We are using Indicator kriging in Geostatistical Analyst to develop a map of the probabilities of nitrate concentrations in groundwater exceeding a specified threshold. The resulting surface appears very blotchy with unnatural looking areas of differing probabilities (rectangles, triangles, etc.). When we use ordinary kriging to create a surface of the estimated nitrate concentrations, the surface appears much more naturally varying as one would expect. Can anyone offer any insight on why indicator kriging produces results like this and potential solutions. Thanks.
Have you tried to normalize the data prior to running the indicator krig? And how many samples are you including in the analysis?