Grade estimation with samples of different length

Samples from different sampling campaigns usually have distinct sampling lengths. All the available information should be incorporated when constructing a grade model. However, the variations in length among the samples must be considered during estimation. We propose to perform kriging using samples of different support. The kriging system is built using average covariances to account for the difference in support between the samples. The technique is applied in a mineral deposit. The resulting grade model is validated with visual inspection and swath plots. Cross validation was used to compare the proposed approach against kriging using point covariances, which disregards the difference in support between the data. Results show that the use of average covariances in the kriging system improved both the accuracy and the precision of the estimates.