What kriging technique to use for environmental data

Discussion created by ttamba on Jul 3, 2014

We are actually working on the estimation of potential soil contamination based on the volume of contaminated soil on the territory characterized using geostatistical approach. By performing exploratory data analysis, we have realized that they are far from being normally distributed (Gaussian); even with a log normal transformation.
I'd like to know what kriging technique is better suited to predict a reliable grid interpolation based on data from environmental, that are "almost never normally distributed"?
I tried the techniques of ordinary and simple kriging with different methods proposed by Geostatistical Analyst, but never get a reliable prediction, even statistically.

So my post sould be a question as well a discussion.

Thank you for your help and your advice.