This question already has an answer here:
I have regularly spaced multibeam data. I'm not sure if the data has been subject to some interpolation already as it was downloaded from the UK hydrographic office. I have emailed to find out how the data has been treated but I'm making the assumption that the xyz points that I have are surveyed points and not interpolated values. I have interpolated using Ordinary Kriging and have produced a good DEM.
I also have a similar dataset however it is not regularly spaced (appears to be raw xyz) and has had no prior treatment aside from removing noise.
With the second multibeam dataset would I be right in my belief that creating a TIN model would preserve elevation values most efficiently?
My reckoning is that if I subsample the data to 2m spacing (the same as the first multibeam dataset) and Krig it that I will be removing surveyed points and estimating them with Kriging which makes no sense and does not honour the elevation efficiently.