CMIP6 Data Recommendations for Southern Ocean Study

COSIMA Biogeochemistry
, , , ,
1

Hi,
I hope you’re well. This is Rony from Curtin University in Perth, WA. I’m reaching out to seek your advice on selecting a suitable Earth System Model for studying the Southern Ocean.

My area of interest spans from 30°S to 85°S and across all longitudes (-180° to 180°). I’ve heard that the ACCESS model was originally developed with a focus on this region, so I wanted to know if it would be a good fit for my study.

Specifically, I’m looking for CMIP6 model outputs that:

  • Have a spatial resolution of at least 0.25 degrees

  • Provide projections through to the year 2100

  • Include the following variables: SST, MLD, shortwave radiation (SWR), wind speed (WS), and sea level anomaly (SLA)

Could you kindly share your insights or suggest the most suitable models for this purpose? Also, if you could point me to reliable sources or links where I can access such datasets, I’d be very grateful.

Thanks so much for your time and support.

1 Like
2

Hi Rony, Welcome to the ACCESS Community Forum!

I don’t have a specific answer to your question, but am wondering, do you want to run your own experiment, or analyse the output from existing CMIP6 experiments? I think the latter, but wanted to make sure.

1 Like
3

Thanks for your reply. You’re absolutely right — I’m looking to use some of the output from existing CMIP6 experiments.
We’ve developed a machine learning model to predict future phytoplankton dynamics (up to the year 2100), and now we need to feed it with input variables from CMIP6. Our aim is to use the most suitable model output from CMIP6 that aligns with our model’s training inputs as well as the SO region.

4

Ahh, right, well that is interesting. @pearseb might be interested in your plans and give you some hints about data to look for.

I don’t know if you’re familiar with his research, but @tylerrohr22 has done some ML work with ocean BGC, and also did a huge study comparing CMIP6 BGC models so might be someone worth reaching out to.

ACCESS-NRI also maintains an intake catalogue which can be used as a convenient method to programmatically find and access CMIP6 data

and there are some training materials for how to use the catalogue

1 Like
5

Hi Rony. Earlier this year, FishMIP published a paper looking at how well 11 CMIP6 models reproduced past conditions for sea surface temperature, sea ice concentration and phytoplankton biomass at the surface. The evaluation only includes one of your variables of interest, but the paper may still be useful: https://doi.org/10.1029/2024EF004849.

1 Like
6

Thank you so much for your kind reply. I am waiting to receive more insights from them @tylerrohr22 and @pearseb

7

Thanks. I was present at FishMIP workshop last month, had a good discussion with so many biogeo modellers there. Got some insights like they are using GFDL and IPSL. Not sure why they don’t use ACCESS although the region they have used, kind of mix. Anyway, Now, I am looking for the best products for Southern ocean

8

Does the ACCESS model provide ocean data at 1/4-degree resolution? @Aidan

9

There are some 1/4-degree BGC studies, some of which weren’t for CMIP.

I realised this was in the General category, so I’ve moved it to COSIMA > Biogeochemistry hoping it will catch the eyes of some others.

In particular @aekiss, @matthew.chamberlain, @PSpence and @adele-morrison who might be able to steer you in the right direction.

10

Heyya,

Sorry for the late reply.

Do you need the BGC (e.g. Phyto biomass) from the models to train your ML algorithm or are you tuning historical model physics to historical Remote sensing BGC properties then projecting?

The reason ACCESS hasn’t been used as a FishMIP forcing so much is until recently the SO wasn’t iron limited so there were issues with P biomass. That has now largely been fixed, but I don’t think anything higher res than 1 degree haas been run with the new version. But if you just need the physics there are definitely runs around that the others tagged will be able to point you towards.

If you do need the model BGC and DONT need .25 degrees there is a lot of public CMIP6 data available. Much of which was assessed/validated in the paper Denisse linked. So for the SO you could use that as a road map to pick a model. I have post-processed data sets for SSP585 projection of MLD, P-Biomass, SST for all those model too if that is helpful. But not any of the metocean stuff. Earth grid system federation probably has them all hosted publically at 1 degree though (just takes some elbow grease to download and process).

If you do need BGC + .25 degree + future projections that might mean coordinating new runs on ACCESS/NCI or ID’ing another study that has and reaching out to collaborate (ie see if they will share the data).

3 Likes
11

Not sure I can add much to what has already been said. It looks like all you need are physical variables, and given that you want 0.25º, that limits the kinds of models that are available to you in the CMIP6 archive. You will only be selecting from a few. Given that, I’d just take all of them if I were you.

Pearse

2 Likes
12

Thanks so much for the detailed explanation — it was really helpful.

Just to clarify my use case: I’m working with a machine learning model to reconstruct and project chlorophyll-a concentrations in the Southern Ocean. For the reconstruction, I’ve used satellite-derived products, and now I aim to use the same model to project future chlorophyll-a concentrations.

To do that, I need to feed the model with five key physical variables:

  • Sea Surface Temperature (SST)
  • Mixed Layer Depth (MLD)
  • Shortwave Radiation (SWR)
  • Wind Speed (WS)
  • Sea Level Anomaly (SLA)

I don’t require chlorophyll from the models, as I’m already using satellite-based Chl-a for training. What I need are reliable historical and future projections of the above physical drivers, ideally under both high and low emission scenarios.

Currently, I’m using CMIP6 data at 1° resolution, but higher-resolution data (e.g., 0.25°) would be even more beneficial if available.

If there are any ACCESS or other model runs (even non-BGC) that include these variables at higher resolution, I’d greatly appreciate your guidance. I’m also planning to extract MLD, so any help on that would be useful as well.

Thanks again for your support!

13

Thanks again for the advice — I really appreciate it.

Yes, I understand that only a few models provide physical variables at 0.25°, so I’ll definitely try to include all those that are available in the CMIP6 archive. At the same time, I’m still not entirely sure about their suitability for the Southern Ocean region. If 1° resolution data works well for this region, I’m happy to stick with that.

I’m not planning to use all models — I’m considering a maximum of three or four, such as ACCESS, GFDL, IPSL, and UKESM, under both high and low emission scenarios. As far as I know, 1° data is available for all of them, and GFDL might also have 0.25° ocean data.
Rony

14

Hey @ronygolder_curtin. Just to let you know I have marked this outofscope for direct ACCESS-NRI support, but hopefully you have already got some useful information, and if you need any more assistance the community can assist further.

1 Like