Aspects of spatial data analysis in exploration and evaluation of mineral resources.

The data collected in projects of exploration and evaluation of mineral resources is spatially dependent. It is in this context that its analysis should be carried out. The calculation of the Geostatistical semivariance function in as many directions and lags as the data allows permits to identify different degrees of data's internal organization by using iso-semivariance diagrams. Structures within the data are outlined by these diagrams together with their relative position, dimensions and orientations. The technique is arrived at through the development of several computer programs that calculate the semivariance function, transform co-ordinates and produce contourings. Trials on simulated and geochemical exploration data show the close relationship between data's structure and derived functions. The diagrams are also used to assess goodness-of-fit in polynomial data modelling and to develop a computerized procedure for data filtering as predictive models of spatial variability. This procedure consists of the locally fitting of covariance-weighted regression models. Its applicability as a new method of ore reserve estimation is demonstrated with bore hole data from a porphyry copper and a stratabound base metal deposit. Limitations to the technique are imposed by inadequate sampling patterns or impossibility for defining realistic iso-semivariance diagrams. A practical application of alternative conventional Kriging methods in global evaluation of a Bauxite deposit is also presented. From these results, the application of probability analysis in financial appraisal of mining ventures clearly defines the deposit as an important and feasible project.