Kriging neighbourhood analysis by slope of regression and weight of mean – evaluation with the Jura data set

Abstract

Kriging search neighbourhood is an important parameter for ordinary kriging and can be determined using weight of the mean in simple kriging to define a minimal neighbourhood, followed by use of ordinary kriging slope of regression of theoretical true grades on estimated grades, to find when sufficient samples are included in the search to reach required precision and minimise conditional biases. This testing is often referred to as kriging neighbourhood analysis (KNA). The Jura data set is used to examine this approach using searches of one to four rings of samples to estimate Cd and Co. Performance comparison suggests KNA may overstate benefits of one search over another. Domaining of samples, assumptions of stationarity and sampling accuracy are likely to be far more important in kriging estimation than KNA.