A better interpolation method for creating Chloride Concentration contour map

[ShuhongYan]

Hi, I am not very familar for spatial analysis knowledge. I have around 30 wells that collected sample Chloride concentration data and I want to create a contour map based on these sample data. 

Does anyone know a better way to create this contour map? I know Kriging is the best way, but I only have 30 points data which is too few for Kriging. 

Any other interpolation method recommended to replace Kriging like IDW, or natural neighbor? Or any way to add more sample data technically based on 30 sample points so that Kriging can be applied?

I have Spatical Analyst and 3D Analyst license in ArcGIS Desktop. I also heard of Geostatistical analyst tool, but even  thought I can understand the analyzed result in the tool or transform the sample data, I think the number of sample data is still the problem. I think it's a normal situation for lots of people since it's difficult to acquire over 100 well data in a small study area.

Any help would be really appreciated!

Shuhong

[SteveLynch]

Shuhong

Please post an image of the output generated using IDW and Natural Neighbor.

Thanks

-Steve

[ShuhongYan]

Hi Steve,

This is the output by using IDW (classified by natural breaks for symbolization)

This is the output by using Natural Neighbor (classified by natural breaks for symbolization)

[DanPatterson_Retired]

Why don't you create Thiessen polygons/Voronoi diagram of the pattern you have given the irregular spacing of your points (and highly clustered in some areas.  It might give you a better spatial picture initially

[ShuhongYan]

I created a Thiessen polygon as you said and I understand the theory of Thiessen Polygons, but how can I use this output polygon to create the contour map?

[DanPatterson_Retired]

maybe contours aren't the best way to represent the data and that should be considered as sell

[SteveLynch]

What do you need the contours for?

From what I see in the IDW/Natural Neighbor output is that there are very steep gradients which will produce many contours.

The best way to decide which interpolation suits your data best is to do validation and/or cross validation.