Modeling An Orebody Having Lognormal Grade Distribution ?Introduction

Lognormal grade distributions are definitely the rule rather than the, exception in many types of deposits, including precious metals, uranium, and many others. The "normal theory" geostatistical techniques do not work well in these situations because the occasional high grade values, perhaps several orders of magnitude larger than the median, can overwhelm the rest of the data, resulting in variograms and Kriged estimates which are unrealistic or even ridiculous. Robust techniques for variogram modeling have been proposed, and truncation of high sample values can produce more reasonable Kriged estimates, but the technique of working with log transformed data appears to be the best method of handling lognormal data. This paper will discuss primarily the two parameter lognormal case, with some discussion of the three parameter lognormal, but no discussion of other methods of transforming data, such as disjunctive Kriging.