The Mine Environment Neutral Drainage (MEND) Program is developing tools for
prediction of waste rock dump leachate quality. The first objective of this study was to
evaluate a recently proposed empirical approach for predicting concentrations of
metals in waste rock dump leachate primarily using pH (Morin and Hutt 1993). The
method has previously been successfully applied at two mines. The second objective
was to investigate refinements to the approach.
Five waste rock piles were selected for the study. Vangorda Plateau (Yukon Territory)
and Sullivan (south eastern British Columbia) mines are volcanogenic massive
deposits. The Cinola project, Queen Charlotte Islands, British Columbia was a previous
MEND study of small test waste rock piles at a proposed sediment-hosted epithermal
gold deposit mine. Mine Doyon is a gold vein deposit located between Val D’Or and
Rouyn, Quebec. Eskay Creek is a stratiform and stratabound gold and silver deposit
located in northwestern British Columbia. Usefulness of the data sets was limited by
missing data, variable detection limits and lack of associated flow information (where
applicable).
The first step involved examination of histograms for each variable and calculation of
regression equations for pH and conductivity against all other parameters. The study
confirmed the utility of the empirical approach. Element concentrations were generally
negatively correlated with pH but positively correlated with conductivity. Geochemical
evaluation of the trends using the equilibrium solution speciation model MINTEQA2
was not useful. However, evaluation of regression equations for sulphate and element
concentrations showed good correspondence with predicted geochemical behaviour,
consistency with site mineralogy and strong similarities between sites suggesting
common mineralogical controls.
The major problems encountered with the empirical models were outliers and
excessive positive skewness, variable detection limits, non-normality of residuals,
departures from linearity and sub-populations. Several refined data screening methods
were evaluated to address these problems, however, the effect on estimates of
regression parameter is minimal. Alternatives to least squares regression and
separation of data according to sub-populations can be considered.
The second step involved investigation of several multivariate techniques: multiple
regression, Principal Components Analysis (PCA) and Cluster Analysis. Due to the
excellent inter-correlation of many parameters, multiple regression does not increase
the predictive power of bivariate regressions. PCA and Cluster Analysis have no
predictive power but are useful as initial data screening tools to restrict the number of
bivariate regressions required to model leachate chemistry.
