Matrix Approach in Spin-up for CLM5

[wvsi3w]

I have checked the forum for the "Matrix Approach" and couldn't find anyone trying it only this thread which is a suggestion to the paper related to that approach:

Spin Up Convergene Check

Hi all, I am working on accelerated spin up for CLM and was wondering if anyone has input on what they typically check for to determine convergence to a steady state. I have been looking at TOTVEGC, TOTECOSYSC, and NBP at a gridcell and "global" level, but also curious if anyone looks at...

bb.cgd.ucar.edu

I was wondering if anybody tried this Matrix approach (SASU) for their spin-up as it has shown a significant decrease in computation time and it would be amazing to know the details they have in the steps toward it.

Also, the link in the paper that refers to the software in GitHub is not working (https://github.com/chrislxj/ctsm/tree/cn-matrix_v3) if you know the correct link let us know.

Thanks a lot.

 

[oleson]

This matrix approach is now implemented as an optional spinup option in CTSM (ctsm5.2.014).
The github link you referenced is a personal github account and I'm not sure why it doesn't work.

 

[wvsi3w]

This matrix approach is now implemented as an optional spinup option in CTSM (ctsm5.2.014).
The github link you referenced is a personal github account and I'm not sure why it doesn't work.

Dear Keith, thanks a lot for your response,

Oh, I understand, do you think I can implement it in my version (CLM5 - CESM2.1.3) on my own, or is it a cumbersome process? because I can't switch to CTSM due to the details of my project, I would appreciate it if you point me toward the process of implementing Matrix Approach in CTSM so that maybe I could put it in CESM2.1.3 (CLM5) too (or maybe I am being very delusional and that is outside of my capabilities :) ).

 

[wvsi3w]

by "pointing toward the process" I mean if there is a link, document or any step-by-step written thing available for it

 

[oleson]

I think it would be extremely difficult to independently implement it in CESM2.1.3. There may be a version available that is close to that. I'd contact the author of the paper mentioned above (Liao).

 

[wvsi3w]

Dear @erik
I have contacted the authors from the Matrix Approach and they suggested that maybe you can help me with this question that Keith and I have discussed above.

To summarize, I was looking for a version of CESM (close to 2.1.3) that the Matrix Approach has been implemented in it so that maybe I can copy that into my version and set it up. As Keith also mentioned this is extremely hard to do as it was a big project itself, thus we think if someone has already did and can share it, it would be very beneficial.
The reason why I am dont think I can switch to CTSM is that there are several differences between my version and any ctsm version and the process I took to configure, run, and test the consistency of the model took a lot of time, also the mere objective of my project needs CESM as a whole.

P.S. Do you have access to the github link from the paper that is currently not working? (https://github.com/chrislxj/ctsm/tree/cn-matrix_v3)

 

[slevis]

@wvsi3w I will answer part of the question:
matrixcn went into the clm in ctsm5.2.014
That's a development version in the model's development branch, so it's contents are not supported. You are welcome to use at your own risk, and we are unable to help.

 

[wvsi3w]

@wvsi3w I will answer part of the question:
matrixcn went into the clm in ctsm5.2.014
That's a development version in the model's development branch, so it's contents are not supported. You are welcome to use at your own risk, and we are unable to help.

Thanks for your response,
I understand. The reason I asked that is because of the response I got from the authors suggesting to ask Erik if there are any cesm versions (rather than ctsm). Thanks again. I will stick to AD and ND method for spin up.

P.S. The conversation is attached.

 

[erik]

@wvsi3w

The branch itself was deleted. But, we did save tags of versions along the way of bringing it into the latest CTSM. A tag that's close to release-clm5.0.37 for CESM2.1.5 is: branch_tags/cn-matrix_v3.n04_ctsm1.0.dev080. You might be able to take that version and get it working in cesm2.1.5. I suspect it would be a lot of work, but if your goal is to get a version of matrix to work in cesm2.1.5 I suggest starting by looking into doing that. As @slevis points out we can't really help, you'll have to figure out a lot on your own...

Best of luck...

 

[wvsi3w]

@wvsi3w

The branch itself was deleted. But, we did save tags of versions along the way of bringing it into the latest CTSM. A tag that's close to release-clm5.0.37 for CESM2.1.5 is: branch_tags/cn-matrix_v3.n04_ctsm1.0.dev080. You might be able to take that version and get it working in cesm2.1.5. I suspect it would be a lot of work, but if your goal is to get a version of matrix to work in cesm2.1.5 I suggest starting by looking into doing that. As @slevis points out we can't really help, you'll have to figure out a lot on your own...

Best of luck...

Thanks a lot.
I will reconsider this.
Your responses are much appreciated.