Page 208 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS

210 Chapter 7

over the latter technique because the objective of prospectivity mapping is to delineate
small prospective zones with high prediction-rates. On the contrary, the slightly poorer
performance of the prospectivity map derived via application of multi-class index
overlay modeling compared to that of prospectivity map derived via application of
binary index overlay modeling, with respect to prospective areas occupying 15-40%,
indicates the caveat associated with binarisation of spatial evidence because this is prone
to both Type I (i.e., false positive) and Type II (i.e., false negative) errors. That is, in the
binarisation process, the classification (or ‘equalisation’) of evidential values to either 1
(rather than less than 1 but greater than 0) or to 0 (rather than greater than 0 but less than
1) means that some evidential values are ‘forced’ to become favourable evidence even if
they are not (thus, leading to false positive error) and that some evidential values are
‘forced’ to become non-favourable even if they are not (thus, leading to false negative
error).
Thus, in addition to its flexibility of assigning evidential class scores, multi-class
index overlay modeling is advantageous compared to binary index overlay modeling in
terms of suggesting uncertain predictions. It is therefore instructive to apply both of
these two techniques together instead of applying only either one of them. A
disadvantage of both of these techniques is the linear additive nature in combining
evidence, which does not intuitively represent the inter-play of geological processes
involved in mineralisation. We now turn to fuzzy logic modeling, which, like the
Boolean logic modeling, allows integration of evidence in an intuitive and logical way
and, like the multi-class index overlay modeling, allows flexibility in assigning
evidential class scores.

Fuzzy logic modeling
Fuzzy logic modeling is based on the fuzzy set theory (Zadeh, 1965). Demicco and
Klir (2004) discuss the rationale and illustrate the applications of fuzzy logic modeling
to geological studies; unfortunately, they do not provide examples of fuzzy logic
applications to mineral prospectivity mapping. Recent examples of applications of fuzzy
logic modeling to mineral prospectivity mapping are found in D’Ercole et al. (2000),
Knox-Robinson (2000), Porwal and Sides (2000), Venkataraman et al. (2000), Carranza
and Hale (2001a), Carranza (2002), Porwal et al. (2003b), Tangestani and Moore (2003),
Ranjbar and Honarmand (2004), Eddy et al. (2006), Harris and Sanborn-Barrie (2006),
Rogge et al. (2006) and Nykänen et al. (2008a, 2008b). Typically, application of fuzzy
logic modeling to knowledge-driven mineral prospectivity mapping involves three main
feed-forward stages (Fig. 7-10): (1) fuzzification of evidential data; (2) logical
integration of fuzzy evidential maps with the aid of an inference network and appropriate
fuzzy set operations; and (3) defuzzification of fuzzy mineral prospectivity output in
order to aid its interpretation. Each of these stages in fuzzy logic modeling of mineral
prospectivity is reviewed below with demonstrations of their applications to epithermal
Au prospectivity mapping in the case study area.