Application of a metal solubility model to geochemical survey data
In areas where heavy metals are introduced into or onto land where they would not normally be present at elevated concentrations, then that land could be considered to be contaminated. A simple way of determining the magnitude of contamination by heavy metals is to measure the total metal concentration in the soil. However, this simple measure is a poor way of assessing the potential risks to the environment and human health. A more effective risk assessment can be achieved by analysing the proportion of the total metal that exists in a mobile or bioavailable form, in other words, the metal solubility. Unfortunately metal solubility is more difficult and costly to measure than total metal concentration in the soil. This thesis examines the application of a metal solubility model to geochemical survey data consisting of pH and metal concentrations. The solubility predictions were interpolated in order to produce maps; however, the interpolated data had very high uncertainties. Further analysis showed that pH was the greatest source of uncertainty in the algorithm, contributing the most for lead, with 76% of the uncertainty being due to pH. pH was least influential for copper, contributing 49% of the uncertainty, but pH was the highest contributor in each metal. In order to examine the accuracy of the algorithm without geostatistical influences, a field work study was undertaken to measure metal solubility directly at the original survey sites. This showed that the algorithm was very good at predicting metal solubility at point sources. In order to assess the shortscale spatial variability of pH, and the errors in pH measurements, a second field work project was conducted, measuring the pH on 200 samples from a single field. This work showed that pH does vary across a field, but more importantly allowed a quantification of the uncertainty involved in sampling and measuring pH. Results show that despite the short-scale variability in pH, point predictions are accurate (the average difference between measured and predicted pZn2+ is 6%), xvi and might be of use to land managers. However, interpolating solubility predictions for mapping produces unacceptably high uncertainties (mean values were 188% for Pb, 417% for Cu and 153% for Zn) for land management or the development of policy measures related to soil. Further work could include calculating the measured Pb and Cu solubility and comparing these to the predictions. A study to investigate how pH and Zn2+ vary together across a field would also be of interest.