Metal speciation from stream to open ocean: modelling v. measurement

Environmental contextThe chemical speciation of metals strongly influences their transport, fate and bioavailability in natural waters. Analytical measurement and modelling both play important roles in understanding speciation, while modelling is also needed for prediction. Here, we analyse a large set of data for fresh waters, estuarine and coastal waters, and open ocean water, to examine how well measurements and modelling predictions agree. AbstractWe compiled a data set of similar to 2000 published metal speciation measurements made on samples of fresh waters, estuarine and coastal waters, and open ocean waters. For each sample, we applied the chemical speciation model WHAM7 to calculate the equilibrium free metal ion concentrations, [M] (molL(-1)), amounts of metal bound by dissolved organic matter (DOM), (molg(-1)), and their ratio /[M] (L g(-1)), which is a kind of local' partition coefficient. Comparison of the measured and predicted speciation variables for the whole data set showed that agreements are best for fresh waters, followed by estuarine and coastal waters, then open-ocean waters. Predicted values of /[M], averaged over all results for each metal, closely follow the trend in average measured values, confirming that metal reactivity, and consequent complexation by DOM, in natural waters accord with the expectations of the speciation model. Comparison of model predictions with measurements by different analytical techniques suggests that competitive ligand-stripping voltammetry methods overestimate metal complexation by DOM, and therefore underestimate [M]. When measurements by other methods are compared with predictions, for all metals, reasonable agreement with little bias is obtained at values of >10(-6)molg(-1) DOM, but at lower values of , the model predictions of [M] are mostly higher than the measured values, and the predictions of and /[M] are mostly lower. Research is needed to establish whether this reflects analytical error or the failure of the model to represent natural high-affinity ligands.