[ncl-talk] Regridding from fine Rectilinear data to coarse curvilinear grid

[Yuqiang Zhang]

Hi Dennis,

 

Thanks for your detailed explanation of the “bin_sum”. 

 

I have used the “ESMF_Regridding” functions to regrid Global Emissions data in fine rectilinear  grid to Regional scale in coarser rectilinear grid. I thought there are some interpolations happens during the regridding. Since I am regridding the population data, which needs to sum up the data in fine resolution into a coarse resolution, I hesitate to use the ESMF regridding function again. However, thanks for your advice, I will check whether Example_9 works for my case. I also found the this function “rgrid2rcm”, and I will also double –check about that. 

 

Have a good night! 

 

Regards,

Yuqiang

 

From: ncl-talk-bounces at ucar.edu [mailto:ncl-talk-bounces at ucar.edu] On Behalf Of Dennis Shea
Sent: Wednesday, October 22, 2014 4:50 PM
To: Yuqiang Zhang
Cc: ncl-talk at ucar.edu
Subject: Re: [ncl-talk] Regridding from fine Rectilinear data to coarse curvilinear grid

 

The 'bin_sum' function was developed to process multiple satellite swaths onto a *rectilinear* grid. Each swath's latitude and longitude arrays are typically 2-dimensional (think, curvilinear). Hence: curvilinear ==> rectilinear. More specifically, the ndtooned function is used on the lat2d, lon2d and data array so, internally the data are treated as random. The reason for confining the output to a rectilinear grid is that a simple 'location finding' algorith is used to determine the location of each input  data.

You have 'shoe-horned' the 'bin-sum' function to do just the opposite. Yes, it works BUT you are invoking the function for EACH 'fine resolution' grid point via the double do-loop. EACH invocation requires the underlying code to reinitialize. This imposes a huge penalty.

===

The procedure can go fron one rectilinear grid to another.

An example of going from an input  high ('fine') resolution rectilinear grid to an output coarser rectilinear grid is

http://www.ncl.ucar.edu/Applications/binning.shtml

Example 3


===

It may be best to use ESMF Regridding

http://www.ncl.ucar.edu/Applications/ESMF.shtml

See: Example 9

Likely, 'conservative remapping' is what you want

So, your 'src' (source) information is the rectilinear data.

The 'dst' (destination) is the curvilinear grid.

Note: the weight file need be generated only once. Subsequently, you can reuse it.

Good Luck

 

 

 

On Tue, Oct 21, 2014 at 9:31 PM, Yuqiang Zhang <yuqiangzhang.thu at gmail.com> wrote:

Hi All,

 

I need to regrid the population data from fine rectilinear grid to coarse curvilinear grid. 

Procedure “bin_sum” will not work very efficiently for me, as 

 

Bin_Sum

“                      gbin [*][*] : numeric ,  

                        gknt [*][*] : integer ,  

                        glon    [*] : numeric ,  

                        glat     [*] : numeric ,  

                        zlon     [*] : numeric ,  

                        zlat      [*] : numeric ,  

                        z          [*] : numeric    ”

The targeted output “glon” and “glat” are one-dimension and need to be equals to the rightmost and leftmost dimension of the gbin. However, I need all of them at the same dimension (lat2d, lon2d). What I am doing right now is loop over each grid cell, and then apply the “bin_sum” procedure. It works fine, but just run very slowly. I copied part of my scripts down here:

 

;*****************************************************************

; Variables to hold binned quantities

;*****************************************************************

 gbin  = new ( (/nlat,mlon/), float )

  gknt  = new ( (/nlat,mlon/), integer)

 

  gbin_temp = new ((/1, 1/), float )

  gknt_temp = new ((/1, 1/), integer)

 

  gbin_temp  = 0.0                      ; initialization

  gknt_temp  = 0

 

  printVarSummary(gbin_temp)

;*****************************************************************

;loop over the all the latitude and longitude in target grid

do I = 0, mlon-1

  do J = 0, nlat-1

  bin_sum(gbin_temp,gknt_temp, glon(J,I), glat(J,I), ndtooned(zlon2d), ndtooned(zlat2d), ndtooned(z) )

 

  gbin(J, I) = gbin_temp(0,0)

  end do

  print("zyq finish "+ I)

end do

 

Any one has the idea to improve this? 

 

Thanks for all your time! 

 

Regards,

Yuqiang 

 

 

 

 

Regards,

Yuqiang

 


_______________________________________________
ncl-talk mailing list
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/20141022/70a6784d/attachment.html