Page 302 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS

Data-Driven Modeling of Mineral Prospectivity 305

Fig. 8-24. (A) Epithermal Au prospectivity map of Aroroy district (Philippines) portrayed as
discriminant scores of spatial evidence layers with respect to training set BB of 86 coherent proxy
deposit-type locations (Fig. 8-8) and 9633 non-deposit locations. Polygon outlined in grey is area
of stream sediment sample catchment basins (see Fig. 4-11). The testing set of locations of 13
epithermal Au deposits is shown as reference to the prediction-rate. (B) Fitting and prediction-rate
curves of, respectively, proportions of coherent training proxy deposits (grey dots) and testing
deposits (black dots) demarcated by the predictions versus proportion of the study area predicted
as prospective based on the discriminant scores. The grey and black dots represent classes of
discriminant scores that correspond spatially with certain numbers of training coherent proxy
deposit-type locations (in grey) and certain numbers of testing deposit-type locations (in black),
respectively.

proximity to intersections of NNW- and NW-trending faults/fractures. The patterns of
intermediate and high discriminant scores in Figs. 8-23A and 8-24A are somewhat
similar to the patterns of intermediate and high discriminant scores in Fig. 8-21A and 8-
22A and thus are more-or-less consistent with the conceptual model of epithermal
mineralisation in dilational or extensional settings as depicted in Fig. 6-16.
The fitting- and prediction-rates of the maps of discriminant scores based on training
set AA (Fig. 8-23A) and based on training set BB (Fig. 8-24A) are similar. However, if
20% of the study area is considered prospective, then the fitting-rates of the map of
discriminant scores based on training set BB (Fig. 8-24B) are better than the fitting-rates
of the map of discriminant scores based on training set AA (Fig. 8-23B). This means
that, in mineral prospectivity mapping, using coherent (proxy) deposit-type locations is
better than using randomly-selected (proxy) deposit-type locations. The results also