Page 296 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS
P. 296
Data-Driven Modeling of Mineral Prospectivity 299
locations in training set B than to the discrimination between the random-selected proxy
deposit-type locations and non-deposit locations in training set A. This implies that the
coherent proxy deposit-type locations have stronger spatial association with NNW-
trending faults/fractures compared to randomly-selected proxy deposit-type locations.
In both of the two discriminant models, the contributions of the ‘ANOM’ predictor
variables (i.e., classes of high integrated PC2 and PC3 scores obtained from the
catchment basin analysis) are more-or-less the same but are subordinate to the
contributions of the ‘FI’ and ‘NNW’ predictor variables. In both of the two discriminant
models, the contributions of the ‘NW’ predictor variables (i.e., classes of proximity to
NW-trending faults/fractures) are the most inferior. These results suggest that the
presence of multi-element geochemical anomalies is a more important predictor of
epithermal Au prospectivity in the case study area than proximity to NW-trending
faults/fractures.
Both of the discriminant models based on training sets A and B indicate that (a)
proximity to intersections of NNW- and NW-trending faults/fractures is a more
important control on epithermal Au mineralisation in the case study area than proximity
to either NNW- or NW-trending faults/fractures and (b) proximity to NNW-trending
faults/fractures is a more important control on epithermal Au mineralisation in the case
study area than proximity to NW-trending faults/fractures. These results contrast
somewhat with the implications of the results of the analyses of spatial associations in
Chapter 6 and the data-driven estimates of EBFs earlier in this chapter. Nevertheless, the
multivariate spatial associations depicted by the results shown in Table 8-V are
consistent with the knowledge that the presence and/or proximity to dilational jogs or
zones of extensions at/near either discontinuities or intersections of faults/fractures are
more important controls on hydrothermal mineralisation than faults/fractures alone
(Sibson, 1987, 1996, 2000, 2001). These results underscore the advantage of multivariate
techniques compared to bivariate techniques in terms of simultaneous analysis and
synergistic interpretation of empirical spatial associations between deposit-type locations
and indicative geological features.
If the magnitudes of the standardised function coefficients are compared and
contrasted with each set of spatial evidence rather than among the classes of spatial
evidence, then the two discriminant models (Table 8-V) indicate that epithermal Au
deposits in the case study area mostly occur within (a) about 200 m of NNW-trending
faults/fractures, (b) about 750 m of NW-trending faults/fractures and (c) about 1 km of
intersections of NNW- and NW-trending faults/fractures. These results are consistent
with the empirical spatial associations between epithermal Au deposits and indicative
geological features as quantified via the distance correlation method rather than as
quantified via the distance distribution method (see Chapter 6, Table 6-IX). In addition,
the two discriminant models (Table 8-V) indicate that integrated PC2 and PC3 scores
(obtained via catchment basins analysis; Chapter 5) greater than 0.25 are associated
spatially with most of the known epithermal Au deposits and therefore represent
significant anomalies. The overall results of the application of LDA are therefore