Monitoring of a pilot GFO filter for removal of low-concentration arsenic in water

Arsenic (As) is a metalloid element. It can be converted into species with different valencies or react to form insoluble compounds – e.g. with iron. Its presence in natural waters is caused by leaching from source rocks and sediments (Hering & Elimelech 1995), or uncontrolled human activity (e.g. in agriculture, arsenic is used as a fungicide and insecticide; in the wood industry, for preservation; in the mining industry, for smelting metal ores, etc.). Details of the environmental origins, occurrence, events and impacts on human health of arsenic can be found in Mandal & Suzuki (2002).

Generally, arsenic reaches the human body through the water (Mondal et al. 2013). Arsenic contamination in natural waters can range upward from a few μg/L and is a worldwide problem that affects at least 100 million people (Bhattacharya et al. 2007). The countries affected include Argentina, Bangladesh, Cambodia, Canada, Chile, China, Italy, the UK, and the USA.

Arsenic occurs in several oxidation states (valencies) in the environment (e.g. –III; 0; +III; V) although, depending on the environmental pH and redox potential (Eh), the predominant species are As(III) and As(V). Arsenic (V) is generally predominant in oxygen-rich surface waters while As(III) usually predominates in groundwater (Choong et al. 2007; Giles et al. 2011).

Effects on human health related to the ingestion of water containing high levels of arsenic – the most serious is ‘black foot disease’ – are well documented (e.g. Mandal & Suzuki 2002; Choong et al. 2007). Both the concentration and valencies of the arsenic in any water are critical for human health. In fact, as reported in many studies (e.g. Arienzo et al. 2002; Qu 2008; Zhang et al. 2010), As(III) is more toxic than As(V). The very high toxicity of trivalent arsenic arises from its strong affinity for thiol groups, which inactivates enzymes (Ahuja 2008).

Arsenic can be removed from water using various techniques based on different chemical, physico-chemical or biological principles; details can be found in, e.g., USEPA (2003); Mondal et al. (2006); Choong et al. (2007); Ahuja (2008) and Singh et al. (2015).

Some arsenic removal techniques (e.g. coagulation and flocculation, lime softening, membrane filtration, adsorption on activated carbon) are well known and widely applied around the world, and are known as consolidated techniques. Others, e.g., activated alumina, titanium dioxide, sulfur modified iron – known as emerging techniques – have not yet been applied in water treatment plants (WTPs), even though their effectiveness has been established in some cases.

Since about 2005, the processes involving iron-based sorbents (IBs) have become some of the most extensively applied for arsenic removal. Two IBs are commercially available and widely used: granular ferric hydroxide (GFH, available since 1997) and granular ferric oxide (GFO) such as Bayoxide^® E33 (developed in 2000). Streat et al. (2008a) report both the preparation and physical characteristics of GFH and GFO.

The main advantages of using GFO and GFH are: (i) high potential removal efficiency (up to 98%), (ii) few maintenance, operating or manpower requirements, (iii) absence of waste sludge, (iv) simplicity of spent material disposal (Streat et al. 2008b; Singh et al. 2015), and (v) low treatment cost (Mohan & Pittman 2007). The treatment cost is strongly related to the arsenic concentration in the raw water, because higher arsenic concentrations reduce the life of the material, and the WTP's configuration – e.g., whether the GFO/GFH filter is used for polishing or as a single-stage for arsenic removal. Sorlini et al. (2014) who surveyed Italian WTPs for arsenic removal, indicate that, if the influent arsenic concentration is between 16 and 100 μg/L, the operating cost of IBs is comparable to that of membrane filtration and ion exchange.

High removal rates (>90%) can be obtained with GFO. These are frequently achieved in batch adsorption experiments on solutions spiked to multiple mg-As/L concentrations (Mohan & Pittman 2007). Few if any studies were aimed at investigating GFO performance in removing low arsenic concentrations – i.e., 10 to 20 μg/L. In many areas of Pianura Padana, arsenic concentrations in groundwater are close to the maximum contaminant level (MCL) of 10 μg/L established under Italian Drinking Water Legislation (Legislative Decree 2 February 2001, n. 31), so a long term experiment aimed at investing GFO performance under those conditions could be useful for many WTP managers.

On this basis, a pilot GFO filter was installed at a WTP treating an average flow of 5 m³/h, for 700 inhabitants. The WTP, in the Province of Pavia, northern Italy, has a treatment configuration typical of many small plants that have been operating since about 1980s and were designed for the simultaneous removal of ammonia, iron, and manganese by biofiltration, without a specific arsenic removal stage. The pilot plant, fed continuously with water from the biofilters (mean arsenic concentration 9.6 μg/L), was monitored (with arsenic speciation) for 19 months.

In the full-scale plant, contaminants are removed by a pre-aeration stage (in-line injected air) and two biofilters (BFs) filled with sand (80%) and pyrolusite (20%). Pyrolusite consists essentially of manganese dioxide (MnO₂) and its use was introduced in BFs in 2014 for arsenic control (Mohan & Pittman 2007; USEPA 2011). Following the methodological approach reported in Sorlini et al. (2015a), a monitoring campaign was conducted and two problems arose: (i) arsenic removal in the BFs was below 10% because the pyrolusite worked mainly as an oxidizer, and (ii) the mass of pyrolusite in the BFs fell due to backwashing. In agreement with the WTP managers, it was decided to investigate the suitability of an additional arsenic removal (polishing) stage, to keep its concentration as low as possible – following the WHO guidelines (2011). Of the arsenic removal technologies available, GFO was chosen because the process is simple (low maintenance and manpower requirements) and reliable. It is also sludge-free so there are no waste disposal costs.

As can be observed from the Figure, the disinfection stage is not in operation in order to minimize the negative effects that can arise from the use of disinfectants such as chlorine dioxide (ClO₂) and, consequently, to keep the WTP's process scheme as simple as possible. In fact, if ClO₂ is used for final disinfection, specific removal stages for chlorine disinfection by-products removal (e.g. activated carbon filtration) could be provided (Sorlini et al. 2015b, 2015c).

Bayoxide^® E33 was chosen from the commercially available iron-based materials. Its main characteristics are listed in Table 1, while those of the water treated are reported in Table 2. Bayoxide^® E33 (E33, below) was chosen because: (i) it is less sensitive to high pH values (>7) than GFH (the water's pH is close to 7.6, the pH_zpc of the GFH); ii) it shows better As(V) removal than GFH (AWWA 2005), and, the concentration of competing ions like phosphate, vanadium, and silicate is low or close to the interference threshold reported in literature (>0.5 mg-/L; >50 μg-V/L; >13 mg-SiO₂/L at pH = 7.5) (AWWA 2005; Jeong et al. 2007; Nguyen et al. 2007; Zeng et al. 2008).

Table 2

Influent parameters to the GFO pilot filter monitored during the test

Parameter .	Units .	Average concentration .
pH	–	7.6 ± 0.3
Ammonia		0.28 ± 0.2
Antimony	μg/L	<0.5*
Arsenic (total)	μg/L	9.6 ± 1
Arsenic (III)	μg/L	0.65 ± 0.5
Arsenic (V)	μg/L	8.95 ± 1
Calcium	mg/L	16.3 ± 0.2
Chloride	mg/L	1.0 ± 0.1
Iron	μg/L	40 ± 30
Magnesium	μg/L	4 ± 1
Manganese	μg/L	11 ± 10
Orthophosphate (as ⁠)	mg/L	0.80 ± 0.3
Silica (as SiO2)	mg/L	24 ± 2
Sodium	mg/L	11.7 ± 0.3
Sulfate	mg/L	0.30 ± 0.1
Vanadium	μg/L	<0.5*

Parameter .	Units .	Average concentration .
pH	–	7.6 ± 0.3
Ammonia		0.28 ± 0.2
Antimony	μg/L	<0.5*
Arsenic (total)	μg/L	9.6 ± 1
Arsenic (III)	μg/L	0.65 ± 0.5
Arsenic (V)	μg/L	8.95 ± 1
Calcium	mg/L	16.3 ± 0.2
Chloride	mg/L	1.0 ± 0.1
Iron	μg/L	40 ± 30
Magnesium	μg/L	4 ± 1
Manganese	μg/L	11 ± 10
Orthophosphate (as ⁠)	mg/L	0.80 ± 0.3
Silica (as SiO2)	mg/L	24 ± 2
Sodium	mg/L	11.7 ± 0.3
Sulfate	mg/L	0.30 ± 0.1
Vanadium	μg/L	<0.5*

Note: antimony (Sb) and vanadium (V) – marked with ‘*’ – were always below the limit of detection. Variation from the mean is indicated by ‘ ± ’.

View Large

The GFO pilot filter consisted of a Plexiglas column (diameter: 3 cm; height: 80 cm) containing 42 g of E33 operating under continuous down-flow conditions. Water leaving the BFs was fed into the pilot plant at 1 L per hour using a metering pump. The empty bed contact time in the filter was 5 minutes and no backwash was implemented during testing.

Water quality monitoring was conducted by taking an average of one sample per week throughout the 19-month test from both the influent (IN) and effluent (OUT) streams – see Figure 1. A total of 152 samples were analyzed. Almost all analyses were carried out according to Italian standards (APAT-IRSA/CNR 2003), but Standard Method n. 3125 (APHA AWWA & WEF 2012) was used for total arsenic, as well as As(III) and As(V). This was done because, due to its neutral charge at neutral pH levels (6–9), As(III) is hard to remove from water by chemical speciation or adsorption (USEPA 2003; Giles et al. 2011). Arsenate (As(V)) can be removed readily due to its positive charge. Because of this consideration, the As(III) concentration in the samples was measured by determining the total arsenic (As_tot) remaining in water filtered through SERDOLIT^® resin, which is selective for As(V) (McNeill & Edwards 1995; Sharma et al. 2016).

Finally, in line with the procedure reported in the citations as well as in Henry et al. (1979) and Henry & Thorpe (1980), the water's As(V) was calculated as the difference between the concentrations of As_tot in the unfiltered sample and that of the filtered sample containing only As(III); (As(V) = As_tot – As(III)).

The main characteristics of the GFO pilot filter effluent are reported in Table 3. During the test, the pH of both influent and effluent remained essentially constant at 7.6. As this is between 6 and 9, it is within the recommended range for GFO treatment (USEPA 2003), so removal occurred in the optimal treatment range.

Table 3

Effluent parameters from the GFO pilot filter monitored during the test

Parameter .	Units .	Average concentration .
pH	–	7.6 ± 0.3
Ammonia		0.26 ± 0.2
Antimony	μg/L	<0.5*
Arsenic (total)	μg/L	See Figure 4
Arsenic (III)	μg/L	See Figure 4
Arsenic (V)	μg/L	See Figure 4
Calcium	mg/L	16.3 ± 0.2
Chloride	mg/L	1.0 ± 0.1
Iron	μg/L	45.1 ± 40
Magnesium	μg/L	4 ± 1
Manganese	μg/L	8 ± 8
Orthophosphate (as ⁠)	mg/L	0.64 ± 0.3
Silica (as SiO2)	mg/L	23 ± 2
Sodium	mg/L	11.7 ± 0.2
Sulfate	mg/L	0.28 ± 0.1
Vanadium	μg/L	<0.5*

Parameter .	Units .	Average concentration .
pH	–	7.6 ± 0.3
Ammonia		0.26 ± 0.2
Antimony	μg/L	<0.5*
Arsenic (total)	μg/L	See Figure 4
Arsenic (III)	μg/L	See Figure 4
Arsenic (V)	μg/L	See Figure 4
Calcium	mg/L	16.3 ± 0.2
Chloride	mg/L	1.0 ± 0.1
Iron	μg/L	45.1 ± 40
Magnesium	μg/L	4 ± 1
Manganese	μg/L	8 ± 8
Orthophosphate (as ⁠)	mg/L	0.64 ± 0.3
Silica (as SiO2)	mg/L	23 ± 2
Sodium	mg/L	11.7 ± 0.2
Sulfate	mg/L	0.28 ± 0.1
Vanadium	μg/L	<0.5*

Note: antimony (Sb) and vanadium (V) – marked with ‘*’ – were always below the limit of detection. Variation from the mean is indicated by ‘ ± ’.

View Large

Comparison of the concentrations of ammonia, magnesium, calcium, chloride, sodium, iron, manganese, sulfate, and antimony reported in Tables 2 and 3, shows that no significant variations were detected during the test.

Among the species that can potentially affect arsenic adsorption, only silica (average feed concentration = 24 mg-SiO₂/L) and orthophosphate (average = 0.80 mg-/L) exhibited concentrations close to the maxima permitted. The concentration of silica was substantially the same in the influent and effluent, confirming that it was not adsorbed. The concentration of vanadium was always below its detection limit and so no further consideration was given to it.

Although the phosphorus compromised the E33's removal efficiency, the media did remove a useful amount of the arsenic up to 40,000 Bv. After this, the E33 was essentially exhausted, technically (C/C₀ = 0.85), and media replacement was necessary. In order to investigate the complete exhaustion of the E33, the experiment was extended to 140,000 Bv. From Figure 3 it can be seen that E33 exhaustion was reached at about 80,000 Bv, when C/C₀ = 0.90.

The Bv value achieved in the tests suggests that long operating cycles and low operating costs can be achieved in plants operating with groundwater arsenic concentration close to the MCL and low concentrations of other compounds that can affect the adsorption, i.e. interferents (e.g. silica, phosphorus, vanadium, etc.).

As can be seen in Figure 4, the ratio of As(III)/As(V) in the influent remained broadly constant (0.30) during the test but dropped to 0.18 in the effluent. Comparison of the As(III) concentration trends in the influent and effluent waters shows that the average concentration in the latter is about half that in the former (1.0 vs 2.3 μg/L). This indicates that As(III) is first converted in As(V) and then adsorbed by the media mainly in this form, confirming the findings of Mondal et al. (2013). Moreover, it can be also seen that the oxidation capacity of the E33 does not change during the experiment, in fact, the amount of As(III) oxidized is substantially constant throughout the test.

Comparing the trends of As reported in Figure 4, it seems that the arsenic is mainly adsorbed by the media as arsenic (V); in fact, in the first 5,000 Bv all As(V) contained in the feeding water has been removed while As(III) is found in the effluent.

In order to investigate the prevalent species removed by the E33, four different media response periods (periods that can be assumed homogenous with respect to operating conditions) were analyzed: 0–5,000; 5,001–20,000; 20,001–82,000; 82,001–140,000 Bv. In these the removal efficiencies have been calculated for the two most important arsenic species and the results are reported in Table 4.

For both species – As(III) and As(V) – the highest removal efficiencies were obtained in the first 5,000 Bv, after which removal decreased progressively. In the last removal period (82,001–140,000 Bv), As_tot removal efficiency was very low (9%) confirming that the E33 was effectively exhausted (C/C₀ > 0.9) and needed to be replaced.

With respect to the high removal efficiencies found for As(III), it must be underlined that it was not possible to distinguish the contributions of the adsorption and oxidation processes, so, removal must be considered as the sum of those contributions and some adsorption of As(III) cannot be excluded. This confirms reports by Mohan & Pittman (2007).

When the experiment ended, the adsorption capacity of the GFO was estimated as 0.57 μg-As_tot/L.

Among IBs, GFO is available for arsenic removal in water treatment. In many parts of Pianura Padana arsenic occurs in the groundwater at concentrations of 10 to 20 μg/L, slightly above the MCL of 10 μg/L.

A GFO pilot filter was installed and monitored for 19 months as a polishing stage in a WTP. The mean influent arsenic concentration in the pilot plant was reasonably constant during the tests at about 9.6 μg-As_tot/L.

Other chemical species present in the groundwater that could potentially have affected arsenic adsorption were orthophosphate (0.80 mg/L) and silica (24 mg/L). The results show that only orthophosphate affected arsenic removal, especially in the first 20,000 Bv, when competitive adsorption occurred with arsenic. Nevertheless, the GFO was able to remove some arsenic up to 40,000 Bv, showing a total adsorption capacity of 0.57 μg-As_tot/mg_GFO. Beyond this point, the GFO can be considered exhausted technically (C/C₀ = 0.9) and GFO replacement must be considered. The results also show that As(III) was removed at a low rate and its average concentration in the effluent was negligible.

Economically, GFO filter installation could be attractive because the operating costs – related to the As concentration in the influent – are comparable to those of other physico-chemical processes.

The authors would like to thank ASMortara S.p.A. (Mortara, PV, Italy) who sponsored the tests. Maria Cristina Collivignarelli planned and supervised the research, Sabrina Sorlini was responsible for data analysis and report drafting, and Matteo Canato carried out data collection and analysis, and also drafted reports. Barbara Marianna Crotti assisted in data collection and sample analysis.