The four individual grids are all pixel grids and have 12.5 m spacing. They do not, however, have boundaries that follow a common X + n * dx pattern. For instance, AP_14636_FBD_F6850_RT1_cutted.dem.grd starts at x = 771676.6875 and AP_14636_FBD_F6860_RT1_cutted.dem.grd starts at 759256.6875. The difference is 12420 meters but that shift is not divisible by 12.5. So one is phase shifted with respect to the other, i.e. the second grid’s west is not an integer number of 12.5 meters from the other grids west. Etc.
Now, none of that should lead to any crash of other issues, so I will look at that, but this is why you are getting those messages. One grid’s nodes do not overlay with the other grid’s nodes hence resample is required.
Found the cause of the crash in grdsample, and fixed in master. Thanks for providing a good test case. In general, if you are able to avoid using a -R that forces resampling you would be better off. Resampling as in this case leads to loss of the highest frequencies which surely is not desirable.
Yeah,actually I calculated every range of the subgrid and resampled it before use grdblend(without changing the spacing),and avoid to resample in grdblend.
The only disadvantage is too inconvenient.
As you can see here:
