ORCHIDEE vs O-CN data

[viatkin]

Hi Song,

Thanks lot for working on ORCHIDEE model and bringing it to a very good shape! When I tried to run you model though, I was confused, because downloading script gets O-CN files. I know that there are several versions of ORCHIDEE in TREDNY and CMIP6, and O-CN is one of them. Could you clarify, which model do we have labeled as ORCHIDEE-V2 in the framework code? What is the difference between OC-N and another ORCHIDEE that is present in TRENDY? Also, I had an issue downloading O-CN files - it seems rh and ra are no longer downloadable form the server.

Thanks!

[songwang1]

Hello Viatkin, sorry for the late response. Actually, I also met problems with the downloading script. Hence I download the ORCHIDEE output data from somewhere else to my computer. The model I worked on is the O-CN model that came from the paper of Zaehle doi:10.1029/2009GB003521. The O-CN contains the nitrogen constrain effect although I didn't consider the N cycle in our working group. The version I used to constrain parameters is S6 in Trendy, it contains the ra and rh data.

[viatkin]

Hello Song,

I didn't quite get what is S6 in Trendy? I see that there is no rh and ra data in OCN, but it is in ORCHIDEE, ORCHIDEE-CNP, ORCHIDEEv3 and ORCHIDEEv3_0.5deg (see screenshot of the trendy.ex.ac.uk repository) Did you use one of those? Can you send me your rh and ra files somehow?

I added some functions to the end of your model_specific_helpers.py and new file test_helpers.py (necessary for tests in the framework, and also for my comparison plotting). Please be careful: I changed file naming in the get_global_mean_vars function: to match TRENDY naming convention they are now called: "OCNS2...", e.g. "OCN_S2_cLitter".

In test_helpers.py I created a variable epa_opt and set it to epa_opt=epa_0. Are epa_0 already optimized? Or you have optimized parameters (after data assimilation) in epa_opt.csv? If you epa_opt are actually different, could you put the optimized parameters to this variable in test_helpers.py?

Thank you!

[songwang1]

Hello Viatkin, I used the dataset in "OCN", but there are the data of ra and rh in OCN.

maybe I can send them to you somehow.

actually, the epa_opt are optimized parameters after DA, but the matching degree is not very well, I have tried a lot but it still didn't work very well.

[viatkin]

Hello Song,

Thank you! It's ok if parameters are not perfect yet, we will run more data assimilation anyway! Regarding ra and rh data, where are you downloading them from? I am downloading form trendy.ex.ac.uk, and for OCN (as well as for most other models) there are only 3 scenarios there: S1, S2 and S3. Is there another server for TRENDY data? Can you send me a link? Or you can upload ra na d rh files here: https://drive.google.com/drive/folders/1W3rEb4fSzg0KGEQ8OR1U6UFv0v4bf8ic?usp=sharing

Best, Kostia

[songwang1]

Hi Kostia, Actually, maybe you can try this one: H: sftp://trendy.ex.ac.uk U:trendy-v8 Password: gcp-2019 P:22 IF THIS doesn't work, I will upload the data to where you suggested.

Best, Song

[viatkin]

Hi Song,

I see now! It is TRENDY 8, while we are working on TRENDY 9 (User: trendy-v9, Password: gcb-2020). I'm downloading these files from TRENDY 8 now. But I'm a bit confused, did you use all data from TREDNY 8 or only ra and rh? Did you use all data from S6 or only ra and rh? (we are mostly working on S2 data)

Thanks! Kostia

[songwang1]

Hello Kostia,

I think I may make a mistake with the output data. I will rewrite the code about the cost function part and use the right data set later. Thanks for your reminder.

Best, Song

[viatkin]

No worries, that happens:) I'm now editing your model's files to implement latest changes we had in the framework. Please wait for my commit to work on the model. The problem is, what data shall we use. In TRENDY v9 S2 (this is the focus of the project now) OCN does not have rh and ra. ORCHIDEE, ORCHIDEE-CNP, ORCHIDEEv3 and ORCHIDEEv3_0.5deg have all files: ra, rh, cVeg, cLitter, cSoil, cProduct, mrso, tas, npp. So, I was thinking, if model structure is not very different from OCN, should we do one of those ORCHIDEE versions instead? What do you think? Do you know if the structure is too different?

[viatkin]

Please see this commit for fixed and remaining issues: 44d4deea0aebcd62982a707c412da02dfd3d7f13

[songwang1]

Thanks for your comments, I will fix them then.

[viatkin]

I improved the parameters so that the output is reasonable (still imperfect initial values). Commit: e6d03e05855afdedbbea6b61570f7b2bb8118c2b Data assimilation can now run with 30 days time step. Still need to remove unnecessary parameters, improve initial values and run DA until the fit is good.