[ncl-talk] esmf_regrid fails to regrid at some locations

[Lijun Diao]

Hi, Mary:

 

Thanks for your suggestion. After plotting the coordinates of the model grids, I found that all the regridded missing points are out of the model domain. So plotting the model domain by box lines as I did before is not an accurate way to represent the model domain. 

 

--------------------------------------

Lijun Diao

University of Houston

 

From: Mary Haley [mailto:haley at ucar.edu] 
Sent: Wednesday, March 29, 2017 12:31 PM
To: Lijun Diao <ljdiao at gmail.com>
Cc: ncl-talk at ucar.edu
Subject: Re: [ncl-talk] esmf_regrid fails to regrid at some locations

 

Lijun,

 

It's hard to tell what the problem is without looking at both the source and destination grids. I don't know what your original data looks like, or where the grid points fall. It would be easier if you could provide me with the two NetCDF files so I can run the script here.

 

What I usually do is plot both the original data and the regridded data on separate plots, and then add the source grid and the destination grid as a series of points so I can see more clearly where they overlap. Depending on whether the data itself has missing values, this could also affect how it is regridded.

 

You can use gsn_coordinates to plot markers at locations where your original or regridded data is missing and/or non missing. This way you don't need to do all the "ind" calls yourself. See the fourth thumbnail image of example #7 here:

 

http://www.ncl.ucar.edu/Applications/plot_data_on_map.shtml#ex7

 

 

To use gsn_coordinates, replace this code:

 

  ; add polymarkers at missing value locations                

  co_regrid2D = CO_regrid(10,0,:,:)

  indices     = ind_resolve(ind(ismissing(ndtooned(co_regrid2D))), dimsizes(co_regrid2D))

  dims        = dimsizes(indices)

  npts        = dims(0)

  Lon_group   = new(npts,float)

  Lat_group   = new(npts,float)

 

  do n = 0,npts-1

    Lon_group(n) = lon2d_AIRS(indices(n,0),indices(n,1))

    Lat_group(n) = lat2d_AIRS(indices(n,0),indices(n,1))

  end do

 

  pmres                     = True

  pmres at gsMarkerColor       = "black"

  pmres at gsMarkerIndex       = 16

  pmres at gsMarkerSizeF       = 0.02

 

  gsn_polymarker(wks, plot, Lon_group, Lat_group, pmres)

 

with this code (UNTESTED):

  ; add polymarkers at missing value locations                    

  mkres = True

  mkres at gsMarkerIndex            = 16     ; filled dots

  mkres at gsMarkerSizeF            = 3      ; you may need to adjust this

  mkres at gsnCoordsNonMissingColor = "black"

  mkres at gsnCoordsMissingColor    = "red"

  gsn_coordinates(wks,plot,co_regrid(10,0,:,:),mkres)

 

Meanwhile, let me know if you can provide the data files. You can use our ftp site:

 

http://www.ncl.ucar.edu/report_bug.shtml#HowToFTP

 

Thanks,

 

--Mary

 

 

 

 

On Mon, Mar 27, 2017 at 9:36 AM, Lijun Diao <ljdiao at gmail.com <mailto:ljdiao at gmail.com> > wrote:

Hi, NCL talk forum:

 

I am using ESMF_regrid to regrid my modeled CO (dimentioned Lev(15)*Row(299)*Col(459)) to AIRS L2 swath locations (dimentioned 45*30). I attached my code test.ncl for the  regridding and plotting. I found some of the regridded output are missing values. Then I plotted those missing value locations (black dots in the attached figure). The red dashed box defines my model domain. I understand when the satellite swath is  out of the model domain, the regridded CO will be missing values since no model grids will correspond to those locations. But as you can see, some black dots are actually inside my model domain, especially the upper right ones, which are relatively far away from the edges of the model domain. In another words, the esmf regrid program can not regrid accurately for some of the swath locations that are inside the model domain. It doesn’t make sense to me. What is the reason?

 

Thanks,

 

 

--------------------------------------

Lijun Diao

University of Houston

 


_______________________________________________
ncl-talk mailing list
ncl-talk at ucar.edu <mailto: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/20170329/5de9209d/attachment.html