[MITgcm-devel] averaging in diagnostics vs. exf

[Martin Losch]

Hi there,
me again, this turns out to be more complicated than I thought. Here's my matlab code to reproduce the averaging in diagnostics:

tmp=(.125+epsi)*atemp(:,:,ii+1)+0.75*atemp(:,:,ii+2)+(0.125-epsi)*atemp(:,:,ii+3);
d=EXFatemp(:,:,:,ii+2)-273.15-tmp;

(atemp: daily forcing as read from the forcing file
EXFatemp: diagnostics variable, daily averages)

rms(d) is the smallest (and the difference in 2D is just noise) when epsi=300/86400;
my deltaT=1200, the averaging period is 86400, so it seems that either the time interpolation in exf or the averaging in diagnostics are introducing an offset of a quarter of the time step. From bad experience with exf, I am guessing the time interpolation in exf is doing this (o:

Is this a problem? Is my atempstartdate not correct and I am expecting the wrong result?

Martin
On Jul 12, 2011, at 5:00 PM, Martin Losch wrote:

> sorry about that,
> it should 0.125,0.75,0.125 instead of 0.25,0.5,0.25. Looks much better now (not perfect, though).
> 
> Martin
> 
> On Jul 12, 2011, at 4:48 PM, Martin Losch wrote:
> 
>> Hi there,
>> 
>> this question is almost too embarrasing to be posted at devel:
>> I want to check the surface forcing that I am using and for that I compare what goes into the model with the diagnostics, for example air temperature. I have daily forcing (and I do not use any interpolation within exf)
>> cat data.cal
>> &CAL_NML
>> TheCalendar='gregorian',
>> startDate_1=19580101,
>> startDate_2=000000,
>> calendarDumps = .TRUE.,
>> 
>> cat data.exf
>> [...]
>> useExfYearlyFields= .TRUE.,
>> [...]
>> #
>> atempstartdate1=19480101,
>> atempstartdate2=120000,
>> atempperiod=86400.0,
>> #
>> [...]
>> and finally in data.diagnostics:
>>   frequency(3) = 86400.,
>>   fields(1,3) = 'EXFatemp',  
>>   filename(3) = 'run10ncf_dailyEXF',
>> 
>> I expect that EXFatemp should be the same as my input (let's call it "atemp") after approprate averaging, e.g. in matlab notation for a specific year:
>> EXFatemp(:,:,2) = 0.25*atemp(:,:,1)+0.5*atemp(:,:,2)+0.25*atemp(:,:,3)
>> (the factors are a result of linear interpolations between noon values) but unfortunately that is not the case (be several degrees). What am I missing? Problem in my data.diagnostics file, or in data.exf?
>> 
>> Martin
>> 
>> 
>> 
>> _______________________________________________
>> 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