With several boreholes localited, we could determine a spatial interpolation of different geological layer. Would it be possible to determine the errors associated with the interpolation. I used ArcMap 10.0
...that's what I'd also suggest; specifically:
You can select single points (boreholes) by random; you can do that using the tool "Create random points" in the "Data management" folder of ArcToolbox. Within this tool, give your borehole points as "Constraining Feature Class (optional)", so that the randomly created points will be a subset of the existing ones. Thus, you can split your dataset in a "training" and a "validation" subset. Do the interpolation with the training subset, and check the resulting raster values at the points of the validation subset.
Here's the ArcGIS reference for the "create random points" tool:
http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//00170000002r000000
Cheers
Tobias
Stuart and Tobias, thank you very much for your answers. I will follow the tutorial...
I can get back to work
Best regard
All the above is sensible, but why not use kriging? Failing that use IDW with leave obe out cross validation to find the most sensible exponent .
Two observationse
1. Nearest neighbor interpolation is not based on any theory so you can't derive any error results
2. The suggestion to use a form of cross-validation is reasonable but it is based on a number of unstated assumptions. There are a variety of ways to split the data set but no theory to justify a claim that one is better than another, the results you get will depend on the original data set, the same technique with a different set of borehole data could be very different, any claims you might make about errors are likely dubious.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography