[MITgcm-support] obcs_apply_ptracer

[Martin Losch]

Taka,

what is the use of an open boundary point right next to a land point?  
It won't do anything ... except cause trouble. One of the "bugs" in  
obcs is that there *must* not any topography gradient across an open  
boundary, and this is also true for coastlines. Can I suggest to  
introduce code at the beginning of the run to check for such a case  
and issue a warning/error message, e.g. in obcs_readparms.F or  
obcs_check.F?

Martin



On 9 Apr 2008, at 14:11, ito at atmos.colostate.edu wrote:
> Hi Martin,
>
>> if (i,j_obs) is an open boundary point, then (i,j_obs-1) is the first
>> interior point. Only if both points are "wet" then maskS(i,j_obs) >
>> 0; you can do this check too with maskC, but you'd have to check if
>> maskC(i,j_obs) and maskC(i,j_obs-1) are > 0.
>
> I noticed that obcs_apply_ptracer can introduce zeros into wet  
> points (at
> least it does to my runs) when S/W masks are equal to 0 on the  
> boundary
> but C mask is nonzero at the tracer point.  When I replace _maskS and
> _maskW with maskC (of the tracer point), it does not introduce  
> zeros into
> wet points.
>
> Consider a case near coastline where you have wet ob point at  
> (i,j_obs)
> but (i,j_obs-1) is on land. In this case maskS(i,j_obs) is zero and
> obcs_apply_ptracer puts 0 into the wet boundary point.  When carbonate
> chemistry solver is applied to the wet point with DIC=0, it  
> produces NaN
> leading to a crash.  So I think it would make more sense to use  
> maskC for
> tracer mask.
>
> Taka
>
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support