Geoscience Reference
In-Depth Information
0.1
0.08
0.06
0.04
0.02
0
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
−
0.02
−
0.04
−
0.06
−
0.08
−
0.1
Posterior Probability
Fig. 5.27
Kolmogorov-Smirnov plot for WLR Experiment 5 (Source: Agterberg and Bonham-
Carter
2005
, Fig. 6)
extremely high values of variables would either be eliminated or severely restricted.
Although discretization implies loss of information within a given study area,
reduction to binary form has the advantage of transportability in that weights
derived for one study area may become applicable in other areas. Coloring the
posterior probability map resulting from WofE (or another method), e.g. by using
red for the largest 2 % class as before. Such a colored map retains validity if there
are undiscovered deposits in the study area that have the same weights as the known
deposits. Consequently, the posterior probability provides target areas for further
exploration where the colors (e.g. red) indicate relatively high probability of
occurrence but where no or few deposits have been discovered.
A suitable method for measuring the performance of mineral potential maps
consists of attempting to predict occurrences of deposits that are deleted from the
data set. A jackknife type approach of successively omitting individual deposits
was proposed by Chung and Agterberg (
1980
). A different jackknife-based
approach will be discussed in Sect.
14.1
. Alternatively, mineral potential maps
can be constructed without use of deposit data in parts of the study area and
Ideally, a historical hindsight study can be performed in which a mineral potential
map is evaluated against new discoveries, after a number of years of subsequent
A difference plot as shown for WLR in Fig.
5.27
can be subjected to an
approximate Kolmogorov-Smirnov test for goodness of fit. This test checks for
the largest difference (in absolute value) between observed and expected
Search WWH ::
Custom Search