Up to Specific issues in real coupled models
Dear All, Take my greetings. I wish to couple the CLM-4 Land surface model with Seoul National University (SNU) AGCM. Both models have there own executable and name list files. As a try I installed the OASIS coupler with pgi compiler and openmpi. After that I did run the tutorial ok. Now how to start to couple of CLM-4 and SNU AGCM ?
Concerning the coupling of your real models, you should have a look to the document : https://oasis.cerfacs.fr/wp-content/uploads/sites/114/2021/08/GLOBC_TR_Maisonnave_cesm_oasis_2011.pdf Then, the best way to start coupling CLM-4 and SNU AGCM is to begin to couple separately each model with a toy model using OASIS3-MCT. To test the coupling OASIS3-MCT/CLM-4, you should adapt the model2 (atmosphere) of the tutorial to your atmosphere grid and coupling fields, compile it and run it with your real land model, CLM-4. To test the coupling OASIS3-MCT/AGCM, you should adapt the model1 or model2 of the tutorial to the grid of your land and to your coupling fields, compile it and run it with your real atmospĥere model AGCM. The different coupling fields as well as the characteristics of the coupling are described in the namcouple file. Be awared that CLM already uses MCT as an internal coupling library. If necessary, we can send to you some tools developped by Eric Maisonnave that rename the MCT modules used by OASIS3-MCT. Best regards, Laure
Hi, I am coupling the Coupled-Ocean-Atmosphere-Wave-Sediment Transport (COAWST) model to the CICE sea ice model. COAWST uses MCT internally, but I would like to couple to CICE using OASIS3-MCT. Could I also get a copy of the tools by Eric Maisonnave that renames the MCT modules used by OASIS3-MCT? Regards Paul
Dear Paul, Could you send to us your email, with your request, using the "Contact" link, so I can send to you the tools ? Thanks, Best regards, Laure
Dear Laure, Thank you very much for your kind help. I have tested the CLM4 and SNU models with model1 and model2 separately without changing the code inside to exchange values of the variables. Only one model is running with openmpi for specified number of processor(s). Although it should run for the command with ':' as below:- $MPIRUN -np $nproc_exe1 ./snu < SYSIN : -np $nproc_exe2 ./clm runjob.err Please let me know how to solve. Also please send me the tools to email address (ashfaq226@gmail.com). With regards, Ashfaq
Dear Ashfaq, I must test openmpi on my computer as I usually work with mpich. Make sure that you compile your real models and the toy models with the same compiler and library. Thanks for you email, I will sent to you the tools of Eric Maisonnave. Best regards, Laure
Hi, Could I also get the tools of Eric Maisonnave? (paul.budgell@imr.no) Regards Paul
Dear Laure, I am continuously trying to compile and run SNU AGCM with example models. As a first step, I tried the example as it is with the *_TP files copied, compile and run. But the following erreor comes: ' oasis_io_read_avfile:av2_FSENDOCN:NetCDF: Variable not found' Please help me urgently to resolve this error. The SNU AGCM is compiled with OASIS3-MCT and running fine without variable exchanging, I will be waiting for your reply and then become successful hopefully. Ashfaq
Dear Ashfaq, You can read the conversation https://www.cerfacs.fr/site-oasis/forum/oa_main.php?c=83 where I explain what is written in the debug files. In fact these fields av*_ are related to the 4 optional arguments of oasis_put used to perform bicubic interpolation or second order conservative remapping. They do not need to be in the file, the message is just a warning. Maybe we should write ': Warning : optional Variable not found' . Best regards, Laure
Dear Laure, There is another error msg: 'oasis_io_read_avfile ERROR: file missing f2avg.nc'. It seems not optional. Please have your time to reply in email or in forum. Regards, Ashfaq
Dear Ashfaq, This file f2avg.nc is linked to the LOCTRANS operations with the OUTPUT field FSENDOCN. The message 'oasis_io_read_avfile ERROR: file missing f2avg.nc' is misleading as the model do not abort. We did not decide yet if this file should be mandatory or not, as if it is, it can be a big constraint for the user. On the other side, if the model aborts, you know exactly what you are doing. I think I will modify the message until we decide what rules we prescribe. So it should not be a problem, the toy tutorial should run. Best regards, Laure
Dear Laure, Many thanks for sending ideas. In fact, the problems were occurring in the previous OASIS3-MCT versions. I have downloaded the latest OASIS3-MCT and the toy models are working fine and successfully exchanging variables. Now for the real model: I could compile the SNU-AGCM source codes as 2nd model and trying to change the namcouple accordingly. The grid sizes for SNU-AGCM is 128 by 64 with 20 vertical levels. The main program contains MPI-INITIALIZATION and all other variables definitions and initialization. Where should I put the OASIS initialization? For testing purpose I put in the middle and it seems work. It will be great if you let me know your opinions. Best regards, Ashfaq
Dear Ashfaq, It is very difficult to answer to your question by mail. To be able to see where you must implement the oasis routines in your real models, you must compare the model1.F90 and model2.F90 to the model1.F90_TP and model2.F90_TP in the toy tutorial, or even better you must modify yourself model1.F90 and model2.F90 to couple them with oasis. Then it will be much easier to couple your real models. I also remember you that there are twice a year some trainings organized at Cerfacs on OASIS. Best regards, Laure
