Visualized Geostatistical Analysis Of Nome Offshore Gold Placer Deposit Using ARCGIS - A Case Study

Abundant beach and offshore resources of gold deposits are demonstrated along the coast of Alaska, and Nome is one of the richest places. Considerable amounts of mineral resources data were collected and stored in geodatabase of ArcGIS. In searching for the best model for the interpretation of the Nome offshore gold placer deposit and accurate estimation of the reserve, the data were processed and visualized using ArcGIS. The gold concentration data collected from the Nome offshore deposit are of characteristics such that: (1) near lognormal distribution of data, (2) high positive skewed distribution, (3) a small number of extreme values dominating the reserve estimation, and (4) apparent anisotropy of the variogram. These characteristics make the development of an interpolation model for the gold deposit a challenging task. In this study, it is determined that five geostatistical methods provided by ArcGIS with visualization capability, i.e., Inverse Distance Weighted (IDW), Ordinary Kriging (OK), Ordinary Kriging with lognormal transformation (OK-log), Simple Kriging (SK) and Indicator Kriging (IK) are suitable candidates for the development of an interpolation model. The five algorithms are evaluated and compared with each other in the study. Cross-validation technique is employed to examine the accuracy of the five approaches. It is determined that the OK and SK algorithms are not suitable for this dataset because of the difficulty of fitting the semivariogram model despite their lower RMSE (Root-mean-square error) and the higher RMSS (Root-mean-square Standardized). The OK-log algorithm does not provide satisfactory prediction either, because it places too much weight on the role of extreme values resulting in overestimation and extremely high RMSE. Reserve estimation of each blocks is conducted using each of the interpolation methods and compared with the reserve calculated using the conventional polygon method. The IWD and IK algorithm seem to provide estimations more agreeable with the conventional polygon method.