[MITgcm-devel] seaice ridging: adjoint

[Patrick Heimbach]

Hi Martin,

it’s global_ocean.cs32x15/ and you have to test input_ad.seaice (i.e. tr_run.seaice) “by hand” since there’s no reference output for tlm.

p.

On Apr 24, 2014, at 4:30 AM, Martin Losch <Martin.Losch at awi.de> wrote:

> Hi Patrick,
> 
> this is indeed puzzling, because my changes should not be related to seaice_solve4temp.
> 
> But I cannot reproduce your results. How (or where) do you get the NaN’s with the TLM? 
> I tried ./testreport -tlm -t lab_sea and everything looks good. 
> 
> What may have happened is that my changes have screwed up some initializations that are only required by TAF/TAMC.
> 
> Martin
> 
> On Apr 23, 2014, at 11:11 PM, Patrick Heimbach <heimbach at MIT.EDU> wrote:
> 
>> Hi Martin,
>> 
>> I haven’t looked in detail yet, but what’s suspicious is that the TLM gives NaN’s.
>> Before your checkin, the TLM values were highly accurate (order 10^-8).
>> Also NaN’s when I switch off seaice dynamics.
>> 
>> The TLM NaN’s in S/R seaice_solv4temp.F
>> a bit after assigning t1 = tsurfLoc(I,J)
>> 
>> which then gets to mm_log10pi = -aa1 / t1 + aa2
>> the TLM of which is NaN (I guess t1 = 0).
>> Not sure why this is since that routine has not changed in 2 years.
>> 
>> If I switch to useMaykutSatVapPoly the TLM looks ok.
>> 
>> I’ll try to look more closely, but perhaps you have an idea.
>> 
>> -Patrick
>> 
>> 
>> 
>> On Apr 23, 2014, at 8:44 AM, Martin Losch <Martin.Losch at awi.de> wrote:
>> 
>>> OK, done, thanks,
>>> 
>>> The adjoint works but gives slightly different numbers. for SEAICE_ITD, there is no adjoint, as far as I can see.
>>> 
>>> Martin
>>> 
>>> On Apr 23, 2014, at 12:38 PM, Patrick Heimbach <heimbach at MIT.EDU> wrote:
>>> 
>>>> Hi Martin,
>>>> go for it. I’m around, so can take a look.
>>>> p.
>>>> 
>>>> On Apr 23, 2014, at 5:04 AM, Martin Losch <Martin.Losch at awi.de> wrote:
>>>> 
>>>>> Hi Patrick,
>>>>> 
>>>>> in preparation for real seaice ridging code, I am moving the regularization part (most of “PART I”) out of seaice_growth into a new routine.
>>>>> 
>>>>> Unfortunately this changes the adjoint gradient values (making the agreement with the FD gradient a little worse). All forward runs are OK:
>>>>> G D M    C  A  F
>>>>> e p a R  o  d  D
>>>>> n n k u  s  G  G
>>>>> 2 d e n  t  r  r
>>>>> 
>>>>> Y Y Y Y 16> 8<16 FAIL  global_ocean.cs32x15.seaice
>>>>> Y Y Y Y 16> 7<16 FAIL  global_ocean.cs32x15.seaice_dynmix
>>>>> Y Y Y Y 16> 5<16 FAIL  lab_sea  (e=0, w=18)
>>>>> Y Y Y Y 16> 6<16 FAIL  lab_sea.noseaicedyn
>>>>> 
>>>>> The taf_ad.log is the same with and without my modifications (let alone line numbers and the extra routine), so no extra. I have no idea where to look. Can I check in my changes anyway and we have a look together?
>>>>> 
>>>>> Martin
>>>>> _______________________________________________
>>>>> MITgcm-devel mailing list
>>>>> MITgcm-devel at mitgcm.org
>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>>> 
>>>> 
>>>> ---
>>>> Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
>>>> MIT | EAPS 54-1420 | 77 Massachusetts Ave | Cambridge MA 02139 USA
>>>> FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach
>>>> 
>>>> _______________________________________________
>>>> MITgcm-devel mailing list
>>>> MITgcm-devel at mitgcm.org
>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>> 
>>> 
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>> 
>> 
>> ---
>> Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
>> MIT | EAPS 54-1420 | 77 Massachusetts Ave | Cambridge MA 02139 USA
>> FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach
>> 
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel


---
Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS 54-1420 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1588 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20140424/0d3202f8/attachment.p7s>