I suspect this has an easy answer, and is probably something I’ve already done (!), but the solution is eluding me. Everything, input/output, is in gridline registration.
The starting oceanic input grid has a 0.2° resolution with -R-180/179.8/-89.8/89.8 bounds (1800x899).
I need to transform it to a grid with 0.25° resolution with -R-179.875/179.875/-89.875/89.875 bounds (1440x720).
So far, I’ve been trying to do this via grdsample, but I can’t seem to get exactly the bounds I need. It comes back with -R-179.875/179.625/-89.625/89.625. I guess grdsample can’t “wrap-around” my input grid to provide longitude 179.875. How can I fix that?
I even tried using surface, which provides the -R needed, but it interpolates values for cells that are located over land (that can be masked), but it also interpolates non-land cells that were missing (invalid) in the original grid.
Summarizing, there are a list of issues regarding a couple of methods I’m attempting to use:
- How best to extend the grid to the poles and to east-most edge for use in grdsample, or
- How to create and apply a mask based on the original grid to the grid interpolated by surface?
Or is there some really simple approach that totally escaped me?!
should it not be -R-179.75/179.75/-89.75/89.75 with 0.25 degree resolution and gridline registration? the “.875” decimal part seems off.
Not sure what it “should” be. I need to match this:
mn_201810.grd: Gridline node registration used [Geographic grid]
mn_201810.grd: Grid file format: nf = GMT netCDF format (32-bit float), CF-1.7
mn_201810.grd: x_min: -179.875 x_max: 179.875 x_inc: 0.25 (15 min) name: longitude n_columns: 1440
mn_201810.grd: y_min: -89.875 y_max: 89.875 y_inc: 0.25 (15 min) name: latitude n_rows: 720
grdsample does not let extrapolate. That is, you cannot go from 179.8 to 179.875
-R-179.75/179.75/-89.75/89.75 is not global. It misses the poles and one column.
But the longitudes wrap around. How can I trick GMT to treat it that way? In other words, the column for longitude -180 should be the same as +180.
I guess one could dump the original file with grd2xyz, copy the values for the -180 longitudinal column, paste it to the end of the file, while changing the longitude to +180, and finally run that back through xyz2grd & then through grdsample.
I’m not so worried about the missing pole values.
-179.875/179.875
is global at 0.25 and grid registration. It’s the same as -180/180 (same number of columns) and pixel registration. It’s just that you cannot do it with grdsample. Can’t follow what are the issues with the surface reinterpolation.
Background on the surface interpolation:
The original grid (at 0.2° resolution) has invalid (NaN) values in non-land oceanic Arctic regions.
Surface will interpolate values for every cell of a grid. It’s trivial to mask the land values.
After interpolating with surface, how can the newly interpolated cells be re-set to match (almost) those NaN-valued non-land cells found in the original grid?
Have you tried surface -M0c?