Page 258 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS

Data-Driven Modeling of Mineral Prospectivity 261

(e.g., Carranza, 2004b). However, like the multivariate spatial data signatures of deposit-
type locations, the multivariate spatial data signatures of proxy deposit-type locations
are, to a certain extent, dissimilar or non-coherent. So, the two-stage methodology for
selection of coherent deposit-type locations is also demonstrated in selecting coherent
proxy deposit-type locations.
The issue of selecting non-deposit locations can be addressed by considering the
following three selection criteria (Carranza et al., 2008b). Firstly, in contrast to deposit-
type locations, which exhibit non-random spatial patterns (see Chapter 6), non-deposit
locations must be random (or randomly selected) so that their multivariate spatial data
signatures are likely non-coherent. Point pattern analysis (Diggle, 1983; Boots and Getis,
1988) can be applied to evaluate degrees of spatial randomness of selected non-deposit
locations (see Chapter 6 for application to deposit-type locations). Secondly, random (or
randomly-selected) non-deposit locations must be distal to (or located far away from)
deposit-type locations under study because locations proximal to deposit-type locations
likely have similar multivariate spatial data signatures to deposit-type locations and thus
probably do not qualify as non-deposit locations. Point pattern analysis (Diggle, 1983;
Boots and Getis, 1988) can be applied to determine the minimum distance from every
deposit-type location within which there is 100% probability of a neighbour deposit-type
location. In some cases, the criterion of ‘minimum distance with 100% probability’ may
leave insufficient locations for random selection of non-deposit locations, so one may
consider a lower probability distance. (For the case study area, non-deposit locations are
randomly selected beyond 2.2 km of any deposit-type location; within this distance from
any deposit-type location there is 90% probability of a deposit-type location (Fig. 8-
1A).) Thirdly, the number of distal and random non-deposit locations must be equal to
the number of deposit-type locations, because the latter locations are rare. This criterion
applies especially when logistic regression is used, as in the analysis of coherent deposit-
type locations (see below). The use of equal number of ‘zeros’ (e.g., non-deposit
locations) and ‘ones’ (e.g., deposit-type locations) in logistic regression is optimal when
the latter is rare (Breslow and Cain, 1988; Schill et al., 1993). In cases of rare ‘ones’,
King and Zeng (2001) aver that the information content contributed by the independent
variables used in logistic regression starts to diminish as the number of ‘zeros’ exceeds
the number of ‘ones’. (For the case study, 117 distal and random non-deposit locations
are selected in order to match the total number of deposit-type and proxy deposit-type
locations. Two sets of distal and random non-deposit locations are generated (Fig. 8-4)
in order to demonstrate the reproducibility and robustness of the methodology of
selecting coherent deposit-type locations.)
We now turn to the two-stage analysis of coherent deposit-type locations.

Analysis of mineral occurrence favourability scores at deposit-type locations
This section describes and discusses the first stage in selecting coherent deposit-type
locations and coherent proxy deposit-type locations. This involves deriving mineral
occurrence favourability scores (MOFS) of spatial data with respect to deposit-type and