Correlation, Method of Least Squares, Linear Regression and the General Linear Model - Geomathematics: Theoretical Foundations, Applications and Future Developments - page 135

T

N

K

V

Fig. 4.15 Copper potential map for Abitibi area based on combined model with polynomial terms
added to the geological and geophysical explanatory variables. Its contour values near Val d'Or
exceed those in Fig. 4.7e-g (Source: Agterberg 1974 , Fig. 126)

The geomathematical mineral resource prediction project in the Abitibi area
which was commenced in 1967 ( cf . Sect. 4.3.1 ) on the premise that the large
amounts of data available for many hundreds of gold deposits including a few
hundred gold mines (almost exclusively past producers) could be fruitfully used to
develop methodology that would be applicable to other types of deposits. The Kidd
Creek mine, which is on a very large volcanogenic massive sulphide deposit, had
already been discovered and prospecting for new deposits in the Abitibi area
had shifted almost entirely to large copper and zinc deposits instead of lode gold
deposits that could no longer be profitably mined because the price of gold had been
kept at 35 dollars per troy ounce. It turned out that spatial frequency of occurrence
(number of deposits per unit of area) and sizes of the lode gold deposits are only
partially controlled by mappable geological and geophysical variables. In situations
of this kind it can be assumed that the regional mineralization processes were deep-
seated and not mainly controlled by the depositional environments.
The problem of uneven geographical distribution of amounts of mineable metal
across a region can be partially overcome by including terms in the regression
equations that are functions of geographic location in the Abitibi area as described
in Agterberg ( 1970 , 1971 ), and Agterberg and Kelly ( 1971 ). An example for the
large copper deposits used for example earlier in this chapter is shown in Fig. 4.15 .
The stepwise regression resulting in Table 4.4 and Fig. 4.7 was repeated after
adding 44 new variables to the 55 explanatory variables used before. These
additional variables were functions of the geographical coordinates of all cells in
the region. Together these functions form an octic trend surface ( cf . Chap. 7 ). The
contour pattern of Fig. 4.15 is similar to the patterns obtained previously (e.g.,
Fig. 4.7 ) except in the southern Val d'Or-Noranda area. This would suggest that,
relatively, there occurs more copper in this vicinity than elsewhere in the Abitibi
area. Reddy et al. ( 1991 ) used logistic regression to predict base-metal potential in
the Snow Lake area, Manitoba, using explanatory variables that included geo-
graphic locations. A two-stage least-squares model for the relationship between