interpolation to a grid with missing data on it

I have a netcdf file with 2000 unstructured points, 1800 of them have missing data and only 200 have data on it. (in.nc)

I want to create a regular lat lon grid and for that I use:

cdo -remapdis,grid.txt,4 in.nc grid.nc

the problem is that remap considers a missing value point as a valid point to interpolate and if all the 4 nearest points are missing value, it puts missing value in the interpolated grid point.

The only way I can avoid this, is to convert all the 2000 values to text, remove the 1800 nodata values, and convert the 200 good values to netcdf again, which is not a smart way to do it...

Is there a way to remove the missing data from the netcdf, or teach remap not to use missing data in the interpolation?

Regards

Joao

Replies (7)

RE: interpolation to a grid with missing data on it - Added by Karin Meier-Fleischer 5 months ago

What about setting the missing values to zero before remapping?

cdo -remapdis,grid.txt,4 -setmisstoc,0 infile outfile

RE: interpolation to a grid with missing data on it - Added by joao ferreira 5 months ago

I cannot change missing data to 0 as you suggest... There is no physical reason to put precipitation=0 on those points... The resulting interpolation grid would have those 0 ingested everywhere...

RE: interpolation to a grid with missing data on it - Added by Karin Meier-Fleischer 5 months ago

Maybe you have to set the environment REMAP_EXTRAPOLATE='off' (default is True) before you call cdo (see https://code.mpimet.mpg.de/projects/cdo/embedded/cdo.pdf#subsection.2.12.4).

export REMAP_EXTRAPOLATE='off' cdo -remapdis,grid.txt,4 infile outfile

Without the data and grid file I can only guess.

RE: interpolation to a grid with missing data on it - Added by joao ferreira 5 months ago

thanks

I don't believe it's a extrapolation problem...

I'm sending the files for a smaller area (73 points on which 50 of them are missing data)

The problem is the same...

When someone performs an interpolation to obtain a grid, it does not make much sense to obtain missing data on the interior of that grid...

Of course, if I do:

cdo -remapdis,grid.txt,73 input.nc output.nc
ncatted -a featureType,global,d,, output.nc

I won't have missing data in the output file, but I don't think this is the correct way to proceed...

Download all files

grid.txt (121 Bytes) grid.txt
input.nc (48.9 KB) input.nc

RE: interpolation to a grid with missing data on it - Added by Karin Meier-Fleischer 5 months ago

export REMAP_EXTRAPOLATE='off'
export CDO_GRIDSEARCH_RADIUS=0.001

cdo remapdis,grid.txt,4 input.nc output.nc

RE: interpolation to a grid with missing data on it - Added by Karin Meier-Fleischer 5 months ago

Btw, the grid.txt file does not contain the area of all stations in input.nc.

The grid defined in grid.txt is the area where x- and y-axis grid lines both are drawn. The upper right where cut off.

plot_area_and_stations.png (44.9 KB) plot_area_and_stations.png

RE: interpolation to a grid with missing data on it - Added by joao ferreira 5 months ago

The data outside the domain is not important

export CDO_GRIDSEARCH_RADIUS=0.001
what do you mean?
I don't think this is correct...
Only search in a 0.001 degrees radius? Like that I won't find any valid data...

what i did was:

cdo -setmisstoc,-999 input.nc dumb_nomiss.nc

cdo -f nc -gtc,0 dumb_nomiss.nc mask.nc

cdo -reducegrid,mask.nc dumb_nomiss.nc input_nomiss.nc

cdo -remapdis,grid.txt,4 input_nomiss.nc out.nc

(1-7/7)

Project

General

Profile

CDO