[MITgcm-devel] negative H in calc_r_star.F

[Menemenlis, Dimitris (3248)]

Jean-Michel, could you advise on r-star and shelfice.  We'd really like to be able to use both together, if possible.

Martin, I am guessing that your simple experiments may not have worked because you were using open
boundaries, and I think that r-star and obcs do not yet work together.  The CS510 configuration is global.

Another possibility is that you had negative H's in your configuration.
The negative H comes from computing the thickness of the wet cell wall between two adjacent grid boxes.
It is computed as: MIN( Ro_surf(i-1,j), Ro_surf(i,j) ) - MAX( R_low(i-1,j), R_low(i,j) )
There is no check to see whether the above gives a negative number, so if your bathymetry and ice shelf
topography has a slope such that it generates negative H's, the code will stop.

By coincidence, in the original (cube86) configuration we did not generate negative H's so the code run OK.
When we switched from hFacMinDr=50 to hFacMinDr=0, however, we started causing a bunch of negative H's.

Those are guesses.  Let's wait to hear from Jean-Michel about this.

Thanks

Dimitris Menemenlis

On May 29, 2012, at 2:11 AM, Martin Losch wrote:

> I think you can still use this code if it works. I just could never make it work for my simple experiments. And you Eta looks OK to me (Eta is not the issue anyway). I don't know anything about the negative H.
> 
> Martin
> 
> On May 25, 2012, at 9:12 PM, Menemenlis, Dimitris (3248) wrote:
> 
>> Ooops!
>> 
>> This means that for past three years we have boldly been going where we had no business of going :-(
>> 
>> http://ecco2.jpl.nasa.gov/data1/cube/cube86/input/ETANjan08.jpg
>> is a figure of ETAN on the Antarctic face of the cube86 global CS510 configuration after 16 years of integration.
>> 
>> Apart from fact that one can end up with negative H at edges of grid cells, which was not the case for cube86,
>> what other aspects of the existing "r_star + shelfice"  solutions are likely to be problematic?
>> 
>> Shall we add a stop or a warning to shelfice_check.F when r_star is enabled?
>> 
>> Dimitris Menemenlis
>> 
>> On May 25, 2012, at 12:52 AM, Martin Losch wrote:
>> 
>>> I am not aware, that shelfice can be used with r-star! 
>>> 
>>> M.
>>> 
>>> On May 24, 2012, at 9:30 PM, Dimitris Menemenlis wrote:
>>> 
>>>> Jean-Michel and Martin, is it allowable to have negative H in calc_r_star.F  when using pkg/shelfice ?
>>>> 
>>>> When we went from hFacMinDr=50 to hFacMinDr=0 in order to better represent the ice shelf cavity,
>>>> we end up with following situation:
>>>> 
>>>> WARNING: r*FacW < hFacInf at       1 pts : bi,bj,Thid,Iter=   1   1   1        10
>>>> e.g. at i,j=   9  99 ; rStarFac,H,eta =  0.019415 -1.499939E-01  1.689013E-01  1.252815E-01
>>>> STOP in CALC_R_STAR : too SMALL rStarFacW !
>>>> 
>>>> I am pretty sure that H is non-zero in center of cell - Hong is double checking.
>>>> ... but shouldn't H at the edges of the cells be set to zero when
>>>> MIN( Ro_surf(i-1,j), Ro_surf(i,j) ) - MAX( R_low(i-1,j), R_low(i,j) )
>>>> is negative?