Some ideas about CISM

[18729351715@163_com]

Hi all,I study the CISM documentation and the ice flow equation . There are some ideas :1.The CESM_ice_sheet's user guide says: currently,it is not possible to have CISM run in other locations except GreenLand . But If I want to study other locations ,such as the Tibet Plateau in China . The resolution or the equations of ice flow or other reasons cause that impossible ?2.Now,I make the "fglcmask" "flndtopo" "fsurdat" and horiz_grid_file :"fracdata" about the Tibet Plateau using myown data set . Then I create newcase : ./create_newcase -case test -res CLM_USRDAT -compset IG -mach bnu . The "./cesm_setup" and "./tset.build" succeeded . But when I run it , there is a message in the cesm.log "(shr_sys_abort) ERROR: (seq_domain_check_grid)  incompatible domain grid coordinates". And this is my env_run.xmlAnd any suggestions would be appreciated!

 

[sacks]

 Due to grid and structural constraints, we can only run CISM over Greenland for now.  Additionally, we have no experience running dynamic ice code for, essentially, glaciers which are mostly sub-gridscale. If you're just interested in the surface mass balance, and not in the icesheet dynamics, then it should be possible to do something like what you're trying to do.If you still want to proceed with that, and want help with #2, then please provide more information about what version of CESM you're using and the exact steps you have taken. Also: Are you trying to run with a new grid, or an existing grid resolution but just changing where CISM is running? If you're running with an existing grid resolution, then I didn't think you needed to change fracdata.Bill

 

[18729351715@163_com]

Thanks Bill .   1. Due to grid and structural constraints .If you can tell me more details about that , I'll very thank you !2.The version of CESM 1_2_1 and CLM 4.0 .The grid I using is 0.01 degree ,so I use my own fracdata and my own surfdata .3.Then when I run the CISM in other locations (Tibet Plateau) and using exist grid (FV1) .But I have not the ice grid file .which like the gland5UM.So any suggestions would be appreciated!

 

[sacks]

I can't tell from what you have posted if the problem is related to the glc/cism setup specifically, or is due to the setup of land with a new grid. You can get more information by looking earlier in your log file: you should see something like '--- checking glc/sno domains ---'. See which domain check caused the problem.Running CISM (as in an IG compset) at a new land model resolution (other than f09, f19 and T31) is not something we have done from this code base. This will become easier with the CESM2 release (planned for this December). So I am not able to give a lot of support with this.More importantly, though, and addressing your question: My understanding is that it is difficult to get CISM working in a scientifically reasonable way for a new domain. You need to do more than just set up a new grid file: there are various parameters (some spatially varying) that need to be set properly for a new domain. In addition, my understanding is that CISM is only meant to be used for large icesheets, like Greenland, Antarctica and other paleoclimate icesheets. I'm not sure of the details regarding mountain glaciers; I'll let others add more here if they have anything more to say about that. Therefore, unfortunately we are not able to provide support for running CISM over anything besides Greenland in this version of the model.

 

[18729351715@163_com]

And if I only want to study the surface mass balance (the glaciers of Tibet Plateau) using my own dataset and a new grid . The compset what should I choose?

 

[sacks]

Just studying the surface mass balance of these mountain glaciers should be possible. You would use an IG compset, like you have been doing.You have two choices. The easiest thing to do is to keep the CISM grid over Greenland, but produce surface mass balance over the Tibetan Plateau as well. To do this, you would use an fglcmask file that includes both Greenland (like the standard one) and Tibet. You could then look at the surface mass balance in each land grid cell. The CLM history field QICE gives the gridcell-average surface mass balance (using whatever glacier elevation class(es) exist in that grid cell). It can also be useful to turn on the CLM history field QICE_FORC (using the CLM namelist option hist_fincl1, hist_fincl2, etc.): this gives the surface mass balance for every elevation class separately. With this approach, you won't get surface mass balance on the glacier grid. But since you're running CLM at such high resolution, this is probably okay.The other approach would be to make a CISM grid for your Tibetan region - replacing gland5UM. This would be trickier. The benefit would be that you get surface mass balance regridded to your CISM grid, but again: I think the benefits for this high-resolution CLM case will be minimal. If you do this, you should set evolve_ice = 0 in user_nl_cism so that CISM doesn't try to do ice dynamics: In this case, CISM will just perform the downscaling / regridding of surface mass balance.One other important thing for your case: When you created a new surface dataset, did you specify '-glc_nec 36' to mksurfdata.pl? The standard is 10, but for these high mountain regions, 10 is not sufficient, and you should use 36: this creates an elevation class for every 200 m elevation band. 

 

[18729351715@163_com]

Thanks, BillAdrees your replies ,my understanding is that 1.CISM can only simulate the Greenland's ice dynamics currently  . For other locations , the SMB can be calculated but CISM can not do the ice dynamic. Correct ?2.We can only run CISM over Greenland for now ,and the land model resolution must be f09,f19,T31. In the future ,we can use new land model resolution but also only over Greenland . Yes?Thank you very much again!

 

[sacks]

1.CISM can only simulate the Greenland's ice dynamics currently  . For other locations , the SMB can be calculated but CISM can not do the ice dynamic. Correct ?Yes, this is correct2.We can only run CISM over Greenland for now ,and the land model resolution must be f09,f19,T31.Yes, in terms of what is supported without needing significant modifications.  In the future ,we can use new land model resolution but also only over Greenland . Yes?Yes, that will be the case for CESM2.

 

[18729351715@163_com]

 

[lipscomb@lanl_gov]

CISM solves a transport equation of the form dH/dt + (del * u_2d*H) = B, where u_2d = (u,v) is the 2D velocity, H is thickness, and B = mass balance. For the SIA solver, u_2d is inserted into the transport equation analytically, wherease for the higher-order solver, u_2d is found by solving a Stokes approximation for the momentum balance.  The vertical velocity w is not computed directly, but can be diagnosed from (u,v) together with incompressibility, del*u_3d = 0.  You can think of w at a given layer interface as the vertical velocity of that interface relative to a fixed Eulerian grid.  Does that answer your question?

 

[18729351715@163_com]

Thank you very much about the answer. But can you give me some papers about the vertical velocity ,which relative to a fixed Eulerian grid. I don't fully understand. And any information would be appreciated!

 

[lipscomb@lanl_gov]

Offhand, I don't know of any papers that address this topic specifically.  I'd suggest working through the equations in Section 3.2.3 and Section 4.1 of the CISM documentation, and maybe drawing some pictures.  See especially Eq. (4.7) - (4.10). The vertical velocity w(z) is defined by Eq. (4.8). However, I think there's a typo in this equation--the lower bound of the integral should be the bed b rather than the surface h.  I'll fix this in the next version of the documentation.

 

[18729351715@163_com]

Thanks . But in the CISM documentation , for this equation dH/dt + (del * u*H) = B ,where u is the vertically averaged ice velocity , not (u,v) 2D velocity. 

 

[lipscomb@lanl_gov]

That's correct.  To be consistent with the CISM notation, I should have writtendH/dt + (del * U*H) = B - M,where U = is the vertically integrated 2D velocity, B is the surface mass balance (positive for accumulation) and M is the basal melt rate (positive for melting).  For the Glide SIA dycore, U can be computed from the underlying equations by integrating over the flow factor A in each layer, and then the above transport equation is solved for H.  For the Glissade higher-order dycore, CISM computes (u,v) in each layer and then solves the above transport equation for the ice thickness in each layer.  In either case, w(z) is a diagnostic obtained by (4.8). 

 

[18729351715@163_com]

Thanks very much for your detailed answer!! But the "U = is the vertically integrated 2D velocity" , adrees your replies ,my understanding is that the U is 2D velocity , which has two directions : vertical and horizontal . Correct ?  Thank you very much again!

 

[lipscomb@lanl_gov]

That's not correct.  The vertically integrated 2D velocity vector U is a vector in the xy plane, whereas w is the vertical (z) component of velocity.  We can write U = (U,V), where U and V are the vertically integrated velocity components in the x and y directions, respectively.  The full 3D velocity at a given point is u = (u,v,w).  (I'm using bold font here to denote vectors.) 

 

[18729351715@163_com]

Hi Bill . There are several confusion may need your help . 1. In the CESM Land Ice Model User's Guide , Section 4.4 , the SMB calculated in the CLM and the surface temperature are downscaled by GLINT from the global grid (lat-lon grid) to local ice sheet grid . I know the method used is an adaption of bilinear interpolation and is termed "mean-preserving interpolation" , but this process of downscaling is not more detailed explanation. I want to understand fully about the precess and look forward to your reply.2. In the CESM Land Ice Model User's Guide , Section 5.2 , the atmospheric surface temperature , ... are downscaled from the mean gridcell elevation to the glacier_mec column elevation using a ... . The downscale here and that downscale above , are they the same thing? If not , when the  glc_nec=10 , what about the[/i] mean gridcell elevation [/i]? 3. The fields downscaled by GLINT from the global grid to  local ice sheet grid only are SMB and the surface temperature .Correct ?Looking forward to your detailed reply and any suggestions would be appreciated! Thang you very much again!!!

 

[sacks]

1. Glint uses bilinear interpolation together with linear vertical interpolation, as described in the user's guide. This is not mean-preserving. CESM2 will use mean-preserving interpolation; this will be described in the technical documentation released with that model version around the end of this year.2. No, these are not the same downscaling. The downscaling in section 5.2 is a lapse rate-based adjustment.3. Yes

 

[18729351715@163_com]

Hi Sacks: I'm so sorry to disturb you again . Now I want to study the surface mass balance of mountain glaciers (Tibetan Plateau) using CESM1_2_1 and CLM4.0 . But the land model resolution must be f09,f19,T31 . Why ? This is too coarse to meet my needs. If I want to use my own data (0.1 degree ) and a new resolution to study the SMB (only the glaciers of Tibetan Plateau) , is that OK ? Any suggestions would be appreciated and look forward to your details !Thank you very much again!!! 

 

[sacks]

This limitation is due to input datasets that only exist at certain resolutions as well as limitations of the remapping between CLM and CISM in that version of the code. If you can wait until the CESM2 release (expected in December), these limitations will be removed there.