Apologies to Heather for rather hi-jacking her thread.....
Thanks to Xander for providing that link to the thesis on hydro surveys in a shipping channel.
I took some time to read through it.
Despite reams of x-validation stats and what-not, it took till page 80 odd to actually see some picture of what the actual interpolations looked like.
If the author hadn't actually placed a polygon on each map showing where the "channel" was supposed to be, it would be hard to spot! So, from a visual perspective, non of the techniques discussed did any sort of reasonable job. If fact, awful.
I was particularly amused by the universal kriging version.
This highlights a particular problem with traverse based data like this. High density of sample points along traverse lines, and a large distance between lines (very high aspect ratio). Non of the traditional interpolation techniques are good at making sense of this sort of data.
Perhaps Dan would like to make a separate thread for these sorts of discussions.