search

ESMValGroup / ESMValCore

ESMValCore: A community tool for pre-processing data from Earth system models in CMIP and running analysis scripts.
https://www.esmvaltool.org
Apache License 2.0
42 stars 38 forks source link

non-fixed "fx" files - causes discontinuity in timeseries between experiments #1474

Open ChrisJones-MOHC opened 2 years ago

ChrisJones-MOHC commented 2 years ago

I have discovered that not all “fx” files are constant for a given model on ESGF. Specific example is for sftlf (land fraction) for NorESM2-LM model data below. This leads to discontinuity in global timeseries if sftlf is used to weight variables such as land carbon stores or fluxes

This is not a ESMVal error, but a danger with data which is useful for users to be aware of

Example – I have not examined all occurrences of sftlf but browsed a few for models where I have spotted an issue. The following for example show that NorESM2-LM has at least 3 different sftlf files. Pasted below are the filenames (including version dates) and the global average. (CNRM-ESM2-1/piControl/r1i1p1f2/fx/sftlf/gr/v20181115/sftlf_fx_CNRM-ESM2-1_piControl_r1i1p1f2_gr.nc also appears to differ from other CNRM-ESM2-1 files)

CMIP6/CMIP/NCC/NorESM2-LM/historical/r2i1p1f1/fx/sftlf/gn/v20190920/sftlf_fx_NorESM2-LM_historical_r2i1p1f1_gn.nc 28.340188318499596 CMIP6/CMIP/NCC/NorESM2-LM/historical/r3i1p1f1/fx/sftlf/gn/v20190920/sftlf_fx_NorESM2-LM_historical_r3i1p1f1_gn.nc 28.340188318499596 CMIP6/CMIP/NCC/NorESM2-LM/piControl/r1i1p1f1/fx/sftlf/gn/v20210118/sftlf_fx_NorESM2-LM_piControl_r1i1p1f1_gn.nc 28.340188318499596 CMIP6/CMIP/NCC/NorESM2-LM/historical/r2i1p1f1/fx/sftlf/gn/v20190920/sftlf_fx_NorESM2-LM_historical_r2i1p1f1_gn.nc 28.340188318499596 CMIP6/CMIP/NCC/NorESM2-LM/historical/r3i1p1f1/fx/sftlf/gn/v20190920/sftlf_fx_NorESM2-LM_historical_r3i1p1f1_gn.nc 28.340188318499596 CMIP6/CMIP/NCC/NorESM2-LM/piControl/r1i1p1f1/fx/sftlf/gn/v20210118/sftlf_fx_NorESM2-LM_piControl_r1i1p1f1_gn.nc 28.340188318499596

CMIP6/CMIP/NCC/NorESM2-LM/1pctCO2/r1i1p1f1/fx/sftlf/gn/v20190815/sftlf_fx_NorESM2-LM_1pctCO2_r1i1p1f1_gn.nc 28.34260542007538 CMIP6/CMIP/NCC/NorESM2-LM/historical/r1i1p1f1/fx/sftlf/gn/v20190815/sftlf_fx_NorESM2-LM_historical_r1i1p1f1_gn.nc 28.34260542007538 CMIP6/CMIP/NCC/NorESM2-LM/1pctCO2/r1i1p1f1/fx/sftlf/gn/v20190815/sftlf_fx_NorESM2-LM_1pctCO2_r1i1p1f1_gn.nc 28.34260542007538

CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp126/r1i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp126_r1i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp370/r1i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp370_r1i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp370/r2i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp370_r2i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp370/r3i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp370_r3i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp126/r1i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp126_r1i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp370/r1i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp370_r1i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp370/r2i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp370_r2i1p1f1_gn.nc 29.352518609708902 CMIP6/ScenarioMIP/NCC/NorESM2-LM/ssp370/r3i1p1f1/fx/sftlf/gn/v20191108/sftlf_fx_NorESM2-LM_ssp370_r3i1p1f1_gn.nc 29.352518609708902

bouweandela commented 2 years ago

Thank you for opening the issue @ChrisJones-MOHC! I have transferred this issue to the ESMValCore repository because that is the tool that reads the recipes and finds the corresponding data.

zklaus commented 2 years ago

Thank. Please also note the following erratum: https://errata.es-doc.org/static/view.html?uid=0b9d744d-be7c-0152-89fc-ca64ab2dbe40

To quote:

all the sftlf and sftgif data that have version numbers v20191108 and earlier should use those from the abrupt-4xCO2 experiment of the LM and MM versions respectively.

ChrisJones-MOHC commented 2 years ago

thanks Klaus - useful to know this has been flagged there.

Although it's easier said than done, because unless you know this the tool defaults to using the experiment you've selected. Is it possible to redact bad files rather than leave them in place and hope everyone reads the errata?

zklaus commented 2 years ago

Yes, that is possible. However, it must be done by the modeling groups.

ChrisJones-MOHC commented 2 years ago

understood! I have emailed the contact email in the netcdf files, noresm-ncc@met.no

zklaus commented 2 years ago

Cheers, @ChrisJones-MOHC. Let us know how it goes.

ChrisJones-MOHC commented 2 years ago

update - NorESM-LM will retract incorrect sftlf files. https://github.com/NorESMhub/noresm2cmor/issues/162 refers

in addition - use of the following line in ESMVal pre-processor forces the tool to use a correct (in this case abrupt-4xCO2) file: fx_variables: [{short_name: sftlf, mip: fx, exp: abrupt-4xCO2}]

YanchunHe commented 2 years ago

Hi, just want to mention, seems that the abrupt-4xCO2 by NorESM2-LM is not correct neither, as the whole dataset has been retracted, merged and republished. By anyhow, the republished v20210118 sftlf of abrupt-4xCO2 used a wrong version.

As mentioned in the https://github.com/NorESMhub/noresm2cmor/issues/162 issue, the v20210118 version will be retracted from ESGF, including the abrupt-4xCO2.

A correct version for NorESM2-LM piControl, historical and amip will be published later. And all other dataset that will not be retracted will be correct, hopefully.

rswamina commented 2 years ago

@bouweandela and @zklaus - What is the best way to record the outcome from this discussion on what files to use? Should users facing this problem search the issues or should we be documenting this somewhere?

YanchunHe commented 2 years ago

@bouweandela and @zklaus - What is the best way to record the outcome from this discussion on what files to use? Should users facing this problem search the issues or should we be documenting this somewhere?

Could you leave this thread open, so that we can wait and summarise the discussion, until all relevant datasets are retracted and republished as mentioned in https://github.com/NorESMhub/noresm2cmor/issues/162.

YanchunHe commented 2 years ago

Now the wrong version of sftlf dataset has been retracted, and republished with correct version of v20191108, for NorESM2-LM, experiments, piControl, historical and amip;

The Errata is: https://errata.es-doc.org/static/view.html?uid=cbf20fbf-bed9-e4fb-9f8f-d62573ef5ed0

for ESMVal pr-processor, one can use, e.g., fx_variables: [{short_name: sftlf, mip: fx, exp: piControl}]

We may close this issue now. Thanks for the patience!