Mechanistic modelling offers a means of simulating the speciation and solubility of trace metals in soils. The WHAM/Model VI model has previously been used to simulate pH buffering and Al solubility in acid soils, and metal partitioning in very organic soils, but has not previously been applied to agricultural soils. Here we have extended the WHAM/Model VI framework to include surface complexation to oxides, and cation exchange, and applied it to batch titrations of limed agricultural soils contaminated by emissions of Cd and Zn from a metal smelter. In contrast to previous studies on forest soils, model predictions were most sensitive to the size of the geochemically ‘active’ soil Ca pool. Following optimization of this pool the model reproduced trends in pH and major cations well. Blind predictions of soil metal concentrations (Cu, Zn and Cd), from estimates of the active soil metal concentrations obtained by extraction with 0.22 m HNO3, were mostly very reasonable. Where predicted metal solubility was different, the model could be fitted to the data by optimizing the size of the active metal pool. In some cases the optimized metal pool was unrealistically large, which indicates a possible deficiency in the way the model considers binding activity and competition. Organic matter was the dominant binding phase in these soils. These results support the contention that speciation modelling has great promise in providing a holistic description of ion chemistry in soils for both major and trace elements.