Efficacy of selected phosphorus sorbing materials (PSMs) to enhance the orthophosphate sorption capacity of filter socks

Agricultural phosphate (PO43−) is a major cause of water quality compliance failures. Filter socks (FS) are a sediment‐control method proven to be effective at removing sediment and sediment bound‐P. Within the water and wastewater treatment industries phosphorous sorbing materials (PSMs) are widely used to remove reactive PO43−. Combining PSMs with FS provides an opportunity to concomitantly remove sediment, sediment bound‐P and PO43− from agricultural runoff. A column experiment was undertaken to evaluate the efficacy of selected PSMs to remove PO43− across a range of PO43− concentrations (0.08, 0.26 and 1.3 mg/L) and contrasting FS fill media (PAS 100 Compost and Woodchip). The results indicate that Ca‐enriched ferrihydrite removed up to 99% of PO43−, and was the only PSM to reduce PO43− to below the target value of <0.05 mg/L. An initial cost‐benefit analysis indicates that PSM‐enhanced FS are a viable option to remove PO43− from agricultural runoff.


Introduction
In England, only 83% of eligible waterbodies are currently meeting their EU Water Framework Directive (WFD) targets (Environment Agency, 2015). Agricultural diffuse pollution contributes 70% of sediments, 60% of nitrates (NO 3 − ) and 25% of phosphates (PO 4 3− ) to receiving waters (National Audit Office, 2010). This has resulted in increased ecological degradation through eutrophication, flooding, sedimentation and turbidity within the watercourses (Tilman et al., 2002).
However, methods to mitigate orthophosphate losses associated with agricultural runoff in-field are limited and include lime/alum application, reducing PO 4 3− inputs, establishment of hedges and shelterbelts, buffer strips and land-use change (Deasy et al., 2008;Newell Price et al., 2011;Schoumans et al., 2014). Consequently, the majority of elevated PO 4 3− in waterbodies is remediated by water treatment works (Defra, 2012). Removal typically involves filtration, precipitation and adsorption using metal ions (Yeoman et al., 1988;De-Bashan and Bashan, 2004;Nancharaiah et al., 2016). Phosphorus sorbing materials (PSMs) have been successfully utilised outside of the water treatment industry (Penn et al., 2017) in the form of surface runoff confined bed filters and surface and subsurface tile drain P removal structures. PSM when combined with filter socks could also prove incredibly beneficial to the agricultural sector for in-field PO 4 3− control. Commonly used PSMs include synthesised alumina (Mekhemer et al., 2000;Karim et al., 2011), synthesised ferrihydrite (Michel et al., 2010;Wang et al., 2013), synthesised goethite (Torrent et al., 1990;Rakovan et al., 1999), natural or synthesised calcite (Suzuki et al., 1986;Li et al., 2017), and clays, such as montmorillonite and illite (Gimsing and Borggaard, 2002;Ooi et al., 2017). Additionally, PSMs are widely used in on-site wastewater treatment systems, such as infiltration beds, constructed wetlands, filterwells and tile drain P removal structures (Cucarella and Renman, 2009;Penn et al., 2017). These PSMs, though researched are not widely utilised outside

Efficacy of selected phosphorus sorbing materials (PSMs) to enhance the orthophosphate sorption capacity of filter socks
A. L. Cooke 1 , R. W. Simmons 2 , R. J. Rickson 2 & R. Sakrabani 2 of the water treatment industry, but could prove incredibly beneficial to the agricultural sector for in-field control of PO 4 3− . To be compliant with WFD targets, the removal of sediment, sediment bound-P and PO 4 3− is necessary from the UK agricultural sector. This is in line with Rules 1 and 6 of the Farming Rules for Water, introduced in the United Kingdom in 2018; namely 'Pollution Risk' and 'Preventing Erosion' (GOV.UK, 2018). Filter socks have been found to effectively remove significant amounts of sediment and associated sediment bound-P from runoff (Faucette et al., 2008(Faucette et al., , 2009Cooke et al., 2017). Filter socks are tubular geotextiles that are typically filled with compost or woodchip and are placed in-field at prescribed intervals, upslope of riparian buffers or in concentrated flow paths to intercept runoff. Filter socks work by: (a) acting as a temporary bund, allowing the runoff to initially pool upslope whereby sand and silt particles deposit; and (b) reducing the velocity of runoff as it passes through the filter sock, allowing the silt and clay particles to deposit within the filter sock fill media.
However, filter socks are currently not effective at removing PO 4 3− . Combining PSMs with filter socks provides an opportunity to concomitantly remove sediment, sediment bound-P and PO 4 3− from agricultural runoff. The objective of this initial study was to critically evaluate the potential efficiency of natural, and by-product PSMs to remove PO 4 3− from runoff, in order to meet a pre-defined water quality target of 0.05 mg/L. This target was chosen as it is representative of the targets of many UK waterbodies, including the Lugg case study catchment (Cooke et al., 2017). Designated waterbodies (e.g. Special Areas of Conservation) have PO 4 3− targets which are particularly stringent (<1 mg/L) to enable the habitats or species for which they are designated to be effectively protected. However, with a short time frame, and a lack of funding and appropriate knowledge of the catchments being operated in, the relevant authorities are struggling to meet these targets.
PSMs were selected based on their potential phosphate sorption capacity as indicated by the literature, their availability, and their potential cost with regards to future adoption.
The selected PSMs were mixed with typical filter sock fill media namely PAS 100 QP Compliant compost and woodchip. These were placed within a series of column experiments, whereby the PSM and fill media were exposed to a range of PO 4 3− solutions of known concentration. Volume of PO 4 3− solutions applied, mass of P applied and retention time were recorded to facilitate the determination of PSE. Column experiments were chosen as they allowed mixing of the PSMs with the fill media, and due to their comparability with real-life conditions as compared with batch experiments (Helfferich, 1995;Loganathan et al., 2014).

Materials
Compost and woodchip were used in this study as the main filter sock 'fill media' into which the selected PSMs were mixed. This was to evaluate whether the PSM could effectively sorb PO 4 3− from both the applied PO 4 3− solution as well as and critically, sorb PO 4 3− released from the compost and woodchip fill media during 'flush events' (Faucette et al., 2008). The study was not intended to replicate flow through a filter sock rather, the results would be used to identify potential PSMs that could then be applied within a filter sock context. Synthesized Fe and Al-based PSMs are widely utilised in the water industry with PO 4 3− sorption occurring through ion and ligand exchange (mononuclear and binuclear bidentate coordination bonds), and surface precipitation (Ippolito et al., 2011;Loganathan et al., 2014;Gérard, 2016). Sorption is most effective at lower pH values, pH < 5, (Hingston et al., 1968;Cornell and Schwertmann, 2003), and consequently PSMs that perform better at higher pH values, such as Ca and Mg, are sometimes used in conjunction to maximise PO 4 3− removal. However, synthesised materials are currently not suitable for widespread use in the agricultural sector for PO 4 3− control due to their comparatively high cost and inaccessibility as compared to waste and by-product PSMs. Consequently, the PSMs investigated in this study were natural ferrihydrite-rich ochre (F Ca -O) and goethite-rich ochre (G-O), kaolin, fullers earth and a commercially available proprietary product. They were selected based on their availability and applicability to the UK agricultural sector, as well as their well-documented phosphate sorption capacity.
Detailed physico-chemical properties of the compost and woodchip fill media and the PSMs were established through standard laboratory and mineralogical analysis ( Table 1). The compost and woodchip used within the treatments were not air dried to avoid them becoming hydrophobic, and were not wetted up, so that the target moisture content of <60% was maintained (Alexander Associates Inc., 2006). The compost was hand sieved prior to the experiment to meet the target particle size distribution, which maintains runoff flow-through rates of 15-38 L/min, whilst promoting sediment deposition (Alexander Associates, 2006). Similar specifications for woodchip are not available, however, hand sieving was also undertaken to ensure a particle size distribution comparable to that of compost.

Column trial methodology
The bulk mineralogy of the PSMs indicates that the proprietary product, kaolinite and Fullers Earth (are dominated by, clinoptilolite (65.8%), kaolinite (95.5%) and smectite (64.0%), respectively ( Table 1). The calcium-enriched ferrihydrite ochre (FCa-O) consists of 46.6% ferrihydrite and 50.4% calcite a stable polymorph of calcium carbonate. In contrast, the goethite ochre is dominated by goethite (84.7%) with 14.1% calcite (Table 1). Specific surface area is in the order kaolinite > G-O > Fullers Earth > F Ca -O with values of 3213, 945, 730 and 588 m 2 /kg, respectively. All PSMs were air-dried and mixed thoroughly with the compost or woodchip fill media at the proprietary product application rate (20 kg/m 3 ). The fill media-PSM mixtures were packed into columns (10 cm length x 6.8 cm diameter PPE pipe with the base covered by fabric gauze) to a bulk density of 700 kg/m 3 , replicating the desired bulk density of fill media in filter socks. Once packed, the columns were then clamped to a frame and leachate collected ( Figure S1 of Supporting Information). Each column had between 182 and 191 ml of PO 4 3− solution applied to achieve treatment saturation, based on the particle density and total porosity of the fill media-PSM mixtures (Table S1-2 of Supporting Information and Equations 1-4).
Fill media saturation (Supporting Information; Table S2) was achieved with deionised water (control), or the PO 4 3− solution as a function of the treatment. The PO 4 3− concentrations investigated in this study were selected based on values associated with runoff from the dominant erodible soils in the Lugg case study catchment (Cooke et al., 2017): • 'Low': 0.08 mg/L • 'Medium': 0.26 mg/L • 'High': 1.30 mg/L Anatase These PO 4 3− concentrations are all above the water quality target for the Lugg catchment of <0.05 mg/L, and were used to calculate the PO 4 3− load (as mg P) applied to the treatments (Supporting Information; Equation 4). This load was 0.015 mg P for the 'Low' PO 4 3− solution and ranged from 0.047 to 0.050 mg P and 0.237 to 0.251 mg P for the 'Medium' and 'High' PO 4 3− solutions (Supporting Information Table S3).
All treatments were replicated in quadruplicate. Two applications of PO 4 3− solution or deionised water (the control) were applied, 48 hours apart; Flush 1 and Flush 2, representing the mean time between rainfall events experienced in the Lugg catchment field trials (Cooke et al., 2017). Solutions were evenly applied to the top of the columns, using the surface gauze to minimise preferential flow paths. The time taken for leachate initiation and leachate volume at 30 seconds, 1, 2, 5, 10, 20, 40, 60 mins postapplication was monitored, and ambient air temperature was recorded, as the temperature can influence PSM performance (Bai et al., 2017). Leachate samples were filtered using < 0.45µm syringe filters prior to analysis. For Flush 1, between 57-63% and 34-63% of the volume of solution applied to the compost and woodchip PSM blended treatments was leached within 60 mins of solution application. This also indicates that between 37-43% and 37-66% of the volume of solution applied to the compost and woodchip PSM blended treatments were retained for 48hr prior to the application of Flush 2. Similar results were observed for Flush 2 (Supporting Information Table S4).

PO 4
3− removed (PR) by the PSM treatments was calculated using (Equation 5): where s is the mean (n = 4) leachate PO 4 3− load (mg) for the control (un-amended Compost and Woodchip) treatments where deionised water was applied; t is the mass of PO 4 3− applied to each treatment (Supporting Information To facilitate the adoption of PSM-enhanced FS technology, and to assess the potential uptake of the PSMs within an agricultural context, an initial treatment cost-benefit was determined following (Equation 7).
where PCM is the PO 4 3− load removed per gram of column material. PR is PO 4 3− removal (Equation 5); Q is the mass of column fill media (254g); and V is the mass of PSM (0 g or 7.25 g).
Dividing by 1 gives the amount of material required for 1 mg P removal. Literature material costs (MC) were averaged per gram of PSM material.

Results
The leachate PO 4 3− concentration (Table 2) varied significantly between and within compost and woodchip treatments. FCa-O was the most successful PSM, regardless of fill media, and consistently met the water quality target of <0.05 mg/L. However, for both compost and woodchip fill media, across all PO 4 3− applications and flushes, kaolinite, proprietary product and Fullers Earth PSMs were ineffective at reducing leachate PO 4 3− concentrations, compared to the non-PSM control (Table 2). This is reflected in the phosphate sorption efficiency (PSE) which was used to quantify treatment performance for both Flush 1 and Flush 2 (Tables 3  and 4).

Leachate concentrations
Flush 1 Significantly higher leachate PO 4 3− concentrations were found for Flush 2 than Flush 1, owing to the release of fill media PO 4 3− into the solution retained by the treatments during the 48hr period between the Flush 1 and Flush 2 events. Additionally, some PSM adsorption sites will have been occupied during Flush 1.  3− compared to the control and consequently are recorded as negative PSE. Values within each row followed by the same letter are not significantly different following One-Way ANOVA and post hoc Fisher LSD analyses.

PSM performance
The compost FCa-O treatment consistently recorded significantly higher PSE (Table 3) (Table 3). Additionally, no significant differences in PSE were observed between Flush 1 and Flush 2 (Table 4).
Though less consistent than the compost FCa-O treatment, the compost G-O treatment was repeatedly associated with significantly higher PSE, when compared to compost amended with proprietary product, kaolinite and Fullers Earth (Table 3). Furthermore, the compost G-O treatment demonstrates a significant reduction in PSE between Flush 1 and Flush 2 (Table 4), suggesting the partial saturation of PSM sorption sites. Few significant differences in PSE were observed between the compost proprietary product, kaolinite, Fullers Earth amended treatments and un-amended compost control, and were associated with negative efficiency values (Table 3), indicating higher leachate PO 4 3− . This is likely due to the partial saturation of PSM sorption sites, and PO 4 3− desorption from the compost fill media. Both the compost proprietary product and kaolinite amended treatments are considered ineffective, as they consistently demonstrate no significant difference in PSE as compared to the un-amended compost control. Significant differences between Flush 1 and Flush 2 for the un-amended compost control, and proprietary product and kaolinite amended compost treatments (Table 4) are due to negative Flush 2 PSE, indicating that greater PO 4 3− is released from the fill media than is sorbed by the PSM.

Leachate concentrations
For the Flush 1 Low PO 4 3− application, the woodchip G-O treatment was associated with a significantly lower leachate PO 4 3− concentration, compared to all other treatments. For the Flush 1 Medium PO 4 3− application, leachate PO 4 3− concentration is in the order woodchip FCa-O = G-O = Fullers Earth < proprietary product = kaolinite = non-PSM woodchip control (Table 2). In contrast, at High PO 4 3− application, the woodchip FCa-O treatment was associated with significantly lower leachate PO 4 3− concentrations, compared to all other treatments.
For Flush 2, across all solution applications, the woodchip FCa-O and G-O treatments are associated with the lowest leachate PO 4 3− concentrations; however, the water quality target was not achieved (Table 2). This is due to the significantly higher leachate PO 4 3− concentrations associated with the W fill media. Following the addition of deionised water, mean un-amended compost and woodchip treatment leachate PO 4 3− concentrations for Flush 1 and Flush 2 were 1.3 and 19.8 and 1.1 and 54.0 mg/L, respectively (for brevity, data not shown).

PSM performance
Except the Flush 1 Medium PO 4 3− application, PSE for woodchip FCa-O and G-O treatments is significantly higher than all other treatments (Table 3). Significant differences in PSE were observed for woodchip FCa-O and G-O treatments (Table 4) between Flush 1 and Flush 2. For the woodchip G-O treatment this is due to significantly lower PSE at Flush 2 than Flush 1. For the woodchip FCa-O treatment, Flush 2 PSE is significantly higher than that of Flush 1.
The PSE of woodchip treatments amended with proprietary product and Fuller Earth and the un-amended woodchip control is regularly significantly higher than the woodchip kaolinite treatments. No significant differences in PSE were observed between the Flush 1 and Flush 2 results for woodchip amended with proprietary product, Fullers Earth, kaolinite, or the un-amended woodchip control (Table 4).

Efficacy of the PSMs to remove PO 4 3−
Across both fill media, FCa-O was the most efficient PSM at PO 4 3− removal, with PSE regularly exceeding 90% and leachate PO 4 3− concentrations of < 0.05 mg/L achieved on three occasions. This level of PSE could help to fail waterbodies realise their water quality targets.
The water quality target was not achieved by the G-O treatments, which had significantly lower PSE than FCa-O at Flush 1 (Low, Medium and High) and Flush 2 (High) when mixed with compost fill media, and at Flush 1 (High) when mixed with woodchip fill media. However, G-O still provided 31.8-75.2% PSE and may be preferable in locations where FCa-O is expensive and/or unavailable.

PO 4 3− adsorption by G-O is likely to occur through similar mechanisms to FCa-O (Tejedor-Tejedor and
However, goethite is more commonly found than FCa-O in ochre deposits as it is more stable (Chesworth, 2008), and therefore may be more widely available.
The kaolinite, proprietary product and Fullers Earth PSMs were significantly less efficient at sorbing PO 4 3− than FCa-O and G-O, and had highly variable results, consistent with the literature (Gimsing and Borggaard, 2002;Gérard, 2016). The kaolinite treatments regularly exhibited negative PSE results. Similar results were observed by Krumina et al. (2016) who indicated negative values may originate from the saturation of the kaolinite with PO 4 3− , followed by removal of the labile fraction following additional PO 4 3− application. Furthermore, during the experiment, fine (<0.002 mm) kaolinite particles were regularly observed leaching out of the columns, unlike that of all other PSMs, resulting in there being less kaolinite for PO 4 3− adsorption. This is likely due to the smaller particle size of the K (99% <0.002mm (Table 1)) which did not allow retention within the fill media.
The Fullers Earth treatments were more consistent across flushes and PO 4 3− applications with only one negative PSE result observed. Consistency of performance is attributed to its relatively large SSA (Table 1) as well as the presence of smectites, namely montmorillonite. These have a 2:1 layer structure with a weak bond between layer sheets, and an unbalanced charge, causing them to be more chemically reactive (Odom, 1984). Consequently, the Fuller Earth provides more reactive surface sites than kaolinite for PO 4 3− adsorption (Uddin, 2017).
The proprietary product is commercially available and used for the control of PO 4 3− in the runoff. When compared to the non-PSM control, the proprietary product treatments often provide statistically similar or significantly less PO 4 3− sorption. When compared to other PSMs, FCa-O and G-O provide highly significant order of magnitude increases in PSE. Kaolinite, Fullers Earth and the proprietary product were all associated with variable results especially when compost was used as the fill media.

Fill media
Generally, the results indicate a more consistent PSE from the woodchip, compared to compost fill media PSM treatments, as there were less negative PSE results reported. This is likely due to the longer retention time of the PO 4 3− solution within the woodchip fill media which retained, on average, 41% more of the solution in Flush 1, and 44% more in Flush 2, compared to the compost fill media treatments (Supporting Information Table S4). This increased retention time increases contact time between the PO 4 3− solution and the PSMs, allowing increased PO 4 3− adsorption. However, this also potentially increases the desorption of PO 4 3− from the woodchip fill media. This is demonstrated by the significantly higher leachate PO 4 3− concentration at Flush 2 (54.0 mg/L) for the un-amended woodchip treatment, following the addition of deionised water, compared to Flush 1 (1.10 mg/L).

Applicability of the PSMs evaluated for the agricultural sector
This research has highlighted the efficiency of FCa-O and G-O PSMs for PO 4 3− removal when combined with both compost and woodchip fill media in filter socks. To ensure widespread uptake for diffuse pollution control, both fill media and PSMs need to be cost-effective and easy to access. Table 5 highlights the affordability and accessibility in the United Kingdom of the various materials used within this study. Furthermore, it frames their cost per mg of PO 4 3− removed. This can be related to specific water quality targets and the amount of PO 4 3− removal required to meet these targets.
Both compost and woodchip fill media are widely available. Compost production is increasing annually as household recycling rates increase (WRAP, 2017) with an annual production of 3.2 Mt of PAS 100 QP Compliant compost (WRAP, 2017) and average cost of £75/t. Woodchip is less accessible than compost, as it is generally imported. National production data are limited as woodchip is often produced as waste and reused on site. The mean cost per 1 t bulk bag in 2017 was on average £87, as established from market analysis of suppliers (n = 5).
The PSMs FCa-O, G-O and kaolinite are most widely available within the United Kingdom, with kaolinite produced in the southwest, United Kingdom, with current reserves in excess of 60 Mt (Cornwall Council, 2012). It is widely accessible and affordable to purchase, both direct from suppliers and online. However, the cost per mg of P removed  is more than that for FCa-O, G-O and Fullers Earth. Furthermore, kaolinite was often associated with low and negative PSE.
Natural FCa-O and G-O are widely available within the United Kingdom as 'ochre', found as a precipitate within mining landscapes, mine water treatment schemes and water treatment plants (Hancock, 2004;Sapsford et al., 2015;Shepherd et al., 2016). The cost of ochre per mg of PO 4 3− removed is as little as £0.01 (Table 5). This is particularly low as currently ochre is not commercially available to buy. The Coal Authority pays to dredge and landfill ochre from their mine water treatment facilities, as it currently does not have an identified 'end-of-waste' certificate. This product certification would need to be undertaken before ochre could be widely accessible and utilised. Synthetic ferrihydrite and goethite is available, but would have financial implications on the cost per mg of P.
The majority of Fuller Earth is imported from central Europe and/or the USA as UK production has ceased owing to the working out of existing planning consents (British Geological Survey, 2006). Nevertheless, Fuller Earth is a relatively low-cost PSM even when import costs are accounted for (£0.18 per mg of PO 4 3− removed). Fuller Earth may, therefore, be considered a viable low-cost PSM for PO 4 3− control where PO 4 3− removal <0.2 mg/L is required. Due to its comparatively low PSE, Fuller Earth may need to be applied at rates >20 kg/m 3 in order to provide consistent and longer term PO 4 3− removal. The proprietary product is the least cost-effective of the PSMs, costing £11 per mg of PO 4 3− removed, even before import costs are considered. Furthermore, as it is currently only widely accessible within the United States, it is not practicable for widespread uptake, especially by the UK agricultural sector where many farmers have low financial margins.
To ensure the commercial viability of the PSM-enhanced filter socks, the cost of utilising the PSMs evaluated needs to be explored further in the context of the site-specific application of PSM-enhanced filter socks.

Conclusions
(1) The most effective and consistent PSM performance was given by FCa-O and G-O with PSEs of up to 99% across both the compost and woodchip fill media, even with the inherently high PO 4 3− load released by the woodchip fill media. This is consistent with other studies (Wang et al., 2013;Moreira et al., 2017). FCa-O consistently removed >50% of the PO 4 3− and was the only PSM which reduced the leachate PO 4 3− concentration to below the water quality target (<0.05 mg/L).
(2) FCa-O and G-O were associated with costs of circa £0.01 per mg of PO 4 3− removed. Consequently, these are recommended for further investigation to optimise their PSE and to be incorporated within filter sock across a range of agricultural runoff scenarios. However, they are currently considered a waste material and end-use certification needs to be obtained.
(3) Fullers Earth was also a relatively economic (£0.18 per mg of PO 4 3− removed) and efficient PSM, with a PSE up to 53.4%. Given that its PO 4 3− removal efficiencies were significantly less than those of FCa-O and G-O, it is recommended that the use of Fullers Earth as a PSM within filter socks is limited to situations where lower removal efficiencies are sufficient to meet water quality targets, or as an interim product whilst FCa-O and G-O end-of-use certifications are obtained. Overall, Fullers Earth has proven to be an important PSM that could be widely utilised in the agricultural sector.
(4) Kaolinite though widely available, was not an efficient or consistent PSM with widely varying results (removal efficiencies ranging from −30.2 to 23.5%), and a higher cost (£0.23 per mg of PO 4 3− removed), compared to FCa-O, G-O and Fullers Earth. Similarly, the proprietary product was significantly more expensive (£11.00 per mg of PO 4 3− removed), and is not widely available in the United Kingdom. Consequently, it is not recommended that kaolinite or the proprietary product be used as PSMs in diffuse pollution control measures in the UK agricultural sector where removal efficiency, cost and reliability are fundamental requirements.
(5) Combining the PSMs FCa-O and/or G-O PSMs with filter socks provides an opportunity to concomitantly remove sediment, sediment bound-P and orthophosphate from agricultural runoff. This has direct implications for the management of water quality. More research is required to establish the most effective application rate of FCa-O, G-O and Fullers Earth PSMs, for specific PO 4 3− concentrations, flows and removal efficiency targets.
To submit a comment on this article please go to http:// mc.manuscriptcentral.com/wej. For further information please see the Author Guidelines at wileyonlinelibrary.com