Slow sand filtration: recent research and application perspectives

N. J. D. Graham* and M. R. Collins

ABSTRACT

Slow sand filtration (SSF) has been widely used in the field of water treatment for over 100 years and while important features of the process have been studied in detail, many aspects of the process remain poorly understood. In addition, modifications to the process and pre-treatment methods are required to enhance SSF performance to meet more demanding treated water quality objectives and reduced costs. This paper highlights some of these aspects and refers to recent research that has investigated the nature and role of the biomass in the process, the benefits of applying surface fabric layers, pre-ozonation and GAC sub-layers, and the development of a deterministic-type mathematical model to simulate the SSF process. In all cases, the research has provided useful insights but much futher work is required to consolidate the findings and bring benefits to the application of SSF in practice.

Keywords Slow sand filtration; biofiltration; fabric layers; pre-oxidation; GAC amendments; process modelling

INTRODUCTION

The process of slow sand filtration (SSF) has been widely used in the field of water treatment for over 100 years but while important features of the process have been studied in detail, many aspects of the process remain poorly understood. This is partly explained by the greater attention that has given to the more widely applied rapid filtration process, but also because of the inherent complexity of SSF. Much is known of the ecology and fundamental dynamics of the filter process in qualitative terms (eg. common biological species and their interactions), but a comprehensive quantitative description of the process remains to be established.

In the 1990's slow sand filtration was the subject of a series of international conferences (Graham, 1988; Collins and Graham, 1994; Graham and Collins, 1996), professional guidance manuals (AWWA, 1991; ASCE, 1991), and literature reviews (eg. Lambert and Graham, 1995), which provided an extensive body of reference material and details of ongoing research studies. Since then, there has continued to be a steady flow of publications in the scientific literature concerned with various aspects of SSF research and plant operation (e.g. Gimbel et al., 2006). Broadly, these have been in four areas of SSF application, namely: drinking water treatment where the SSF is the principal treatment step (eg. for small community water supplies after rudimentary pre-treatment); drinking water treatment where the SSF is a secondary process following extensive prior treatment (Rachwal et al., 1996); the tertiary treatment of wastewaters (eg. Muhammad and Morris Hooke, 2003), particularly for effluent re-use; horticultural applications, such as the disinfection of recirculating nutrient solutions (Garibaldi et al., 2003). Only developments in drinking water applications will be described in this paper.

The key areas of general interest in SSF include pre-treatment, process mechanisms, treatment performance, modelling, and process enhancement. Associated with these are specific limitations or gaps in our current knowledge, such as the following:

i) a comprehensive, quantitative description of the fundamental process mechanisms;

ii) the relationship between the influent water quality and the nature of the SSF schmutzdecke;

iii) the removal of natural and synthetic organic substances;

iv) predicting filter run time;

v) methods of increasing filtration rates and filter run time;

vi) enhanced cleaning technologies.

Aspects of SSF performance that continue to be the subject of study are the nature of the biomass (in the schmutzdecke and filter bed), the mechanisms of treatment, the impact and role of animals, the removal of pathogens and specific substances (e.g. pesticides, pharmaceutical and endocrine disrupting compounds), and the benefits of covering and media amendments. In view of the broad range of the above topics, particular aspects of previous, and potentially future, research will be referred to here. These are: the nature and development of the biomass in SSF; the application of fabric layers to support the schmutzdecke; the effect of ozone as a pre-treatment for SSF; the incorporation of GAC sub-layers for enhanced organics removal; the development of a deterministic process model.

BIOMASS

The high degree of water treatment achieved by SSFs is partly explained by the slow filtration rate (0.1 - 0.3 m/h) and fine effective size of the sand (0.1 - 0.3 mm), but is also attributed to biological processes in the layer of material that accumulates above the sand surface (schmutzdecke) and within the upper layers of the sand bed. The schmutzdecke is also believed to be primarily responsible for the progressive increase of head loss observed during filter operation. Despite its current and historical importance as a water treatment process, the fundamental biological composition and mechanisms affecting water purification and head loss development during SSF runs remain poorly defined (Haig et al., 2011). Microbial biomass in the schmutzdecke and filter sand bed has been quantified using a range of microbiological methods (e.g. Duncan, 1988; Yordanov et al., 1996), but the different approaches used to measure biomass concentrations and the inconsistent units and sampling intervals adopted have confounded inter-study comparisons of biomass development and behaviour during SSF. Most reports are of single measurements of the net biomass production at the end of a filter run, prior to cleaning, and these emphasise the significant variability apparent in schmutzdecke and sand biomass accumulation in operational slow sand filters. The collection of representative samples of schmutzdecke and sand material during filter operation is difficult in practice, and the lack of a simple routine method for measuring microbial biomass are probable reasons for the limited amount of field-scale investigation of the biological mechanisms of SSF. Detailed analyses of biomass growth in the schmutzdecke and within the sand bed during filter operation would improve understanding of the complex and fundamental interactions between the biological and physico-chemical processes operating in the filter and enable the development of mechanistic models for SSF operational management (e.g. prediction of head loss rate, run time, frequency of sand cleaning and renewal).

One recent study has considered the development of microbial biomass in the sand and schmutzdecke layer in pilot and full scale slow sand filters, operated both with and without light shading (Campos et al., 2002). Through the use of random sampling and specially developed analytical methods, it has been observed that the interstitial microbial biomass in an uncovered sand bed increases with time (Fig. 1.1) and decreases with sampling depth. In contrast, there was only a small accumulation of sand biomass with time in a parallel covered filter, but no relationship was apparent between biomass concentration and depth.

Biomass accumulation and thickness of the schmutzdecke layer from the uncovered filter bed were highly variable and showed no consistent patterns of spatial or temporal development (Fig. 1.2). The substantial spatial variability of the schmutzdecke was in contrast to the relatively uniform patterns of biomass growth observed in the sand. It was speculated that microbial biomass in the sand of uncovered filters is largely related to carbon inputs from photosynthetic activity in the schmutzdecke and involves mechanisms that spatially distribute carbon substrate from the schmutzdecke to the sand. Since TOC and DOC removals were similar in both covered and uncovered filters (~20%) it is believed that relatively small biomass populations are sufficient to remove residual labile carbon in the influent water. As expected, there was no discernible development of a schmutzdecke layer in the covered filter.

FABRIC LAYERS

Previous research has shown that SSFs protected by surface layers of synthetic fabrics are capable of operating at significantly longer run times compared to un-protected SSFs (up to 8 times as long; Graham and Mbwette, 1991). Although a typical SSF bed is up to 1 m in depth, the majority of the filtration processes are concentrated in the upper 2-3 cm of the sand bed. With the use of fabric-protected slow sand filtration, and through careful selection of fabric type and specification, the filtration processes can be concentrated in the fabric layers of 2-3 cm thickness. Hence, solids penetration through to the sand can be prevented. As a consequence of the porosity of fabrics being significantly greater than that of sand, for example 90% compared to 45-50%, the rate of fabric blocking is much lower than the rate of sand blocking with unprotected slow sand filters. Therefore the rate of head loss development during the filter run period is reduced which leads to an extension of the filter run time in between filter cleaning operations. As well as extended run times, fabric-protected slow sand filtration can also eliminate the need for sand removal, and off-line cleaning, through the prevention of solids penetration by the fabrics. Therefore, an extended filter run time and reduced maintenance effort with fabrics can result in a substantial reduction in filter operational costs. Previous research at Imperial College (Mbwette, 1989; Graham and Mbwette, 1991) has demonstrated that by using six layers (3 cm) of a moderate-to-high density polypropylene fabric placed on the surface of slow sand filters, the filter run times can be extended by up to a factor of eight, and regularly by a factor of four, compared to a conventional slow sand filter. After a filter run, the fabrics can be removed and in the case of synthetic fabrics, can be cleaned and reused. A limitation of the use of synthetic fabrics is their cost, which can considerably increase the capital costs of a slow sand filtration project, especially if the importation of the fabrics is necessary. In view of the widespread non-availability of synthetic non-woven fabrics in developing countries, and the above-mentioned high costs of purchase and importation, there is a need to evaluate the potential use of natural fabric materials that may be locally available and produced.

A preliminary evaluation of the use of filtration fabrics made from natural materials, as a method for protecting slow sand filters in developing countries using indigenous materials and locally available textiles technology, was carried out some time ago (Luxton and Graham, 1998). The properties of the materials are shown in Table 1.1, with the natural materials (jute, abaca, sisal and flax) compared to a synthetic non-woven fabric (FiberTex). Coir was also included owing to its high resistance to biodegradation. The results of pilot scale tests showed that a combination of four layers of jute below a single layer of coir (3 cm total thickness) could achieve a doubling of the SSF run time. In addition, solids penetration through the fabrics into the sand was prevented and the fabrics displayed no significant deterioration. These reults were very encouraging but the development of natural fabrics as a SSF enhancement still requires further investigation and field demonstration in order to confirm their potential.

OZONE PRE-TREATMENT

As principally a filtration process, SSF is able to remove only a minor proportion of the total influent organic matter, commonly expressed as colour and DOC, by sorption and microbial degradation. It has been reported that DOC removal may range between 5 and 40%, but typically less (9-15%) for humic-type DOC (Lambert and Graham, 1995). Similarly, the removal of trace natural and synthetic compounds is generally poor (e.g. pharmaceuticals; Kuhlmann et al., 2006). However, in some cases the extent of treatment can be substantial depending on the presence of an active schmutzdecke and the properties of the specific compound, such as polar aromatic sulphonates (Eichhorn et al., 2002) and microcystins (Grutzmacher et al. (2002).

When ozonation is employed prior to SSF, the beneficial impact on the overall removal of natural and synthetic organic substances can be considerable, mainly through the combined effects of direct oxidation and increased biodegradability. Ozone causes substantial structural changes to humic substances which include: a strong and rapid decrease in colour and UV-absorbance due to a loss of aromaticity and depolymerisation; a small reduction in TOC (eg. 10% at 1 mg03/mgC); a slight decrease in the large apparent molecular weight fractions, and a slight increase in the smaller fractions; a significant increase of the carboxylic functions; and the formation of ozonation byproducts (Graham, 1999). The ozonation by-products have been reported to be mainly aldehydes (formaldehyde, acetaldehyde, glyoxal, methylglyoxal) and carboxylic acids (formic, acetic, glyoxylic, pyruvic and ketomalonic acids) (Camel and Bermond, 1998); in addition, glyoxalic acid and hydrogen peroxide have been identified as fulvic acid by-products. Ozone can also disrupt algal cells leading to an increase in assimilable organic carbon (AOC) (Muller et al., 2003).

Whilst it is generally believed that the SSF can remove the additional biodegradable organic material and by-products generated by the ozone, only a small number of studies have confirmed this. A summary of the removal of DOC by pre-O3 and SSF has been presented previously, and is given in Table 1.2; it can be seen that the removal values are significantly greater compared to SSF without pre-O3 (i.e. 16-18%). Ozone by-products have been identified in a few studies and comprise low molecular weight aldehydes whose formation increased with ozone dose, but measured concentrations (total <100 µg/l) were reduced to below detection limits by subsequent SSF (Graham, 1999).

The precise effect of ozone on the ecology of SSF is neither well understood nor well researched. A few studies of this aspect have been carried out (e.g. Yordanov et al., 1996) in which various techniques have been used to characterise the schmutzdecke biomass. These have provided some evidence that pre-O3 has an effect on the surface microbiology of SSFs. The indication of an increased bacterial population on the filter surface seems to be consistent with the higher levels of BDOC in the influent water after ozonation. Further evidence, indirectly, of the impact of pre-ozonation on the fundamental behaviour of the schmutzdecke is the effect the ozone has on filter headloss development and thus, filter run time. A summary of some experiences with pre-O3 on SSF run time is shown in Table 1.3. It is clear that the influent water quality and seasonal effects are important and these may relate to the level of nutrients, temperature and algal activity.

The ozone pre-treated SSFs in Portsmouth (NH, USA) were observed to remove 40-70% of the UV absorbance (254 nm) and the THM formation potential (THMFP) compared to the 10-15% removals for the conventional slow sand filters (Eighmy et al., 1993). As can be seen in Figures 1.3 and 1.4, a significant fraction of these elevated UV and THMFP removals (up to half of the overall removals) were observed in the water column above the sand media surface along with significant increases of water column bacterial populations compared to the non-ozonated SSFs.

Ozone pre-treatment were also observed to cause dramatic increases in filter headloss development causing filter run times to be less than ½ to ¼ of the control SSFs. These elevated headloss levels occurred in the schmutzdecke due to accelerated accumulations of biomass. In separate pilot filter runs, the rate of headloss development in pre-ozonated SSFs was significantly reduced by using larger sized sand media (up to 0.70 mm effective size) without a measurable reduction in treatment performance.

The primary ozonation by-product (OBP) observed in the pilot filter runs on two different water sources were formaldehyde. Production of the formaldehyde appeared to be a function of the ozone: DOC ratio. Formaldehyde was readily removed by the biological slow sand filtration process, as were acetaldehyde and glyoxal. In virtually all cases, SSF effluent OBPs measured in this study were below detection limits.

GAC AMENDMENTS

GAC placed as a sub-layer in the sand media of a slow sand filter can significantly enhance the removal of organic precursors as quantified by dissolved organic carbon (DOC). Research on GAC amended SSFs began at the University of New Hampshire back in 1987. A more mechanistic understanding of DOC removal by the enhanced SSF began in the mid 1990s (Page, 1997). In a comparative pilot study between a control SSF and two GAC amended SSFs (7.5 cm and 15 cm GAC sub-layers), DOC removals reached pseudo steady-state removals of 12%, 28% and 46%, respectively, after 200-300 days of continuous operation as shown in Figures 1.5 and 1.6. Removals of the biodegradable fraction of DOC, i.e. BDOC, were roughly the same for all three SSFs suggesting, as expected, that the elevated removals of DOC by the GAC amended SSFs were due to adsorption. Thusly, the pseudo steady-state removals by adsorption were 16% and 34% for the 7.5 cm and 15cm GAC amended SSFs, respectively.