Re: center_finite_diff_n missing values

> I think Example 6 may have a documentation error.
>
> The following
>
> dTdX(:,:,nl,:) = center_finite_diff_n (T(:,:,nl,:), dX , True,0,0)
>
> should be
>
>
> dTdX(:,:,nl:nl,:) = center_finite_diff_n (T(:,:,nl:nl,:), dX ,
> True,0,3)
>
> ----
> dTdX(:,:,nl:nl,:)
> 0 1 2 3 <== dimension number
>
> the nl:nl forces the array to remain 4D so the dimension numbering
> stays the same.
>
> ---
> Let us know if this is the issue.
Nope, that didn't fix it.
>
> +++++
> Below ... if hus is already (time,lev,lat,lon) there would be no need
> to do this: hus(time|:,plev|:,lat|:,lon|:)
It's not. I need it in a different order for another operation.
>
>
> On 12/1/11 4:17 PM, Nicole Schiffer wrote:
>> My goal is to calculate the 1-D moisture flux (mfu, mfv)and then use
>> that to figure moisture convergence.
>>
>> I was trying to follow Example 6 from the documentation
>> (http://www.ncl.ucar.edu/Document/Functions/Built-in/center_finite_diff_n.shtml).
>>
>> I switched the dimension of n1,n2 because I want to look at divergence
>> rather than vorticity, as in an earlier post
>> (http://www.ncl.ucar.edu/Support/talk_archives/2010/2835.html).
>>
>> I also tried it using lat and lon without the looping or conversion from
>> degrees to distance just to get the function to work (see code below).
>> NCL still sees missing values in the u data and not in the v data (and
>> hence tries to calculate the gradient between real values and 1e+20).
>>
>> mfu(:,:,:,n2) =
>> center_finite_diff_n((hus(time|:,plev|:,lat|:,lon|:)/g)*ua(:,:,:,:),lon,False,0,2)
>>
>> mfv(:,:,:,:) =
>> center_finite_diff_n((hus(time|:,plev|:,lat|:,lon|:)/g)*va(:,:,:,:),lat,False,0,2)
>>
>>
>> -Nicole
>>
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> * Nicole Schiffer
>> * Graduate Research Fellow (Dept. of Energy)
>> * Department of Atmospheric Sciences
>> * University of Illinois, Urbana-Champaign
>> * Email: nschiff2 [at] illinois [dot] edu
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>
>>
>> On 12/1/11 4:56 PM, Mary Haley wrote:
>>> Hi Nicole,
>>>
>>> Why are dx and dy calculated the way they are?
>>>
>>>> dx = 6378388*cos(0.0174533*lat(n2))*dlon
>>>> dy = 6378388.*0.0174533*dlon
>>>
>>>
>>>
>>> Did you mean to subscript "lat" with n2? You use "n2" as a longitude
>>> subscript
>>> in the center_finite_diff_n call, so I'm wondering if there's an error
>>> there...
>>>
>>> Something doesn't look quite right with the above code. I'm guessing
>>> one or
>>> both of dx or dy are incorrect.
>>>
>>> --Mary
>>>
>>> On Dec 1, 2011, at 10:31 AM, Nicole Schiffer wrote:
>>>
>>>> I'm using center_finite_diff_n on two 4D array (time, pressure, lat,
>>>> lon) with missing values. With one array, the u-wind, it handles
>>>> missing
>>>> values fine. With the v-wind array, it doesn't. Both arrays have the
>>>> same metadata (missing and fill values). Any ideas why v-wind is a
>>>> problem?
>>>>
>>>> Code snippet:
>>>> dlat = 2.5*0.0174533
>>>> dlon = dlat
>>>> do n2=0,nlon-1
>>>> dx = 6378388*cos(0.0174533*lat(n2))*dlon
>>>> mfu(:,:,:,n2) =
>>>> center_finite_diff_n((hus(time|:,plev|:,lat|:,lon|n2)/g)*ua(:,:,:,n2),dx,False,0,2)
>>>>
>>>>
>>>> end do
>>>> dlon = 2.5*0.0174533
>>>> do n1=0,nlat-1
>>>> dy = 6378388.*0.0174533*dlon
>>>> mfv(:,:,n1,:) =
>>>> center_finite_diff_n((hus(time|:,plev|:,lat|n1,lon|:)/g)*va(:,:,n1,:),dy,False,0,2)
>>>>
>>>>
>>>> end do
>>>>
>>>> Thanks,
>>>> Nicole
>>>>
>>>> --
>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>> * Nicole Schiffer
>>>> * Graduate Research Fellow (Dept. of Energy)
>>>> * Department of Atmospheric Sciences
>>>> * University of Illinois, Urbana-Champaign
>>>> * Email: nschiff2 [at] illinois [dot] edu
>>>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>>>
>>>> _______________________________________________
>>>> 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 Thu Dec 1 19:20:20 2011