Arbejdsrapport fra Miljøstyrelsen, 7/2005 – Rensning af arsen i en traditionel vandbehandling vandværker

Arsenic Removal by Traditional Water Treatment

Summary and conclusions

Background

Recently, the maximum permissible concentration limit for arsenic in drinking water produced by Danish waterworks was lowered to 5 µg/l. As a consequence, several Danish waterworks now exceed the maximum concentration limit. This requires enhanced knowledge of the possibilities of arsenic removal by traditional Danish drinking water treatment plants.

Traditional Danish drinking water treatment consists, firstly, of an aeration/stripping step. Subsequently the aerated water is led through sand filters, and finally it is distributed to the consumers.

The purpose of the present report is to investigate the possibilities for optimizing the removal of arsenic by fairly simple upgrading initiatives, while at the same time maintaining the traditional Danish drinking water treatment in its existing form. As a starting point, the optimisation solutions imply changes in the manner of which the waterworks are operated, e.g. variation of filter velocities, frequency of back washing or addition of iron as part of the water treatment. At the same time, the investigation aims to provide enhanced knowledge of the chemical and physical processes for controlling arsenic removal in sand filters.

Localities

About 8 % of all waterworks in the County of Storstrøm distribute drinking water with concentrations of arsenic above the 5 µg/l concentration limit. The present investigation involves two waterworks in the County of Storstrøm, i.e. Elmevej Waterworks and Holmegårdsvej Waterworks. These waterworks are both operated by Fensmark Waterworks. Both waterworks carry out a traditional Danish drinking water treatment. During the present investigations, arsenic concentrations of 5 and 15 µg/l in the treated water from Holmegårdsvej and Elmevej Waterworks, respectively, have been measured.

The wells of the two waterworks extract water from an artesian Danien limestone aquifer. The extracted raw water is reduced and arsenic concentrations, predominantly As(III), range between 20 and 32 µg/l. Depth specific sampling in two of the waterworks' wells showed insignificant arsenic concentration variations with depth across the water yield zones of the wells.

Arsenic chemistry and sorption of arsenic to iron oxides

Arsenic is a redox sensitive substance and is present in groundwater and in water treatment processes in a reduced state as As(III), or oxidized state as As(V). In natural groundwater, arsenic predominantly exists in inorganic complexes.

Arsenic is removed during water treatment by sorption to iron oxides. The iron oxides form as soluble iron in the raw water precipitates. The iron oxides appear as red-brownish precipitates on the sand surfaces. Several studies have shown a positive correlation between the removal of arsenic and the removal of iron. The higher the amount of precipitated iron oxides during water treatment, the greater the percentage of arsenic removed. Thus, the addition of soluble iron during water treatment can increase the removal efficiency.

The binding of arsenic to iron oxides may occur as an adsorptive process (surface adsorption) to preformed iron oxides or arsenic may co-precipitate – i.e. arsenic may be incorporated in the oxides during their formation. Earlier studies have demonstrated co-precipitation to better utilize the iron than adsorption to preformed iron oxides.

Logically, when arsenic reacts with iron oxides, the concentration of arsenic in the iron oxides rises. The literature describes an aspect of much importance for the present report, namely, that such a reaction between arsenic and iron oxides does not proceed until all arsenic in solution is depleted. Rather, the reaction proceeds until a certain equilibrium situation is reached. At this equilibrium the concentration if arsenic in the water is in equilibrium with the concentration of arsenic bound to the iron oxides. A further increase in the arsenic concentration in the solution is mirrored by a further increase in the concentration of arsenic in the iron oxides. A new equilibrium situation will emerge at which the dissolved arsenic concentration is elevated relative to the previous equilibrium situation. Likewise for the iron oxide bound arsenic concentration, this will be elevated in the new situation relative to the previous. The general superior perspective is that the higher the dissolved equilibrium arsenic concentration, the higher the iron oxide-bound equilibrium arsenic concentration.

Likewise, one may imagine the reverse reaction. If arsenic containing iron oxide is brought into contact with arsenic free water, arsenic will be released from the iron oxides until the equilibrium situation is reached. Thus, the reaction between arsenic and iron oxides is a reversible reaction characterized by the dissolved arsenic and the iron oxide-bound arsenic equilibrium concentrations being positively correlated.

Many naturally occurring substances can compete with arsenic for iron oxide surface sites and thereby potentially reduce arsenic removal efficiency. A review of studies on such competitive behaviour indicates that especially silicate and phosphate can significantly reduce the extent of arsenic sorption to iron oxides. Groundwater concentrations of calcium and magnesium, on the other hand, enhance sorption of arsenic to iron oxides. Thus, in water treatment processes both sorption inhibitors as sorption enhancing substances should be taken into account.

Many previous studies show that arsenic in the oxidized form as As(V) is sorbed significantly more efficiently to iron oxides than As(III). However, in reduced groundwaters, arsenic preexists in the reduced form as As(III). Consequently, for effective removal of arsenic during water treatment, As(III) must be oxidized to As(V) before or simultaneously with the iron precipitation.

Both abiotic and biotic arsenic oxidation processes are known. For water treatment processes, more studies in the literature show manganese oxides to be effective oxidants for As(III). As for iron oxides, manganese oxides are formed by precipitation of soluble manganese from the raw water, and manganese oxides do make up a portion of the filter sand surface precipitates.

New scientific results from USA suggest that simultaneous As(III) oxidation and removal, independently of the occurrence of manganese oxides, may be possible in water treatment processes. Lee et al. (2003) describes the method which includes the addition of iron in the form of Fe(VI). The investigations conducted by Lee et al. (2993) show that Fe(VI) is an effective oxidant for As(III). Products of the oxidation reaction are Fe(III) and As(V). Subsequently As(V) may co-precipitate with the Fe(III).

Strategy of present investigations

Firstly, a characterization of the chemical composition of the water through the treatment processes was conducted. Following this, sand filter pore water samples were collected within distances of 5 – 10 cm along vertical profiles.

Likewise, filter sand samples were collected. These were analysed by a five-step sequential extraction method to produce the vertical distribution of precipitated arsenic, iron and manganese in the sand filters. In other experiments, filter sand samples were brought into contact with groundwater free of arsenic in order to establish the mobility of arsenic bound to the filter sand precipitates.

Subsequently, the effect of residence time on removal efficiency was investigated by varying filter velocities. Also pore water samples were collected at times where filters were out of operation.

Finally, a full scale test with iron addition between serially connected filters was conducted.

Results

Waterworks in the County of Storstrøm show a positive correlation between the concentrations of iron in raw waters and the arsenic removal efficiencies. In agreement with this, removal efficiencies of arsenic at Holmegårdsvej and Elmevej Waterworks of 45 % and 75 % and raw water iron concentrations of 0.2 and 1.7 mg/l, respectively, were determined.

In the raw water the predominant part of the arsenic exists in the reduced oxidation state as As(III). In the reaction chamber a small part (2 – 5 %) of the raw water As(III) is transformed into As(V), while about 50 % of the iron is oxidised. The predominant part of the arsenic removal occurs in the filtration step, whereas just up to 5 % is removed prior to filtration. Arsenic removal occurs in the upper part of the filters simultaneously with the complete transformation of As(III) into As(V). Parallel to this, residual iron is oxidised.

In the lower part of the filters, a constant concentration level of As(V) occurs. For Holmegårdsvej and Elmevej Waterworks, the respective concentrations are app. 5 and 15 µg/l. The 15 µg/l arsenic concentration at Elmevej Waterworks is independent of filter velocity for filter velocities up to 4.5 meter/hour. Serial connection of filters does not enhance removal efficiency relative to the existing single filtration.

Iron, manganese and arsenic are vertically and homogeneously distributed in the waterworks filters. The precipitated amounts of these substances correspond to several annual productions of water. Filter sand arsenic concentrations of 0.2 and 1.3 g As/kg dry filter sand were determined for Holmegårdsvej and Elmevej Waterworks, respectively. Calculations show significant differences between the concentrations of arsenic in the iron oxide coatings at the two waterworks. Thus, at Holmegårdsvej Waterworks, the calculated arsenic concentration is 9,700 µg As per gram of iron oxides. At Elmevej Waterworks the iron oxide bound arsenic concentration was calculated at between 12,700 and 13,400 µg per gram of iron oxides. Comparison of arsenic concentrations in water and iron oxides at each of the two waterworks reveals a positive correlation between the iron oxide-bound arsenic concentration and the dissolved arsenic concentration in the lower part of the filters during operation.

The experiments conducted with filter sand samples show arsenic to be a mobile component of the filter sand coatings. 40 % of the total filter sand arsenic content is mobilised by wet extraction with a solution of phosphate. When samples of filter sand are immersed into arsenic-free groundwater, arsenic is released too.

Addition of 3 – 4 mg iron/l as Fe(II) between two serially connected filters at Elmevej Waterworks reduced the dissolved arsenic concentrations from 15 to below 5 µg/l in the bulk water above the filter sand of the last filter. This reduction in arsenic concentration occurs simultaneously with the precipitation of a significant part of the added iron. However, as the water infiltrates and runs through the filter sand, the dissolved arsenic concentrations increased to the initial 15 µg/l, after the precipitation of residual iron. This is due to the release of mobile arsenic from the filter sand.

Conclusions

In the filters of the two waterworks, the removal efficiency of arsenic is controlled by equilibrium between arsenic in the water and arsenic bound to the filter sand coatings. High filter sand concentrations of arsenic imply a lower boundary for the current removal efficiencies of arsenic. Hence, as the filter sand concentrations of arsenic are lowest at Holmegårdsvej and highest at Elmevej Waterworks, arsenic concentrations of the treated water are also lowest at Holmegårdsvej (5 µg/l) and highest at Elmevej Waterworks (15 µg/l).

The filter sand contains a pool of arsenic, which can be mobilised. Consequently, even if other abstraction wells with low arsenic concentrations were introduced, the water would be contaminated during water treatment by the release of mobile arsenic from the filter sand.

High filter velocities may cause lowering of removal efficiency if the residence times become insufficient for equilibrium conditions for As(V) sorption to the filter sand coatings to be reached. At Elmevej Waterworks the removal efficiency is controlled by equilibrium for filter velocities up to 4.5 m/hour.

In agreement with literature studies, the results indicate arsenic sorption to iron oxides to be significantly more efficient for As(V) than for As(III). Oxidation of arsenic presumably occurs on the surface of manganese oxides precipitated on the filter sand surfaces.

As the water runs through the filters, arsenic concentrations are probably reduced primarily due to co-precipitation of As(V) with iron, and secondarily due to adsorption to pre-existing iron oxides.

Based on the investigations at Fensmark Waterworks it is not likely that the arsenic removal efficiencies can be improved by implementing simple changes in the operation conditions.

Recommendations

Either the filter sand will have to be replaced by new material, or the current pool of mobile arsenic will have to be reduced enough to produce equilibrium concentrations of arsenic below 5 µg/l.

The arsenic removal efficiency may be improved by bypassing the reaction chamber. This would minimise the loss of soluble iron prior to oxidation of arsenic into As(V). In addition, addition of iron before filtration may be required. In this connection with it should be ensured that As(III) is transformed into As(V) prior to the oxidation and precipitation of the added iron, or that As(III) oxidation and iron precipitation occur simultaneously.

Concerning an effective oxidation of As(III), problems may be encountered as a consequence of a possible filter sand replacement, by which the current pool of manganese oxides will be removed. Therefore, the establishment of a water treatment that includes double filtration (i.e. serially coupled sand filters) and addition of iron in between the two filtration steps may be necessary. For the first sand filter, filter sand from the current filters (e.g. at Holmgårdsvej Waterworks) can probably be reutilised, since this sand has proven its ability to oxidise effectively As(III) into As(V). For the second sand filter, new filter sand or filter sand that does not desorb arsenic to equilibrium concentrations above 5 µg/l, should be used.

Lee et al. (2003) describe a method by which the relatively expensive establishment of double filtration may potentially be needless. The authors add Fe(VI) instead of (or as a supplement to) Fe(II). Firstly, the added Fe(VI) may oxidise effectively As(III) into As(V) and subsequently, as the iron precipitate, the produced As(V) may co-precipitate.