Influence of filter age on Fe, Mn and NH4+ removal in dual media rapid sand filters used for drinking water production.

Rapid sand filtration is a common method for removal of iron (Fe), manganese (Mn) and ammonium (NH4+) from anoxic groundwaters used for drinking water production. In this study, we combine geochemical and microbiological data to assess how filter age influences Fe, Mn and NH4+ removal in dual media filters, consisting of anthracite overlying quartz sand, that have been in operation for between ∼2 months and ∼11 years. We show that the depth where dissolved Fe and Mn removal occurs is reflected in the filter medium coatings, with ferrihydrite forming in the anthracite in the top of the filters (< 1 m), while birnessite-type Mn oxides are mostly formed in the sand (> 1 m). Removal of NH4+ occurs through nitrification in both the anthracite and sand and is the key driver of oxygen loss. Removal of Fe is independent of filter age and is always efficient (> 97% removal). In contrast, for Mn, the removal efficiency varies with filter age, ranging from 9 to 28% at ∼2-3 months after filter replacement to 100% after 8 months. After 11 years, removal reduces to 60-80%. The lack of Mn removal in the youngest filters (at 2-3 months) is likely the result of a relatively low abundance of mineral coatings that adsorb Mn2+ and provide surfaces for the establishment of a microbial community. 16S rRNA gene amplicon sequencing shows that Gallionella, which are known Fe2+ oxidizers, are present after 2 months, yet Fe2+ removal is mostly chemical. Efficient NH4+ removal (> 90%) establishes within 3 months of operation but leakage occurs upon high NH4+loading (> 160 µM). Two-step nitrification by Nitrosomonas and Candidatus Nitrotoga is likely the most important NH4+ removal mechanism in younger filters during ripening (2 months), after which complete ammonia oxidation by Nitrospira and canonical two-step nitrification occur simultaneously in older filters. Our results highlight the strong effect of filter age on especially Mn2+but also NH4+ removal. We show that ageing of filter medium leads to the development of thick coatings, which we hypothesize leads to preferential flow, and breakthrough of Mn2+. Use of age-specific flow rates may increase the contact time with the filter medium in older filters and improve Mn2+ and NH4+ removal.

Meta-omics profiling of full-scale groundwater rapid sand filters explains stratification of iron, ammonium and manganese removals.

Rapid sand filters (RSF) are an established and widely applied technology for groundwater treatment. Yet, the underlying interwoven biological and physical-chemical reactions controlling the sequential removal of iron, ammonia and manganese remain poorly understood. To resolve the contribution and interactions between the individual reactions, we studied two full-scale drinking water treatment plant configurations, namely (i) one dual-media (anthracite and quartz sand) filter and (ii) two single-media (quartz sand) filters in series. In situ and ex situ activity tests were combined with mineral coating characterization and metagenome-guided metaproteomics along the depth of each filter. Both plants exhibited comparable performances and process compartmentalization, with most of ammonium and manganese removal occurring only after complete iron depletion. The homogeneity of the media coating and genome-based microbial composition within each compartment highlighted the effect of backwashing, namely the complete vertical mixing of the filter media. In stark contrast to this homogeneity, the removal of the contaminants was strongly stratified within each compartment, and decreased along the filter height. This apparent and longstanding conflict was resolved by quantifying the expressed proteome at different filter heights, revealing a consistent stratification of proteins catalysing ammonia oxidation and protein-based relative abundances of nitrifying genera (up to 2 orders of magnitude difference between top and bottom samples). This implies that microorganisms adapt their protein pool to the available nutrient load at a faster rate than the backwash mixing frequency. Ultimately, these results show the unique and complementary potential of metaproteomics to understand metabolic adaptations and interactions in highly dynamic ecosystems.