Volume 6, Issue 6 e70010
ORIGINAL RESEARCH
Open Access

Impact of flocculated and softened particles on UV254 inactivation of indigenous spores

Judith Straathof

Judith Straathof

Department of Civil, Environmental and Geodetic Engineering, The Ohio State University, Columbus, Ohio, USA

Department of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands

Contribution: Data curation, Formal analysis, Validation, ​Investigation, Visualization, Methodology, Writing - original draft, Writing - review & editing

Search for more papers by this author
Zuzana Bohrerova

Zuzana Bohrerova

Ohio Water Resources Center, The Ohio State University, Columbus, Ohio, USA

The Sustainability Institute, The Ohio State University, Columbus, Ohio, USA

Contribution: Conceptualization, Supervision, Methodology, Writing - review & editing

Search for more papers by this author
Natalie M. Hull

Corresponding Author

Natalie M. Hull

Department of Civil, Environmental and Geodetic Engineering, The Ohio State University, Columbus, Ohio, USA

The Sustainability Institute, The Ohio State University, Columbus, Ohio, USA

Correspondence

Natalie M. Hull, Department of Civil, Environmental and Geodetic Engineering, The Ohio State University, Columbus, OH 43210, USA.

Email: [email protected]; [email protected]

Contribution: Conceptualization, Supervision, Funding acquisition, Validation, Methodology, Project administration, Writing - review & editing

Search for more papers by this author
First published: 22 December 2024

Deputy Editor: Vanessa L. Speight

Associate Editor: Graham A. Gagnon

Abstract

US regulatory ultraviolet (UV) disinfection credit is typically granted when turbidity is ≤1 NTU. However, studies show turbidity does not always correlate well with UV dose responses. This study examined the impact of worst-case high turbidity scenarios at drinking water treatment plants on UV254 inactivation of indigenous spores from unfiltered source water and unsettled flocculation and softening steps. Flocculated water (turbidity = 6.49–164 NTU) had the lowest dose response with a significantly lower Geeraerd-tail maximum inactivation rate (kmax = 0.021 cm2/mJ) and higher residual population density (Nres = 7.081 SFU/mL). Raw source water (kmax = 0.027 cm2/mJ, Nres = 1.168 SFU/mL, turbidity = 0.978–215 NTU) and softened water (kmax = 0.030 cm2/mJ, Nres = 0.216 SFU/mL, turbidity = 318–495 NTU) had similar dose responses despite significantly different water quality. Particle size and the degree of particle-associated spores best explained the differences in dose responses. Almost all spores were associated with flocculated particles instead of free-floating, which increased tailing and negatively impacted UV inactivation. Based on regulatory reduction equivalent dose bias factors and UV sensitivities of spiked Bacillus subtilis spores, Cryptosporidium would be 4-log inactivated in these raw, flocculated, or softened waters if UV transmission were ≥65%, 90%, or 80%, respectively, even though turbidity was grossly >1 NTU. Depending on particle characteristics, partial inactivation credit when turbidity is >1 NTU should be considered to avoid high-tier violations while still protecting public health.

Article Impact Statement

Particle size and the degree of association between microorganisms and particles have the greatest impact on UV disinfection. Turbidity did not correlate well to UV dose responses.

1 INTRODUCTION

Ultraviolet (UV) light is widely used to disinfect drinking water and treated wastewater by inactivating pathogens. Advantages of UV disinfection over conventional chemical disinfection include effective inactivation of protozoan parasites and negligible disinfection by-product formation (Choi & Choi, 2010; Dotson et al., 2012; Masschelein & Rice, 2002). UV disinfection is commonly performed on filtered water to reduce the amount of particles that can interfere with UV exposure to pathogens (Soleimanpour Makuei et al., 2022) and with turbidity levels below 1 NTU (USEPA, 2006). Turbidity is used as a water quality indicator and can be easily, inexpensively, rapidly, and accurately measured (USEPA, 2006). As climate change drives an increase in the frequency and intensity of extreme weather events, high turbidity events of source waters will increase in frequency and magnitude due to droughts, rainfall, snowmelt, sediment supply, and soil erosion (Hannoun et al., 2022; Lee et al., 2015; Mi et al., 2019; Mukundan et al., 2018; Zhang et al., 2013). More frequent high turbidity events and lower source water quality pose challenges for unit processes at drinking water treatment plants (DWTPs), especially for smaller water utilities and utilities with aging infrastructure that may be less adept or funded to handle these challenges. Sometimes filters are operated under upset conditions which create opportunities for particle breakthrough to occur (Cantwell & Hofmann, 2008).

Although UV disinfection is especially effective against chlorine-resistant protozoa, including Cryptosporidium parvum and Giardia muris (Betancourt & Rose, 2004), DWTPs in the United States do not get protozoan inactivation credit when the combined filter effluent turbidity exceeds the maximum value of 1 NTU or when the 95th percentile monthly turbidity measurements are greater than 0.3 NTU during UV disinfection (Code of Federal Regulations §141.551). Not meeting these regulations is considered a tier 1 violation, which requires immediate notice to the public from the utility. However, there is inactivation when turbidity is above 1 NTU (Amoah et al., 2005; Baldasso et al., 2021; Clancy et al., 2000; Passantino et al., 2004; Templeton et al., 2009). Tier 1 violations should only be used when human health is at risk. Small utilities especially could benefit from a more accurate and updated turbidity regulation value or tier violation value because small water systems with financial difficulties are more likely to have violations (Eskaf, 2015). Safe consumption is the focus of DWTPs, but it must be done in a cost-effective manner that strengthens the public's trust in the utility: regulations and violation guidelines should reflect that.

Although UV disinfection is regulated by turbidity, the UV system is controlled by flow and UV irradiance and water quality is monitored by UV absorbance/transmission (UVA/UVT). The current regulations are in place to account for potential water quality and particle shielding that can lead to microbial survival during UV disinfection. Particles in the water matrix can impact the effectiveness of UV disinfection by scattering, blocking, or absorbing UV light (Christensen & Linden, 2003) and by shielding microorganisms embedded within a particle from UV light (Emerick et al., 1999; Örmeci & Linden, 2002; Parker & Darby, 1995; Qualls et al., 1983, 1985). Although regulations stipulate that turbidity values should remain below 1 NTU to avoid violations, inactivation has been widely observed above 1 NTU (Amoah et al., 2005; Baldasso et al., 2021; Clancy et al., 2000; Passantino et al., 2004; Templeton et al., 2009) (Table 1).

TABLE 1. Impact of turbidity and particles on microorganisms during ultraviolet (UV) inactivation.
Literature source Water type Water matrix change Target organisms Microbial method Extraction method UVA (cm−1) or UVT (%) at 254 nm Turbidity value (NTU) Particle size (μm) Impact on UV inactivation
Amoah et al. (2005) Untreated lake water Concentrate lake particles Seeded Giardia muris, Cryptosporidium parvum Mouse infectivity assay Centrifugation Not specified 0.3–20

Majority

5–25

No (when turbidity <10 NTU)
Baldasso et al. (2021) Filtered surface water and synthetic water Added kaolinite and humic acids to water for lab tests MS2 phage Double-layer agar enumeration NA 69%–96% for field tests, 77%–100% for lab tests 0.09–5 for field tests 0–5.2 for lab tests Not specified No
Batch et al. (2004) DWTP effluent (GW, lake, river) Dechlorinated Seeded MS2 coliphage Top agar assay NA <0.4, except one utility ~0.7 0.3

Majority

<10, but >10 present

No
Bohrerova and Linden (2006) Wastewater effluent 100, 41, 20 μm nylon net filtered Seeded Mycobacteria Spot plating NA 0.03–0.16 (0.45 with spiked mycobacteria) 1.00–15.56

Majority

0.06–4.5, but on average at least 15 particles/mL >41

Yes (when self-aggregates or particles are >41 μm)
Cantwell et al. (2010) Unfiltered surface water NA Total coliforms and total aerobic spores MF (coliforms), agar plate enumeration (spores) NA Mostly ~90% (max 97%, min 82%) 0.13–17.9, 99% <5.6 <10, 10–100 Yes (when particles >10 μm)
Caron et al. (2007) Two surface waters Unfiltered and 8 μm filtered Indigenous aerobic spore-forming bacteria MF—Barbeau method Chemical extractant, blending at 8000 rpm for 4 min River 1: 88%–94%, River 2: 40%–55% 10–35, <1.5 5–10, 10–20, >20, majority 5–10 Yes (amount of particles >8 μm correlated when trying to achieve >2 log inactivation)
Carré et al. (2018) Activated-sludge effluent Addition of mixed liquor from the aeration tank Total coliforms, Escherichia coli, enterococci MPN NA

Unfiltered: 57%–71.6%

Filtered: 70.2%–71.2%

0.6–23.3 >2, 2–5, 5–25, >25 Yes (strong correlation between tailing and particles >25 μm)
Christensen and Linden (2003) Raw water (DWTP) Filtered to make size fractions, diluted to vary turbidity NA (tested UV dose) NA NA <40 μm: 0.044–0.27, 54%–90% (IS 0.021–0.23, 58%–95%), <11 μm: 0.044–0.25, 57%–90% (IS 0.014–0.18, 66%–97%), <5 μm 0.051–0.27, 53%–89% (IS 0.035–0.24, 57%–92%)

<40 μm: 0.057–16.2

<11 μm: 0.057–10.1

<5 μm: 0.046–10.1

0–5, 0–11, 0–40 Yes (when turbidity >3 NTU, can account for it with proper IS absorbance measurement)
Clancy et al. (2000) DI water, backwash recycle water NA Seeded C. parvum oocysts Mouse infectivity assay NA Not specified DI: <1, backwash water: up to 11 Not specified No
Darby et al. (1993) Activated sludge secondary effluent Unfiltered and filtered (0.9 mm sand filter) Total coliforms MTF/MPN NA

Unfiltered: 78.0 (4.4)%

Filtered: 75.6 (6.9)%

Unfiltered: 3.8 (1.5)

Filtered: 1.1 (0.4)

Peaks at 1 and 35 Yes (but filtering increased UV disinfection efficacy)
Emerick et al. (1999) Wastewater influent Unfiltered, 80 and 11 μm filtered Total coliforms MTF, fluorescent microscopy NA Activated sludge: 69.4%–76.2%, Trickling filter: 68.6%, Aerated lagoon: 51.6%, Facultative lagoon: 27.2% Not specified Up to 200 Yes (particles >10 μm increase tailing effect)
Guo and Hu (2012) DWTP (pre/post coagulation and filtration) Alum addition MS2 bacterio-phage Double-layer agar enumeration NA Not specified 1–5 Not specified Yes (inactivation rate increased with alum addition)
Jolis et al. (2001) Filtered secondary effluent Alum and polymer addition, 8 μm filtered Total coliforms Not specified NA Not specified 0.2–4.8 Wide range, but majority <7 Yes (when particles >7 μm)
Kollu and Örmeci (2012) DI water Alginate, calcium, latex particles Seeded E. coli MF Vortex (45 s) Not specified Not specified 1, 3.2, 11, 25, 45 Yes (at high doses only)
Li et al. (2009) Treated wastewater NA Indigenous aerobic spore-forming bacteria MF Compared none to homogenization at 13,500 rpm for 1.5 min with dispersal buffer Not specified Not specified Up to 80 Indigenous ASFB were not associated with particles to a degree that impacted UV disinfection
Liu et al. (2007) River water and WTP process water (floc/coag) NA Seeded E. coli MF None

River: 0.24–0.34

Process: 0.31–0.38

River: 12–32

Process: 5.3–17

River: 0.5–8.1

Process: 5.4–38.2

No (river water)

Yes (when floc particles are present)

Madge and Jensen (2006) Wastewater (2 WWTPs) NA Fecal coliform MPN Chemical extractant with blending on some samples

Pre-chem A: 59.7 (4.5)%, B: 38.2 (7.3)%

Post-chem A: 39.6 (3.9)%, B: 24.5 (4.7)%

A: 5.4 (1.3)

B: 8.0 (2.9)

Not specified Yes (when particles >20 μm)
Mamane and Linden (2006a) Simulated drinking water Addition of montmorillonite clay, NOM, alum, varied pH Seeded Bacillus subtilis spores Pour plating Blending Not specified 0.46–17.8 (additions of 0, 5, 10 NTU clay particles) Average up to 2.47, most average ~1 Yes (<0.3 log inactivation decrease for spore-clay aggregates)
Mamane and Linden (2006b) Natural raw water, simulated drinking water Montmorillonite added to simulated drinking water Seeded B. subtilis spores Pour plating NA 5 NTU clay + spores: ~0.28 (direct) ~0.17 (IS) Natural: 6.3–15.8, simulated: 0, 5, and 10 0.5–10.5 Yes (when turbidity >5 NTU)
Oppenheimer et al. (2002) Raw, unfiltered water (DWTP) NA Seeded C. parvum, G. muris, MS2 coliphage Mouse infectivity assay, MS2: ATCC 15597 B1 and E. coli (Adams, 1959) NA Not specified 0.65–7.00 NA No
Örmeci and Linden (2002) Secondary wastewater effluent Unfiltered and 5 μm filtered Non- & Particle-associated coliform MF EGTA extraction, filtration, blending Not specified Not specified 2–5, 5–10, >10 (majority 2–5) Yes (inactivation rate of non-particle > particle-associated coliform)
Parker and Darby (1995) Secondary wastewater effluent NA Total coliforms, fecal coliforms MTF/MPN Blending and sonication with chemical extractant

Unfiltered: 60.9%–72.4%

Filtered: 66.5%–73.9% (avgs)

1.30–3.95 Majority 1, particles >10 also present, peaks at 1, 8, and 35 Yes (particle-associated total coliforms are shielded)
Passantino et al. (2004) Natural unfiltered water and DI water Addition of montmorillo-nite clay, algae to DI water Seeded MS2 bacterio-phage Double agar layer method, E. coli host (Adams, 1959) NA 0.075–0.125 (74%–84%) 3.0–12 Not specified No
Pichel et al. (2021) DI water and natural surface water Addition of dust and humic acid sodium salt Seeded MS2 bacterio-phage and E. coli Double agar layer method, E. coli host for MS2, and MF for E. coli NA Not specified 0–40 NA

No (when turbidity below 20 NTU)

Yes (from humic acid)

Qualls et al. (1983) Secondary wastewater effluent Unfiltered, 8 and 70 μm filtered Total coliforms, fecal coliforms MPN NA

Unfiltered: 0.179 (0.031)

Filtered: 0.161 (0.029)

Unfiltered: 4.9 (1.8)

Filtered: 2.1 (1.4)

<8, <70, >70 Yes (when particles >8 μm)
Qualls et al. (1985) Secondary wastewater effluent Unfiltered, 10 and 0.45 μm filter Total coliforms MF NA Unfiltered: 0.233 (0.073) 1.9–14 (unfiltered) <0.45, <10, >10 Yes (when particles >40 μm, no significant impact for particles <10 μm)
Soleimanpour Makuei et al. (2022) 2 DWTPs prior UV disinfection, raw intake NA Seeded MS2 bacterio-phage, E. coli host Pour plating (plaque counting) NA Raw intake: ~0.29–0.48, A: ~0.04–0.07, B: ~0.03–0.06 Raw intake: 1–3.5, A: 0.38 (0.09), B: 0.61 (0.33) Raw intake: peaks at 0.69 and 5.17, A: 2–6, peak at 4.95, B: NA Yes (30 min mixing to encourage potential particle-microorganism association)
Templeton et al. (2009) 2 DWTPs with worst-case scenario conditions Alum, activated silica, chemical coagulants Total coliforms MF Chemical extractant with blending (20,000 rpm, 3 min) A: 48%–81%, B: 73%–81% Up to 2.9 Mainly 2–10, <~5% particles >10 No
Winward et al. (2008) Treated wastewater effluent, treated gray water NA E. coli and total coliforms, Enterocci MPN (E. coli and total coliforms) and MF (Enterocci) Blending (4000 rpm for 60 s) Gray: 47%, wastewater: 57%, treated gray: 62% (avgs) Gray: 18, wastewater: 10, treated gray: 6 Wide range, mainly 1–1233 for gray water Yes (due to larger particle size)
This study Raw, flocculated, softened water from DWTP NA Indigenous spores and seeded B. subtilis spores Pour plating Manual shaking See results section See results section See results section Yes (impact due to larger particles and degree of microorganism particle association)
  • Abbreviations: ASFB, aerobic spore-forming bacteria; DI, deioized; DWTP, drinking water treatment plant; EGTA, egtazic acid; GW, groundwater; IS, integrating sphere; MF, membrane filtration; MPN, most probable number; MTF, multiple tube fermentation; NA, not applicable; NOM, natural organic matter; UVA, UV absorbance; UVT, UV transmission; WTP, water treatment plant; WWTP, wastewater treatment plant.

Shielding of microorganisms in surface water during UV inactivation has mainly been attributed to particles ≥7–10 μm (Cantwell et al., 2010; Caron et al., 2007; Jolis et al., 2001; Madge & Jensen, 2006; Qualls et al., 1983). However, in some cases, particles >10 μm did not impact UV inactivation (Amoah et al., 2005; Bohrerova & Linden, 2006; Kollu & Örmeci, 2012). Particles >20 and 40 μm provided more coliform shielding than particles <20 μm even though there were more smaller particles (Qualls et al., 1985). Attempts at determining a definitive particle size threshold impacting UV disinfection are further complicated by the degree of association between particles and microorganisms. Free-floating microorganisms are considered to be easily disinfected, while self-aggregates and particle-associated microorganisms attached to or enmeshed within particles are affected by shielding and are considered difficult to disinfect (Caron et al., 2007; Mamane & Linden, 2006a; Örmeci & Linden, 2002), resulting in the majority of tailing effects (Tan et al., 2017). The degree of association is dependent on particle size, structure, amount, surface charge, and hydrophobicity (Cantwell & Hofmann, 2008; Templeton et al., 2008), with zeta potential having a more significant impact on tailing than turbidity and absorbance (Soleimanpour Makuei et al., 2022). UV light can still inactivate microorganisms if light-accessible pathways are present through scattering/reflectance (Emerick et al., 2000). Based on literature research (Table 1), it is expected that other water quality characteristics than turbidity will be a better predictor of UV inactivation in water with particles.

Experiments are commonly performed with indicator organisms due to the impracticality of using pathogens in laboratory experiments. Drawbacks of using Cryptosporidium oocysts include high costs and difficulty in producing and analyzing oocysts (Ryan & Hijjawi, 2015), low reproducibility of oocyst assays (Clancy et al., 1994), and low initial oocyst concentrations in raw or treated drinking water (Brown & Cornwell, 2007; Karanis et al., 2006). Indigenous spores from aerobic spore-forming bacteria (ASFB) can serve as a conservative surrogate for pathogenic protozoan Cryptosporidium (oo)cysts because spores are slightly smaller, share similarities including an isoelectric point below pH 3, neutral to strongly negative zeta potentials, and have glycoproteins on the exterior surface (Bradford et al., 2016), and are more resistant to UV inactivation (Mamane-Gravetz et al., 2005; Mamane-Gravetz & Linden, 2004, 2005). Aerobic bacterial spores are approximately 0.8 μm in width and range from 1 to 2 μm in length while Cryptosporidium oocysts are 3.5–6 μm in size (Bradford et al., 2016; Feng et al., 2003). Indigenous aerobic bacterial spores are also more resistant than lab-strain spores, such as Bacillus subtilis ATCC 6633 (Mamane-Gravetz et al., 2005; Mamane-Gravetz & Linden, 2004, 2005). Bacillus species spores are 5–50 times more resistant to UV inactivation than their vegetative cells (Setlow, 2001). Indigenous spores will undergo the same mixing process as other pathogens at DWTPs. Multiple studies indicate that seeded indicator bacteria may be easier to inactivate than indigenous bacteria in unfiltered waters because seeded bacteria have a lower degree of association with particles (Caron et al., 2007; Mamane & Linden, 2006a; Örmeci & Linden, 2002). Seeded and indigenous microorganisms may also have variable relative proportions between free suspension and particle attachment if simulated lab mixing, timing, and particle characteristics are not the same as at the DWTP. In one study, 30% of seeded C. parvum and Giardia lamblia attached to particles during the first few minutes of mixing but increased to 75% after 24 h (Medema et al., 1998).

More research is needed on UV disinfection during worst-case scenarios at DWTPs to accurately reevaluate the 1 NTU turbidity threshold for receiving no credit for UV disinfection and triggering the highest tier violation, and to explore potential alternative water quality parameters for evaluating regulatory compliance for UV disinfection efficacy of protozoan pathogens. This study builds on previous research to more closely approximate potential real-world worst-case scenario conditions by investigating UV inactivation of indigenous aerobic bacterial spores in DWTP source and unfiltered treatment process waters with extremely high and variable turbidity rather than simulated water. The novel aspect of this study includes the use of indigenous spores rather than seeded bacteria, as well as turbidity and particles that are naturally present from the flocculation/coagulation and softening process, as we expect that the source of turbidity and particles will have an impact on UV disinfection and therefore it is crucial to use water collected from the DWTP. The objective of this research was to determine the impact of filtration failure before or after coagulation, flocculation, and softening on low-pressure (LP) UV254 inactivation of indigenous spores in surface water, and to investigate relationships between the inactivation and various water quality characteristics.

2 METHODS

2.1 Drinking water treatment plant

Water samples were collected from Dublin Road Water Treatment Plant (DRWP) in Columbus, OH (Figure 1). DRWP treats raw surface water from the Griggs and O'Shaughnessy Reservoirs on the Scioto River. After passing through rotating screens, aluminum sulfate [XAl(SO4)2∙12H2O] (alum) is added for coagulation and flocculation and sodium hydroxide, sodium carbonate (soda ash), and hydrated lime are added for softening.

Details are in the caption following the image
Abbreviated treatment schematic of Dublin Road Water Treatment Plant (DRWP) with sample collection locations labeled for raw surface water (raw), unsettled flocculated water (floc), and unsettled softened water (soft). Image modeled after City of Columbus (2022).

2.2 Water sample collection

On each collection date from April 2019 to January 2020, three water types were collected. Two (biological duplicate) sterile 1-L plastic bottles (Nalgene, PP) were filled with raw surface water, flocculated water collected from the end of the flocculated rapid mixing channel before settling, and softened water collected from the end of the softened rapid mixing channel before settling (Figure 1). Flocculated water and softened water were collected as grab samples with a dipper. Raw source water was collected in the treatment plant from their sampling tap. For most water quality measurements, technical duplicates were taken from each biological replicate, resulting in four total replicates per water type. The only exception was January 6, 2020 when one biological replicate of each water type was collected, with three technical replicates. Alum dosing to flocculated water averaged 90 ppm and varied from 60 to 93 ppm during July (Figure S1). Samples were transported in a cooler with ice and stored at 4°C. To maximize turbidity consistency with the DRWP and between analyses, all samples were stirred for 125 rpm for at least 60 s before measurements and UV exposure (Text S1 and Table S1).

2.3 Physical chemical water quality analysis

The LAMBDA 950 UV/Vis Spectrophotometer was used to take UV absorbance measurements from 200 to 350 nm. Measurements were taken according to the instrument manual. A one-centimeter path length quartz cuvette (Azzota Corp Q204) with four clear windows was filled with 3.2 mL of sample (or total liquid volume if diluted with deioized [DI] water) and rinsed with DI water between measurements. Direct absorbance through the samples was measured, and “true” absorbance accounting for reflectance was measured with the cuvette placed against the outside of the 60 mm integrating sphere (IS) with an 8° wedge at the reflectance port on a 24-mm platform. To calculate “corrected” UV absorbance, IS “true” absorbance was subtracted from direct absorbance (Mamane & Linden, 2006b).

The Malvern Panalytical Mastersizer S was used to measure the particle size distribution of the samples. Measurements were taken according to the instrument manual, with a refractive index of 0.4. Fifty milliliters of sample was dispersed at 1200 rpm (the lowest mixing rate for the wet sample dispersion unit). DI water was used in between each measurement to flush the container and tubing. To ensure the accuracy of particle size analysis, potential outlier data was identified and omitted (Text S2). Various metrics were used to analyze particle size distribution: percent of particles present at specific size ranges, the 10th, 50th, and 90th percentile particle diameter, and the De Brouckere (D[4,3]) and Sauter (D[3,2]) mean diameters. D(v, 0.1) or D10 is the particle diameter size at which 10% of total particles are smaller than it. D(v, 0.5) or D50 is the median particle size, at which 50% of total particles are smaller than the D50 value. D(v, 0.9) or D90 is the particle diameter size at which 90% of total particles are smaller than it. De Brouckere mean diameter is the volume moment mean and reflects the size of particles which constitute the bulk of the sample volume and is most sensitive to the presence of large particles. D[3,2] or the Sauter mean diameter is the surface area mean and is most sensitive to the presence of fine or small particles. For all particle size measurements, the results are plotted with the value of the right edge of the bin size (default bin sizes) on the x-axis.

Turbidity measurements were taken according to standard method 2130B (Rice et al., 2012). The Mirco 100 Turbidimeter (HF Scientific Inc.) was used. Twenty-five milliliters of sample was placed in the clear sample cell that was rinsed 3 times with the respective sample. The sample cell was also rinsed with DI water in between each water type.

Total suspended solids (TSS) were measured according to standard method 2540D (Rice et al., 2012). Glass-fiber filter disks (0.7 mm, Whatman 1825-047) were used along with aluminum weighing dishes (Fisherband, 08-732-102).

Dissolved organic carbon (DOC) was measured according to the 5310B combustion-infrared method (Rice et al., 2012) and the instrument manual for the Shimadzu TOC–VCSN analyzer. Clear borosilicate glass bottles were washed with DI three times, sealed with aluminum foil, and baked at 550°C for at least 4 h. 25–30 mL of sample was filtered through a 0.45-μm Polypropylene membrane syringe filter (Foxx Life Sciences, 37B-3216-OEM). DI water was measured between each water type. The instrument was set to do a 50-μL injection three times of each sample cell, with 2 minutes of sparging, and 1.5% acid injection.

UV absorbance normalized by DOC is called specific ultraviolet absorbance (SUVA), and is a measure of aromaticity and has been positively correlated with DOC hydrophobicity and molecular weight (Peacock et al., 2014). It provides a general characterization of the natural organic matter (NOM) in the water. A high SUVA value indicates that there is a large portion of humic matter in the sample. SUVA was calculated by dividing the UVA at 254 nm by DOC.

2.4 UV inactivation experiments

2.4.1 Enumeration

Pasteurization was conducted to enumerate only aerobic bacterial spores and not vegetative cells of ASFB. Samples were placed in a sterile plastic tube incubated at 35–37°C for 30 min, and then pasteurized for 15 min at 65°C in a water bath (Barbeau et al., 1997; Mamane-Gravetz & Linden, 2005). Samples were placed on ice until enumeration when 1–10 mL sample was inoculated on a 100 mm petri dish (VWR, 25384–342) after which approximately 25 mL of sterile nutrient agar (Tryptic Soy Broth [TSB] [BD, 211825] containing 1.5% agar [BD, 214010]) at 55°C was poured over it. This pour-plating method maximized spore recovery from samples compared to vacuum membrane filtration, swirl, and spread plating (Text S3 and Table S2). All tubes were shaken manually for approximately 3 s to ensure that particles were suspended to maximize spore recovery (Table S3).

To determine the approximate proportion of indigenous spores attached to size fractions of flocculated particles, flocculated water samples with low, medium, and high turbidity (November 20, 2019, June 4, 2019, and June 18, 2019, respectively) were filtered with 100 μm (pluriStrainer), 70 μm (pluriStrainer), 12 μm (Whatman, 7060–2516), and 1.2 μm (Scientific Tisch, SF17970) filters. UV absorbance, particle size, and spore counts were measured for each size fraction. This experiment was repeated once for raw surface water and softened water.

2.4.2 UV inactivation

UV exposure was performed on samples collected from July 2019 to January 2020 according to standard methods (Bolton & Linden, 2003) using a bench scale UV quasi-parallel collimated beam apparatus with four 6-W low-pressure lamps emitting nearly monochromatic light at 254 nm (6 W Cnlight Co., Ltd, UV Linear Germicidal Lamp, ozone free) with a shutter (Figure S2), excluding raw surface water collected on August 27, 2019, when pump failure contaminated raw water. The distance between the sample surface and the UV lamps was measured each time and used to calculate the divergence factor. An ILT5000 Research Radiometer was used to measure the irradiance at the center of and across the petri dish to determine the petri factor. The average fluence rate in the water is equal to the radiometer reading at the center of the petri dish at the surface water level multiplied by the petri factor, reflection factor, water factor, and divergence factor. The average fluence is equal to the average fluence rate multiplied by the exposure time. Twelve milliliters of sample was placed in a 60-mm petri dish with a flea stir bar under the quasi-parallel beam on a magnetic stirring plate. The exposure times for pre-determined fluence 0, 10, 20, 40, 80, 160, and 200 mJ/cm2 were calculated by dividing the UV fluence by the average UV irradiance. Timing started when the shutter was pulled out and ended when the shutter was returned. Exposures were performed in technical duplicates on biological duplicate sample bottles for each water type.

2.4.3 Seeded B. subtilis

For water collected on January 6, 2020, B. subtilis subsp. Spizizenii (ATCC 6633) were spiked into each water type to compare the dose responses for indigenous spores to commercial lab strains. ATCC 6633 was reconstituted according to the ATCC method. After plating, a colony was scraped and added to a flask with 25 mL of TSB from which freezer stocks were made by centrifuging the overnight culture (shaking 24 h at 30°C) and washing with sterile DI water and resuspending in 20% glycerol and TSB. Freezer stocks (1 mL each) were stored at −80°C. To propagate spores, one freezer stock was added to 25 mL of 2 × SG medium (Leighton & Doi, 1971) at 35°C for 72 h, shaking at 180 rpm in a baffled flask, achieving a spore concentration of 108 SFU/mL. This was spiked into each water type and 1× phosphate-buffered saline (PBS) targeting an initial B. subtilis spore concentration of 106 SFU/mL for UV inactivation as described in Section 2.4.2 and enumeration as mentioned in Section 2.4.1.

2.5 Statistical analysis and modeling

The Geeraerd model (Geeraerd et al., 2000) is a mechanistic inactivation kinetic model that can include biological parameters like shouldering and tailing. It has been used for UV inactivation (Rattanakul & Oguma, 2018). Below is the log-transformed Geeraerd model:
log 10 N F N 0 = log 10 [ 1 10 log 10 N res N 0 · e k max F · e k max S F 1 + e k max S F 1 e k max F + 10 log 10 N res N 0 ] $$ \kern-6.25em {\log}_{10}\left(\frac{N_F}{N_0}\right)={\log}_{10}\left[\left(1-{10}^{\log_{10}\left(\frac{N_{\mathrm{res}}}{N_0}\right)}\right)\cdotp {e}^{-{k}_{\mathrm{max}}F}\kern10.25em \cdotp \left(\frac{e^{k_{\mathrm{max}}{S}_F}}{1+\left({e}^{k_{\mathrm{max}}{S}_F}-1\right)\bullet {e}^{-{k}_{\mathrm{max}}F}}\right)+\kern1px {10}^{\log_{10}\left(\frac{N_{\mathrm{res}}}{N_0}\right)}\right] $$ (1)
where N0 is the initial spore concentration (SFU/mL), NF is the spore concentration after UV inactivation (SFU/mL), F is UV exposure in fluence (mJ/cm2), kmax is the maximum inactivation rate (cm2/mJ), Nres is the residual population density (SFU/mL), and SF is the shoulder length in units of fluence (mJ/cm2). This equation has three phases: log-linear, shoulder, and tailing. To derive reduced models without shouldering (Equation S1) or tailing (Equation S2), set SF or Nres equal to zero, respectively. For plotting, the equations were modified from log survival to log inactivation (Equations S3–S5).

Analysis was performed on uncorrected and corrected dose responses. Uncorrected dose responses used direct spectroscopy for the absorbance measurement. Corrected dose responses had modified fluences based on the corrected spectroscopy (IS absorbance value subtracted from direct absorbance value). Excel add-in tool GInaFiT (Geeraerd et al., 2005) was used to obtain Geeraerd inactivation model parameters for dose responses (Equation 1; Equations S1 and S2). Because GInaFiT only allows for log survival, the model parameters were then used to calculate the dose responses as log inactivation (Equations S3–S5).

Pearson-wise correlations were performed between model parameters and water quality characteristics. Pearson's r-values were interpreted as follows: a strong correlation had an r-value >0.7, a moderate correlation had an r-value between 0.4 and 0.7, and an r-value <0.4 was considered a weak association. The R-functions “aov” and “TukeyHSD” were used to perform analysis of variance (ANOVA) on model parameters and post hoc testing to determine significant differences. A nested ANOVA was used to test the difference between corrected and uncorrected model parameters.

3 RESULTS

3.1 Water characteristics

During UV exposure experiments from July 9, 2019, to January 6, 2020, the average turbidity ranged from 0.978 to 215 NTU for raw surface water, 6.49 to 164 NTU for flocculated water, and 318 to 495 NTU for softened water (Figure 2a). Because turbidity and TSS measurements (Figure 2b) had a significant positive relationship for each water type (p-values < 0.05, R2 > 0.9, Table S4), and labor- and resource-intensive TSS measurements were discontinued. Because regulations are based on turbidity, turbidity measurements continued. Initial indigenous spore concentration results showed that all three water types had similar trends, but flocculated water and raw surface water had consistently higher values than softened water (Figure 2c). Flocculated water had slightly higher initial spore concentrations than raw surface water, possibly due to spores attaching to the flocculated particles before settling and allowing for spores to be more concentrated at the end of the flocculation channel where samples were collected.

Details are in the caption following the image
(a) Turbidity, (b) total suspended solids, (c) initial indigenous bacterial spore concentration, (d) dissolved organic carbon, (e) specific ultraviolet absorbance at 254 nm, and (f) direct absorbance at 254 nm of samples of unsettled flocculated water (floc), raw surface water (raw), and softened water (soft) at Dublin Road Water Treatment Plant (DRWP). Points represent the average across four total technical replicates from both biological replicates for all plots, except for plot (c) where points represent the average across four to eight total technical replicates from both biological replicates for each type. Error bars represent the standard error of the mean.

DOC results show that unlike TSS and turbidity, raw surface water had the highest DOC concentration and softened water has the lowest DOC concentration (Figure 2d). At times, SUVA for flocculated water and raw surface water followed similar trends (Figure 2e), but not as closely as turbidity. Softened water SUVA varied less than raw surface water and flocculated water SUVA.

Direct absorbance and corrected absorbance measurements were taken from wavelengths 200–350 nm (Figures S3 and S4). Softened water absorbance at 254 nm varied less than flocculated water and raw river water (Figure 2f). There were no visible trends between the three water types.

Raw surface water particle size distributions varied between sampling dates and often had no or very few particles which made measuring difficult (Text S2) and therefore it sometimes may appear as if raw surface water had larger particles than flocculated or softened water (Figures 3 and 4; Text S2). Flocculated water had larger particles with a narrower distribution than softened water (Figure 3). The size range most commonly containing the highest percentage of particles was significantly greater (t-test p-value = 0.0305) for flocculated water (30.5253–35.5618 μm) than softened water (26.2020–30.5252 μm).

Details are in the caption following the image
Percent of total particles for (a) raw river water, (b) flocculated water, and (c) softened water. Lines represent the average of four total technical replicates from two biological replicates for each water type on each collection date. The dates in the legend are in YYYYMMDD format. Full range (up to 879 μm) particle size distributions (percent and cumulative) and one sample date with each replicate plotted are in Figures S5–S7.
Details are in the caption following the image
(a) 10th, (b) 50th, and (c) 90th percentile particle diameter in samples of flocculated water, raw river water, and softened water at Dublin Road Water Treatment Plant (DRWP). (d) The Sauter mean diameter (D[3,2]) and (e) the De Brouckere mean diameter (D[4,3]) on an abbreviated graph in samples of flocculated water, raw river water, and softened water. Points represent the average across four total technical replicates from both biological replicates. Error bars represent the standard error of the mean.

On most sample dates, flocculated water had the highest D10 diameter, indicating a lower percentage of small particles than raw surface water and softened water (Figure 4a). Overall, flocculated water had the largest median particle diameter (Figure 4b). D90 was usually similar for flocculated and softened water (Figure 4c).

Flocculated water had a significantly greater Sauter mean diameter than softened water (Figure 4d), which agreed with the D10 results (Figure 4a) that softened water had proportionally more smaller particles than flocculated water. Flocculated water and softened water had similar De Brouckere mean diameter values, but the values were higher for flocculated water (Figure 4e).

Pearson correlation coefficients between water quality characteristics are in Table S5 along with p-values (Table S6). Turbidity did not have strong correlations with UVA and particle size characteristics.

3.2 Spore association and absorbance of particle size fractions

For three flocculated water samples (low turbidity [6.49 ± 0.248 NTU], medium turbidity [73.6 ± 1.58 NTU], and high turbidity [430 ± 9.00 NTU]), approximately 27.7% (±5.28%) of indigenous spores were associated with flocculated particles ≥12 μm (Figures S8 and S9a–c). Approximately 72.2% (±5.28%) of indigenous spores were either associated with flocculated particles between 1.2 and 12 μm or free-floating. Approximately 0.153% (±0.188%) of indigenous spores passed through the 1.2-μm filter. No indigenous spores were present in the flocculated water 0.45 μm filtrate. t-Tests were performed between spore concentrations in the flocculated water filtrates (Text S4), showing similar results for each type of turbidity. Absorbance and corrected absorbance scans are in Figure S10. The filtrate experiment was also performed once on softened water and raw surface water (Figure S9d,e). In raw surface water, all indigenous spores were either free-floating or associated with particles between 1.2 and 12 μm in size (Figure S9d).

3.3 UV inactivation of spores

UV exposure experiments of indigenous spores showed that raw surface water and softened water had a similar dose–response, while flocculated water had a lower dose response (Figure 5a–c). The absorbance measured with the IS was used to correct the dose received based on the actual exposure time. Correcting UV fluence to account for reflection increased the overall fluence for all water types. Analysis of variance was performed on this data set for both corrected and uncorrected dose responses for each water type to determine the statistical significance of trends observed in the supermodels. Since ANOVA results confirmed that there were no significant differences between all corrected and uncorrected model parameters for each water type (Tables S7 and S8), only uncorrected dose responses are shown.

Details are in the caption following the image
Dose–responses of indigenous spores in (a) raw surface water, (b) unsettled flocculated water, and (c) unsettled softened water. For every collection date, each type of water was collected with two biological replicates. During ultraviolet (UV) exposure, each biological replicate had two technical replicates at every dose. Dose replicates were plated in two technical replicates. The average was calculated between the biological replicates. Error bars represent the standard error of the mean of two biological replicates.

3.3.1 Modeled dose responses

Dose–responses were modeled per sample date and across all sample dates. In this paper, the model across all sample dates resulting from the average of each biological replicate for each water type and collection date is called the super model (Figure 6). Goodness-of-fit (root mean square error, R2) values showed similarities between the three Geeraerd models (shoulder, tailing, shoulder + tailing) (Table S9). However, because modeled shoulder lengths were small to negative, the tailing-only model was chosen. Previous research on indigenous spores also reported no shouldering (Mamane-Gravetz & Linden, 2004). When comparing differences in Geeraerd model parameters, flocculated water had the lowest kmax (maximum inactivation rate) and highest Nres (residual population density, or tailing parameter) (Figure 6; Table 2), indicating slower inactivation kinetics and more surviving spores post UV irradiation. ANOVA results confirmed significant differences for flocculated water compared to both softened water and raw surface water for kmax (p-values: 5.57 × 10−5 and 4.79 × 10−3, respectively) and Nres (p-values: 4.34 × 10−5 and 6.43 × 10−3, respectively). There was not a significant difference between raw surface water and softened water. All p-values are in Table S8.

Details are in the caption following the image
The lines represent the super Geeraerd-tail model for each water type. Super models were calculated from the averages of each biological replicate for each water type and collection date. UV, ultraviolet.
TABLE 2. Super model parameters for each water type and standard deviation, calculated with the averages of each biological replicate for each water type and collection date.
Water type kmax (1/fluence) Nres (SFU/mL) N0 (SFU/mL)
Raw 0.027 ± 0.004 1.168 ± 1.887 134.0 ± 1.186
Floc 0.021 ± 0.004 7.081 ± 1.561 200.6 ± 1.134
Soft 0.030 ± 0.003 0.216 ± 1.297 15.50 ± 1.124

3.3.2 Determining reduction equivalent dose of Cryptosporidium

The reduction equivalent dose (RED) is the UV dose calculated by entering the log inactivation measured during full-scale reactor testing into the UV dose–response curve that was derived through collimated beam testing for the challenge microorganism tested (USEPA, 2006). When using an indicator microbe as a surrogate for a pathogen, the UVDGM uses RED bias values to account for the impact of different inactivation kinetics between challenge and target microorganisms on real-world full-scale UV reactors (USEPA, 2006). During unideal hydraulic conditions, the log inactivation of more sensitive microorganisms will be noticeably lower than more resistant microorganisms. Because the difference between indigenous spores and Cryptosporidium was above the RED bias values in the UVDGM, we spiked less resistant B. subtilis lab spores into the three water types to calculate the RED for Cryptosporidium for these conditions to estimate the level of disinfection of Cryptosporidium in those DRWP water types.

Results from spiking B. subtilis spores into raw river water, flocculated water, and softened water collected January 6, 2020, which contained indigenous spores at 1648 ± 33, 2370 ± 45, and 65 ± 7 SFU/mL, respectively, and sterile PBS showed again that the flocculated water had the lowest dose response (Figure 7). Although the Geeraerd shoulder-tail model fitted slightly better (Table S10), Figure 7 shows the Geeraerd tail model for consistency with Figure 6. Shouldering was expected for lab-type B. subtilis spores and agreed with previous research (Mamane-Gravetz et al., 2005; Mamane-Gravetz & Linden, 2004; Sommer & Cabaj, 1993). Model parameters confirmed that flocculated water had the lowest kmax and the highest Nres (Figure 7; Table S10). Flocculated model parameters kmax and Nres significantly differed from raw surface water, softened water, and PBS model parameters (Table S11). PBS and raw surface water had similar kmax values, but significantly different Nres values. Softened water had the lowest Nres value. Water quality values for the spiked water are in Table S12.

Details are in the caption following the image
Average dose–responses and Geeraerd tail models for indigenous spores and Bacillus subtilis ATCC 6633 spores spiked into flocculated water, raw surface water, softened water collected on January 6, 2020, and phosphate-buffered saline. Error bars represent standard deviation values between the technical dose replicates.

The UV sensitivity of B. subtilis and indigenous spores was calculated by dividing the fluence of 40 mJ/cm2 by the value of log inactivation at that fluence from the spiked Geeraerd model for each water type. The UV sensitivity values of B. subtilis and indigenous spore in raw surface, flocculated, and softened water were 12.7, 26.4, and 18.2 mJ/cm2/log I, respectively. Values from “Table G.1. RED Bias values for 4.0-log Cryptosporidium Inactivation Credit as a Function of UVT and UV Challenge Microorganism Sensitivity” from the UV Disinfection Guidance Manual (2006) and the calculated UV sensitivity of B. subtilis and indigenous spores in each water type based on the Geeraerd shoulder-tail model kmax value were used to calculate the theoretical dose that Cryptosporidium would have experienced in full-scale reactors in these raw surface, flocculated, and softened waters (Table S13). The RED bias values are a function of challenge organisms, target pathogen UV sensitivities, and UV transmission. The tables provide bias factors for UVT above 65%, but the UVT during this experiment was below 65% (Table S12). After accounting for RED bias, Cryptosporidium would receive a theoretically calculated fluence above 22 mJ/cm2, which is the dose required for 4-log inactivation, in raw surface, flocculated, and softened water if the UVT is ≥65%, 90%, or 80%, respectively (Table 3). Since UVT was below 65% in the three water types, a modeled regression on the data in Table 3 (Figure S11) determined that the theoretically calculated dose for Cryptosporidium in raw surface, flocculated, and softened water would be 17.6, 6.08, and 15.03 mJ/cm2, respectively, based on a 40-mJ/cm2 fluence and UV absorbance values from January 6th samples pre-spiking.

TABLE 3. The theoretically calculated dose (mJ/cm2) for Cryptosporidium in each water type based on the ultraviolet (UV) sensitivity of Bacillus subtilis and indigenous spores in the spiking experiment for a UV transmission (UVT) range of 65%–100% and a fluence of 40 mJ/cm2.
Water type UVT (%)
≥98 ≥95 ≥90 ≥85 ≥80 ≥75 ≥65
Raw 35.4 32.3 29.2 27.8 26.8 26.1 25
Soft 34.2 29.9 25.8 23.7 22.5 21.4 20.1
Floc 33.1 28.4 23 20.2 18.7 17.5 16.1

3.4 Relationships between water characteristics and UV inactivation

Impacts of water quality measurements on dose responses are visualized with different point sizes in the appendix (Figures S12–S17) and statistically quantified in Tables S14 and S15, with p-values in Tables S16 and S17.

For raw surface water, Nres and N0 had more significant strong correlations to water quality than kmax. Nres and N0 did not have strong correlations with most of the particle size characteristics (D10, D50, D90, sauter mean diameter, DeBrouckere diameter). For kmax, corrected absorbance at 254 nm and DOC had the strongest positive correlations. For the uncorrected model, kmax is negatively related to the particle size characteristics. For the corrected model, the negative correlation became stronger and significant between kmax and the particle size characteristics (Tables S14 and S15).

In flocculated water, Nres and N0 had more significant and strong correlations than kmax. Nres and N0 were strongly correlated with TSS, turbidity, DOC, absorbance at 254 nm, and SUVA. N0 also had a strong, positive correlation with IS absorbance at 254 nm. For both corrected and uncorrected models, the particle size water quality characteristics had mostly negative correlation with the model parameters (Tables S14 and S15).

For softened water, Nres had the most significant positive correlations for non-particle size water quality characteristics, including SUVA, absorbance, and IS absorbance at 254 nm. The correlation with turbidity was significant, but it was only 0.50. N0 had strong positive correlations with particle size characteristics. Although correlations were moderate, kmax is negatively related with particle size characteristics. Overall, there were less strong correlations for kmax and Nres for softened water than for raw surface water and flocculated water (Tables S14 and S15).

When comparing the water quality characteristics to the model parameters independent of water type, a lot of the strong, positive correlations disappeared (Tables S14 and S15 [rows labeled “All”]). The strongest correlations were between N0 and absorbance and IS absorbance at 254 nm. Nres and N0 had the most significant correlations compared to kmax. For both corrected and uncorrected models, the particle size water quality characteristics had mostly moderately negative correlations with kmax.

For all three water types, kmax had a negative relationship with particle size characteristics. Larger particles lead to a lower kmax. Raw surface non-particle size water quality characteristics had stronger correlations with the Geeraerd model parameters than flocculated and softened water. For both softened and flocculated water, Nres had stronger positive correlations with non-particle size water quality parameters than kmax, including turbidity, direct UVA254, and SUVA.

4 DISCUSSION

4.1 Overall impacts of water quality on UV inactivation

Notably, model parameters and dose responses were repeatable for a given water type despite extremely varying water quality conditions and initial spore concentrations over the sampling period, both for indigenous spores and spiked B. subtilis spores, and even between different water types: raw and softened water. The final supermodel parameters combined data from experiments with extremely variable water qualities and times of the year, with other unmeasured factors that could have impacted results, yet are able to provide a reproducible estimate of UV inactivation. Climate change impacts on consistency, quantity, and quality of source waters for drinking water treatment making it especially important to understand UV inactivation kinetics under challenging scenarios. Informing regulatory changes to properly account for disinfection when turbidity is >1 NTU could be especially useful for small or aged utilities that may not be as equipped to handle highly variable water qualities.

The dose responses of both indigenous spores and spiked spores in flocculated water were significantly different than dose–responses in raw surface water and softened water. Neither turbidity, absorbance, nor any particle size parameter had a definitive impact on dose–response parameters across water types. Many water quality characteristics, including turbidity, have been shown to correlate poorly with UV inactivation or UV absorption (Cantwell & Hofmann, 2011; Soleimanpour Makuei et al., 2022). While some studies confirm absorbance as the most important process of quality control and has a relationship with dose–response (Tan et al., 2017; Wright et al., 2011), others show UVA and turbidity do not explain results (Farrell et al., 2018; Soleimanpour Makuei et al., 2022). Amount, size distribution, surface charge, chemical nature and structure of the particles' charge have been shown to be better predictors for potential shielding of microorganisms during UV disinfection than turbidity (Farrell et al., 2018; Liu et al., 2007; Mamane & Linden, 2006b; Qualls et al., 1985; Soleimanpour Makuei et al., 2022; Templeton et al., 2005), which would explain the significantly lowered UV dose response in flocculated water compared to softened and raw surface water and why softened particles did not have the same impact on UV inactivation as flocculated particles.

The maximum specific inactivation rate (kmax) is usually the main parameter used to describe dose–response kinetics. A negative correlation was observed between kmax and particle size characteristics for all water types, indicating that larger particles were associated with a lower kmax. While Carré et al. (2018) reported a strong correlation between inactivation rate constant and turbidity, UV254 transmission, and TSS, and Loge et al. (1996) stated that UV transmission and suspended solids concentration significantly impacted tailing, other studies did not report a correlation between inactivation and those water quality characteristics (Darby et al., 1993; Madge & Jensen, 2006). Instead, many studies reported a correlation between inactivation, particle size, and particle amount (Caron et al., 2007; Carré et al., 2018; Emerick et al., 1999, 2000; Jolis et al., 2001; Liu et al., 2007; Madge & Jensen, 2006; Qualls et al., 1985; Soleimanpour Makuei et al., 2022; Winward et al., 2008), which agrees with our results. Additionally, particle size data did not correlate well with turbidity or TSS, which was also noted by other studies (Cantwell et al., 2010; Qualls et al., 1983). Two studies reported that when testing on the impact of NOM on UV inactivation, it was the presence of humic acid that had a significant negative effect on UV inactivation and not turbidity up to 20 NTU (Baldasso et al., 2021; Pichel et al., 2021), which supports our results: For all three water types, higher SUVA values significantly correlated with a higher residual population density.

4.2 Particle size, structure, and interaction with microorganisms

The presence and nature of flocculated particles could still explain the difference in UV inactivation between flocculated water and soft and raw surface water. Particle size distributions demonstrated that flocculated water had larger particles that were also more narrowly distributed than softened water. Larger particles have a stronger negative impact on UV inactivation, often through tailing, than smaller particles (Carré et al., 2018; Madge & Jensen, 2006; Winward et al., 2008). Studies, where turbidity (up to 20 NTU) did not significantly impact UV inactivation, reported small particle sizes (Amoah et al., 2005; Cantwell et al., 2010; Pichel et al., 2021; Templeton et al., 2009), indicating that larger particles instead of higher turbidity is the main driver for lowered UV inactivation.

Results indicated that spores were free-floating in the raw surface water but predominantly particle-associated in the flocculated water. Based on the significant difference in the initial spore concentration between flocculated and softened water, a majority of the indigenous spores were likely associated with flocculated particles rather than free-floating and settled in the flocculation basin while the small number of free-floating spores continued to the softening channel. The free-floating spores were most likely present as self-aggregates rather than individual cells due to the extremely low concentration in the 1.2-μm filtrate for all three water types. It is easier to disinfect free-floating microorganisms and spores than particle-associated microorganisms (Emerick et al., 2000; Mamane & Linden, 2006a; Örmeci & Linden, 2002). Additionally, although softened water and raw surface water had similar dose–responses, the initial concentration of indigenous spores in softened water was extremely low. This can lead to high variability at high UV doses.

The flocculated particles also have a different structure than softened particles. Different particle composition characteristics, such as surface charge and porosity, can impact the effect on UV disinfection. In agreement with our study, Liu et al. (2007) noted that even at a turbidity of 32 NTU, surface water particles essentially had no influence on spiked Escherichia coli UV inactivation, while the presence of flocculated particles in lower turbidity waters led to significantly lower E. coli UV inactivation. Additionally, Templeton et al. (2005) reported that humic acid flocculated particles enmeshed and protected viral surrogates at extremely high turbidity levels, while inorganic kaolin clay particles did not provide protection under the same turbidity conditions. Flocculated particles have a porosity and structure conducive to trapping microorganisms (Gorczyca & Ganczarczyk, 2001), while softening particles do not. Recent studies show a strong correlation between surface charge of particles and negative impact on tailing (Farrell et al., 2018; Soleimanpour Makuei et al., 2022; Tan et al., 2017). Additionally, two studies determined that even a low concentration of NOM (as humic acid) had a significant detrimental effect on UV disinfection while turbidities below 5 NTU and 20 NTU did not (Baldasso et al., 2021; Pichel et al., 2021). UV dose–response differences between flocculated and softened water were not explained by turbidity and are more likely due to differences in particle size and structure.

4.3 Translation to practice

The spiking study is likely representative because the combined spiked B. subtilis and indigenous spores (mixed) had similar dose responses across the three water types compared to indigenous spores only across all three water types. Flocculated water had a significantly lower maximum inactivation rate and a significantly higher residual population density. The mixed maximum inactivation rate is one order of magnitude greater than the maximum inactivation rate of the indigenous spores only for all three water types because B. subtilis is a lab-type spore, which is less resistant than wild spores (Mamane-Gravetz et al., 2005; Mamane-Gravetz & Linden, 2004, 2005), and likely mostly free-floating (Mamane-Gravetz & Linden, 2005) and therefore easier to disinfect than spore-particle aggregates (Farrell et al., 2018; Mamane & Linden, 2006a; Örmeci & Linden, 2002).

Because RED bias values are derived from real-world worst-case UV reactors, the calculation is considered an overly conservative estimate. Additionally, calculated RED for Cryptosporidium is linked to UVT, even though our results show that turbidity and UVT were not well correlated with changes in inactivation kinetics, making RED even more conservative. Calculated RED for Cryptosporidium would achieve ≥4-log inactivation when UVT is above 65%, 90%, or 80% in raw surface, flocculated, or softened water, respectively. However, 4-log Cryptosporidium inactivation in water with worse UVT values may be predictably achieved if other impacts of water quality factors, such as particle size or degree of microorganism-particle associations, are quantified. If regulations are adjusted in a way that accounts for inactivation that occurs past 1 NTU, small systems with UV disinfection can work more efficiently especially as climate change will degrade source waters.

5 CONCLUSION

Although turbidity can be a good indicator of water quality, it is not the most mechanistically precise water quality parameter to regulate UV disinfection. Raw surface water and softened water had very different water qualities including turbidity, but their indigenous spore UV dose–responses were similar to each other, and the UV inactivation of spiked spores in both raw and softened water was similar to inactivation in lab water. In this and other studies, particle size may be a better indicator of UV inactivation than turbidity. Particle size and the degree of particle-associated microorganisms impacted the maximum inactivation rate and tailing, especially in flocculated water which was negatively impacted by adverse water quality while raw surface and softened water were not. Particle-associated microorganisms are more difficult to disinfect with UV. However, with very poor water quality, there was still significant and repeatable inactivation of indigenous spores and seeded B. subtilis. This means significant and predictable inactivation would be expected of the more sensitive C. parvum (oo)cysts even under variable water quality including very high turbidity conditions with unsettled particles. With climate change driving lower and more variable water quality of source waters, current UV disinfection regulations may be too conservative by focusing only on turbidity and/or UVT when particle size distribution and aggregations should also be considered to more accurately predict levels of disinfection and protect public health without unnecessary utility violations.

AUTHOR CONTRIBUTIONS

Judith Straathof: Data curation; formal analysis; validation; investigation; visualization; methodology; writing – original draft; writing – review and editing. Zuzana Bohrerova: Conceptualization; supervision; methodology; writing – review and editing. Natalie M. Hull: Conceptualization; supervision; funding acquisition; validation; methodology; project administration; writing – review and editing.

ACKNOWLEDGMENTS

This study was supported by funding from the Water Resources Institute 104(b) Program through the Ohio Water Resources Center. We would like to thank Dr. Merle de Kreuk and Dr. Jan Peter van der Hoek for their comments and feedback on the thesis portion of this journal article.

    CONFLICT OF INTEREST STATEMENT

    The authors declare no conflict of interest.

    DATA AVAILABILITY STATEMENT

    The data that support the findings of this study are available from the corresponding author upon reasonable request.