Page 257 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS

260 Chapter 8

SELECTION OF COHERENT DEPOSIT-TYPE LOCATIONS FOR MODELING
Every mineral deposit, even if classified into a deposit-type, is unique and has
characteristics that are, to a certain extent, dissimilar to other mineral deposits of the
same type. It follows that multivariate spatial data signatures of deposit-type locations
are, to a certain extent, dissimilar or non-coherent. Because modeling of mineral
prospectivity involves ‘fitting’ (i.e., establishing spatial associations between) a map of
D with several maps of spatial data of X i evidential features, dissimilarity (i.e.,
heterogeneity or non-coherence) of multivariate spatial data signatures of deposit-type
locations can undermine the quality of a data-driven model of mineral prospectivity.
Carranza et al. (2008b) have shown that uncertainties of a data-driven model of mineral
prospectivity can be reduced and that fitting- and prediction-rates of a data-driven model
of model prospectivity can be improved by using a set of coherent deposit-type locations
(i.e., with similar multivariate spatial data signatures).
A two-stage methodology for selection of coherent deposit-type locations is
explained and demonstrated here (after Carranza et al., 2008b): (1) analysis of mineral
occurrence favourability scores of individual spatial data sets with respect to deposit-
type and non-deposit locations; and (2) analysis of deposit-type locations with similar
multivariate spatial data signatures. This two-stage methodology is demonstrated in the
case study area (Aroroy district, Philippines). Before doing so, let us address first the
issues of (a) increasing the number of locations of the target variable in a study area if it
is considered and/or found insufficient (e.g., 13 as in the case study area here) to derive,
depending on the method (Tables 8-I and 8-II), a proper (e.g., statistically significant)
data-driven model of mineral prospectivity and (b) selecting non-deposit locations
required in the analysis of coherent deposit-type locations and in the application of
multivariate methods of data-driven modeling of mineral prospectivity (Table 8-II).
The issue of increasing the number of locations of the target variable can be
addressed by considering locations immediately around each of the known deposit-type
locations as proxy deposit-type locations. This consideration is based on the assumption
that locations immediately around known deposit-type locations are also probably (albeit
weakly) mineralised. Thus, given a map of D partitioned into equal-sized unit cells, the
eight unit cells surrounding the unit cell representing a deposit-type location can be
considered proxy deposit-type locations. (For the case study area, there are 104 (i.e.,
13×8) proxy deposit-type locations.) Alternatively, an optimum distance buffer around
each deposit-type location can be sought via point pattern analysis (Boots and Getis,
1988; Rowlingson and Diggle, 1993) such that (a) there is zero probability of a
neighbour deposit-type location within a buffered deposit-type location and (b) based on
knowledge that mineralisation is a very rare geological phenomenon, the total number of
unit cells representing deposit-type and proxy deposit-type locations forms a very small
percentage (say, 1%) of the total number of unit cells in a study area (Carranza and Hale,
2001b, 2003; Carranza, 2002). The application of proxy deposit-type locations reduces
artificial spatial associations between evidential data and deposit-type locations
(Stensgaard et al., 2006), which usually occur when the number of the latter is small