I am trying to do interpolation in a coastal area with data points in several rivers emptying into a bay and on opposite sides of peninsulas.
I have created a coastline layer to use it as a barrier but cannot figure out how to do this in ArcGIS 10.
I do not see a barriers object anywhere in any of the kriging tools.
I tried just doing the kriging then clipping it to the land, but I get unrealistic results where data points close together but separated by land create problems.
I see online documentation for doing this in geoprocessing scripts in 9.3 but nothing similar for 10.
Best Answer
Do the rivers run generally in a specific direction? If so, you could use directional binning along the centerlines of the rivers to constraint the empirical semivariograms only to point pairs running along the length of the rivers. (See here for more details: http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#/Exploring_directional_autocorrelation/0031000000m2000000/)
Another option would by to partition the point samples into separate datasets and run your kriging on each set separately. This might be more appropriate anyway, since each river might have different statistical models to them; but would not work if your sample size is small for each river.
Do you need to use a geostatistical method, or would a deterministic method work? I think you might be able to achieve what you want using Diffusion Interpolation with Barriers. You could use your land as an extremely high cost barrier. http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//00310000002s000000.htm
Also, here is Diffusion Interpolation with absolute barriers as a geoprocessing script tool. Your coastline would make a perfect absolute barrier. http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#/Diffusion_Interpolation_With_Barriers/003000000005000000/