Re: missing data problem

Do *not* ever perform a spectral analysis on interpolated data.

Use
http://www.ncl.ucar.edu/Document/Functions/Built-in/wgt_areaave.shtml

for the region you want ... with appropriate weighting.
Use opt=0

Good luck

Ioana Colfescu wrote:
> Thank you very much Mary and Dennis for the help.
>
> ' what do u have for x ? ' - I just realized since the spectrum needs a 1-D series I ll need to average and have the global mean for SST indeed so once I average the problem of missing values is gone.
>
> On the other side, as is written even in what Mary send me - about the interpolation - seems that if I interpolate I get a very different spectrum from the uninterpolated values right ? Mathematically actually once I interpolate the spectrum might have fundamental differences from the uninterpolated one so the only way to 'solve' my problem is using the global average for SST I think and not include other inteprolated values.
>
> ----- Original Message -----
> From: "Dennis Shea" <shea@ucar.edu>
> To: "Ioana Colfescu" <colfescu@cola.iges.org>
> Cc: NCL-talk@ucar.edu
> Sent: Tuesday, March 9, 2010 10:27:51 AM GMT -05:00 US/Canada Eastern
> Subject: Re: missing data problem
>
> specx_anal works on a one-dimensional series
>
> function specx_anal (
> x [*] : numeric, <=== one-dimensional
> iopt [1] : integer,
> jave [1] : integer,
> pct [1] : numeric
> )
>
> What are you using for "x" ?
>
> your SST ... Is this a global average or regional average series?
>
> It is unlikely you want the power spectrum at every grid point ...
> However, if that is what you want, you will have to use a loop
>
> do nl=0,nlat-1
> do ml=0,mlon-1
> if (all(.not.ismissing(sst(:,nl,ml)))) then
> dof = specx_anal (sst(:,nl,ml),iopt,jave,pct)
> ;whatever
> end if
> end do
> end do
>
>
>
> On 03/09/2010 07:34 AM, Mary Haley wrote:
>> I think you will need to interpolate to fill in the missing values. See:
>>
>> http://www.ncl.ucar.edu/Support/talk_archives/2008/0867.html
>>
>> On Mar 8, 2010, at 3:40 PM, Ioana Colfescu wrote:
>>
>>> Sorry for asking an incomplete question : The error message is
>>> 'missing values not permitted' and is given when I try to use the
>>> built in function 'specx_anal' . I checked the documentation and it
>>> seems that there are not missing values admitted when using this
>>> function indeed. Since there is SST data , what can one do in case you
>>> only have over ocean but nothing on continents and you still want to
>>> calculate the spectrum ?I think is possible to calculate a spectrum
>>> for such values but I'm not sure how NCL deals with this problem .
>>>
>>> ----- Original Message -----
>>> From: "Mary Haley" <haley@ucar.edu>
>>> To: "Ioana Colfescu" <colfescu@cola.iges.org>
>>> Cc: ncl-talk@ucar.edu
>>> Sent: Monday, March 8, 2010 3:13:29 PM GMT -05:00 US/Canada Eastern
>>> Subject: Re: missing data problem
>>>
>>> There's not enough information to debug this problem.
>>>
>>> Can you tell us what the exact error message is? How are you
>>> calculating the spectrum?
>>>
>>> Providing your script may help us debug the problem, but sometimes we
>>> need to look at your data too.
>>>
>>> --Mary
>>>
>>> On Mar 8, 2010, at 12:36 PM, Ioana Colfescu wrote:
>>>
>>>> Hi,
>>>>
>>>> I try to calculate a spectrum for temperature field ( output from
>>>> CAM ).
>>>> The problem is that when I try to plot it I get the message that
>>>> the spectrum can't be made because of missing data.
>>>> Is there any solution for this problem ? Could anyone suggest me
>>>> what should I do about it what should I replace the data with or
>>>> what other thing I could do ?
>>>>
>>>> Thank you
>>>> _______________________________________________
>>>> ncl-talk mailing list
>>>> List instructions, subscriber options, unsubscribe:
>>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Tue Mar 9 09:52:04 2010