gap between developers and evaluators of models.
- How to demonstrate that ACCESS does the best job for Australia.
Detailed discussion notes:
- Harun: 1. Emissions and international commitments. 2. Bushfires/floods/extremes - how will they change?
- Gab: Carbon economy will become a big thing. Natural emissions, sequestrations - carbon taken up and released by the landscape.
- Ramona: There is a huge gap between model output and on-the-ground decision making. We don’t get funding to fill this gap. Often we get funding to produce more data but not any for translating into decision making. NESP is good from this point of view - mapping the pathway to impact. Collaboration is key (instead of fed/state split of responsibilities). This way can focus the best people on the right tasks.
- Harun: We should focus on one step in the chain. Separate global model and regional models/downscaling.
- Gab: Much of the expertise you need to do CMIP projections is the same as for regional modelling.
- Christine: Two issues: simulation of drivers, and simulation of impacts of drivers. And interaction between the two. It’s complicated.
- Gab: Validation is required for regional-scale modelling as well. One missing piece of infrastructure is evaluation systems for each step in the chain.
- Ramona: National partnership on climate change projections is trying to do this, but it’s a “coalition of the willing”.
- Kelsey: What improvements can be offered over CMIP6?
- Ramona: Christine’s work on rainfall and teleconnections (with NESP) is useful here.
- Christine: Huge spread of model responses is the main issue.
- Matt: Looked at CMIP simulations around the Antarctic shelf - important because warming sets speed of ice-sheet melt. These outputs are used by ice-sheet modellers as forcing. The projections are biased and have problems. By going to higher resolution these problems are fixed. There are gaps in polar processes (e.g. ice sheet melting) that affects global climate responses and tropical processes. CMIP community may not realise this importance. Important to remember that Antarctic is near us and of first-order importance for projections. CMIP participation would allow us to have scientific inputs.
- Gab: Around science, it’s about what we have the capacity to add and what we can do in the timeline.
- Kelsey: From safe to radical approaches, what is the spectrum of options?
- Matt: Community-based responses. Could have two submissions, one risk averse and other more aspirational.
- Harun: Important to start early. Note we got advancements out of CMIP6 over CMIP5 - mostly incremental but some large like the carbon cycle. What to include in CMIP7? Science improvements (parameterisations etc) and perhaps going to slightly higher resolution. E.g. from 1 deg. ocean to 1/4 deg. ocean, and to ~60 km resolution in the atmosphere.
- Gab: Hard to know what changes will result in improvement until you put everything together.
- Harun: Gap between model developers and model evaluators. Need to close this gap - means having usable feedback between groups.
- Kelsey: On problems that can be addressed: what about the hot model problem?
- Harun: This is about climate sensitivity, ACCESS ESM is in the middle of the range of outputs.
- Gab: In most cases, what is diagnosed is change in the model anyway.
- Matt: Global temp is most important for policy.
- Gab: Understanding what the model doesn’t do well is how we go about fixing things. Understanding the landscape with bias etc is how we can understand the ability for the model to make inferences.
- Matt: MIPS bring the scientific community in AU into collaborations and give outcomes for Australia. - Gab: Participating can leverage other scientific advances into our own modelling system.
- Matt: There is a whole community who will use the model output but won’t contribute to the science in the model.
- Ramona: If we are saying ACCESS is the best for Australia we have to demonstrate it.
- Harun: We have an existing system for validation and showing performance.
1 Like
