Experiment Proposal: ACCESS-ESM1.5 Pacific Warming AMIP Experiments II

Experiment title :bell:: ACCESS-ESM1.5 Pacific Warming AMIP Experiments

Summary :bell::
Following our recent Pacific zonal SST warming pattern AMIP experiments in ACCESS ESM1.5, we plan to run a set of new experiments with the hemispheric warming uncertainty and its interactions with plausible zonal warming patterns. These experiments will provide a complete understanding of the role of Pacific warming uncertainty in the future climate.

Scientific motivation:

Observed SST trends over recent decades indicate that the tropical western Pacific warmed more than the eastern Pacific, leading to a strengthened equatorial zonal SST gradient (Lee et al. 2022). In contrast, the majority of CMIP models simulate a reduced zonal SST gradient, with more warming in the eastern Pacific (Cai et al. 2021; Wills et al. 2022; Bai et al. 2023). This discrepancy represents one of the major sources of uncertainty in future tropical climate projections as global atmospheric circulation and regional hydroclimate strongly depend on Pacific variability.

Resolving the causes of this disagreement between observation and climate models will require continued advances in the models and improved understanding of the relative roles of externally forced change and internal climate variability on future Pacific warming pattern evolution. However, an important and more immediate question is how uncertainty in the Pacific warming pattern translates into uncertainty in future rainfall projections. Quantifying this relationship would provide valuable information for policymakers by identifying the extent to which regional hydroclimate projections depend on plausible alternative Pacific warming scenarios.

To address this question, we will perform a set of atmosphere-only ACCESS-ESM1.5 experiments by varying the tropical Pacific warming pattern from the observed La Niña-like trend to the El Niño-like warming projected by many CMIP6 models. In addition, we will investigate uncertainty associated with hemispheric warming asymmetry and its interaction with alternative Pacific zonal warming patterns. Together, these experiments will quantify the climate impacts of plausible future Pacific warming states and provide new insight into the physical drivers of uncertainty in global and regional hydroclimate projections under climate change.

Experiment Name :bell:: pac_warm_amip_esm1.5
People :bell:: Abhik Santra, in collaboration with Shayne McGregor and others.
Model: ACCESS-ESM1.5
Configuration: AMIP-style simulations with prescribed SST perturbations
Initial conditions: /g/data/k10/sza565/access-esm/restart/restart_1994
Run plan: 21 years per experiment, with 5 ensemble members for each experiment.
Simulation details: A set of sensitivity experiments will be conducted using prescribed SST patterns with varying meridional SST gradients. Ideally, four distinct gradient magnitudes will be tested in each case. In addition to that, a combination of both zonal and meridional gradient experiments will be conducted to better understand the contribution of the predominant modes of the Pacific warming pattern uncertainty.
Total KSUs required :bell:: Based on our earlier CPU usage, we need ~1.5 MSU (15 experiments x 5 ensemble members x 21 years) in Q3 2026.
Total storage required :bell:: 10.5 TB (15 experiments x 680GB) on gdata/lg87.
Storage lifetime :bell:: next 2 years, at least.
Long term data plan :bell:: model outputs will be transferred to MDSS (tape).
Outputs: standard monthly variables will be stored on /g/data/lg87/sza565/Pac_Warming_expt/
Restarts: will be available every 5 years.
Related articles:

Analysis:
To be updated.

Conclusion:

Quantifying the rainfall uncertainty associated with Pacific warming pattern uncertainty would provide valuable information for policymakers by identifying the extent to which regional hydroclimate projections depend on plausible alternative Pacific warming scenarios, thereby improving the interpretation and use of climate projections for adaptation and risk assessment.

Hi @abhik, thank you for submitting this experiment proposal. I’ve discussed with the co-chairs a little and we would be happy for you to use the proposed SU to run the experiments.

The required storage would take /g/data/lg87 quite close to its quota however, and we would like to discuss whether it would be possible to reduce the required storage. Will all of the atmospheric variables produced by the configuration be required for your analysis? If not, we’d be happy work with you to customise the STASHC file and cut out some of the excess ones.

Thanks,
Spencer

Hi @spencerwong, thanks for approving the experiment proposal.

We currently archive only the monthly and daily output variables from our simulations, while all sub-daily output files are deleted after the run.

I agree that the model output size could be further reduced by customising the STASH configuration to retain only a selective set of monthly and daily variables. Your help on this would be greatly appreciated.

Thanks, Abhik

Hi @abhik, that sounds like a good plan. We can also remove the sub-daily outputs from the STASHC so that they won’t have to be deleted later on.

When you get some time, would you be happy to look through some sample atmospheric output files produced by your configuration, and send me a list of the variables you would like to keep? I can then assist with modifying the STASHC file.

Thanks,
Spencer

Hi @spencerwong, I have prepared a list of required monthly and daily variables from the AMIP experiments in the attached PDF file. Could you please update the STASH accordingly and let me know?
Thanks, Abhik.
AMIP.ESM1.5.Monthly.Daily.Output.Variables.pdf (149.2 KB)

Hi @abhik, apologies for the delay in getting back with an updated STASHC. I’ve set up a STASHC file based on the listed variables and added a copy here. I’d recommend doing a test run to confirm it contains all the required outputs. For the monthly maximum and minimum air temperature at 1.5m, it would be easiest to calculate it from the daily maximum and minimum values.

With the changes, the configuration will produce 4.8G of output per year, and so a total of ~7.5TB. Please note that 4GB of that output comes from the daily 3D variables.

Let me know if you have any questions about adding the file to the configuration

Hi @spencerwong, thank you for your help with the output variables. I can see that the previous STASH configuration generates around 6.5G of monthly and daily data per year in netCDF format. Although your STASHC configuration outputs only 78 variables, the required storage space remains high.

Could you clarify whether the storage requirements were estimated based on the original pp files or after netCDF convertion?

If the estimates are based on the pp files, would it be possible to convert the pp output directly to netCDF? Earlier, you provided a script for offline pp-to-nc conversion. Having a similar conversion at runtime would make data handling and storage management more convenient.