Incorporation of metal speciation data into QICAR models and application to data-poor technology-critical elements

In a previous article, we developed quantitative ion character-activity relationships (QICARs) to relate the intrinsic properties of a metal to its acute toxicity towards freshwater aquatic organisms. These predictive tools were developed for a set of data-rich training elements and then applied to a representative selection of technology-critical elements (TCEs). The toxicity of the TCEs was reasonably well predicted, with most values located within the 95% prediction intervals. In this work, we have extended this approach to use the calculated metal speciation. Linear free energy relationships were used to estimate some of the needed thermodynamic constants. Using this information, we expressed the concentration resulting in a 50% effect level value as free metal activities and performed regression analyses. For the training metals, the determination coefficients slightly increased compared with those obtained using the total dissolved metal. As before, the log-transformed composite value of the covalent index (χm2r) was the best predictor of their acute toxicity towards algae and daphnids (χm = metal's electronegativity; r = ionic radius). However, for the TCEs, the regressions were much poorer, particularly when the predicted free metal ion concentrations were very low (e.g., < 10-18 M). We suggest that this result reflects the distinctive speciation of these metals, where (i) the free metal ion is present only at vanishingly low concentrations (the calculation of which is problematic) and (ii) in all but one case (Au(CN)2-), the metal's calculated speciation is dominated by neutral polyhydroxo species (e.g., Au(OH)30, Ge(OH)40…). In our view, this result does not undermine the use of QICARs. Rather, the use of QICARs revealed that free-ion activity could be inadequate for predicting the toxicity of the studied data-poor metals.