search

plesager / ece3-postproc

Suite of processing tools for EC-Earth3 output
5 stars 8 forks source link

fix timeseries plots #15

Closed etiennesky closed 6 years ago

etiennesky commented 6 years ago

currently timeseries plots are unusable: for the atmosphere, years overlap instead of being sequential, and nemo plots are just unreadable.

plesager commented 6 years ago

I have not tried to run the timeseries in a very long time. If I remember, the nemo part is very much a subset of barakuda. I suspect nobody is using it? Initially t does look like a good idea to quickly get some usual timeseries plot, but (I guess) people have their own plot routine that are more flexible (e.g. comparing several experiments in one plot, ....)

etiennesky commented 6 years ago

I agree that the nemo part might not be so important (since barakuda does it better), bu the atmosphere plots are interesting and is quick.

oloapinivad commented 6 years ago

I agree that the Nemo part is deprecated now since barakuda is a comprehensive extension of it.

However as long as I remember at ISAC we have a working version, at least for the atmospheric part. If you let me a few days I will talk with @abalanzino that has run it for a while, we can try to pull our local version (not in GitHub yet).

etiennesky commented 6 years ago

Attaching here plots of atmosphere (msl) and ocean (sst) to illustrate the problem.

msl_a0wy

sosstsst_a0wy

etiennesky commented 6 years ago

Although nemo is deprecated, Barakuda is not so easy to setup so if we could have some key variables plotted it would be helpful.

abalanzino commented 6 years ago

Paolo, I confirm we have a working version of timeseries!

oloapinivad commented 6 years ago

@abalanzino worked on this last week, we are almost ready to pull our updated version. We added a -u USERexp flag also for this.

Some examples:

schermata 2018-03-08 alle 14 02 36 schermata 2018-03-08 alle 14 02 25
etiennesky commented 6 years ago

brillant! nice work!

etiennesky commented 6 years ago

brilliant! with your changes I can get the atmospheric and ocean timeseries.

The only problem is the AMOC value which are 0, but this is because somehow cdfmoc sets depthw to 0 on all levels... any idea? I am using cdftools-3.1

etiennesky commented 6 years ago

I am still having problems with the maxmoc, cdfmoc sets depthw to 0 .

I tried both cdftools 3.0.1 and 3.0.2. What version are you using @abalanzino @oloapinivad ? Thanks!

abalanzino commented 6 years ago

We are using cdftools_3.0.

plesager commented 6 years ago

Since nothing is documented, and I'm not in the mood to read more code today, let me ask: 1) do we expect plots or just netcdf files? I only have the later and no hint in the log about the former. 2) I have a lot of problem warnings:

2a) The program seems confused by the year range. I keep reading while processing 1950-1969:

PROBLEM: we cannot find /nfs/home/users/sager/EC-Earth/diag//timeseries/n_05/ocean/n_05_1951_1969_time-series_ocean.nc !!!

2b) and is clueless about amoc:

Doing AMOC *** File /lustre3/projects/CMIP6/sager/rundirs/n05/post/mon/Post????/n_05_1969_moc.nc not here! Skipping variable zomsfatl !

etiennesky commented 6 years ago

Hi @plesager

  1. there is a small netcdf file, some svg plots and a index.html 2a). no clue... 2b) hiresclim2 has to create the *moc.nc file which is currently in the extras section

I have some fixes coming I will let you know.

etiennesky commented 6 years ago

If you merge my latest MR moc is computed in nemo_post.sh even without nemo_extra=1 .

Somehow cdftools does not read properly the depthw variable in mesh_zgr.nc supplied with this tool , resulting in a 0 MOC. With cdftools-3.0.1 I had to do the following command in the DATADIR:

ncrename -v gdepw_0,gdepw_0_bak mesh_zgr.nc

@plesager I suggest you delete you post and diags and re-run hiresclim2 and then timeseries.

timeseries, like barakuda, requires you to delete the timeseries files it creates if something goes wrong.

In any case I think the log you mention in 2a) is not really a bug, it just means it didn't find the timeseries file so it starts from 0.

With this latest MR I am able to plot atmosphere and ocean timeseries.

abalanzino commented 6 years ago

We had no problem with AMOC values (we are using cdftools release 3.0). I'm sorry, but we didn't write the timeseries code. It results from an original code of Laurent Brodeaux (a code which had to be adjusted).

plesager commented 6 years ago

Got it to work on Rhino (see dcca94e, which also removes a lot of white space in the code and the log). The bug: it assumed there was no underscore in the EXPID. Which I had in my experiments. On the other hand, no problem with cdftools-4 to get the AMOC. Still need to test on cca.

plesager commented 6 years ago

It is working on cca. Bug fix (unbound BASE_YEAR_INI), and easy check for success at the CLI have been committed (be6694b) along the cca config.