Finding CMIP5/PMIP3 & CMIP6/PMIP4 past1000 experiments

Earth System
,
1

I am trying to figure out which of the PMIP3/PMIP4 past1000 experiments have replicas on Gadi. I think there are some on /g/data/al33 but I can’t find any documentation of which models have the past1000 experiments stored.

In the past, I think this could be done using hh5/clef but that doesn’t seem to work from xp65(?) Is there some new version of this? That is, something that will list if (and where) any past1000 experiments are replica-d on Gadi.

And if there are some that are not - do we know if there is a way to request the data?

Thanks,

Georgy

2

Hi @georgyfalster ,

I don’t know much of the current status of things but I might be able to give some information (for you to see if that’s useful information…)

First, you won’t be able to request any new CMIP5 data, so what you can find under al33 is the max you can have. For CMIP6 data, I’m not sure if you can request more data, that’s a question for NCI.

For finding the data, I don’t think clef is supported anymore (if it is I don’t know who supports it). So I went to the source doing a search on ESGF MetaGrid

I searched on (activity_id = PMIP) AND (experiment_id = past1000) (the activity is chosen under the “General” tab and experiment is under “Identifiers”). Once the search is done, the “Source ID” and “Institution” (under identifiers) will show you only the models and institutions that have the past1000 experiment. Using that you can then see if we have the data under /g/data/oi10. It’s a bit slow and manual but there aren’t that many models that have published the past1000 experiment. So it’s manageable.

From a quick look, it doesn’t seem we have much of them under oi10 (didn’t check all). So if you’d like to request some, please contact NCI and ask them what they need you to provide to them if they are ready to download the data.

I hope it helps.

3

Thanks @clairecarouge. It is helpful to know that I should look under oi10.

Around 10 models published PMIP3 past1000 experiments, but as far as I know only four submitted to the PMIP4 past1000 (and at least one of those is potentially problematic - hence why I am still interested in PMIP3!). So I guess it will be a combination of this manual approach and then getting variables in myself where necessary.

Only other question: do you know who at NCI I would ask about data to see if they are willing to at least add any missing PMIP4 experiments? Or just the general NCI helpline?

4

Hi @georgyfalster. Just to add to @clairecarouge’s answer, NCI maintain Intake-ESM datastores for their CMIP5 and CMIP6 replicas and these are included in the ACCESS-NRI Intake Catalog. So you could use that to find (and load) what is available.

From a quick look, according to the Intake-ESM datastores there are no “past1000” experiments in the CMIP6 replica, but there are in the CMIP5 replica:

import intake

cat = intake.cat.access_nri

# cmip6_past1000 is an empty datastore
cmip6_past1000 = cat["cmip6_oi10"].search(experiment_id="past1000")

# cmip5_past1000 is a datastore with 2196 datasets
cmip5_past1000 = cat["cmip5_al33"].search(experiment="past1000")

To find which models are included, you could then do:

cmip5_past1000.unique()["model"]

which returns ['GISS-E2-R', 'IPSL-CM5A-LR', 'MIROC-ESM', 'MPI-ESM-P', 'CCSM4', 'FGOALS-gl', 'bcc-csm1-1', 'MRI-CGCM3'].

You can use the same tool to find out which variables, frequencies etc are available for each model, and to eventually load the data you’re after (if it’s there) - see the NCI docs and Intake-ESM docs.

3 Likes
5

To contact NCI, it is always through the help email. They triage from there.

1 Like