re-converting 1D to 3D array.

NCL users,

 I'm still very raw in my knowledge with ncl, and was wondering if anyone
can help me out.

I'm currently using precipitation data with MM5 (from EGS), and trying to
mask a specific lat and lon point, with the ultimate purpose of outputting
data within that specific grid, and truncating the data to where when I end
up concatenating other files, I am well below the 2gb threshold.

Steps leading to retrieving the desiredvalues in 1D looks fine, and the
output file is well below the original input file size. However, when I
reconvert the 1D to 3D, the file is back to the same size as the input file,
and it looks like missing values are inputted into the output file, which
doesn't make the file any smaller.

Is there a way to reconvert the 1D array to 3D array and keep only the
precip values we want, without the missing values(keep in mind that the
associated metadata needs to be there as well)? A simple onedtond doesn't
look like the solution, and I'm misunderstanding the coding.

The script is as follows. Any help is appreciated.

;*****************************************************
; Read original NetCDF file.
;*****************************************************

  fin = addfile ("pr_MM5I_1979010103.nc","r")

;******************************************************
; Upload netCDF file variables, set dimensions
;******************************************************

  prc = fin->pr ; obtain pr(time,xc,yc) values. Dimension is in
3D
  lat = fin->lat
  lon = fin->lon
  time = fin->time
  Lambert_Conformal = fin->Lambert_Conformal

  dim = dimsizes(prc)
  ntime = dim(0)
  nlat = dim(1)
  nlon = dim(2)

  latlonmask2 = prc(:,:,:) ; assign dim size to masking var
  latlonmask2 = 0 ; initialization

;**************************************************************
; Indicate coordiate you want to analyze. Make sure you
; check the NetCDF file for correct lat lon specification!!
;
; Mask through time, xc, and yc for entire precip variable to
; analyze specifc regions
;**************************************************************

  lat1 = 37
  lat2 = 47
  lon1 = 269
  lon2 = 279

  do k = 1, ntime-1
   do j = 1, nlat-1
    do i = 1, nlon-1
      if (lat(j,i).ge.lat1 .and. lat(j,i).lt.lat2 .and. lon(j,i).gt.lon1
.and. lon(j,i).le.lon2) then
         latlonmask2(k,j,i) = 1
      end if
    end do
   end do
  end do

desiredvalues = mask(prc, latlonmask2, 1)

;********************************************************************
; Convert the masked region to a 1d array, which allows us to then
; use the ncl function "ind" to keep only the data from the region
; we are interested in looking at. This also keeps all the
; metadata associated with each precip value.
;*******************************************************************

desired1d = ndtooned (desiredvalues) ; convert to 1d array
indices = ind (.not.ismissing(desired1d)) ; keep precip values that
have no missing data.
output1d = desired1d(indices) ; all non-missing values
desiredvalues = onedtond(output1d,dimsizes(desiredvalues))
delete (desired1d) ; clean up if you wish.
delete (indices)

;******************************************************************
; Create a new nc output file
;******************************************************************

   system("/bin/rm -f region_output.nc") ; remove any pre-existing
file
   fout = addfile("region_output.nc","c")
   filedimdef(fout,"time",-1,True) ; Makes time an UNLIMITED
dimension (as suggested by NCL)

   fout->prc = output1d

   fout->lat = lat
   fout->lon = lon

end

-- 
Sho Kawazoe
Graduate Student Meteorology
Iowa State University
shomtm62@iastate.edu

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Tue Mar 9 16:55:23 2010