The Improved Detection of Geochemical Soil Anomalies by Multiple Regression Analysis of Biochemical Data

Multiple regression analysis was used to predict elemental concentrations of copper and nickel in soils from consideration of the concentration of these elements in vegetation in an ultrabasic complex in New Zealand. Regression equations were computed from sixteen independent variables involving concentrations of ten elements in plant ash, and various physical factors such as slope and altitude of the site, and height of the tree. One elemental and one physical variable were significant for nickel at the 90 per cent level of probability and five elemental and two physical variables were significant for copper at the same level. Use of the regression equations explained 32·5 per cent and 26·5 per cent of the variance in the respective relative accumulations (amount in plant divided by amount in soil) of copper and nickel.Predicted concentrations of copper and nickel in soils agreed more closely with the observed values in the soils than did the levels of these elements in plant ash.Although the results may specifically apply only to the test area, the principles should have universal application.