Removals of cryptosporidium parvum oocysts and cryptosporidium-sized polystyrene microspheres from swimming pool water by diatomaceous earth filtration and perlite-sand filtration

Numerous outbreaks of cryptosporidiosis have been reported to be associated with swimming pools in the United States (USA), United Kingdom (UK), Australia, and other countries over the past 20 years (Lisle & Rose 1995; Puech et al. 2001; Karanis et al. 2006; Hlavsa et al. 2011; Lu et al. 2013). In August 1988, the first outbreak of cryptosporidiosis associated with a swimming pool in the UK was recognized (Joce et al. 1991). In 1992, a large increase in the number of cases of cryptosporidiosis was reported in the USA (McAnulty et al. 1994). Due to Cryptosporidium oocysts being extremely tolerant to free chlorine at levels typically used in treated recreational water venues, this pathogen poses a significant risk to bathers, as exemplified in large outbreaks in New York in 2005 (more than 4,000 cases of cryptosporidiosis), in Utah in 2007 (nearly 2,000 cases), and in the Dallas, TX area in 2008 (at least 378 cases) (Rolfs et al. 2008; Cantey et al. 2012).

The Centers for Disease Control and Prevention (CDC) recommends that swimming pools in the USA maintain a free chlorine residual of 1–3 mg/L for effective disinfection (CDC 2012). At 1 mg/L of free chlorine, achieving a 3 log (99.9%) inactivation of Cryptosporidium could require almost 11 days assuming a 3 log Ct at pH 7.5 [free chlorine concentration (C) × exposure time (t)] of 15,300 mg/L·min (Shields et al. 2008). Thus, other treatment techniques (e.g., filtration, UV irradiation) are necessary to reduce Cryptosporidium in treated recreational water venues under normal operating conditions. However, swimming pool sand filters in the USA are not very effective at removing Cryptosporidium parvum oocysts and Cryptosporidium-sized (4.5 μm) microspheres (<63%, or 0.43 log) (Amburgey 2011; Lu & Amburgey 2016).

Media such as diatomaceous earth (DE) and perlite can be used to precoat filters and improve filtration efficacy. DE is a naturally occurring, soft, siliceous sedimentary rock and can be applied to filters in the form of a very fine white to off-white powder. Perlite is a naturally occurring amorphous volcanic glass that, when heated sufficiently, has the unusual property of greatly expanding in volume with a concomitant drop in density (10- to 35-fold). While often used in horticultural and insulation applications, perlite is also widely used as a filter aid. Amburgey et al. (2012) reported that DE and perlite precoat filters demonstrated pathogen removal efficiencies of 2.3 to 4.4 log (or 99.4–99.996%) in a small-scale 946 L model swimming pool.

Intentionally interrupting the water flow in a precoat filter by turning off the filter pump allows the precoat media and collected contaminants to detach from the filter septum and become intermixed. When the filter is restarted, the media reattaches to the filter septum, forming a new (intermixed) matrix. This process can lengthen the media lifespan. However, because interruption of water flow changes the media and pathogen distribution within the filter, it might impair pathogen removal or facilitate the release of pathogens previously trapped in the filter. The effects of flow interruption have not been sufficiently studied (Amburgey et al. 2012).

Additional research is needed to optimize DE filter performance in removing Cryptosporidium oocysts in pilot- and full-scale pools. Parameters to be investigated included water flow rate, mass/depth of DE, and operational practices like filter flow interruption.

The current study investigated Cryptosporidium-sized microsphere removals with different amounts of precoat (i.e., DE or perlite) in a pilot-scale swimming pool in order to determine an appropriate amount of precoat for full-scale pool operation. Cryptosporidium-sized microsphere and C. parvum removals by DE filtration and perlite-sand filtration in a full-scale swimming pool were evaluated to verify the filtration performance. The impact of flow interruption on DE filter performance was evaluated in pilot-scale and full-scale pools to determine its impact on the removal of Cryptosporidium from the pool.

For pilot-scale study, duplicate experiments were conducted with triplicate samples (50 mL influent and 500 mL effluent samples were collected); for full-scale study, triplicate samples were collected (500 mL influent and 1 L effluent samples were collected).

Fluoresbrite™ Carboxylate YG polystyrene microspheres (Polysciences, Inc., Warrington, PA, USA) with diameter of 4.5 μm were used as a model particle surrogate for C. parvum oocysts in both the pilot- and full-scale studies because they are virtually identical to Cryptosporidium oocysts in size, shape, and density. For microsphere analysis, influent and effluent samples were mixed by vortexing and hand shaking for at least two minutes each before analysis. Samples were passed through 3.0 μm pore size polycarbonate track-etched filters. Each polycarbonate filter was mounted on a glass microscope slide with polyvinyl alcohol-DABCO solution, covered with a glass cover slip, and enumerated under an epifluorescence microscope (Lu & Amburgey 2016).

Heat-inactivated C. parvum oocysts were also used in full-scale experiments. In the perlite study, oocysts produced in experimentally infected bovine calves were purified from calf feces by sucrose and cesium chloride gradients and stored in phosphate-buffered saline (PBS) with antibiotics and 0.01% Tween 20 for approximately 7.5 months. In the DE study, oocysts were purchased from Waterborne, Inc. (New Orleans, LA, USA) and stored at 4 °C in PBS with antibiotics for approximately four months. Oocysts in both studies were heat-inactivated at 55 °C for 1 hour prior to seeding.

For C. parvum analysis, filter influent and effluent samples were concentrated using two centrifugation steps at 4,000 × g at 4 °C for 15 minutes followed by two microcentrifugation steps at 15,000 × g at 4 °C for 5 minutes. All bottles and tubes were rinsed with a Tween 80/Antifoam A solution (0.01% and 0.001% [v/v], respectively) between transfers to prevent oocysts from sticking. The entire resuspended pellet was transferred onto one or more Crypto SuperStick™ slide wells (Waterborne™, Inc., New Orleans, LA) and stained with Easy Stain™ (TCS Biosciences Ltd, Buckingham, UK) per the manufacturer's instructions. C. parvum oocysts were detected and enumerated using an epifluorescence microscope under 200 × magnification (USEPA Method 1623, 1999) (USEPA 1999).

DE density was 320 kg/m³, and the permeability was 3.6 μm². The filter surface area was 2.23 m². The amount of precoat was defined as the mass of DE loaded divided by the filter surface area. A DE slurry was prepared by mixing dry DE media into water at a volumetric ratio of 1:2 and mixed well in a 10 L plastic container. The slurry was delivered through a skimmer, circulated through the system, and collected on the filter septum resulting in final amounts of 0.5, 0.7, and 1.0 kg·DE/m². Pool water was pumped through the filter and circulated at filtration rates of 3.6 m/h, 5 m/h, and 6 m/h.

To assess pilot-scale DE filtration efficacy, microspheres (final concentration of 150 microspheres/mL in the pool water) were fed over 60 minutes directly before the filter pump with a peristaltic pump (Model 505 Di, Watson Marlow, Wilmington, MA, USA). Triplicate samples were collected over the 60-minute time period (at 10 min, 30 min, and 50 min), and triplicate samples were taken each time to assess variability. Flow interruption was conducted by restarting the filter after a 5-minute, 15-minute, or 30-minute interruption of filter flow. Influent samples were collected in 50 mL polypropylene centrifuge tubes, and effluent samples were collected in 500 mL glass bottles. All samples were stored at 4 °C and analyzed within 48 hours.

The filter surface area was 2.2 m². Dry DE media was delivered through a skimmer, circulated through the system, and collected on the filter septum, resulting in final density of 0.7 kg DE/m³. Pool water was pumped through the filter and circulated at a filtration rate of 3.6 m/h.

To assess full-scale DE filtration efficiency, approximately 2.5 microspheres/mL (10⁸ total) microspheres and 1.5 oocysts/mL (6 × 10⁷ total) were then mixed and seeded to the pipe directly before the circulation pump over a 60-minute interval (as shown in Figure 2). Triplicate samples were collected 30 minutes after seeding began. Influent (500 mL) and effluent (1 L) samples were collected in glass bottles containing Tween 80 (0.01% [v/v]), Antifoam (0.001% [v/v]), and sodium polyphosphate (0.01% [v/v]), to prevent oocysts from sticking to surfaces. Flow interruption was conducted by restarting the filter after a 15-minute interruption of filter flow and then collecting influent and effluent samples as described above. All samples were stored at 4 °C prior to analysis.

Experiments were carried out using 30 cm of sand with or without different amounts of perlite added on top of the sand. Table 1 shows sand and perlite characteristics as well as filter operation details. The filter surface area was 0.018 m². Each perlite slurry was prepared by mixing perlite dry media with water at a volumetric ratio of 1:2 in a 10 L plastic container. The perlite slurries were delivered through a skimmer, circulated through the system, and collected on the filter surface resulting in final amounts of 0.24, 0.37, 0.49, 0.61 kg·perlite/m². Pool water was pumped through the filter at a filtration rate of 37 m/h.

To assess pilot-scale perlite-sand filtration efficiency, microspheres (influent concentration of 150 microspheres/mL in the pool), were fed over 60 minutes directly before the circulation pump. Microspheres were seeded 15 minutes before each sample collection in order to ensure microspheres could be present in the filter effluent during sample collection. The theoretical hydraulic detention time of the piping and filtration systems was approximately 3 minutes.

Perlite (Tech-Flo 2000X/SwimBrite, IIG, LLC, Brunswick, GA, USA) permeability was 3.21 μm², and the particle size range was between 2 and 200 μm. A control experiment with the sand filter was conducted with sand media only. Dry perlite media were delivered through a skimmer, circulated through the system, and collected on the sand filter surface resulting in a final amount of 0.5 kg·perlite/m². Pool water was pumped through the filter at a surface loading rate of 34 m/h.

To assess full-scale perlite-sand filtration efficiency, approximately 2.5 microspheres/mL (10⁸ total) microspheres and 1.5 oocysts/mL (6 × 10⁷ total) were then mixed and seeded to the pipe directly before the circulation pump over a 60-minute interval (as shown in Figure 2). The remaining experimental procedures were as described in ‘Full-scale’ in the subsection ‘DE experimental procedure’.

Box plots are used to present the C. parvum oocyst and microsphere removal efficiencies. The box plots depict removal data in five metrics: the smallest observation (or sample minimum), lower quartile, median, upper quartile, and largest observation (or sample maximum). Removal of microspheres through each filtration process was expressed either as percent removal or in terms of the logarithmic reductions (base 10). Log10 removals are presented as the log10 of the ratio of the influent and effluent concentrations. Log removals that incorporated non-detects (i.e., no particles detected in filtrate) are prefixed with the ‘ > ’ symbol and based on a calculated value that assumed one particle was present in the sample volume analyzed. A two-tailed t-test was performed to compare removals of C. parvum oocysts and microspheres before and after flow interruption (Microsoft Excel 2010, Redmond, WA, USA). The t-test significance level applied was α = 0.05. In addition, analysis of variance (ANOVA) was performed to compare microsphere removals by different amounts of DE (Microsoft Excel).

Removals of microspheres after flow interruption with a 5-minute stop were 1.7 ± 0.16 log (97.1% to 98.6%) for 0.5 kg DE/m², 2.1 ± 0.21 log (98.7% to 99.5%) for 0.7 kg DE/m², and 1.8 ± 0.17 log (97.7% to 98.9%) for 1.0 kg DE/m². No significant differences were observed between the removals before and after flow interruption at a filtration rate of 5 m/h with a 5-minute interruption of the filter flow for the different amounts of precoat as determined by a paired t-test (p = 0.36 for 0.5 kg DE/m², p = 0.38 for 0.7 kg DE/m², and p = 0.06 for 1.0 kg DE/m²).

Perlite-sand filtration provided roughly a 2 log improvement in the removal of microspheres as compared to the high-rate sand filtration control. The increased removal could be attributed to the small size of perlite (and the correspondingly small pores). The effective size of the sand was more than 28 times larger than that of the perlite. The removal of microspheres in a precoat media filter occurs by straining through the pores in the filter bed, or by sedimentation of microspheres in the media pores (Letterman & Yiacoumi 2010).

The initial pressure loss was 20 kPa for the sand-only filter prior to perlite addition, but the pressure increased as the amount of perlite increased. Initial filter influent pressures for the 0.24 kg·perlite/m², 0.37 kg·perlite/m², 0.49 kg·perlite/m², and 0.61 kg·perlite/m² conditions at a filtration rate of 37 m/h, were 55 kPa, 55 kPa, 62 kPa, and 69 kPa, respectively.

Removals after flow interruption were not significantly decreased by 15-minute stop of the DE filter at a filtration rate of 3.6 m/h (t-test p = 3.9). Removals obtained from full-scale filters after flow interruption were slightly higher than those at the pilot-scale.

Microsphere removals by the full-scale perlite-sand filter (having a thin perlite layer overlying the sand) were 2.7 log (99.8%) (data not shown), compared with 2.4 log (99.7%) in pilot-scale trials at 0.5 kg·perlite/m². Replicate samples were lost due to an unplanned backwash. Data for C. parvum oocyst removal could not be collected because oocysts were mistakenly seeded at a level that was too low (resulting in bulk water concentration <0.001 oocysts/mL). The full-scale pool could not be studied further, so only microsphere data were collected for the perlite-sand filter.

The results indicate that sand filters are relatively ineffective for C. parvum oocyst and microsphere removal from swimming pools, with removals in the range of 0.1–0.4 log (20% to 63%). The DE precoat filter and perlite-sand filter improved removals significantly. At least 0.7 kg DE/m² was required to achieve 2 log (99%) of microsphere removals at 5 m/h indicated by pilot-scale studies. When DE filtration was performed with 0.7 kg DE/m² and a filtration rate of 3.6 m/h, greater than 2.7 log (99.8%) of C. parvum oocysts and microspheres were removed from full-scale swimming pool water. Pilot-scale results for DE filters showed filter restarts with flow interruption times of 15 minutes led to decreased removals of C. parvum and microspheres at the same filtration rate, but this effect was not observed for the full-scale pool filter.

Adding a layer of perlite on the top of the sand filter increased microsphere removal compared with a sand filter only. The pilot-scale results showed average microsphere removals in the pilot study were 2.3 log (99.4%), 2.4 log (99.7%), and 2.5 log (99.8%) for 0.37 kg·perlite/m², 0.49 kg·perlite/m², and 0.61 kg·perlite/m², respectively. Filter influent pressure increased since the perlite was captured at the surface of sand in the filter. Perlite-sand filtration (0.5 kg·perlite/m²) in the full-scale study removed a similar level of microspheres [2.7 log (99.8%)].

The data from this study demonstrate that under the tested conditions the removal of Cryptosporidium-sized microspheres and C. parvum oocysts from swimming pool water can be significantly increased above typical sand filter levels through the use of a DE precoat filter or by applying a perlite layer of at least 0.37 kg·perlite/m² to a sand filter. Pool operators and pool water treatment engineers may want to consider using DE or perlite to improve the performance of pool filters for removing Cryptosporidium oocysts and other particulates.

The authors wish to thank the National Swimming Pool Foundation (NSPF), Arch Chemicals, Inc (now Lonza Group Ltd) and other companies whose products were tested, the National Natural Science Foundation of China (41403090), the Fundamental Research Funds for the Central Universities (2017QNA35), and Amber Khanzada for her assistance with pool maintenance in the full-scale studies. The use of trade names and names of commercial sources is for identification only and does not imply endorsement by the CDC or the US Department of Health and Human Services. The findings and conclusions in this report are those of the authors and do not necessarily represent those of the CDC.