Simple Kriging question-please help!!

[JohnNash]

I am very new to ArcGIS and I am having what seems to be a simple problem that I can't seem to figure out. I am trying to do a kriging of precipitation data. I am using ArcGIS 10.2.There are 9 rain gauges which I have displayed the xy data via a spreadsheet. The spreadsheet has 4 columns where one of them is "precipitation'. The input are the data points, precipitation is the z, and everything else i left as default. However when the output results are generated I get a range of 40.47-40.48 (kriging_shee1)for the rainfall. Another hint or insight in the problem is when i made a similar spreadsheet with the values ranging from 2-7 inches of precipitation the kriging worked. It is whenever I change the precipitation values to higher values I seem to get more of an average value, not the range. Enclosed is a jpg which provides insight. Thank you for any help if anyone can figure this out.[ATTACH=CONFIG]28632[/ATTACH]

[MarcoBoeringa]

John,

I can't help you with the exact cause of your observation, but in general about Kriging and interpolations:

- Since kriging relies on modelling the semivariance, it requires a minimum set of datapoints to get any reliable result. You only have 9 sample points, which is a very low number. I really think your dataset is to small. This is also confirmed by the document I attached. It is a nice introductory document to Kriging and Geostatistics written by Gregg Babish of Environment Canada ("Geostatistics Without Tears"). It states in one of its paragraphs:

"Variograms computed from small samples almost always appear erratic, and in many instances the principal cause is too few data. It has been shown (Webster and Oliver, 1992) that variograms computed on fewer than 50 sample points are of little value and that at least 100 data are needed."

- I would strongly urge you to refrain from using the "Simple Kriging" options of Spatial Analyst and 3D Analyst for any real "scientific" application, as it has only very basic options, and doesn't give much insight in what you are doing. I would urge you to use Geostatistical Analyst instead, which offers far more and reliable (visual, graphical and statistical) feedback on your chosen settings and interpolation parameters.

The Geostatistical Analyst Help and some documents here on the ESRI website, also give good introduction to kriging and geostatistics.

*** EDIT ***
I now noticed there is slightly more recent, and extended, 2006 version of the same paper by Gregg. You can download it from iBrarian:
http://www.ibrarian.net/navon/page.jsp?paperid=15417085