Support ERA5 forcing

[aekiss]

This issue is a continuation of an email discussion on making configurations that support ERA5 forcing, e.g. to assess the impact of the forcing dataset on the sea ice simulation.

ERA5 is available on NCI at /g/data/rt52: https://opus.nci.org.au/display/ERA5/ERA5+Community+Home ~[edit: use /g/data/ik11/inputs/ERA5 instead]~

• the hourly single-level data is probably what we want. This appears to be from the "reanalysis" or "HRES" dataset (as opposed to the lower-resolution "ensemble" or "EDA" dataset).
• It's 1-hourly 0.25° data (higher resolution than JRA55-do, which is 3-hourly 0.5625°), though this has been interpolated from the reanalysis grid, which is 0.28125° (31km) for HRES.
• It's currently only from 1979 onwards but data from 1950 will be available soon (cf. 1958 for JRA55-do).
• JRA55-do has a lot of adjustments relative to JRA55 to reduce drift. ERA5 doesn't have these, so we'd expect greater drift in long climate simulations. But it may be superior to JRA55-do for shorter timescales.
• for more detail see the ECMWF ERA5 documentation and download page

Replacing JRA55-do with ERA5 would require

• [ ] new grid remapping weights files (see here)
• [ ] making sure that we have physically equivalent fields (height and variable)
• [ ] discussion on whether ERA5 fields should be used for initial conditions, e.g. for SST or sea ice concentration (SIC)

and changes to

• [ ] config.yaml
• [ ] atmosphere/forcing.json
• [ ] namcouple

and possibly more, e.g.

• [ ] coming up with liquid runoff data (unless we'd be happy with JRA55-do for that)
• [ ] parameter changes for bulk formulas or whatnot.

Our configurations currently support only JRA55-do forcing: https://github.com/COSIMA/access-om2/tree/master/control but we have some old, unsupported CORE configs here that may be useful as a reference for the changes required https://github.com/COSIMA/1deg_core_nyf https://github.com/COSIMA/025deg_core2_nyf https://github.com/COSIMA/025deg_core_nyf

[aidanheerdegen]

It would be great to get more eyes on this, in particular on the sea ice. In principle, there is nothing standing in the way of equivalent runs being setup at 1/4-degree (and even 1/10-degree), in addition to the RYF runs I've already done.

Chuck up a forum topic and ask for feedback? I personally prefer the way nbviewer renders notebooks, so if it is useful:

https://nbviewer.org/github/rmholmes/cosima-scripts/blob/master/ERA-5/ERA-5_Initial_Analysis-IAF.ipynb

If you do put up a topic, excerpting out a few plots for a bit of eye candy might increase interest/uptake.

In principle, there is nothing standing in the way of equivalent runs being setup at 1/4-degree (and even 1/10-degree), in addition to the RYF runs I've already done.

You're still using the ~winds~ temp and relative humidity at the "wrong" height though, right? Do you want to remedy that before burning resources on larger models like the 1/4 and 1/10th?

Edit: fixed incorrect assertion about wind forcing to reduce possibility of error propagation. See below and thanks @aekiss

[aekiss]

@rmholmes can you add me to the NCI issue ticket? I raised the missing 1979 data almost a year ago (HELP-182126, 4 April 2022) and was told it was an upstream issue, but I'd have thought it would have been fixed upstream by now.

[rmholmes]

@aekiss oh sorry I hadn't seen that you'd asked exactly the same question. I've added you to the ticket and sent another message.

[aekiss]

ERA5 winds are already at the right height (10m), but temp and humidity (from dew point) are at 2m and ideally should be moved to 10m to match. But this doesn't make much difference for temperature; not sure about humidity. There are other outstanding issues too - I'll pull together a summary.

[aekiss]

Here's my attempt at summarising the remaining issues from this crazy-long thread:

1. Compare ERA5 and JRA55-do fields we use to drive the model as was done here to check we have the correct fields and how much they differ
2. double-check conversion of dew point to humidity implemented in this commit - compare to JRA55-do humidity
3. move temperature and humidity from 2m to 10m to match the winds and JRA55-do; but doesn't make much difference for temperature; not sure about humidity
4. Check rain and snowfall - see here
5. remapping issue resolved for 1deg but will probably need to be resolved for 1/10deg runs; there's also a runoff remapping issue at 1/4deg
6. ~missing data for hours 0-6 on 1 Jan 1979 - NCI ticket HELP-187454 (previously HELP-182126)~ - THIS IS NOW SOLVED.
7. NCAR or ECMWF bulk formula?
8. back-compatibility in libaccessom2
9. Unresolved issue with caching code in the ERA-5 libaccessom2 branch.
10. Be aware of several bad input data times/points in ERA-5 winds

[rmholmes]

Thanks @aekiss!!! I just added one extra point on the caching code.

I think it's also important to continue to make a value judgement on additional work in this direction given the possibility of a ERA-5-do product in the future, and the move to ACCESS-OM3/NuOPC framework. But I'm happy that we at least pushed it to the point of something that looks reasonable.

[aekiss]

Yes, it has certainly been more epic than we expected. Hopefully life will be easier in ACCESS-OM3 with DATM and DROF instead of YATM - they support ERA5, and read and remap in parallel.

[aekiss]

@rmholmes can you push your test run as a branch on my config https://github.com/COSIMA/1deg_era5_iaf pls?

[rmholmes]

Yes, but looks like I don't have the permissions to do that? The branch is at https://github.com/rmholmes/1deg_era5_iaf/tree/ryan_testing, or give me permissions?

I should probably also create repositories on COSIMA for https://github.com/rmholmes/1deg_era5_ryf and https://github.com/rmholmes/025deg_era5_ryf?

Finally, given Will's comments this morning it seems like it would be low hanging fruit and productive to setup a 025deg_era5_iaf with JRA55 v1.5 run-off. I can try to do that this afternoon or tomorrow. Is this the right run to take v1.5 forcing from, or this one?

[access-hive-bot]

This issue has been mentioned on ACCESS Hive Community Forum. There might be relevant details there:

https://forum.access-hive.org.au/t/cosima-working-group-meeting-minutes-2023/407/9

[aekiss]

OK I've copied rmholmes/1deg_era5_iaf/tree/ryan_testing here https://github.com/COSIMA/1deg_era5_iaf/tree/ryan_testing

For https://github.com/rmholmes/1deg_era5_ryf and https://github.com/rmholmes/025deg_era5_ryf you can see if it will let you transfer ownership to COSIMA (settings > general > transfer ownership)

atmosphere/forcing.json and forcing in config.yaml look the same in both of those, i.e. how I set them up back in May https://github.com/COSIMA/access-om2/issues/247#issuecomment-1118436234

[aekiss]

Thanks for transferring ownership. This one's private - is that intentional? https://github.com/COSIMA/1deg_era5_ryf

[rmholmes]

Ooops. Fixed it.

I've also setup a 1/4-degree IAF ERA5 (with JRA55 v1.5.0 run-off) (https://github.com/rmholmes/025deg_era5_iaf) and a corresponding JRA55 v1.5.0 run (https://github.com/rmholmes/025deg_jra55_iaf/tree/jra55v150), both to start in 1980. There's no hours to test these now but I'll kick them along at some stage.

[aekiss]

Re. point 1 above, ACCESS-OM3 is using the DATM data atmosphere from CDEPS, which supports ERA5 out of the box. So that might teach us something useful about what fields we should use.

This table suggests the ERA5 input fields are

u10
v10
t2m
skt
d2m
msl
tp    *
cp    *
lsp   *
csf   *
lsf   *
ssrd  *
ssr   *
strd  *
str   *
aluvp
aluvd
alnip
alnid
sshf  *
slhf
ewss
nsss

* are not in the list of ERA5 fields

The ERA5 DATM code is here and exports these fields to the other model components

z
u10m
v10m
wspd10m
t2m
tskn
q2m
pslv
rain
rainc
rainl
snowc
snowl
swvdr
swvdf
swndr
swndf
swdn
swnet
lwdn
lwnet
sen
lat
taux
tauy

[aekiss]

The list of ERA5 input fields in DATM seems to omit visible radiation - or is that included in the * fields?

[ofa001]

Some of those terms in your list @aekiss for the DATM look like a mirror of what the CESM would have been passing so they have converted the ERA5 into CESM look a like so it can run their shortwave scheme for instance over CICE.

[ofa001]

@aekiss The visible radiation is in the shortwave radiation fields swvdr swvdf

[aekiss]

Ah OK - might have to look at the code to see how shortwave is calculated from the input files

[ofa001]

Yes @aekiss I think I need to look at the existing DATM code as well I have only looked at the DOCN and DICE boxes so far and how that fitted with CICE. Like my conversation with Kieran Ricardo the other day the way NUOPC is set up is still new to us and how we thought about things linked together before may need rethinking I guess we need to think how the UM will link to an ACCESS_OM3 MOM/CICE/WW3 as well as the JRA55 and ERA set ups in the data model.

[rmholmes]

I did a quick check of the radiation and precipitation variables (points 1 and 4 at https://github.com/COSIMA/access-om2/issues/242#issuecomment-1487906161). Just for a single day, but looking reasonable to me:

[StephenGriffies]

@rmholmes and others.

Your experiences with ERA5 and JRA55 will be valuable as NCAR, GFDL, and interested others start the process of developing an "ERA5-do". I hope we can count on you for testing new forcing fields when they are ready. Gokhan is in charge at NCAR. He hopes to have something for beta-testing early 2024.

[aekiss]

Great to hear ERA5-do is going ahead @StephenGriffies. I'll be interested in testing this in ACCESS-OM3 when it's ready too.

[aekiss]

Thanks for looking at those field @rmholmes - I agree they don't look very different.

My worry was whether exactly the same things are included in both cases, e.g. the same wavelength bands and diffuse/direct for shortwave, or the same precip types. This might only show up on maybe a 10% level, which is harder to test as the fields might differ by this amount between JRA55 and ERA5 anyway even if they're trying to show exactly the same things. Might require diving into documentation.

[aekiss]

ERA5 has a list of known issues here https://confluence.ecmwf.int/display/CKB/ERA5%3A+data+documentation#ERA5:datadocumentation-Knownissues including bad 10m winds which have tripped up the 0.25° model as discussed here

[aekiss]

NCI also has a table of ERA5 data issues here https://opus.nci.org.au/display/ERA5/Known+Issues

[rmholmes]

I have now finished a 1/4-degree ERA-5 forced IAF run, covering the period from 1980 up to the latest available forcing time of 31st of March 2023. The main problem I ran into is blow-ups under unrealistically large ERA-5 winds (up to 130ms-1) as described in this ACCESS-OM2 issue and this ECMWF post. Getting past them involved scaling down the strength of the wind fields using the offset/scaling system in libaccessom2 and this notebook.

A quick look at the output is looking promising (IAF runs, and RYF runs). I hope to do a more detailed analysis and will put up a forum post at some point asking for input from anyone who is interested.

[aekiss]

Awesome, thanks @rmholmes!

[access-hive-bot]

This issue has been mentioned on ACCESS Hive Community Forum. There might be relevant details there:

https://forum.access-hive.org.au/t/era-5-forced-access-om2-simulations/1103/1

[access-hive-bot]

This issue has been mentioned on ACCESS Hive Community Forum. There might be relevant details there:

https://forum.access-hive.org.au/t/jra-3q-as-a-replacement-for-jra55/1759/8