Spatial Analyst/Interpolation versus Geostatistical Analyst Wizard.

Discussion created by wannaz on Aug 30, 2010
Latest reply on Sep 2, 2010 by wannaz
Dear all,

     I have interpolations to perform of several hundred gridded (global 2x2.5 degrees) datasets that I need as 4000x2000 pixels rasters. I am currently testing a script that loops over all the datasets. I started with the Geostatistical Analyst's Wizard (Ordinary Kriging, default/automatic parameters) on one particular dataset and I obtained a smooth interpolation as shown on the first page of the attached document. The parameters (range, sill, ..) automatically determined by the wizard seemed to work. I then used the Spatial Analyst/Interpolation/Kriging tool, hoping that  parameters (that seem to be optional) would be determined the same way, and I obtained what is displayed on the second page of the attached document. There is clearly a structure appearing. I remember having tuned interpolation parameters for getting rid of such structures years ago (and under MATLAB), but I had to detect them visually before acting, and I cannot do this for all the datasets that have to be interpolated.

So my question: as my script is based on the Spatial Analyst tool (I don't think that the wizard's internal tool(s) is/are available as methods of the Geoprocessor (?)), is there a method(ology) for me to evaluate interpolation parameters the way the wizard does (so I can setup gp.Kriging_sa() the same way)?

The issue for me with any "interpolation artifact" is that I have to perform then a zonal statistics with zones that can be as small as 10km by 10km "squares". I wanted to have at least one pixel center per zone, and for these small zones I can clearly not count on any averaging effect that would smoothen to some extend the artifacts.

Thank you and best regards,