Filter media for basin infiltration: a case study

Managed aquifer recharge (MAR) is a common way of producing and storing groundwater worldwide (SWWA 2016; Sprenger et al. 2017; Stefan & Ansems 2018), where aquifers are recharged with surface water through infiltration basins. During infiltration and percolation, natural organic matter (NOM) and pathogens are removed to varying degrees in the unsaturated and saturated zones (WHO 2016; Sarma 2020; Hägg et al. 2021). Infiltration basins follow certain operational cycles, where the resistance increases in the basins due to the natural compaction of the sand and accumulation of organic and inorganic matter, and microorganisms (Jeong et al. 2018). As a result, infiltration basins require maintenance by emptying and scrapping the top layer of sand. Through this process, water treatment plants (WTPs) experience losses of natural sand material; therefore infiltration basins need additional sand to replace these losses. It is common that the natural sand material from the recharge field is used or that new sand is brought in from other natural glacifluvial deposit sources. In Sweden, it is very common that MAR-plants use sand that follow the Swedish recommendations on slow sand filtration (SFF) sand (Hägg et al. 2018). The specifications for filter media used for SSF vary internationally and is often applied for infiltration basins (Huisman & Wood 1974; Hansson 2000). However, the situation for MAR plants is very different from SSF, where the natural grain size variations in the recharge fields makes it difficult to apply SSF sand recommendations to infiltration basins. Very limited information is published for practitioners when it comes to guidelines and recommendations on sand quality for MAR plants. The purpose of this study was to identify a close to optimal sand quality for infiltration basins based on the natural geological conditions, and apply this reasoning in a case study at Vomb WTP, in south Sweden. The goal is to give practitioners tools and ideas on how to establish an optimal filter sand quality based on their requirements.

The function of filter sand is to have good hydraulic properties and high surface area for retention of organic matter and other contaminants. The hydraulic properties of the filter media is mainly governed by the sand grain size distribution, which is commonly described by grain sizes d₁₀, d₆₀ and the uniformity constant (UC) (Brandt et al. 2016; Crittenden et al. 2017). d₁₀ and d₆₀ are size measures of the granular media determined by sieve analysis, where the sizes of which 10 and 60 percent of the media are finer, respectively. The concept of d₁₀, the effective size (ES), was proposed in 1890 because it gave a good measurement of hydraulic conductivity independent of the overall grain size distribution (Fair et al. 1971). Other grading characteristics, e.g. d₂₀, d₆₀ and UC, have also been shown to affect hydraulic conductivity (Cabalar & Akbulut 2016). The effective size should be as low as possible to achieve adequate treatment and to avoid deep clogging (Huisman & Wood 1974). At the same time, too low effective size could result in inadequate permeability (Ellis & Wood 1985; Cheremisinoff 2019). UC is a measurement of the uniformity of the granular media and is given by the ratio of d₆₀ and d₁₀ (UC=d₆₀/d₁₀). The measurement is useful for filter media that undergoes backwashing to avoid stratification, such as rapid sand filters (Crittenden et al. 2017). For slow sand filters, UC is useful to ensure that the granular media is reasonably uniform with good porosity (Huisman & Wood 1974).

The function of infiltration basins is in many ways similar to SSF, why this could be suitable for MAR application. However, the recommendations for this filter sand varies internationally (Table 1).

Clogging can occur in different places and for different reasons (Song et al. 2020), where retention of inorganic particles tends to occur deeper than organic particles (Hendricks 1991; Wang et al. 2018). The accumulation of organic matter and biofilm development is important for the function of sand filters, especially considering the comparably large pore size of clean SSF sand (about 60 μm) (WHO 2016). Many studies have investigated granular filter media capability to retain fine particles through full-scale and column tests (Siriwardene et al. 2007; Martin 2013; Du et al. 2018). Schuh (1990) showed in a field study that clay particles from turbid infiltrated water were found up to 38 cm of depth in the infiltration basin. Fine particles (<50 μm) have also been found to be retained in and penetrated through columns with fine sand (d₅₀ = 0.5 mm) (Du et al. 2018). The same study showed the infiltration rates had a considerable impact on particle mobility.

Many articles describe the connection between mobilization and retention of particles and the grain size of the media and the particles. Crittenden et al. (2017) describes that a uniform media can retain particles that are down to 15% of the media's diameter, which means that a filter media with a grain size of 0.5 mm can retain particles bigger than 75 μm. Particles entering sand filters are also reported to be retained if the particle size is at least 20% of the filter media's d₁₀-value (Huisman & Olsthoorn 1983). In column trails with filter sand, the clogging phenomena is often described as either surface, mixed or inner clogging. In a column with fine to coarse sand, surface clogging occurred for 75 μm particles, mixed clogging for 40 μm particles, and particles below 13 μm were able to pass through the entire column (Du et al. 2014).

For infiltration basins, it is important that fine particles from the infiltration water (raw water) are retained in the surface and not transported into the aquifer. For this reason, microscreens are often used for pre-treatment for sand filters to reduce and mitigated clogging (Ljunggren 2006). Other pre-treatment methods for removal of algae, organic and inorganic matter include chemical flocculation, sedimentation and membrane filtration (Li et al. 2019; Hägg et al. 2020)The filter media used in infiltration can also be a source for fine particles that could cause clogging issues in the native material (Hansson 2000). The transport of clay particles in aquifers can occur over significant distance and result in reduced permeability (McDowell-Boyer et al. 1986; Akhtar et al. 2021). In Sweden, it has been reported that WTPs have chosen to use unwashed granular media in their basins, resulting in deep clogging due to fine particles in the sand (SWWA 2007). The importance of washing filter sand used in infiltration basins to remove fine particles cannot be understated, and has been set as recommendations in several rapports, handbooks, and articles (Huisman & Wood 1974; Visscher 1990; Hendricks 1991; Sundlöf & Kronqvist 1992; Jabur & Mårtensson 1997; Johansson 2003; Jönsson & Wikström 2003; Sternö 2005; JTI 2016; Ten States Standard 2018). The reported recommended standards for silt concentrations have been in the range of 0.5–1.0 g/L (Brandt et al. 2016), which is equivalent to about 0.03–0.06% by weight.

Granular media for filters need to be resistant to shear forces to avoid weathering (Ratnayaka et al. 2009), therefore the media should mainly consist of quartz and feldspar and free from mica mineral (Hansson 2000; SWWA 2011). Micas are considered weathering minerals due to their sheetlike structure (Fetter 2014) and easily break from mechanical stress (Göransson 2014), which is also an issue in road construction (Hellman 2013). For MAR plants, the issue of chemical and mechanical weathering mica minerals is the creation of clay particles (Höbeda 1987; Göransson 2014; Price & Velbel 2014; Johansson 2015), which can result in clogging.

Calcium and magnesium concentrations should be below 2% (calculated as carbonate) to avoid cavities in the filter media (Huisman & Wood 1974). The requirements for acid solubility for SSF sand varies. The Swedish standard is maximum 1% weight reduction after samples have been soaking in a 10% hydrochloric acid solution for 2–3 hours (SWWA 2011). Internationally, it is common to measure a 5% weight reduction over a 24-hour period in concentrated hydrochloric acid (40%) (Huisman & Wood 1974; Ten States Standard 2018). The recommendations for filter media is that it should be washed and free from any contaminants (Huisman & Wood 1974; Hansson 2000; SWWA 2011; Ten States Standard 2018). However, it is very difficult to find natural deposits without any traces of unwanted contaminants, e.g., chloride, arsenic, and iron. Useful information on soil contamination has been published by the Swedish Environmental Protection Agency (EPA), where the agency has set guidance limits on metal concentrations in potentially contaminated soil intended for sensitive land use (as mg/kg total solids), e.g., parks and residential areas (Swedish EPA 2016). It is recommended that a site-specific limit is developed for each site and the values for Vomb WTP with the general limits are noted in Table 2.

Other recommendations for SSF sand from the Swedish Water Works Association (SWWA) are the following:

• Turbidity (after wash) ≤4 FNU

• Acid solubility ≤1%

• Iron concentration ≤0.1%

• Humic substances ≤1,000 color (mg Pt/L)

• Chemical oxygen demand (COD_Mn) ≤0.3 mg/g.

The infiltration basins follow certain seasonal and operational cycles, where newly renovated and refilled basins are taken into operation in late autumn-winter to ensure that the bottom of the basins are covered by water and a bioactive layer is created before the algal blooms in late summer (Li 2020). Once the development of the biofilm creates too much resistance (bioclogging), usually around 1–2 years, the basins are emptied and skimmed (about 5 cm of the top layer). The basins are then restarted and continue to be in operation until the resistance is too high. After a few years, once a basin has been skimmed about 2–5 times, the basin is renovated by excavating the top 50 cm of the basin. The sand is washed, mixed with stored native sand, and put back into the basin. After that, the newly renovated basin is ready to be returned to operation.

In this study, 114 samples from the infiltration basins were collected for sieve analysis (Table A1, supplementary material). The samples were taken at varies depths in the infiltration basins and were analysed at Vomb WTP. The sieve sizes were 2.00, 1.60, 1.25, 1.00, 0.80, 0.63, 0.50, 0.40, 0.315, 0.25, 0.20, 0.125 and 0.071 mm. The results were compared with literature data and operational experiences to create a filter media with viable properties for infiltration basins. With this information, sand with viable and optimal properties were suggested.

The distribution showed that the low and high flow rate basins had some overlap in the large fraction range and no overlap in the smaller fraction range. This result concurs with Fair et al. (1971) that stated that the effective size was the determinant factor for conductivity. Another way to illustrate the effect of grain sizes on infiltration rates can be done by calculating the hydraulic conductivity (Hazen 1911) and specific grain surface area (Crittenden et al. 2017) (Figure A1, supplementary material). The basins with a permeability that is too low also showed high total levels of fine grain sand (<0.2 mm) (Table 3). The UC and d₆₀ do not seem to be a good indicator for basin permeability within this dataset, which is why the finer grain sizes might be more suitable for infiltration rate predictions. However, statistical verification would require a larger sample size. The 5 basins that showed too low hydraulic conductivity had an effective size, d₁₀, up to 0.21 mm. Out of the 54 basins, only two have showed too high infiltration rates, d₁₀=0.41 and 0.87 mm. Note that the basin with d₁₀=0.41 comes from one sieve sample 1.5 m below the surface and might not be the sand layer causing high infiltration rates.

The grain size distribution of the glacifluvial deposits at Vomb WTP consists of finer material than many other recharge fields used in Sweden. Samples from filter media taken from basins at Swedish MAR plants showed an effective size range of 0.27–0.43 mm (Hansson 2000), which is why SSF sand with the Swedish standard is likely more viable for these WTPs.

The sand properties for the basins need to be based on the literature data and experiences from other MAR plants, and applied to the natural conditions in the recharge field. The optimal method would be to adapt the grain sizes of the filter media for each basin. However, this is not possible for practical reasons. Instead, the goal is to find a filter media with acceptable grain size distributions for the whole recharge field. According to the WHO, the filter media for MAR should have equal or slightly coarser grain size distribution than SSF sand, which would assumably imply d₁₀ ≥ 0.15 mm. Furthermore, in sand filters, each sand layer should have less than 4 times the effective size (d₁₀) of the sand layer above it. If applied to Vomb WTP, the effective size of the filter media used in the basins should be d₁₀ > 0.21 mm, based on the coarsest native glacifluvial deposits (d₁₀ = 0.87 mm). Based on operational experience, the basins with too low infiltration rates have an effective size of d₁₀ ≤ 0.21 mm. This gives a good indication that new filter media for Vomb WTP should have an effective size of d₁₀ ≥ 0,22 mm. The upper limit for the effective size is more difficult to establish because of the wide range of grain sizes in the recharge field. A starting point that could be applied is that the filter media should be finer than the native sand media to avoid clogging in the aquifer (Huisman & Wood 1974). To set the upper d₁₀ limit as high as possible to avoid limiting the access from suppliers and at the same time make sure that the sand would be adequate for a majority of the basins. A d₁₀ value higher than the median effective size of the field was chosen for this reason. In this case, d₁₀≤0.38 mm, which is equivalent to the coarsest Swedish recommendation for SSF sand (Table 1). An optimal range has also been set to d₁₀ 0.22–0.30 mm, making it more suitable for all infiltration basins, increasing the protection of the aquifer from clogging and reducing high infiltration rate in the early stages of newly started basins.

The uniformity coefficient (UC) does not seem to have a significant impact on permeability according to the literature and collected data. However, the recommendation from the WHO is that the UC value should be within a similar range of the surrounding sand layers. The collected UC values are very broad, around 1.1–5.2, and a majority of the UC recommendations are mostly around 1.5–3 with a couple of exceptions extending it to about 5. Thus, a suitable choice is likely around 1.5–3, which is also relatively close to the median UC-value of the glacifluvial deposits (UC=2.3). An acceptable alternative is possibly to allow for UC values up to 5. The coarser material only inflates the cost of the filter media and does not have any meaningful impact on water treatment, therefore these grain sizes should be limited. The Swedish recommendation on filter media used for SSF also include an upper cut-off for grain sizes above 2.0 mm, which is omitted in the international recommendations. It is possible that the effective grain sizes and UC values are sufficient, and that this recommendation is unnecessary. However, the natural glacifluvial deposits have a broad grain size distribution and the grain size cut-off is needed to determine sieve sizes when maintaining and washing the sand supply. The median value for the coarser grain sizes (>2.0 mm) in the recharge field 4.6%, which would result in some losses if this cut-off was chosen. The purpose of adding filter media in MAR, as in the case for Vomb WTP, is also to ensure adequate distance to the water table and thereby achieving sufficient unsaturated zones. For these reasons, an argument can be made that the upper grain size limit could be extended to 3.0 mm with a potential minor loss in treatment efficiency. However, these losses are estimated to be insignificant due to the long retention times before groundwater extraction, so the upper cut-off limit of ≤1% for grain sizes over 3 mm will be chosen. If higher treatment efficiencies are desirable, a 2.0 mm upper cut-off limit could be set to ≤1%. Based on site-specific conditions and considerations, the upper cut-off limit could vary.

The d₁₀ value will be the controlling factor for the finer fractions of the media, and therefore no value will be set for fractions <0.2 mm. The fine silt and clay fractions (<0.0071 mm) are very important to limit to avoid deep clogging and reduced permeability of the aquifer. As reported by the SWWA, MAR plants have experienced lowered permeability in their recharge fields due to particles originating from unwashed filter media. The recommendation for SSF sand containing silt particles was 0.03 and 0.06 wt. %. For practical reasons, this limit will be set to 0.1% with an optimal value of 0.05%. An infiltration basin at Vomb WTP needs around 15,000 tons of sand, which means that this limit still allows up to 15 tons of silt and clay particles.

The mineral composition and the limits for contaminants for filter media in rapid and slow sand filters in Sweden differs from international recommendations, and there are no reasons why these recommendations should be less strict for filter media used in MAR plants. It is clear that filter media should mainly consist of quartz and free from mica minerals and contaminants. The measurements for acid solubility differ in Sweden compared to international standards, and it is not clear how they relate to one another. However, both methods aim to quantify magnesium and calcium concentrations. For the purpose of this study, the Swedish standard will be used. The recommendations from the SWWA (SWWA 2011) and the Swedish EPA for the remaining contaminants for the filter media will be used (Table 4). To ensure that the filter media is washed with fresh water and that chloride concentrations does not increase in the recharge field, a calculated value has been presented. Higher levels of chloride, between 50 and 100 mg Cl⁻/L, can be found in several coastal areas in Sweden (HaV 2018), and has been used as the limit in this study. Based on a field capacity (the water volume that stays attached to the grains) of 25% for the filter media (Fetter 2014) and washing water with an chloride concentration of 50 mg/L, the resulting maximum chloride concentration on the media is 6.9 mg Cl⁻/kg.

This study investigated suitable filter media properties for infiltration basins used for MAR, and applied this reasoning in a case study at Vomb WTF, a MAR plant in south Sweden. The filter media recommendations focused on protecting the aquifer from contamination and deep clogging and ensuring that clogging particles are retained in the upper layers of the filter media. Also, choosing filter media with lower permeability than the natural deposits improves the unsaturated conditions for infiltration basins in the early operational stages before physical and biological clogging occurs. This was achieved by choosing a filter media that was finer than the natural glacifluvial material and coarser than the media found in infiltration basins with low permeabilities. The effective grain size (d₁₀) range was coarser than the international recommendations for SSF sand, which concurs with the recommendations from WHO for filter media used in MAR. Based on the results, the finer grain material (<0.2 mm) and the effective size (d₁₀) seem to be the dominant factors for predicting permeabilities of filter media. The coarser grain sizes do not contribute to higher treatment efficiencies, therefore limiting these fractions are preferred for filter media for SSFs. However, the conditions for MAR-plants are different and it is possible to extend the upper grain size limit depending on site-specific considerations.

The supply of crushed rock is vast, which makes it a sustainable alternative to natural sand material. The material has been used as filter media for wastewater treatment (JTI 2016) and media for coarse filtration of corrosive liquids (Cheremisinoff 2001). There have also been bench-scale experiments that investigated the viability of crushed rock for drinking water supply (Sternö 2005; da Silva et al. 2021). However, there is very little published on it being used full scale. Because the grain surface structure differs from that of natural sand, there are several uncertainties related to its properties; the main being, the material's resistance to mechanical weathering and the changes in hydraulic properties once compacted. Other considerations include the development of the biofilm on the grains, the capacity to retain organic matter and the effect of mixing native natural sand on its hydraulic properties. However, there are no reasons why this material could not be utilized as filter media for water production, provided that the material fulfils the requirements.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.