I am developping a 3D model of soil texture and would like to give more weight to the interpolation horizontally than vertically. I have a data set of 300K points of soil texture in mm, with data points at every meter interval and at top of intervals. Dataset is based on borehole information.
A - what would be the most suited Filter for this, Resample to Image, Gaussian Resampling, structure vs unstructured grid?
B-how to give more weight horizontally to the interpolation?

I think the Gaussian Resampling filter is closest to what you want.

Gaussian Resampling does support anisotropic splatting kernels, but to only way to orient the kernels is by a normal field, which would be a pain to add to your data. A quick a dirty way to get what you want would be to use the Transform filter to squish your data horizontally, then apply Gaussian Resampling, and then apply Transform again to the result to spread the grid back out to normal size.

So, for example, assuming by â€śhorizontalâ€ť you mean along the x-axis and you want the sampling to have twice the spread in that direction, you would first apply the Transform filter with a Scale of [0.5,1,1]. Then you would apply the Gaussian Resampling filter. Then you would apply another Transform filter with a Scale of [2,1,1].