Chapter 14: Treatment processes, filtration and adsorption

14.3 Slow sand filtration

The World Health Organization published Slow Sand Filtration in 1974. It is still in demand and much of its content remains valid so they continue to make it available electronically. WHO (1974) states that under suitable circumstances, slow sand filtration may be not only the cheapest and simplest but also the most efficient method of water treatment. The process requires a lot of land.
Slow sand filtration began in the early 1800s and was developed at regular intervals throughout that century. Its history, purification mechanisms, design, operation and maintenance requirements and other details are described extensively in a report Slow Sand Filtration, published in 1991 (in a period when renewed interest was being taken in the process) by the American Society of Civil Engineers, and in WHO (1974). Also refer to WHO (2004a).
Slow sand filters are not widely used in New Zealand. Examples have included Little River in Banks Peninsula District and Linton Army Camp near Palmerston North. The process produces drinking-water in Apia (Samoa). The Paris water supply (from the River Seine) is treated by slow sand filtration (and other processes) at the Ivry sur Seine plant. It is used there as an organic barrier, particularly to phenols and similar contaminants.
Slow sand filtration (sometimes called biological filtration) operates by two methods:

a surface filter, which processes the water biologically

a deep sand bed, which purifies the water by adsorption and some straining.
Slow sand filters comprise a relatively deep sand bed supported on a layer of graded gravel over underdrains (the sand typically 0.9–1.2 m deep on start-up and not to reduce below 0.6 m before resanding). The sand is finer than the 0.6–2 mm range that is typical in the more common rapid granular media filters, having, typically, a mean particle size in the range of 0.15–0.4 mm. This is similar in size to most beach sand. The water takes several hours to pass through the sand, providing ample time for purification by adsorption of microscopic particles adhering to sand grains; 1 m³ of sand has a surface area of about 15,000 m²!
The filters should operate with a head of 1–1.5 m of unfiltered water above the sand. It is most undesirable that the water level in the filter box should drop below the surface of the filter medium during operation. To eliminate the possibility of this happening, a weir is incorporated in the outlet pipe system. The water sits above the sand for 3–12 hours.
The surface of the sand ripens; that is, a biologically active layer, primarily of algae and bacteria, develops on it, adding a biological process to the sand filtering. Ammonia and nitrite can be oxidised in this layer and the organisms living there strip nutrients from the water too. This surface layer is called schmutzdecke, a German term meaning dirt layer or filter skin. It takes a day or so to develop and, until it does, the filter will not present a proper barrier to microbial pathogens. This layer does not develop on rapid granular media filters because they are backwashed before any significant amount schmutzdecke has had time to develop.

The loading rate (the flow per square metre of filter bed surface area) is low, usually at a constant flow of 100–300 litres per second per square metre per hour, which equates to an equivalent velocity of 0.1–0.3 m/h. The rate may also be expressed as mm/s or m/d. A rate of 0.1 mm/s is equivalent to 0.36 m/h. For protozoal compliance, the DWSNZ state that the filtration rate should not exceed 0.35 m/h, and must be constant. Note that rapid granular media filters (after coagulation) can operate successfully at 30–40 times this rate.

Even at this slow rate, the headloss is typically around 0.1 m when the sand is clean, and increases to about 1.2 m when the sand needs cleaning. This initial headloss is due to the fine size of the sand and the depth of the bed.
There is no backwash system, so all solids captured build up on the surface, with a small amount of penetration into the sand. For protozoal compliance (earning 2.5 log credits), the final water turbidity should be below 0.5 NTU (section 5.10, DWSNZ), and some form of post-disinfection will almost certainly be required in order to achieve bacterial compliance (section 4.3, DWSNZ).
Section 3.6 of WHO (2009) discusses a modification of slow sand filters for household use, calling them biosand filters. These can be used intermittently.

14.3.1 Cleaning

When the bed resistance (headloss) has increased to such an extent that the regulating valve is fully open, it is time to clean the filter bed, since any further increase is bound to reduce the filter output. The top 20–30 mm of sand is scraped off and discarded. Sand removal at small plants can be manual but at larger plants it is more common to use a mechanised system to avoid the large amount of labour required. The water is then turned back on and the filter left to ripen so the schmutzdecke layer can build up enough to provide effective filtration again. If the scraping has been completed before the bed has dried out, ripening should take only 1–3 days. During ripening, the water is recycled, passed to another filter, or passed to waste during this time.
Ultimately, maybe after 20–30 scrapings, perhaps after several years, and before the sand reaches its minimum design depth, topping up with new sand, or full cleaning, or complete replacement, will be required. To accelerate the ripening process after re-sanding, some of the residual bottom sand or scrapings from the surface layer can be placed over the new sand.
A new filter must be run continuously for at least several weeks in tropical climates and longer where temperatures are low (WHO 1974). The time also depends on the nature of the raw water: the cleaner it is, the longer the ripening process will take. As ripening proceeds, there will be a slight increase in the headloss across the bed as the organisms build up, and the formation of a schmutzdecke will gradually become visible. These are signs that ripening is proceeding satisfactorily, but only after comparative chemical and bacteriological analyses of raw water and effluent have demonstrated that the filter is in full working condition may the effluent be directed to the public supply.
A detailed study in the Netherlands found removal of MS2 bacteriophage and E. coli WR1 was strongly dependent on the water temperature and schmutzdecke age (Schijven et al 2013). This model is intended to be incorporated into the Dutch ‘legislatively required quantitative microbial risk assessment’ (QRMA) and reduce the monitoring required for determining removal efficiency of SSF.
Because of the need to ripen cleaned filters, treatment plants need more than one filter, preferably at least four. The weakest point of a sand filter is the edge where raw water may leak past; to minimise this filters should be at least 100 m², preferably double this.
Full records of cleaning operations should be retained.

14.3.2 Monitoring

The primary protozoal compliance monitoring criterion for slow sand filtration is the turbidity of the filtered water. This should remain under 0.5 NTU (section 5.10, DWSNZ). Should it exceed this, the operator should check whether:

the sand bed has been disturbed; in particular, whether the sand has bound together then cracked, or pulled away from the filter walls

the schmutzdecke layer still appears normal. If it has been poisoned or damaged in some other way, it may have died off and be losing material into the sand

the raw water quality has changed abruptly; in particular, whether significant turbidity from clay (rather than coarser material) has been present. An increase in turbidity could be due to heavy rain in the catchment

any operating parameters such as temperature, pH, flow through each bed, headloss, downstream chlorine demand etc has changed significantly. The operator should assess whether this was good or bad news and how it relates to the higher filtrate turbidity. Colder water, for example, should show higher headloss but not necessarily higher turbidity.
If the turbidity of the filtrate from any filter exceeds 0.5 NTU for more than 5 percent of the time, the cause needs to be determined and resolved before the water from that filter can be used. If it is not possible to shut off the supply from the filter during the investigation, boil water notices must be issued. If the filtered water turbidity is greater than the raw water turbidity, there is a real chance that the filter is discharging; if it is not due to a sudden change in raw water quality, shut down the filter, scrape off the top sand and ripen again.
It is recommended that water suppliers establish a control limit for each MAV or operational requirement. Control limits are discussed in Chapter 17: Monitoring. The preventive actions that are to be considered when a control limit is reached are to be documented in the WSP. The purpose of control limits and the preventive actions is to avoid transgressions of the MAV or operational requirement. For example, a control limit for turbidity of the water leaving each filter set at 0.30–0.40 NTU may be prudent.
WHO (1974) considers the following are the basic records that should be kept for each filter:

the date of commencement of each cleaning

the date and hour of return to full service after ripening

daily raw and filtered water levels, and headloss

the filtration rate

raw water and filter effluent quality, ie, colour, turbidity, temperature, E. coli

details of incidents, unusual weather etc.

14.3.3 Aeration

The filtered water may become anoxic as it passes through the sand so may need to be aerated to restore the dissolved oxygen level and remove dissolved carbon dioxide. This is achieved by having a simple weir on the outlet, dropping the filtered water about 1 m vertically. The water passes first through the sand, then through the underdrains and is taken back up to the top of the weir, which is on the same level as the sand surface. This arrangement ensures the water level in the sand does not drop below the surface. If this happens, the surface will dry out, killing the biota, and air will blind off some of the flow paths through the bed. The weir also ensures that the filter operates independently of any level fluctuations in the water above the sand.

14.3.4 Some operating issues with slow sand filtration

Raw water quality: The water going on to the filter should not be too turbid or the filter will quickly overload. Experience will show when this level approaches. Although raw waters up to 100 NTU have been treated successfully for brief periods, 50 NTU is a more realistic upper limit, and optimum purification occurs around 10 NTU. Very turbid raw waters should receive some form of pretreatment.

Controlling algal growth: The amount of algal growth on the surface must be limited, but not prevented. Algae require sunlight so the simplest method of controlling their growth is to limit sunlight where necessary. The Little River units (two in parallel) do not have this feature and have not had this problem.

Cold weather: This can reduce the filter’s effectiveness in two ways. As well as limiting the biological activity in the schmutzdecke layer, low temperatures also increase the headloss by increasing the viscosity of the water. Slow sand filters are recognised as being more suited to, and more efficient in, warm climates. The DWSNZ call for a minimum temperature of 6^oC for protozoal (oo)cyst control. If this temperature is likely to be reached, water suppliers may consider covering the filtration area to limit the cooling effect of wind and frost.

Disinfectants: Chlorine, or any other disinfectant or algicide, should not be added before the water is filtered, because it will kill the organisms in the schmutzdecke layer. Efficiently operated slow sand filters have been demonstrated to be effective in removing protozoal (oo)cysts, as well as bacterial and viral pathogens. However, dosing chlorine after filtration is strongly recommended.