[MITgcm-support] open boundaries!

Martin Losch mlosch at awi-bremerhaven.de
Wed Oct 12 03:03:46 EDT 2005 

Van Thinh,

recently, exp4 has changed a little to include a (relatively) new  
possibility to prescribe open boundary conditions, via external input  
files (in fact this possiblity is almost a year old, but was never used  
in the verification experiments, only in
http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm_contrib/heimbach/ 
obcs_prescribe
(as Stephano has already pointed out, what follows is in part similar  
to what Stephano has already written earlier:).
In general, open boundaries are applied *as if* they were Dirichlet  
BCs, that is, at the end of a time step the fields get (re-)set to the  
boundary values, that are computed or read in in obcs_calc.F. You can  
either modify obcs_calc.F, as Matt has suggested or create input files  
as now shown in exp4. If you want v.Neumann conditions, you'll have to  
change obcs_calc.F anyway (or the obcs_apply_*.F routines). There is an  
example how do to this *crudely* for quasi-homogeneous BC for the  
passive tracers (ptracer) in obcs_calc.F or strictly in  
obcs_apply_ptracer.F (the code is commented out). You could in  
principle do this to T/S/U/V as well, but for that you have the  
Orlanski radiation conditions, which are probably more useful (see

I suggest that you update your code with
cvs update -d -P
or checkout a new copy. Then you have a look at exp4 again (it has  
changed a little) and start modifying it SLOWLY, making only one change  
at a time, until it is similar to what you actually want to do. You'll  
find out what changes make the model give you these "very bad results".

Good luck,
Martin

On Oct 11, 2005, at 8:51 PM, Matthew Mazloff wrote:

>  Hi Van Thinh,
>
>  I have no idea why you have bad results.  If you are following what  
> was done in exp4 and only changing the domain and obcs forcing  
> function things should work.  Did you make sure useobcsbalance=.FALSE.  
> and that the frequency is given in radians/second?
>  You will have to give me more info as to what is wrong with the  
> solution if you want more advice.  Perhaps you should print out the  
> prescribed forcing for the first few time steps to see if it is what  
> you want it to be, other things may be going wrong.
>
>  -Matt
>
>
>  Van Thinh Nguyen wrote:
>> Hi Matt,
>>
>> Thanks so much for helping me about OBs.
>> I just would like to have more questions about this problem:
>>
>>  I exactly follow your hints, however I just edited 'obcs_calc.F', in
>> which I wrote my function at OBWu and OBEu
>> (e.g. OBWu(J,K,bi,bj)=Uinflow*cos(futureTime*freq)
>>  where freq=M2=1.4053e-4, Uinflow=0.2
>> I am not sure how the program pickup futureTime here!)
>>  So I've got very bad results, it was something confused!
>>
>> I've also done the same way as exp4, of course with different domain.
>>
>> Could you or someone experienced with OB give me a hint!
>>
>> Thanks a lot,
>>
>> Van Thinh
>> --------------
>>
>> On Thu, 29 Sep 2005, Matthew Robert Mazloff wrote:
>>
>>
>>> Hi Van Thinh,
>>>
>>> As you know, you tell the model you have an open western boundary by  
>>> writing in
>>> data.obcs:
>>>  OB_Iwest=   55*1
>>>
>>> You can then prescribe a u velocity component by adding to this file:
>>>
>>> useOBCSprescribe=.TRUE.,
>>> OBNuFile='UNBC.bin',
>>>
>>> This can have multiple records in it...and then in data.exf you add
>>> obcsWstartdate1     = 19920101,
>>> obcsWstartdate2     = 000000,
>>> obcsWperiod         = 2635200.0,
>>>
>>> which tells the model to switch to the next velocity record every  
>>> 2635200.0
>>> seconds.
>>>
>>> A better way to impliment your function would be to write a new  
>>> routine.  The
>>> model sets the U velocity on the open western boundary to
>>> OBWu.  This is read in when the routine OBCS_PRESCRIBE_READ executes
>>>
>>>  call EXF_SET_OBCS_YZ(  OBWu, OBWu0, OBWu1, OBWufile, 'u'
>>>      I                     , fac, first, changed, count0, count1
>>>      I                     , mycurrenttime, mycurrentiter, mythid )
>>>
>>> Here you could change this call to
>>>
>>>  call VAN_THINHS_FUNCTION(  OBWu, OBWu0, OBWu1, OBWufile, 'u'
>>>      I                     , fac, first, changed, count0, count1
>>>      I                     , mycurrenttime, mycurrentiter, mythid )
>>>
>>> and in the routine VAN_THINHS_FUNCTION, OBWu is set from your  
>>> function.  The
>>> same can be done for the eastern boundary.
>>>
>>> Good luck,
>>> Matt
>>>
>>>
>>>
>>> Van Thinh Nguyen wrote:
>>>
>>>
>>>> Hi Stefano and Patrick,
>>>>
>>>> I took a look on MITgcm Support Archives to find how to set up OBCs,
>>>> however I still confuse in setting up some parameters.
>>>>
>>>> My test case is almost the same as exp4:
>>>>
>>>> Flow in a channel over a bump with
>>>> - north and south boundaries are walls
>>>> - west & east are OBs
>>>> I would like to set the tidal forcing at west OB with velocity  
>>>> component
>>>> U=Uo*cos(omega*t). (Uo=0.025; omega=2*pi/43200)
>>>>
>>>> Could you please give a hint how can I set these inputs?
>>>>
>>>> Thanks so much for your helps,
>>>>
>>>> Van Thinh
>>>> ----------------
>>>>
>>>>
>>>> On Mon, 26 Sep 2005, Stefano Querin wrote:
>>>>
>>>>
>>>>> Hi Van-Thinh,
>>>>> maybe the best thing you can do is to take a look to the files in  
>>>>> the CVS
>>>>> repository (or directly in your /pkg/obcs folder if you have a  
>>>>> recent
>>>>> version of the code). Some routines have short and useful comments  
>>>>> that help
>>>>> the user in the preliminar obcs tests.
>>>>> Basically (and in very few words), you can impose:
>>>>> - constant OBCs (obcs_calc.F sets u, v, w and eta to zero, and T,  
>>>>> S to tRef
>>>>> and sRef);
>>>>> - radiating (passive) OBCs for momentum and tracers following  
>>>>> Orlanski (see
>>>>> Orlanski, 1976);
>>>>> - time varying (active) OBCs using the obcs package (set
>>>>> useOBCSprescribe=.TRUE.,) together with exf and cal.
>>>>>
>>>>> You can also define a sponge (or better, nudging?) layer on the OB  
>>>>> (setting
>>>>> useOBCSsponge=.TRUE., sponge thickness and relaxation parameters)  
>>>>> and/or
>>>>> prescribe the net flow across the OB using the useOBCSbalance  
>>>>> option.
>>>>>
>>>>> A useful example is in:
>>>>>
>>>>> http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm_contrib/heimbach/ 
>>>>> obcs_prescribe/
>>>>>
>>>>> As the old obcs/README said, happy open-boundarying!
>>>>>
>>>>> Cheers,
>>>>>
>>>>> Stefano
>>>>>
>>>>>
>>>>> ----- Original Message -----
>>>>> From: "Van Thinh Nguyen" <vtnguyen at moisie.math.uwaterloo.ca>
>>>>> To: <mitgcm-support at mitgcm.org>
>>>>> Sent: Monday, September 26, 2005 6:26 PM
>>>>> Subject: [MITgcm-support] open boundaries!
>>>>>
>>>>>
>>>>>
>>>>>> Dear MITgcm-supporters & users,
>>>>>>
>>>>>> Could someone please explain for me more detail about how to set  
>>>>>> up an
>>>>>> open boundary conditions. For instance, in exp4:
>>>>>>
>>>>>> when we set in data.obcs:
>>>>>>
>>>>>> "OB_Jnorth=80*-1" to define a face where OB is appied, however I  
>>>>>> still
>>>>>> don't know which boundary condition is appied here (Neumann?).
>>>>>>
>>>>>> And what is the "useOrlanski" in OB option?
>>>>>>
>>>>>> There are probably some docu around in MITgcm users' manuals (I  
>>>>>> actually
>>>>>> took a look but haven't found), if someone know please give me a  
>>>>>> hint?
>>>>>>
>>>>>> Thanks a lot!
>>>>>>
>>>>>> Van Thinh
>>>>>> _______________________________________________
>>>>>> MITgcm-support mailing list
>>>>>> MITgcm-support at mitgcm.org
>>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>>>>>
>>>>>>
>>>>> _______________________________________________
>>>>> MITgcm-support mailing list
>>>>> MITgcm-support at mitgcm.org
>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>>>>
>>>>>
>>>> _______________________________________________
>>>> MITgcm-support mailing list
>>>> MITgcm-support at mitgcm.org
>>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>>>
>>>>
>>> _______________________________________________
>>> MITgcm-support mailing list
>>> MITgcm-support at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>>
>>>
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support

More information about the MITgcm-support mailing list