The researches were done in the southern part of Lubin deposit, which is prepared for exploitation of copper ore. The aim of statistical and geostatistical researches was to describe the variability structure of Pb and Zn contents (% weight), accumulation index of Pb and Zn. The results of the two kinds of deposit sampling in the mine workings were used in the analysis: routine, mine sampling made by geological staff and experimental sampling in small, local scale of observation. The geostatistical modelling of studied parameters using classical semivariogram is very difficult, because of the outliers which mask-random component of metals variation. Easy to interpret and model structure of metal contents variability were obtained due to Inverted Covariance semivariogram. The kriging standard errors of mean metal contens estimation within blocks of horizontal dimensions 1000 x 1000 are unrealistic high, between 50% and 100%. Additional information, originated from the experimental sampling, used in geostatistical modelling do not change substantially the assessment accuracy of metal contents. Similary, cokriging and probability kriging bring accuracy of mean metal contens assessment the reserves of Pb and Zn can be classified as inferred only.