Impact of temperature and water source on drinking water microbiome during distribution in a pilot-scale study

The study addresses how climate change–driven increases in water temperature and shifts in source-water quality affect microbial stability in drinking water distribution systems (DWDS). DWDS experience substantial seasonal temperature variation (often up to ~25 °C), which can stimulate microbial growth and alter community structure, potentially favoring opportunists (e.g., Legionella, coliforms). Concurrent climate impacts on raw water (higher temperature and nutrients) are expected to persist longer, elevating risks of taste, color, and odor issues. Previous work shows that source-water switching can strongly affect DWDS microbiomes, particularly during initial transitions. Given the four growth phases in DWDS (bulk water, suspended solids, biofilm, loose deposits) and continuous exchange between bulk and biofilms, understanding how temperature and water source shape both phases is crucial. The research tests the hypothesis that once a mature biofilm is established (in chlorinated systems), its composition remains stable and is not substantially affected by increased water temperatures, while bulk water may be more sensitive.

Prior studies report seasonal temperature effects on microbial counts (higher HPC and cell concentrations in warmer periods), though relationships with organic carbon are system-dependent. Biofilm communities are influenced by nutrient availability, surface characteristics, and hydraulics; biofilms harbor the majority of biomass and can impact water quality and harbor pathogens. Pilot studies simulating DWDS conditions have shown that in chlorinated systems, early biofilm development is temperature-sensitive, whereas in longer-term, non-chlorinated systems, elevated temperatures did not affect mature biofilm biomass or diversity after initial months. Source-water switching is known to introduce new taxa and alter biofilm communities, with recovery to a new stable state typically within a month. Flow cytometry (including phenotypic fingerprinting) and 16S rRNA sequencing are established tools to monitor microbial dynamics and distinguish source-water types.

A pilot-scale DWDS comprised three identical recirculating loops (each with a 1 m³ non-transparent HDPE tank and 100 m of U-PVC pipe, 80 mm diameter; ~500 L per loop). Loops operated at fixed temperatures: 16 °C (Loop 1), 20 °C (Loop 2), and 24 °C (Loop 3). Flow was 24 L/min and pressure 0.7–0.9 bar. Twice weekly, 500 L was drained and replenished with fresh drinking water, yielding a hydraulic residence time of 7 days. The pilot received chlorinated municipal drinking water whose source alternated between treated groundwater and treated surface water during the 137-day experiment, allowing assessment of source effects. Online flow cytometry (Accuri C6 Plus with onCyt autosampler) measured total cell counts every 8 h using SYBR Green I staining; cytometric fingerprinting was used to assess phenotypic traits. Bulk water was sampled weekly for chemistry (NO3-N, NO2-N, total N, Ca, orthophosphate as P2O5, Fe, NPOC) and on selected days (0, 7, 14, 28, 49, 70, 91, 116, 137) for 16S rRNA gene amplicon sequencing (1.5 L filtered on 0.22 μm). Biofilms were sampled via PVC-U coupons installed in each loop on days 0, 7, 14, 28, 49, 70, 91, 116, 137. Biofilm cells were detached with an electric toothbrush into filtered water and quantified by flow cytometry (Attune NxT; SYBR Green I). Biofilm sequencing was done on days 0, 14, 70, 91, 116, 137 after filtering 12 mL. DNA extraction used DNeasy PowerSoil Pro; sequencing on Illumina MiSeq; reads processed with DADA2 and SILVA v138 for taxonomy. Growth curves (Gompertz model selected after comparison to Logistic and Richards) were fit to flow cytometry time series to estimate specific growth rates and carrying capacities; low-quality fits (R2 < 0.1, negative rates) were excluded. Statistical analyses included Kruskal–Wallis, Mann–Whitney, and PERMANOVA; multivariate ordinations (PCoA, NMDS) and k-means clustering were applied to community data.

• Temperature significantly increased bulk water bacterial cell densities across loops (p = 2.20 × 10^-16, Kruskal–Wallis). Average bulk cell concentrations were on the order of 10^6 cells/mL, with higher values at elevated temperatures. • Source-water switching significantly affected nutrients and bulk microbiology. Surface water feeding increased orthophosphate (71.55 ± 38.40 μg P2O5/L vs 11.04 ± 17.84 μg/L) and NPOC (1.81 ± 0.48 mg/L vs 1.22 ± 0.33 mg/L) relative to groundwater (both p < 5 × 10^-9). Source also influenced bulk cell densities at 20 °C and 24 °C (p = 2.71 × 10^-6 and 2.64 × 10^-6), but not at 16 °C (p = 0.9944). At 24 °C, averages were 3.94 ± 1.41 × 10^6 cells/mL (groundwater) vs 2.89 ± 0.88 × 10^6 cells/mL (surface water). • Growth kinetics from flow cytometry revealed higher specific growth rates with treated groundwater at elevated temperatures (median 0.017 h^-1) compared to surface water (median 0.0069 h^-1) (p = 0.0041, Mann–Whitney). Carrying capacities increased for groundwater at 20 °C and 24 °C (Loop 2 p = 0.0045; Loop 3 p = 0.027). For Loop 2 (20 °C), median carrying capacity was 4.59 × 10^6 cells/mL (groundwater) vs 3.03 × 10^6 cells/mL (surface water). • Phenotypic cytometric fingerprints differed by temperature and source (both p = 9.99 × 10^-10, PERMANOVA). Bulk community composition by 16S sequencing was significantly shaped by source (p = 9.99 × 10^-6), with higher Chitinophagaceae (Sediminibacterium spp.) during surface-water periods; temperature did not significantly affect bulk composition (p = 0.321), though some families varied with temperature. • Biofilm: Temperature did not significantly affect biofilm cell densities (p = 0.727) or community composition (p = 0.667). Biofilm cell densities stabilized from day 70 onward at 3.99 ± 0.64 × 10^8 cells/cm^2, indicating maturity. Source influenced biofilm cell densities in loops at higher temperatures (not composition; PERMANOVA p > 0.1), though some increases may reflect time-to-maturity effects. • NMDS and k-means clustering showed mature biofilms clustered tightly (average within-cluster distance 0.50) compared to bulk (0.86), evidencing higher compositional stability in biofilms. • A core biofilm microbiome was identified and resilient to temperature/source changes: Rhodocyclaceae (Zoogloea spp.; ASV3, ASV14; Methyloversatilis spp., ASV6), Comamonadaceae (Rhizobacter spp., ASV4; Hydrogenophaga spp., ASV18), Xanthobacteraceae (Xanthobacter autotrophicus, ASV13; Bradyrhizobium spp., ASV23), and Sphingomonadaceae (Plot4-2H12 spp., ASV20; Sphingomonas spp., ASV21) with average abundances ranging ~0.6–13.11%. • Bulk communities were dominated by Chitinophagaceae (Sediminibacterium spp., ASV1; mean 31.93 ± 22.29%) and Comamonadaceae (ASV4/5/18). Evidence of bulk-biofilm exchange was observed, with shared ASVs at lower abundance in bulk.

The findings confirm that increased water temperature elevates bacterial cell densities and alters phenotypic traits in the bulk phase, while the biofilm, once mature, remains compositionally stable. Source-water quality exerts a strong control on bulk community composition and, in combination with elevated temperature, increases growth rates and carrying capacities—particularly when groundwater (with different nutrient characteristics) is supplied. These results support the hypothesis that mature biofilms in chlorinated DWDS are resilient to temperature increases and short-term source changes, whereas bulk water is more sensitive to environmental variation. Operationally, producing biostable water close to carrying capacity can limit net growth; however, temperature and source-induced shifts, especially under higher water age, may challenge stability and raise risks of aesthetic or microbial quality changes at consumers’ taps. The observed core biofilm microbiome dominated by Alphaproteobacteria and Betaproteobacteria suggests functional redundancy and robustness (e.g., carbon degradation, biofilm/EPS formation), explaining resilience under varying temperatures and sources. Understanding these dynamics can inform DWDS management under climate change scenarios, emphasizing control of source-water quality and water age to maintain stability.

In a pilot-scale DWDS, elevated temperatures increased bulk bacterial densities but did not enhance biofilm biomass or alter mature biofilm composition. Source-water variations significantly influenced nutrients, bulk cell densities, and bulk community structure, and—together with elevated temperatures—increased growth rates and carrying capacities, especially for groundwater periods. Biofilms reached maturity around day 70 with stable densities and a resilient core microbiome unaffected by temperature or source. Practically, higher temperatures are unlikely to accelerate colonization of new pipes beyond early stages, but shifts in source-water quality can destabilize bulk water, potentially affecting tap water quality. Future work should evaluate source switching impacts once biofilms are stably mature, with higher-frequency sampling and additional timepoints, and assess a broader range of hydraulic conditions and residence times to generalize to full-scale systems.

• The pilot operated with recirculation and a long hydraulic residence time (~7 days) and mixed water types; this worst-case water age and mixing may have amplified temperature and source effects and altered carrying capacities relative to typical distribution conditions. • Twice-weekly partial refreshment and recirculation could increase bulk growth and detachment dynamics compared to once-through systems. • Biofilms matured over time; some observed increases during source changes may partly reflect time effects before day 70. • Sequencing for bulk water used one sample per loop per timepoint, potentially limiting within-timepoint variability assessment. • Pilot-scale configuration, materials, and coupon locations may not fully capture heterogeneity found in full-scale DWDS, though positional tests showed no significant compositional differences. • Free chlorine was reported below the detection limit during routine checks, which may influence generalizability to systems with higher residuals.