Page 283 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS

286 Chapter 8

is also applied to logically integrate the data-driven estimates EBFs of the spatial
evidence maps of epithermal Au prospectivity in the case study area.

Case study application of data-driven EBFs
The objectives of the case study are to illustrate the utility, in mineral prospectivity
mapping, of (a) coherent deposit-type locations and (b) coherent proxy deposit-type
locations. The spatial evidential data sets used in the case study are given in the
introduction section of this chapter. The strategies applied for cross-validation of the
predictive models, portrayed as integrated values of Bel, of epithermal Au prospectivity
in the case study area are given the preceding section.
Calibration experiments were performed not only in deriving data-driven estimates of
EBFs for classes of catchment basin analysis anomaly values (Fig. 8-11) and classes of
proximity to NNW-trending faults/fractures (Fig. 8-13) but also in deriving data-driven
estimates of EBFs for classes of proximity to NW-trending faults/fractures and
proximity to intersections of NNW- and NW-trending faults/fractures. These calibration
experiments resulted in common properly calibrated classes of each of the data sets
(Table 8-IV) with respect to the training sets of 13 known and 11 coherent (out of the 13
known) locations of epithermal Au deposits and 86 randomly-selected and 86 coherent
proxy locations of epithermal Au deposits (Fig. 8-8). The data classes in Table 8-IV are
considered properly calibrated because the plots of the data-driven estimates of Bel
versus the upper limits of the properly calibrated classes with respect to the different
training sets are not noisy (Fig. 8-15) and facilitate recognition of threshold data values
that are associated spatially with the training sets of deposit and proxy deposit locations.
The graphs in Fig. 8-15 indicate that the training deposit and proxy deposit locations are
mostly within 0.3 km of NNW-trending faults/fractures and within 1.5 km of NW-
trending faults/fractures and intersections of NNW- and NW-trending faults/fractures
and that the training deposit and proxy deposit locations coincide mostly with integrated
PC2 and PC3 multi-element scores greater than 0.3. These results are generally
consistent with the results of analyses of spatial associations between the same spatial
data sets and the locations of epithermal Au deposits in the study area (see Chapter 6,
Table 6-IX).
The graphs in Fig. 8-15 show that, for the range of data values that are associated
spatially with the deposit and proxy deposit locations, the applications of the training
sets of coherent deposit and proxy deposit locations result in higher values of Bel than
the applications of the training sets of all deposit locations and randomly-selected proxy
deposit locations. Conversely, the graphs in Fig. 8-15 show that, for the range of data
values that lack spatial association with the deposit and proxy deposit locations, the
applications of the training sets of all deposit locations and randomly-selected proxy
deposit locations result in higher values of Bel than the application of the training sets of
coherent deposit and proxy deposit locations. These results imply that, compared to the
applications of all deposit locations and randomly-selected proxy deposit locations, the