| GRDSAMPLE(1gmt) | GMT | GRDSAMPLE(1gmt) |
grdsample - Resample a grid onto a new lattice
grdsample in_grdfile -Gout_grdfile [ -Iincrement ] [ -Rregion ] [ -T ] [ -V[level] ] [ -fflags ] [ -nflags ] [ -rreg ] [ -x[[-]n] ]
Note: No space is allowed between the option flag and the associated arguments.
grdsample reads a grid file and interpolates it to create a new grid file with either: a different registration (-r or -T); or, a new grid-spacing or number of nodes (-I), and perhaps also a new sub-region (-R). A bicubic [Default], bilinear, B-spline or nearest-neighbor interpolation is used; see -n for settings. Note that using -R only is equivalent to grdcut or grdedit -S. grdsample safely creates a fine mesh from a coarse one; the converse may suffer aliasing unless the data are filtered using grdfft or grdfilter.
When -R is omitted, the output grid will cover the same region as the input grid. When -I is omitted, the grid spacing of the output grid will be the same as the input grid. Either -r or -T can be used to change the grid registration. When omitted, the output grid will have the same registration as the input grid.
Regardless of the precision of the input data, GMT programs that create grid files will internally hold the grids in 4-byte floating point arrays. This is done to conserve memory and furthermore most if not all real data can be stored using 4-byte floating point values. Data with higher precision (i.e., double precision values) will lose that precision once GMT operates on the grid or writes out new grids. To limit loss of precision when processing data you should always consider normalizing the data prior to processing.
By default GMT writes out grid as single precision floats in a COARDS-complaint netCDF file format. However, GMT is able to produce grid files in many other commonly used grid file formats and also facilitates so called "packing" of grids, writing out floating point data as 1- or 2-byte integers. (more ...)
Resample or sampling of grids will use various algorithms (see -n) that may lead to possible distortions or unexpected results in the resampled values. One expected effect of resampling with splines is the tendency for the new resampled values to slightly exceed the global min/max limits of the original grid. If this is unacceptable, you can impose clipping of the resampled values values so they do not exceed the input min/max values by adding +c to your -n option.
If an interpolation point is not on a node of the input grid, then a NaN at any node in the neighborhood surrounding the point will yield an interpolated NaN. Bicubic interpolation [default] yields continuous first derivatives but requires a neighborhood of 4 nodes by 4 nodes. Bilinear interpolation [-n] uses only a 2 by 2 neighborhood, but yields only zero-order continuity. Use bicubic when smoothness is important. Use bilinear to minimize the propagation of NaNs.
As an alternative to bicubic spline, linear spline or nearest neighbor interpolation one can instead send the entire dataset through surface for re-gridding. This approach allows more control on aspects such as tension but it also leads to a solution that is not likely to have fully converged. The general approach would be something like
gmt grd2xyz old.grd | gmt surface -Rold.grd -Inewinc -Gnew.grd [other options]
For moderate data set one could also achieve an exact solution with greenspline, such as
gmt grd2xyz old.grd | gmt greenspline -Rold.grd -Inewinc -Gnew.grd [other options]
To resample the 5 x 5 minute grid in hawaii_5by5_topo.nc onto a 1 minute grid:
gmt grdsample hawaii_5by5_topo.nc -I1m -Ghawaii_1by1_topo.nc
To translate the gridline-registered file surface.nc to pixel registration while keeping the same region and grid interval:
gmt grdsample surface.nc -T -Gpixel.nc
gmt, grdedit, grdfft, grdfilter, greenspline, surface
2019, P. Wessel, W. H. F. Smith, R. Scharroo, J. Luis, and F. Wobbe
| May 21, 2019 | 5.4.5 |