Dissolved zinc (Zndiss) and copper (Cudiss) are a threat to aquatic life, but continue to enter urban waterways largely via stormwater passing over the roofing and cladding materials ubiquitous in the urban environment. Typical stormwater treatment is aimed at removing particulates, so a retrofittable inline downpipe device was developed (the Storminator™) to remove dissolved metals from roof runoff prior to it reaching the stormwater network. The device with its waste seashell media inside had previously been shown to be effective at removing Zndiss and Cudiss (>80% removal for both metals), but the mechanisms of removal were unconfirmed.

Research into metal retention mechanisms by biogenic calcium carbonate (CaCO3) such as seashells had so far been limited to higher metal concentration solutions and longer contact times than are relevant to the Storminator™ system. In addition to this, modelling to predict removal mechanisms carried out prior to this research was largely confined to mathematical correlation, with no basis in potential geochemical causal pathways. Therefore, this study aimed to determine the dominant Zn and Cu removal mechanisms occurring in a Storminator™ style system, using geochemical modelling to augment chemical and spectroscopic methods of analysis.

Geochemical models, such as PHREEQC, can predict when a mineral will be oversaturated in a given solution, or adsorbed to a surface such as hydrous ferric oxide (HFO), by balancing known thermodynamic equations for their formation. Therefore, predictions of how much Zndiss or Cudiss could be removed from solution by those mechanisms were made using PHREEQC. These predictions were validated against Zndiss and Cudiss reductions measured in flow-through column experiments, and evidence of either mechanism was sought through analysis of used shells. Shells which had been exposed to high loads of Zn or Cu were subjected to a sequential extraction procedure (SEP), which is designed to release elements bound in one specific chemical phase at a time. The shells were also analysed by scanning electron microscopy coupled with energy dispersive X-ray (SEM-EDS), to look for evidence of precipitates or adsorption by visual identification of particulates and elemental concentration mapping. Flow-through column results showed reductions of 73%–97% for Zndiss and 55%–82% for Cudiss, for concentrations typical of roof runoff, and that ranged over 2 orders of magnitude (Zndiss ≈ 0.3–3 mg/L, Cudiss ≈ 0.5–3 mg/L). PHREEQC geochemical modelling suggested that no stable Zn minerals were predicted to form at the measured pH, while Cudiss could have been reduced by up to 99% by precipitation of Cu hydroxide carbonate minerals. Further PHREEQC modelling suggests there was insufficient HFO present for adsorption onto this mineral to be a dominant removal mechanism for Zn or Cu.

The sequential extraction of used seashell media released the largest proportions of Zn from the “carbonate” fraction. Cu was predominantly released from both the “carbonate” and “Fe oxides” fractions, though it appeared that larger proportions of Cu and Zn were released in the “Fe oxides” fractions when total Cu or Zn concentrations on the shell were low.

SEM-EDS analyses of used media rarely highlighted ‘hotspots’ of high concentrations of Zn or Cu, instead generally showing low levels uniformly spread through the shell structures. The rare SEM-EDS analyses where Zn was concentrated in visible particles were of shells that had been exposed to high initial Zn concentrations (27 mg/L Zn), and these appeared to be either hydrozincite (a Zn hydroxycarbonate) or a Zn/Na/Al/Si based compound.

Results suggest that adsorption to the calcium carbonate shell surface dominates removal mechanisms for Zn, and that the formation of surface precipitates is likely where influent Zn concentrations are high. The formation of Cu hydroxycarbonates is likely to be the dominant mechanism for Cudiss removal, though the relative importance of aqueous precipitation of such minerals, and their formation on the shell surface via adsorption to the CaCO3, was less clear. The role of adsorption to other surfaces, such as organic material, HFO and aluminosilicates, in the removal Zndiss and Cudiss appeared to increase in importance as dissolved Zn and Cu concentrations decreased.

Implications for the optimisation of a Storminator™ type device include: maximising the shell surface area by minimising the shell fragment size, and increasing the runoff retention time within the device. However, these will be constrained by the hydraulic conductivity requirements of the inline system. Results suggest that the lifespan of the device is likely to be limited by operational factors rather than the availability of adsorption sites, so estimations of lifespan should be based on future field trials. Extension of the device lifespan may be possible by re-packing, or by regeneration of the media with EDTA or a low concentration weak acid, which should be investigated in further study.