Abstarct: Mineral resources/reserves classification and their definition have became a critical subject for mining industry since 1970. Up to now, many classification codes have been published such as JORC, SME, SAMREC and UNFC codes for this issue. Albeit of all efforts made to converge these codes, there is no unique criterion to classify whole ore resource/reserve, which is due to the intrinsic differences among ore deposits. Qualitative (classic) and quantitative (geostatistical) criteria have been proposed to classify ore deposit. Qualitative criteria are more subjective and used in preliminary steps of the deposit evaluation, while in the detailed exploration steps, geostatistical criteria are preferable to be more objective. In this thesis, geostatistical criteria are introduced and a comparison is made among them through a simulated synthetic model to know their weakness and strengths. A crucial point in the application of such criteria is defining the proper thresholds for them. In this thesis a new method is proposed for determination or calibration of threshold of criteria, baxsed on the calculated relative error upon the better known part of deposit. The sensitivity of the geostatistical criteria has been analyzed to change of support, because mineral resources/reserves classifications strongly dependent on selective mining unit. The classification criteria are compared and high-performance criteria which affect less by change of support are introduced. At the last chapter of this thesis a case study performed on Songun copper deposit. Classification of the Songun copper deposit has been done with using the proposed method in this thesis for different defined ores (Leach, Sulfide, and Hypogenous). The optimal threshold has been calculated for each criteria using the blast holes which belong to mined part of Sungon and applied for whole of deposit. It should be mentioned that all the computer codes used for estimation and classification of synthetic model and Songun copper deposit have been done in MATLAB¨ software.