[ncl-talk] EOF analysis with NARR data (LON, LAT values)

[Dennis Shea]

To expand on DaveA's response.
===========================

[1]
You can only use coordinate subscripts and the associated syntax operators
( &, {...}) with 'coordinate variables'
The file you are using has coordinate variables 'time', 'x' and 'y'. The
units of 'x' and 'y' are meters.
If you could specify the appropriate locations in **meters**, you could use
the meters. Hpwever, this is unusual and a bit awkward.

[2]
The file does have' lat(y, x)' and 'lon(y,x)'. These are curvilinear
coordinates. I am sure the graphics portion of the NCL Workshop explained
how to plot 2d curvilinear coordinates:

       lat2d = f->lat
       lon2d= f->lon

       [a] Use the reserved attributes 'lat2d' or 'lon2d'
             There are many examples illustrating this approach.

              t at lat2d = lat2d
              t at lon2d = on2d
or

      [b] Set the 'sfXArray', 'sfYArray'

  res at sfXArray     = lon2d
  res at sfYArray     = lat2d

   Please see the workshop graphics note.

[3]

If you want to subset the array, you must use index (subscript) values.

As noted by Dave, you will have to use efunctions to determin the subregion:


    http://www.ncl.ucar.edu/Document/Functions/Contributed/region_ind.shtm


Good luck



On Wed, Jun 17, 2015 at 3:41 PM, Dave Allured - NOAA Affiliate <
dave.allured at noaa.gov> wrote:

> Jothiganesh,
>
> > 1) Is there a way to do the subscripting in the EOF code
> > by defining lon and lat instead of x and y.
>
> Yes.  Please see the region_ind function.
>
> > 2) Also, the output netcdf file of EOF does not have the
> > lon, lat meta data. Instead it has x and y, which does not
> > overlay when I tried with NARR plot.
>
> I already mentioned the solution.  You need to copy or retain the
> 2-dimensional lat and lon arrays from the input file to the output file.
> Then create a plot with the projection methods in NARR example 4, which
> utilize the 2-D lat and lon arrays.
>
> Alternatively you can copy the Lambert projection parameters and use the
> native grid plot method, as in NARR example 1.  But I recommend against
> this method because it is prone to subtle offset errors.
>
> Can anyone point us to concise basic documentation for how 2-D coordinates
> are structured in a Netcdf file?  The CF website does not include a concise
> explanation of 2-D geographic coordinates.
>
> --Dave
>
>
> On Tue, Jun 16, 2015 at 9:00 PM, Jothi Ganesh Shanmuga Sundaram <
> joshanmugasundaram at mix.wvu.edu> wrote:
>
>> Thanks a lot Dave. It's helpful. I tried running without any coordinate
>> script, it is running and I was able to generate the EOF and EOF-time
>> series netcdf files for the entire domain. I am interested in coordinate
>> subscripting as in the EOF example 1.
>>
>> When I tried to enter the lon lat ranges, the script does not run. It
>> says lon is not a dimension or variable. It accepts x or y, but x and y are
>> some strange values in the data. If I type ncl_filedump file -v lon, I see
>> the values.
>>
>> Here is the ncl_filedump summary.
>>
>> short air ( time, y, x )
>> ..
>> float lat ( y, x )
>> ....
>> float lon ( y, x )
>>
>> 1) Is there a way to do the subscripting in the EOF code by defining lon
>> and lat instead of x and y.
>>
>> 2) Also, the output netcdf file of EOF does not have the lon, lat meta
>> data. Instead it has x and y, which cannot does not overlay when I tried
>> with NARR plot.
>>
>> Here is my script.
>>
>> ;Read the NARR data (Air temperature at 2m)
>> f = addfile ("air.2m.mon.mean.nc", "r")
>>
>> ;Define the parameter, convert the format short to float
>>  p = short2flt(f->air(::12,:,:))
>>
>> ;Calculate the weights
>> w = sqrt(cos(0.01745329*p&y))  ;  p&y works, but if i enter p&lat, it
>> does not work.
>> wp = p*conform(p, w, 1)
>> copy_VarCoords(p, wp)
>>
>> t = wp(y|:,x|:,time|:)            ; x and y works, but if i enter lon or
>> lat, it does not work.
>>
>> ;Number of EOF to be calculated
>> neof = 3
>>
>> ;NCL EOF functions
>> eof = eofunc_Wrap(t, neof, False)        ;EOF (3modesXlonXlat)
>> eof_ts = eofunc_ts_Wrap (t, eof, False)  ;EOF (3modesXtime)
>>
>> ;Export EOF spatial data as netcdfile
>> system ("/bin/rm -f EOFnarr.nc")
>> fout = addfile ("EOFnarr.nc", "c")
>> fout at title = "EOFs of pre 1971-2013"
>> fout->EOF = eof
>>
>> Regards
>> Jothiganesh
>>
>>
>> On Tue, Jun 16, 2015 at 9:58 PM, Dave Allured - NOAA Affiliate <
>> dave.allured at noaa.gov> wrote:
>>
>>> Jothiganesh,
>>>
>>> Yes you can directly read NARR data and compute EOF's without
>>> regridding.  The underlying EOF functions eofunc and eofunc_ts do not use
>>> any spatial coordinates.  They operate purely on numeric arrays.
>>>
>>> The first EOF example uses coordinate subscripting and the lonFlip
>>> function.  If you remove both of these methods from the example, then you
>>> can compute EOF's over the entire NARR grid.
>>>
>>> However, to generate valid plots, you will need to retain the NARR
>>> coordinate metadata, and modify the plot section to use one of the plotting
>>> methods shown on the NARR examples page.
>>>
>>> I recommend one of the projection methods in NARR example 4, because
>>> they use the retained 2-D coordinate grids to directly align the data with
>>> the underlying map.  After you get that working, you can switch back to
>>> Lambert or some other projection of your choice.
>>>
>>> --Dave
>>>
>>>
>>> On Tue, Jun 16, 2015 at 6:39 PM, Jothi Ganesh Shanmuga Sundaram <
>>> joshanmugasundaram at mix.wvu.edu> wrote:
>>>
>>>> Hello,
>>>>
>>>> I use NARR gridded data (349x277 Lambert Conformal Conic grid) in
>>>> netcdf format in NCL. I am able to plot graphics without any problem using
>>>> the existing script in NCL website.
>>>>
>>>> I want to do EOF analysis with NARR data. When I tried to use the EOF
>>>> script available from NCL website to analyze NARR data, it is not working.
>>>> When I checked the data using ncl_file dump the lon, lat formats are very
>>>> different. My question is
>>>>
>>>> 1) Whether I can do EOF analysis in NCL by directly reading NARR data?
>>>> If so, I cannot define the LON LAT value, as NARR has a projected
>>>> coordinates value?
>>>>
>>>> 2) Do I need to regrid the NARR curvilinear to rectlinear data before
>>>> doing EOF? or Is there a way to convert the format of LON, LAT in grid
>>>> without regridding inorder to do EOF analysis in NCL
>>>>
>>>> Thanks for your help.
>>>>
>>>> Regards
>>>> Jothiganesh
>>>>
>>>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20150617/79b7b73b/attachment.html