Uncertainty analysis of geochemical multi-surface models for solid-solution partitioning and speciation of heavy metals in soils

Abstract

Geochemical models are powerful tools to improve our understanding of the behavior of heavy metals in soils. Especially potent are multi-surface models that predict the binding to various soil reactive surfaces. Nevertheless, to date the uncertainty of such models has not been comprehensively evaluated, thus limiting their applicability. We quantified the uncertainty of the combined NICA-Donnan model for organic matter and the CD-MUSIC model for iron oxides in modelled solid-solution partitioning and speciation of cadmium, zinc and copper. We followed a statistical approach where we randomly sampled model parameters and input values from their normal distributions. A random sample of model parameters (N = 1000) was applied to 25 contrasting samples from background and contaminated soils around the world. A local ‘best-case’ uncertainty analysis was done by including measured humic and fulvic organic matter fractions, and the reactive surface area of ferrihydrite. A global ‘business-as-usual’ scenario was included by relying on common assumptions about the reactive surfaces. Model accuracy was evaluated by comparing predicted and measured (0.01 M CaCl2) dissolved metal. The analyses are currently carried out. Relating the resulting uncertainty to soil properties allows for evaluation of the suitability of multi-surface models for certain soil types. Moreover, the analysis will yield practical information regarding how to effectively invest resources to improve model accuracy, i.e. which parameters to optimize or which reactive surfaces to quantify. Overall, our study will improve the applicability of geochemical models to understand the behavior of heavy metals in soils.