Hi all,
At present a number of future climate experiments run for ACCESS-OM2-01 with repeat year forcing + warming trend surface forcing exist. The obvious problem is these experiments miss a lot of important variability in the ocean/climate system.
I’ve raised this idea of doing some ACCESS-OM2-01 runs with “full” future climate surface forcing with COSIMA people at IMAS and seen some interest. I appreciate there’s major issues around experiment design to resolve before we can actually run anything (e.g. variability in surface properties between different CMIP models is rather wide, so do we just choose one CMIP model for surface forcing and accept it might not be realistic; do we do IAF forcing plus warming trend or is there any other options?) but given if we go ahead with this it’ll be a fairly large set of experiments I thought it best to gauge interest in the COSIMA community at large before getting tied down in details.
So, if interested please post here, including some idea of what you would want to use any such model runs for.
Interesting idea. Given the large internal variability, do you think you’d need to run a set of simulations forced with different ensemble members from the same coupled climate model, and/or a set of runs each with forcing from different models?
Thanks for getting this conversation started, interesting idea! Just to mention that there’s a few existing discussions on using OM2-01 for new runs that it may be helpful to be aware of, namely:
So if you do go ahead, feel free to reach out and we can liaise on coordination/best-practice. In general, we are excited to see scientific applications/extensions of OM2-01 using ACCESS-NRI infrastructure.
Hi @Christopher_Roach we had a bit of a crack at such experiments in this study: Abyssal ocean overturning slowdown and warming driven by Antarctic meltwater | Nature
My thinking here was that we wanted to be guided by CMIP multi-model mean (MMM) climate trends, but that anything we do needed to retain storminess, weather-scale variability etc. Applying the MMM to the ocean model would have resulted in no sub-monthly variability. So we just continued the RYF run, but added RCP8.5 multi-model mean trends in T-air, zonal winds and longwave radiation (we could have included other atmospheric variables, but thought best to keep it simple and focus on the main known trending variables). We also added in Antarctic meltwater as we felt that was a critical missing piece of CMIP projections. Which it is.
With Kaitlin Naughten some years back we ended up using a single model (using daily fields), but the downside there is you are wedded to one model’s view of climate change.
I’d be interested to discuss options at a COSIMA meeting sometime, or via zoom with a smaller group of interested people. Cheers, Matt
Hi Matt,
I’m well aware of those future climate+RYF runs (grabbed 'em off Hannah and reran bits of 'em to get daily fields for lagrangian particle tracking). And I was thinking something like that setup but MMM+IAF instead of MMM+RYF… Of cause, that has the problem of assuming variability under historic and future climate will remain similar.
-Chris
Hi Adele,
In the ideal world, yeah, that would be good. But I expect that the compute resources wouldn’t be available (particularly if we want to do it for multiple SSPs), so was thinking (as outlined in my reply to Matt below) multi-model mean trend+historical IAF might be a practical solution.
For now mostly wanna scope out how useful such runs would be for the COSIMA community at large and also sort out if it’s worthwhile doing it with OM2-01 or waiting for OM3-01.
For my work, I don’t think multi-model mean CMIP trend+historical IAF would add much value over multi-model mean trend+historical RYF, given the expense. I do think an ensemble of different single model CMIP forcing could be useful to quantify the uncertainty of future projections, but that is likely too expensive to consider.
I think I agree. Well, for my interests / applications at least. One small variation on the “RYF” approach that we considered was a “Repeat Decade Forcing” spin up to get around the IAF looping back from present day to 1958. It helped when reconstructing the historical ocean warming (see Fig 1 in Drivers and distribution of global ocean heat uptake over the last half century | Nature Communications). That could be extended forward as a projection as well. A kind of hybrid between RYF and IAF, although still with issues (just like the RYF, it locks you onto a particular control state that might not be fully ‘typical’ of a climatological mean state).
For my work this is the crunch, we’er increasingly interested in potential changes in variance (we might certainly expect to see increased synoptic activity over the Southern Ocean). Which as far as I can see means just using the full boundary conditions from, say, a CMIP6 projection rather than “current-atmosphere-plus-trend”. Which I suspect would mean having to run a “historical” run from the same coupled model to keep things conststent