raster resampling gurus?

Discussion created by bruce.simonson on Nov 13, 2013
Hi folks,

I have a large set of 1 meter grid DEMs which I wish to resample to coarser resolutions, for processing with the Hydrology toolset.

I remember reading, somewhere, that cubic convolution is the preferred resampling operation for topographic DEMs.  In other threads, I think I remember reading that going from 1x1 to 3x3 grids is advisable for the cubic resampling algorithm.  Perhaps this has to do with the cubic operation dealing easily, and well, with what I might call tic-tac-toe grids (9 cells, in a 3x3 layout).

So, if I do this in two resamples: 1x1 -> 3x3 -> 9x9, I can see how this would work, if the tic-tac-toe geometry is really relevant to the cubic convolution, to get to 9x9.  And from there, to 81x81, if I want.

However, does this cubic operation work well, and appropriately, if I want a final 5x5 grid?  I mean, going directly from 1x1 to 5x5.  This would apply the cubic operation against a 5x5 grid (instead of a tic-tac-toe geometry of 3x3).  Would this compromise the accuracy of the results? 

If not, that would be great for me (i.e., if the underlying geometry for the cubic operation doesn't have to be a power of 3). My source grids are 1500x1500, and I would like to resample to 30x30, or 50x50 (for two example resampling sizes).  This would allow me to maintain the boundaries of my source grids, and not have to deal with edge effects.   (I.e., I want my final resampled size to divide evenly into 1500x1500).

Does the cubic operation work (well) with source geometries that resample from 1x1 to 5x5, or 30x30, or 50x50, for some examples?


PS:  Don't know if these layouts are particularly helpful, but this is how I imagine things might work.  In a 3x3, the value of the new grid would be a function of the center cell and the surrounding 8 cells; in a 5x5 the value would be a function of the center cell, and the surrounding 24 cells.  Does the cubic operation care?

[x][x][x]      [---]
[x][x][x]      [-X-]
[x][x][x]       [---]

[x][x][x][x][x]  [-----]
[x][x][x][x][x]  [-----]
[x][x][x][x][x]  [--X--]
[x][x][x][x][x]  [-----]
[x][x][x][x][x]  [-----]