A Localized Indicator Kriging (LIK) methodology was developed and applied to the Phoenix gold-skarn deposit, which has been historically difficult to estimate because of its low-grade and high nugget effect. Previous resource models used bias correction factors obtained from analysis of paired blast holes with exploration composites in mined out areas. It was thought that poor reconciliation resulted from various data types. However, theoretical corrections did not work well when applied to mined out areas requiring continual revisions. The new LIK approach appears to be more robust in dealing with the nuggetty, low-grade nature of the deposit and is working well without correction factors.
The Phoenix deposit can be simplified to consist of ?good? host rocks such as the Lower Battle Formation and portions of the Antler Peak formation where high-grade mineralization (> 2.0 g/t) can be reasonably continuous and ?bad? host rocks where high-grade mineralization is isolated and without significant volume. Previous models were challenged by estimation domains being based on structural domains within geologic formations that contain both good and bad host rocks.
There are no reliable geologic criteria for separating good and bad host rock due to the only difference being the apparent continuity of the gold mineralization. As a proxy for geologic control, a probability based approach was developed to distinguish these two sub-domains. For example, within the Battle Formation, a visual inspection concluded that the gold in the bad host rock did not have any apparent continuity above 1.5 g/t, identifying this grade as a probability threshold that was subsequently used to separate the two host rocks. Once the two sub-domains were established, LIK was used to estimate each, where the only difference was the value of the upper indicator class (bad host rock = 1.25 g/t, good host rock = 2.9 g/t). The resulting model produced an almost perfect match with the underlying production data.