The ACCESS-NRI Scientific Advisory Committee is developing an ACCESS Roadmap for the next 10 years. To help guide recommendations on which model development and support should be prioritised, we are starting by thinking about the most important science we want ACCESS to enable over the next decade.
@pearseb and I have drafted the following priorities as a starting point:
We would love your feedback / input on this to take to the Scientific Advisory Committee. We are just as interested in what should be a lower priority as what should be a high priority. The danger of listing everything we could possibly want is that the most important capabilities may end up being less well supported.
We also encourage you to think beyond the current COSIMA community. Are there emerging science priorities, or communities not yet well supported by ACCESS-NRI, that we should be considering?
In particular, we would love your thoughts on:
Do the proposed science priorities capture the main science COSIMA should be enabling in 2036?
What, if anything, is missing or should be changed in the science priorities?
Are the proposed required model capabilities the right capabilities to enable these priorities? What is missing or unnecessary?
Is there anything listed as ‘nice to have’ that you think should be essential, or vice versa?
Feel free to start discussion below, or you can provide feedback privately through this survey link. Please provide feedback by Friday 28th August.
Are the changes in surface waves such high priority (4th item)? I would think it’s not that much of a priority. The fact that they don’t seem to appear in the required model capabilities but in the “nice-to-have” ones seems to agree with that?
Do we want to mention something in the capabilities in the spirit of: ability to utilise modern HPC hardware (like GPUs)?
It is good to see the mention of improved parameterizations and tuning. This is where I see a significant gap in the Australian community, at least outside of the Antarctic margin. I would love to see more of an emphasis on taking the latest and greatest parameterizations, implementing them, and doing the detailed and difficult work of verification and tuning. It might not lead to many papers, but that’s the work that will actually result in a better representation of ocean processes around Australia and in the mid-latitudes and tropics, and thus better climate projections and better science.
10 years is a long time horizon, so I’m wondering if we should look further beyond current capabilities?
Here are some possibilities that I think would be technically feasible, though probably only a subset would be realistic within resource constraints:
The NUOPC coupler is very flexible and could be leveraged more than we currently do. For example:
“Plug and Play” modelling: Consider all ACCESS models (e.g. ACCESS-OM3) as a special case of a universal “ACCESS-everything” model (ACCESS-ESM3+waves+ice sheets), for which any active component can be replaced by a prescribed data model (e.g. JRA55-do instead of UM) or nothing (e.g. waves or BGC turned off). We already do this to some extent (using CDEPS DATM and DROF in OM3 in place of UM in CM3), but this could be extended to allow DOCN instead of MOM6, DICE instead of CICE, DWAV instead of WW3, etc. This approach would permit stepwise construction/deconstruction of a full ESM, for which all other models (e.g. ACCESS-OM3) are considered subsets. This approach would allow
exploration of the impacts of feedbacks between model components, including more/fewer/different components than we currently have in OM3 and CM3
stripped-down partly-coupled models (e.g. active WW3-CICE with DOCN data ocean) to explore particular feedbacks in isolation, and also more efficiently debug and tune components
user-friendly ways to use the ability of CDEPS and the CMEPS mediator to combine and modify data streams and coupled fields, to facilitate perturbation experiments and explorations of the impact of modified feedbacks between any components, e.g.
modifying coupling fields between any components, e.g. switch between two-way and unidirectional coupling, or have coupling which is amplified, attenuated, spatially or temporally filtered, or perturbed in some other way, e.g. blended with climatology
separate CICE instances for NH and SH in global ACCESS-OM3 to permit independent parameters in the two hemispheres
NUOPC cap for Oceananigans, so we can optionally use it in OM3 in place of MOM6
exchange grids for more accurate coupling between components of different resolution (e.g. 8km OM3 with coarser WW3)
explicit icebergs, with grounding and fast sea ice
ocean coupling to dynamic ice sheet
wetting/drying in MOM for shoreline change, e.g. in regional models or sea level rise
other forcing datasets (e.g. CFORCE)
support other BGC models (or even ecosystem models) besides WOMBAT variants
offline BGC for rapid spinup and cheap experiments
Anderson acceleration
atmospheric boundary layer model as halfway point between prescribed forcing and coupling to a full model (UM)
data assimilation? (or is ACCESS-NRI not allowed to work on that?)
AI / ML applications?
Obviously this is an overly long list, but maybe some well-considered subset of these would have a worthwhile payoff in opening up new areas of research. It would need to be driven by scientific priorities.
Thanks for putting this together and providing the opportunity for us to chip in.
As mentioned in today’s COSIMA meeting, I think there is value in mentioning coupling to the atmosphere, even though this list is obviously focusing on the ocean/sea ice. I suggest to add a third science priority highlighting the need to understand the ocean’s role in the coupled climate system, cross-cutting with the ESM WG. Regions of priority should be (i) Australia and (ii) pan-Antarctic.
