16 Multi-stage filtration technology

Gerardo Galvis C; Jorge Latorre M;

Alberto Galvis C.
16 Multi-stage filtration technology

16.1 Introduction

The technology of multi-stage filtration (MSF) presented in this chapter is a combination of

coarse gravel filtration (CGF) and slow sand filtration (SSF). This combination allows the
treatment of water with considerable levels of contamination, well above the levels that
can be treated by SSF alone. MSF retains the advantages of SSF in that it is a robust and
reliable treatment method that can be maintained by operators with low levels of formal
education. It is much better suited than chemical water treatment to the conditions in rural
communities and small and medium-sized municipalities in the South and in more remote
areas in the North. Other treatment processes such as simple sedimentation, sand traps
and screens can precede MSF technology. Wherever possible, terminal disinfection needs
to be included as a safety barrier after the MSF. This chapter provides a summary
description of the components of MSF systems. It gives an overview of indicative cost
implications and ends with a selection guide.

16.2 Slow sand filtration technology

There are some typical operational differences between SSF and rapid filtration (RF) units.
Filtration rates are around 50-150 times lower for SSF. Flow retention periods are about
30-90 times longer for SSF. Filter run lengths are about 30-90 times longer for SSF, and the
surfaces of the SSF units are usually scraped at the end of the filter runs, whereas RF units
are cleaned by backwashing. These differences originate from the most distinctive feature
of SSF, its biological life. The water treatment in SSF is the result of a combination of
physio-chemical and biological mechanisms that interact in a complex way.

Inorganic and organic matter enter the SSF units in the raw water. Photosynthesis gives
rise to another fraction of organic matter. Soluble matter in the sand bed is utilised by
bacteria and other micro-organisms. Zooplankton grazing occurs and respiration of the
entire biomass is continuous.

The principal physical mechanisms contributing to particle removal are surface straining,
interception, transport, and attachment and detachment mechanisms. Physical particle
removal in SSF is not exactly the same as in RF since in RF the particles have previously
been destabilised by chemical coagulants and the biological activity is not so relevant.

Design characteristics of slow sand filtration units

In an SSF treatment plant at least two units should operate in parallel for continuous
supply. A unit basically consists of a structure that contains flow control and drainage
systems, a supernatant water layer and a filter bed (Fig. 16.1).

328
 Chapter 16

Fig. 16.1. Basic components of SSF units with inlet (A) and outlet (B) flow control
A: Inlet valve to regulate filtration rate F: Valve to contact tank or water storage

B: Valve to drain the supernatant layer of water G: Inlet weir

C: Valve for backfilling unit with filtered water H: Calibrated flow indicator

D: Valve to drain the filter bed I: Outlet weir

E: Valve to waste filtered water J: Outlet control valve

Flow control systems

Controlling the flow in SSF units is necessary to maintain the proper filtration rate
through the filter bed and the submergence of the media under all conditions of
operation. Abrupt filtration rate increases should be avoided. Two types of flow rate
control are used, outlet- and inlet-controlled flow.

329
In an outlet-controlled filter the supernatant water level is kept close to the maximum
desired level above the filter bed. To control the flow rate, the outlet valve is gradually
opened to compensate for the increase in the head loss over the filter media. This is the
usual control method in Europe and has been adopted in some of the units built in the
Americas. The storage capacity above the sand bed provides for some equalisation of
the influent water quality, sedimentation of heavier particles, and time for some
biological activity, as well as some buffer capacity.

In inlet-controlled filters any increase in head loss is compensated by an increase in the

height of the supernatant water. Some researchers have found similar performance in
terms of effluent water quality, head loss in the filtering bed, and filter run times for inlet-
and outlet-controlled SSF units run in parallel, with filtration velocities in the range of 0.13-
0.5 mh-1. In the inlet flow control option, the inlet box has several purposes. It provides
flow control, reduces excess energy to protect the filtering bed from scouring, facilitates
flow distribution to the SSF units filtering in parallel, and permits possible overflow.

The drainage system consists of a principal drain with lateral branches, usually constructed
in perforated pipes, brickwork or tiles and covered with a layer of graded gravel and a layer
of coarse sand. The drainage system of SSF has to achieve the following functions:
- Support the filter material and prevent it from being drained from the filter
- Ensure uniform abstraction of the water over the filter unit
- Allow for the backfilling of the filter and drive out possible air pockets

The main drain should discharge the filtered water freely at atmospheric pressure into
the outlet box. A flow indicator is required at both inlet and outlet side of the units to
facilitate operational procedures and to verify water balance, as an indication of possible
water losses in the main filtering boxes. The outlet weir is also necessary to maintain the
supernatant water layer above the maximum level of sand, protecting biological activity,
preventing pressure drops in the filter bed, and ensuring the functioning of the units
independently of the level fluctuations in the contact or storage tanks.

Supernatant water layer

The layer of supernatant water provides the static head necessary for the passage of
water through the sand bed. In a clean bed the initial head loss is usually below 0.1 m
and it gradually increases until the maximum level is reached. In units with outlet
control, variations of the supernatant depth for small systems have been reported in the
range of 0.6-1.2 m. At the Weesperkaspel plant in Amsterdam, where the SSF units deal
with highly pre-treated water, the average supernatant water height is 2 m. Filter
shading may contribute to improve filter runs if significant production of filter-blocking
algae is occurring on the filter skin or in the supernatant water layer, but few definitive
advantages in terms of filtrate quality have been reported.

330
 Chapter 16

Filter bed
The adequate selection of sand includes size grading, characterised by the effective size
diameter d10, and the uniformity coefficient, uc = d60/d10. Huisman and Wood (1974)
advise that d10 should be small enough to produce safe water and to prevent
penetration of clogging matter to such depth that it cannot be removed by surface
scraping. Experiences in the USA report a total coliform removal reduction from 99.4%
at d10 of 0.1 mm to 96% at d10 of 0.6 mm.

Deeper sand beds should result in improved removal of particles. However, due to the
development of the filter skin and the biological activity concentrated mainly in the
upper sand layers, particle removal is more effectively accomplished in this part of the
SSF units. Experimental evidence supports the practice of having a minimum sand depth
of 0.3-0.5 m in the SSF units to achieve more than one log reduction of indicator
bacteria. This is relevant for small systems working with low flow rates (0.1-0.2 mh-1), but
having to filter at higher rates during short periods due to their lower buffering capacity
when one of the units is out of operation.

The sand to be put into the SSF units should be clean and free of clay, earth and organic
material. The presence of dust or fine material produces high initial head losses and
seems to limit the essential development of an active and effective microbial population
in the filter bed. Placing dirty sand in the filter may interfere with the treatment process
and makes it necessary to remove the sand earlier for correct washing.

Operation and maintenance procedures

SSF units must operate continuously, since this contributes to better quality effluents
and a smaller filtration area is required for a given daily water production. Declining-rate
filtration can be applied, but intermittent operation should be avoided, since oxygen
depletion in the bed compromises biological activity. Research carried out in India
reports deterioration of effluent bacteriological quality when filters recommence
operation after 5 hours. In the USA, initial ripening periods in the range of 35-100 days
were identified before the effluents of the SSF units became stabilised for parameters
such as viruses, indicator bacteria and turbidity.

After several weeks or months of running, the SSF unit will gradually become clogged as
a result of the accumulation of inorganic and organic material, including the biomass
that is formed on top of the filter bed. The major increase in head loss occurs in this top
layer. By scraping off this layer, the hydraulic conductivity is restored to the level at the
beginning of the filter run. Classically, this is achieved by scraping the top 1-3 cm of the
filtering bed. After several scrapings, when the filter bed reaches its minimum depth
(0.3-0.5 m), resanding is required.

331
Manual cleaning has been the option for most small SSF units. In general, a high
frequency of scraping is associated with one or more of the following factors: high solids
concentration in the raw water; growth of algae in the supernatant water; small media
grains; low available head; and high water temperature. The filter runs (periods between
scrapings) of small SSF units in the USA range from one week to one year, with the
average about 1.5 months. There, manual scraping labour requirements are in the range
of 1.3-8 (average 4.2) person-hours per 100 m2 of area scraped. The labour requirement
increases significantly when the depth scraped is greater than about 2.5 cm. Filter runs
for small systems could vary in the range of 20-60 days and, for cycles shorter than 1.5
months, labour costs will escalate and operator satisfaction with the plant will diminish.

After scraping the sand surface, a secondary ripening period may be necessary for the
SSF units to recover their previous treatment capacity. Values in the range of 0-10 days
have been recorded for this secondary period. The most important factor affecting the
duration of a secondary ripening period appears not to be the removal of the filter skin,
but the dewatering of the sand bed. The cleaning is best done in warm periods and by
keeping the water table within 10 cm of the sand surface. This procedure ensures that
spirotrichs and peritrichs protozoa are retained in the sand (inoculation); they are
susceptible to desiccation and are unable to re-establish themselves at less than 3°C.

Scraped sand should be washed and stored. After several filter runs this activity leads to
a gradual reduction of the sand bed depth until a minimum value, usually in the range
of 0.3-0.5 m, is reached. Then re-sanding becomes necessary. For resanding, the
remaining sand in the filtering bed should be lifted to become the top portion, with the
stored and washed sand becoming the bottom. In this way the sand on top of the
filtering bed should provide seed organisms to shorten the ripening period. Resanding in
the USA requires around 50 person-hours per m2.

The wet-harrow cleaning technique uses a horizontal and sometimes vertical pressurised
water flow below the sand surface for washing across the filter skin being harrowed,
without dewatering the sand beds. The wash water is passed out via a surface overflow
weir. Shorter cleaning and ripening periods have been recorded with this technique in
the USA, where it is applied in SSF units treating clear raw waters with low turbidity.

Design guidelines
Great differences exist in the application of SSF technology around the world, as it
depends on drinking water quality standards, raw water quality, the type and level of
pre-treatment specified; and the local conditions. These conditions include institutional
development and support capacity to community-based organisations, availability of
materials and financial resources, user income, and willingness to contribute to capital
investment and running costs of the water supply infrastructure.

332
 Chapter 16

Design criteria presented by various authors and based on different experiences and
conditions are summarised in table 16.1. Those recommended by Visscher et al. (1987),
although oriented worldwide, were considered adequate for small systems in the USA,
where the experience with SSF was being re-established. The last column in table 16.1
corresponds to the design criteria proposed by Cinara - IRC based on planning, design,
monitoring and evaluation of over 100 SSF systems built in Colombia and other Latin
American countries.

Table 16.1 Comparison of design criteria for slow sand filtration from various authors

Design criteria Recommendation

Ten states Huisman Visscher, Cinara – IRC
standards and Wood et al. (1997)
USA (1987) (1974) (1987)
Design period (years) Not stated Not stated 10 -15 8 - 12
-1
Period of operation (hd ) 24 24 24 24
Filtration rate (mh-1) 0.08 - 0.24 0.1 - 0.4 0.1 - 0.2 0.1 - 0.3
Sand bed: Initial height (m) 0.8 1.2 0.9 0.8
Minimum height (m) Not stated 0.7 0.5 0.5
Effective size (mm) 0.30 - 0.45 0.15 - 0.35 0.15 - 0.30 0.15 - 0.3
Uniformity coefficient: Not stated <3 <5 <4
Acceptable < 2.5 <2 <3 <2
Support bed. Height including 0.4 - 0.6 Not stated 0.3 - 0.5 0.25
drainage (m)
Supernatant water. Maximum 0.9 1 - 1.5 1 0.75
height (m)
Freeboard (m) Not stated 0.2 - 0.3 0.1 0.1
2
Maximum surface area (m ) Not stated Not stated < 200 < 100

Water quality limitations of slow sand filtration

Slow sand filtration has been recognised as a simple, reliable and efficient treatment
technology and a most effective unit treatment process in improving water quality.
However, SSF does not necessarily remove all harmful substances to the extent required
by relevant drinking water quality standards. Table 16.2 presents typical treatment
efficiencies that SSF can achieve. The reported efficiencies have normally been achieved
in filter units operated at filtration rates in the range of 0.04 and 0.20 mh-1, temperature
above 5°C, and sandbed depths greater than 0.5 m.

333
The efficiencies in table 16.2 cannot always be achieved though, because much depends
on the nature, composition, and concentration of the components in the influent
waters; and the effect of design parameters, and ambient and operating conditions.
Even if high removal efficiencies can be obtained, SSF alone cannot always produce
water of a high standard. Raw water sources in many countries are already so
deteriorated that a combination of treatment processes is required to meet water
treatment objectives or national drinking water standards

Clearly, SSF, like all other treatment processes, is not a panacea for every water quality
problem. In general, two situations can be identified under which SSF presents
limitations:
- Levels of contamination in the raw water may exceed the treatment capacity, or
may result in short filter runs to comply with existing standards;
- Conditions that inhibit or reduce the efficiency of the treatment process.

Levels of contamination that exceed the treatment capacity

Suspended solids or turbidity. The most frequently mentioned limitation of SSF when it is
used as a single treatment step is its inability to treat water with a high level of
suspended solids or turbidity. These solids can create major increases in head loss and
adverse conditions for the biomass active in the filtering bed. Even short peaks of solids
may bury the large number of bacterial predators present in the sand bed and thus
reduce their capacity to remove harmful micro-organisms. This important potential
reduction in biological performance is, however, rarely cited in the technical literature,
despite the fact that it may have a very negative effect on the quality of the treated
water. The literature seems to focus instead on the difficulties of treating water sources
with small particles of a colloidal nature or the impact of high concentrations of
particulate matter on the duration of filter runs.

To prevent high effluent turbidity, frequent blockage of the filter bed (filter runs shorter
than one month) or an environment that is unfavourable for microbiological activity,
upper limits are usually specified for the influent turbidity. The limits vary, however,
between < 5 NTU and < 50 NTU. Furthermore, the majority of the references accept
higher values in the range of 50-120 NTU, provided these are of short duration, i.e. less
than few hours to 1-2 days, though they recognise these high limits as undesirable.
Nevertheless, turbidity alone is not sufficient to identify the limitations associated with
the duration of filter runs.

334
 Chapter 16

Table 16.2 Treatment efficiencies of slow sand filters (Galvis et al., 1992a; Fox et al., 1994; Lambert
and Graham, 1995)

Water quality Performance Comments

parameter or removal
capacity
Enteric 90-99.9% Reduced by low temperatures; increased hydraulic rates; coarse
bacteria and shallow sand beds; and decreased contaminant level
Enteric viruses 99-99.99% At 20°C: 5 logs at 0.2 mh-1 and 3 logs at 0.4 mh-1
At 6°C: 3 logs at 0.2 mh-1 and 1 log at 0.4 mh-1
Giardia cysts 99-99.99% High removal efficiencies, even directly after cleaning (removal
of the filter skin)
Cryptospori- > 99.9% Cryptosporidium oocytes. Pilot scale studies
dium
Cercaria 100% Virtually complete removal
Turbidity < 1 NTU The level of turbidity and the nature and distribution of particles
affect treatment capacity
Pesticides 0-100% Affected by the rate of biodegradation
DOC1 5-40% Mean around 16%. Removal appears to be site specific and
varies with raw water and O&M
UV- 5-35 % A slight, but not significant difference in treating upland and
absorbance lowland water sources. Mean 16-18%
(254 nm) Colour associated with organic material and humic acids.
True colour 25-40% Colour associated with organic material and humic acids. 30%
being the average
UV- 15-80% Colour (°Hazen). Mean 34%, but upland water sources 42% and
absorbance lowland water sources 26%
(400 nm)
TOC2; COD3 < 15-25% Total organic carbon; chemical oxygen demand
AOC 14-40% Assimilable organic carbon. Mean about 26%.
BDOC 46-75% Biodegradable dissolved organic carbon. Mean 60%
Iron, 30-90% Fe levels > 1 mgl-1 reduce the filter runs
manganese

1. DOC = dissolved organic carbon

2. TOC = total organic carbon

3. COD = chemical oxygen demand

335
Turbidity is accepted as an indirect indicator of the presence of particulate matter, because
of its ease of application. This parameter does not always properly reflect the load of solids
that the filter receives though, particularly if the particles are of an organic nature such as
algae. In addition, very few recommendations exist about the maximum load of suspended
solids (SS) an SSF can accept. Technical literature suggests a SS load below 5 mgl-1 but
without evidence related to the impact of this level of SS on SSF units.

Iron and manganese. Bacteria that contribute to the oxidation of iron and manganese are
present in the filter bed. Small quantities of iron deposits improve the removal capacity for
organic components. On the other hand, high concentrations of iron (above 1 mgl-1) may
contribute significantly to the clogging of the SSF unit.

Algae. Algae may grow in rivers, lakes, storage reservoirs, or even in the supernatant of the
SSF. The presence of algae in moderate quantities is usually beneficial for functioning of the
SSF units. Most algae are retained by the SSF, but under certain conditions occasional and
significant algal growth or algal blooms may develop. This massive growth can cause
a quick reduction of the permeability of the filtering bed, greatly reducing the filter run.
Algae may also play an important role in the production of high concentrations of soluble
and biodegradable organic material in the water, which in turn create smell and taste
problems, and contribute to microbial growth in the distribution system. Furthermore, as
a result of photosynthesis, algae may affect the buffer capacity of the water and increase
the pH to levels of 10 or 11. This can result in the precipitation of magnesium and calcium
hydroxides in the sand bed (calcification) and contribute to the obstruction of the filter
bed, increase the effective diameter of the sand, and reduce the efficiency of the process.

Controlling algae is difficult, but possible methods are based on reducing the nutrient
content of the raw water, or creating a storage system or a supernatant environment in
which algae can be controlled by the exclusion of light. This is done by covering the filters.
Before investing in covers for the SSF, it is prudent to check if standard operation and
maintenance procedures are not enough to manage moderate quantities of algae by
occasional harvesting. Different levels have been established for the concentration of algae
and other parameters (table 16.3).

Organic colour and organic carbon. A limitation of SSF is its low efficiency in the removal
of organic colour and organic carbon. In fact, some studies report no removal at all and
others indicate TOC and COD removal in the range of 15-19%. However, there are also
studies reporting COD removals in the range of 50-68%. The discrepancy lies in the diverse
composition of organic compounds, which are grouped together under surrogate
parameters such as COD or TOC. SSF units generally remove between 5 and 40% of
DOC, although the mean value is only 16%, and the difference between upland and
lowland water sources is not significant (data from wide literature review).

336
 Chapter 16

True colour. True colour removal, as colour units of Pt-Co in filtered or centrifuged
samples, includes only colloidal and soluble substances, especially natural organic matter.
The removal of true colour is normally reported to be in the range of 25-30%. Because of
the potential formation of disinfection by-products in the presence of organic material,
low colour levels are desirable. The colour level, however, should not determine the
application of final disinfection, as the risk of acute microbiological contamination is far
more significant.

Heavy microbiological contamination. In some communities the only source available

for water supply may be so heavily contaminated with harmful micro-organisms that
SSF alone will not be able to produce a good quality effluent. Whilst long-term efforts
are directed at protecting catchments, pre-treatment of the raw water may be necessary
before SSF can be properly applied.

Conditions that inhibit or reduce the efficiency of the treatment process

Various circumstances can interfere with the treatment process in the SSF units and
prevent the expected efficiencies from being obtained. Some of these are related to the
short filter runs considered in the previous item. Other important inhibiting conditions
are low temperatures, low nutrient content and low dissolved oxygen content.

Low temperature. Low temperature increases the viscosity of water and reduces the
biochemical activity in the sand bed, affecting the treatment efficiency. E. coli removal may
be reduced from 99 to 50% when the temperature falls from 20°C to 2°C. The strategy in
countries that face cold periods during the year has been to cover the filters or to build them
underground to prevent the freezing of the units and reduce the impact of low temperatures.
This, of course, has considerable economic implications. Reducing the filtration rate is another
way to reduce the impact of low temperature on the treatment process.

Nutrients. The micro-organisms active in the sand bed require nutrients such as carbon,
nitrogen, phosphorus and sulphur for their metabolism and growth. Humic and fulvic
acids are rich in carbon but low in the other elements. This may be part of the
explanation for the low removal of natural colour in SSF treating water sources that are
well protected. In experimental SSF units, adding nutrients has been shown to increase
the biological activity and improve removal efficiency for turbidity and microbiological
contamination.

Dissolved oxygen. When the flow velocities and the dissolved oxygen level in the water
source are low, particularly if this is combined with a high amount of biodegradable
material, the oxygen in the water can be depleted, resulting in anaerobic conditions in
the filter skin. This anaerobic condition in the filter must be avoided because it may
create serious water quality problems such as bad smell and taste, as well as

337
 re-suspension of heavy metals, with aesthetic implications and interference with the final
disinfection stage.

In summary, in spite of the potential of the SSF process illustrated in table 16.2, surface
waters presenting relatively moderate to high levels of contamination could not be
treated directly by conventional SSF units. Far too great a strain would be placed on the
terminal disinfection, limiting its role as a final safety barrier. This is critical in most
developing countries, where the reliability of disinfection is low.

Table 16.3 Some water quality guidelines that permit direct slow sand filtration treatment

Water quality parameters Quality limitations based on references of 1991

Spencer, et al. Cleasby Di Bernardo
(1)
Turbidity (NTU) 5 - 10 5 10
Algae (units/ml) 200(2) 5 mgl-1 (3) 250
True colour (PCU) 15 – 25 5
Dissolved oxygen (mgl-1) >6
Phosphate (PO4) (mgl-1) 30
Ammonia (mgl-1) 3
Total iron (mgl-1) 1 0.3 2.0
Manganese (mgl-1) 0.05 0.2
Faecal coliforms (CFU/100ml) 200

(1) The type of turbidity and the particle distribution may produce changes in the water quality of the effluent of

the SSF.

(2) Both the number and the type of species present in the water source are important. This reference suggests

covered filters.

(3) This limit corresponds with chlorophyll-a in the supernatant water as an indirect measure for the algae content.

16.3 Overcoming the water quality limitations of slow sand filtration

Multi-stage and integrated water treatment concepts take advantage of the great
potential of SSF technology. They have made it possible to overcome many of the water
quality limitations previously identified and to meet drinking water quality requirements.
In practice they are not new concepts as can be seen from the gradual evolution of
water treatment in two important European cities.

London
By the beginning of the twentieth century, SSF was already accepted as a vital barrier in
the provision of safe drinking water in London. A few years later, long-term storage
reservoirs and terminal disinfection with chlorine were incorporated as additional

338
 Chapter 16

treatment steps. Each of these treatment stages was fundamental in contributing to

improve drinking water quality. Nevertheless algal growth in the reservoirs and the
increased load of suspended solids gradually created premature clogging problems in
the SSF units. This problem was overcome in 1923 when the Metropolitan Water Board
introduced its first “rapid” sand filter (without coagulants). This double filtration was used
without major modifications until the 1980s. The gradual microbial improvement of
each step in this four-stage treatment is illustrated in figure 16.2. In the 1990s, to
comply with the requirements of the European Community, the treatment plants were
improved by including ozone treatment and a layer of activated carbon in the filter bed
to increase the biodegradability and the removal of organic compounds and improve
the reliability of disinfection.

Fig. 16.2. Gradual removal of microbes indicating pollution (Escherichia coli) from a half pint glass (284 ml) of
water at each stage of a typical London water treatment plant, based on a 10-year (1961- 1970) average.
(Adapted from Windle-Taylor, 1974)

Zurich
The city of Zurich draws its water from three sources: Lake Zurich, groundwater and
springs. The first treatment plant with SSF began operation in 1871. Gradually other
processes were added due to water quality deterioration and higher water quality
standards setting lower acceptable levels of organic contamination. Today lakewater
provides 70% of the water supply and is treated in two water treatment plants. In 1975,
SSF became the seventh of an eight-stage treatment system comprising: pre-oxidation in
the lake water collectors, coagulation/flocculation, pH adjustment, rapid sand filtration,
ozone treatment, activated carbon filtration, SSF, and disinfection (Fig. 16.3). Velocities up
to 0.7 mh-1 are now applied in the SSF. One of the benefits of SSF in this treatment plant
is to contribute to removal of the organic compounds that support biofilm growth in the
distribution system, reducing the requirements for high levels of disinfectant residuals.

So, SSF continues to be used as a treatment process in large European cities, but today
it is one of the final treatment stages, after quite complex pre-treatment stages. As
a result the SSF units receive water of very good quality. Hence, these systems, with
reliable operation, maintenance, and management conditions, can operate at high
filtration rates of around 0.3-0.7 mh-1. In these European cities the multiple barriers

339
 Fig. 16.3. Flow diagram of the water treatment system in Lengg, Zurich (Adapted from Huck, 1988)

strategy and basic water treatment concepts gradually developed from field experience
and under the pressure of tighter regulations. To extend the possibility of surface water
treatment to rural areas and small towns, these same concepts can be used in
identifying, developing and promoting pre-treatment alternatives in harmony with the
simplicity of operation and maintenance of SSF.

The search for pre-treatment alternatives for small water supply systems
The adequate use of SSF technology in small systems has often been determined by the
availability of good quality water resources, as is apparent from the application of SSF in
the USA. Pre-treatment appears to be the technical link missing from the SSF
technology for small communities with lower raw water qualities. During the last few
decades pre-treatment alternatives have been developed to extend the application of
SSF to poorer water sources without requiring skilled staff, complex mechanical
equipment, or chemical supplies. Some of these methods, such as riverbank filtration
(infiltration wells) and riverbed filtration (infiltration galleries), are oriented towards
improving surface water quality at the abstraction point. Other methods, using plain
sedimentation, are long- and short-term storage, and tilted plate settling. Others are
based on coarse filtration, such as dynamic filtration, and horizontal flow, downflow, and
upflow gravel filtration.

Infiltration wells
One of the oldest pre-treatment techniques is filtration in infiltration wells or riverbanks
along a river or stream (Fig. 16.4). Depending on the surface water quality and the
abstraction soil strata, the abstracted water may be acceptable for direct human
consumption or to be feed water for SSF units. Experiences with river Rhine water showed
that riverbank filtration reduced turbidity from a range of 1-6 NTU to a range of 0.2-0.8
NTU. Trace metals, DOC and COD were also significantly reduced. However some
problems were reported with the re-suspension of iron and manganese oxides when the
oxygen level in the river fell below 1 mgl-1. Changes in sediment transport in the river may
also affect the capacity of the wells. One possible disadvantage of this system is that
changes may occur underground, and can be difficult to remedy by maintenance activities.

340
 Chapter 16

Fig. 16.4. Infiltration wells and infiltration galleries

Source: Wehrle, personal communication; and Galvis and Visscher, 1987

Infiltration galleries
In infiltration galleries or riverbed filtration, water is abstracted using perforated pipes
through the natural riverbed material, or if the permeability is too low, through an artificial
bed of coarse sand and gravel. Riverbed filtration systems include longitudinal and lateral
drain systems, modular sub-sand abstraction, and river dam filtration systems. Flow velocity
through the filtering bed ranges from 0.25-1.5 mh-1, depending on turbidity levels and
effluent requirements. Removal efficiencies up to 98% have been reported for riverbed
dam filtration from rivers with turbidity levels in the range of 48-200 NTU. However, in
a field evaluation the efficiencies were found to be around 20%. This may be due to
difficulties in implementing periodic cleaning or repositioning of the clogged filtering
material, particularly during the rainy season, when the rivers have high flows and high
solids transport capacity. Because clogging of the infiltration area can mean reconstruction
of the riverbed filter or the infiltration area, pre-treatment filtration alternatives completely
separated from the surface water source are receiving more attention.

341
Plain sedimentation
Exposing the water to very slow or non-moving conditions allows suspended matter to
be removed by the action of gravity and natural particle aggregation without the use of
coagulants. This process is called plain sedimentation. Ideally, the clarification efficiency
of a settling basin, for a particular suspension of discrete particles, depends only on S0,
the surface charge (relation between the flow and the settling surface area). In practice,
however, disturbing factors such as turbulence and short-circuiting reduce the effective
settling velocity. Plain sedimentation is described in detail in chapter 15.

Tilted plate settlers

Improved flow conditions in the settling zone (laminar and stable flow) and lower values
of S0 (greater surface area for a given flow) can be obtained in a given conventional
sedimentation tank by introducing parallel plates set a short distance apart (5-10 cm).
To achieve self-cleaning, these plates are tilted or inclined at an angle of 50-60° to the
horizontal. Tilted plate settlers may reduce the required area of a conventional settler
(without plates) by some 65%. They are widely used in chemical water treatment, but
their application with non-coagulated water is very limited. Besides, in small systems, if
area is not a critical issue, this option may have comparable capital costs to
conventional settling, but higher running costs, since more frequent attention and
cleaning is needed because of its lower sludge storage capacity. Tilted plate settlers are
also described in chapter 15.

Prolonged storage basins

Plain sedimentation may have long retention times, measured in days or weeks. In this
case other factors are important, including wind, thermal, and photosynthetic effects.
This usually makes it an expensive solution to be adopted exclusively for water supply
purposes in small systems. A classical goal of storage basins is to provide supplies during
periods of low rainfall in multipurpose projects, and off-channel storage can provide
a source during short-term pollution events. Storage basins can be used as preliminary
treatment. Indeed, for extremely turbid waters, above 1000 NTU, storage provides the
best pre-treatment. In England the water depth in pumped storage reservoirs is typically
about 10-20 m and the theoretical retention time ranges from about 10-50 days. In
London, in long-term storage prior to SSF, turbidity reductions from around 30 NTU to
below 4 NTU have been reported. As shown in figure 16.2, the average E. coli faecal
coliform counts were reduced by 96%. However, the periodic blooms of algae made it
necessary to introduce microstrainers or rapid filters before the SSF units. Management
techniques have been developed to minimise algal blooms and other detrimental water
quality effects in the reservoirs. These techniques include pumping devices to control
the thermal stratification. The potential of long-term storage to protect SSF in small
systems directly or in combination with other treatment steps needs to be evaluated
under local conditions, introducing the possibility of a multipurpose reservoir.

342
 Chapter 16

Coarse media filtration (CMF)

Porous media such as gravel and sand are old water clarification processes with
documented applications in several European countries since the 1800s. Development
and promotion of this technology was interrupted with the arrival of chemical and
mechanised water treatment technologies. Since the 1970s the use of SSF technology in
small WS systems has gained increasing attention because of the potential of CMF to
improve the quality of deteriorating surface waters. During the 1980s it became clear
that CMF was a good option to condition the water before it reached the SSF units,
based on studies conducted in Africa, Asia, Europe, and Latin America. These
technologies and new ones are still being developed.

Coarse media filtration as a pre-treatment step for slow sand filtration

Short-term plain sedimentation may be the first conditioning stage of surface waters
that transport relatively large and heavy particles, such as grit or sand. However, rivers
usually transport a wide range of particles, including those with sizes of less than
10-20 mm. Most streams or small rivers in the tropics have peaks in suspended solids for
short durations, giving a high load on the water treatment system; these can happen in
the absence of the water treatment plant caretaker.

CMF is considered to be a promising pre-treatment technique for small water supply

systems since it is more effective in removing suspended particles than short-term plain
sedimentation and because of its ability to maintain treatment simplicity comparable to
that of SSF. CMF units are easier to operate and maintain than long-term storage
reservoirs and are not dependent on the hydraulic behaviour of streams or rivers, as are
riverbank and riverbed filtration, particularly during the rainy seasons in tropical countries.

Classification of coarse gravel filters

Different CMF alternatives using gravel as the filter media are described in the following
sections and schematically illustrated in figure 16.5. CMF alternatives have been classified
according to the main application purpose and the flow direction as shown in figure 16.5.

Dynamic gravel filters (DyGF)

Dynamic gravel filters include a shallow layer of fine gravel in their upper part and
coarse gravel that covers the underdrains. The water enters the unit and passes through
the fine gravel to the drainage system. With moderate levels of suspended solids in the
source water, the DyGF gradually clogs. If quick changes in water quality occur, the
clogging may be much faster. Eventually the gravel bed will be blocked and the total
water volume will just flow over the clogged surface area to waste, protecting the
subsequent treatment steps that are more difficult to maintain.

343
Depending on the flow direction in the layer of gravel, the second treatment step – the
gravel filters – are called upflow (UGF), downflow (DGF) or horizontal flow (HGF)
systems. A comparative study of these alternatives showed that the option of UGF was
technically and economically preferable over the DGF and HGF, although these also
achieve good removal efficiencies.

Upflow gravel filters (UGF)

Upflow gravel filters consist in principle of a compartment in which the gravel layer
reduces in size in the direction of flow. A drainage system placed on the bottom of the
structure serves to distribute the flow during the filtration period or to drain the gravel
layers during periods of cleaning, discharging the water through the drainage system.
There are two alternatives: upflow gravel filters in layers (UGFL) when the gravel layers
of different size are installed in the same unit and upflow gravel filters in series (UGFS)
when the gravel layers are installed in two or three different units, each having a main
gravel size that decreases in the direction of flow.

Downflow gravel filters in series (DGFS)

Downflow gravel filters (as used Colombia) consist usually of three compartments with
the coarsest gravel in the first unit and less coarse in subsequent units. The functioning
or performance of the DGFS is similar to the UGFS in terms of removal efficiency, but
maintenance is more difficult because the sludge tends to accumulate on the surface of
the first unit. Cleaning is more difficult than for the UGFS units, where the sludge is
accumulated basically in the bottom part close to the drains.

Horizontal-flow gravel filters (HGF)

Horizontal-flow gravel filters consist of at least two parallel modules constructed
basically in three compartments separated by perforated walls. In the beginning this
option was very voluminous, because it did not include a drainage system for hydraulic
cleaning. Nowadays a drainage system is included. Although it is possible to reduce the
size of the units, the activities of operation and maintenance in an HGF are more
demanding in terms of manpower and water consumption. Research on HGF in series
gives promising results in terms of hydraulic performance and the gain is a substantial
reduction in the length of the gravel bed, while maintaining similar efficiency levels as
the conventional HGF.

General considerations
Effluent water quality
Coarse gravel filters (CGF) have normally been specified to produce an effluent with
turbidity < 10-20 NTU, or suspended solids < 5 mgl-1, although the impact of these or
other values on the SSF performance or maintenance is not clearly established. Besides,
other parameters such as high levels of faecal contamination or natural organic matter

344
 Chapter 16

Fig. 16.5. Schematic view of coarse gravel filtration alternatives

(Based on Galvis and Visscher, 1987)

that could limit the application of SSF, are not normally considered as critical factors in
the specification of this technology.

Head loss and flow control

Final head loss in CMF units is small, usually a few centimetres, with a maximum value
around 0.30 m. Because of these low values, CMF units usually have inlet flow control.
The inlet structure should include facilities for energy dissipation, flow control, flow
measurement, and overflow. A well-designed inlet box facilitates the operation and
control of the system. A weir or a raised effluent pipe maintains the water above the
filter bed level. Flow measurement devices are recommended at the inlet and outlet
sides to control the operation and to verify that the filter boxes are watertight. Since the
CMF units in small water supply systems deal with low flow and low pressure values,
some simplified valves, gates, and weirs can be used together with more commercial
hydraulic devices.

Design criteria and filter run time

The main criteria for CMF design have been removal efficiency and head loss related to
particle retention in the filtering bed. Process variables such as particle nature and size
distribution, collector size (dc), filtration rate (v), and filter length (L) determine the filter
run time up to the breakthrough, related to a maximum concentration value (Cm) in
the effluent, or up to a clogging point, related to a maximum head loss (Hm).

345
A qualitative illustration of the impact of some process variables on breakthrough and
clogging in RF and CMF is illustrated in figure 16.6. As predicted by the trajectory
approach in filtration theory, removal efficiencies in coarse filtration will be smaller due
to its greater collector size. This limitation is partly overcome by lower filtration rates
and longer filtering beds in CMF.

Biological activity
Biological activity takes place in the coarse filtration units when they are processing
natural waters and synthetic waters with organic matter or nutrients. Most probably,
with mechanisms similar to those present for SSF, bacteria and other micro-organisms
may form sticky layers in some areas of the filter media or produce exocellular polymers
that contribute to particle destabilisation and attachment. Macro-biological creatures
inhabiting the coarse filters are thought to contribute to the sloughing off of stored
material or biofilm observed. There is evidence of organic matter decomposition during
cleaning procedures of full scale HGF, calling for frequent maintenance of units
susceptible to high biological activity.

Note: Although effluent water quality is expected to be lower in the CMF, its tc (breakthrough) value should be

higher than for the RF

Fig. 16.6. Effects of some process variables on the breakthrough and clogging points in rapid (RF), and coarse
media filters (CMF). (Adapted from Boller, 1993)

Flow conditions and coarse media filtration efficiency

Research was carried out with vertical flow filter columns of 1 m depth filled with gravel
varying from 1-64 mm in size and filtration rates from 0.5-8 mh-1. The turbidity of the
raw water mixture was maintained at around 60 NTU. Good turbidity reductions were
obtained at filtration rates < 2 mh-1. This experience shows that significant solids
removal efficiency is only achieved under laminar flow conditions (see fig. 16.7).

346
 Chapter 16

Further laboratory and field tests with UGF and HGF confirmed that effluents with
a turbidity below 10 NTU were achieved only at filtration rates of 0.5-1.0 mh-1.

The filtering media

The filtering media should have a large surface area to enhance particle removal and
a high porosity to allow the accumulation of the separated solids. Filtration tests with
kaoline clay suspensions revealed that neither the roughness nor the shape of the filter
material had a great influence on filter efficiency. Any inert, clean and insoluble material
meeting the previous criteria could be used as filtering media. Gravel is the commonly
used material, but broken bricks, palm fibre, and plastic material have also been
reported in different experiences. In a review of CMF performance with different filter
media, a filter filled with palm fibre achieved better turbidity removal than a gravel filter.
This is the result of the greater porosity (92% versus 37%), resulting in a lower effective
velocity. However, since the use of palm fibre causes a considerable drop of dissolved
oxygen along with odour and taste problems, this filtering medium has serious
limitations. The use of plastic material may be an alternative, but the uplift forces of the
water have to be overcome.

Fig. 16.7. Influence of flow conditions on coarse filtration efficiency

Source: Wegelin and Mbwette (1989); Wegelin et al., (1991)

Operation and maintenance (O & M).

Operation of CMF units requires a frequent (at least daily) control of the influent and
effluent flow and the quality of the raw and filtered water. Maintenance is associated
mainly with the cleaning process, which tries to restore the initial head loss. To facilitate
maintenance, a minimum of two units should be constructed in parallel. Frequent
cleaning of the CMF units is recommended to limit head loss development and to avoid
operational or maintenance difficulties due to solids consolidation or organic
decomposition inside the filter media. CMF units are cleaned both manually and
hydraulically. Manual cleaning involves media removal, washing and replacement, which
is time consuming and labour intensive. So, hydraulic cleaning facilities for in-place

347
media flushing become a key component of the units to ensure a long-term
sustainability of this treatment technology.

Initially only surface raking was used to clean dynamic gravel filters (DyGF). Later it was
combined with filter bed drainage. Only manual cleaning was initially used to clean HGF
and gradually fast drainage of the filter bed compartments has been incorporated in its
application. Fast or moderate drainage velocities, combined in some cases with some
surface raking, are being applied to maintain DGF and UGF. The area and the height of
the filter boxes should be limited to facilitate both frequent hydraulic cleaning and
eventual manual cleaning.

The drainage system

In the case of DyGF, HGF and DGF, the drainage system collects and provides an outlet for
filtered water during normal operation, as well as for washing water during hydraulic
cleaning by fast drainage. In the case of UGF, the drainage system distributes the water to
be filtered, and collects and provides an outlet for washing water during hydraulic cleaning.
The system may consist of a small trough, a false filter bottom, or perforated pipes or
manifolds. One small trough would have limitations to produce an even flow distribution
across the entire filter bed compartment. A good false bottom would ensure an even water
collection or distribution but imply additional hydraulic structures. A properly designed
manifold should have a good hydraulic efficiency with lower construction costs, although it
requires an additional gravel layer to embed the pipes. The decision between false bottom
and manifolds should be taken after analysing local conditions.

16.4 Considerations about multi-stage filtration

The combination of coarse gravel filtration and SSF is what in this publication is called
multi-stage filtration (MSF).

The MSF technology has received a lot of positive response in Colombia and other Latin
American countries, where over 100 systems are already in operation today. Ten of these
built in Colombia date from the middle 1980s, each producing effluents with low sanitary
risk before terminal disinfection and with low operation and maintenance costs that are
to a large extent covered by the users. They pay a tariff of some 2 USD/month, in
a country with a minimum official salary of some 140 USD. All systems are administered
by community-based organisations with some technical support from sector institutions.

MSF does not compromise the advantages of an SSF system in terms of ease of
operation and maintenance and the production of good water quality. It is an option
that is applicable to many rural communities and small- and medium-sized
municipalities, where treatment with chemical products has very little potential.

348
 Chapter 16

Table 16.4 presents a summary of the considerations concerning MSF treatment and
figure 16.8 shows a layout of MSF with three components, DyGF, UGF and SSF.

The following combinations of CGF and SSF can be made:

DyGF + SSF
DyGF + UGFL + SSF
DyGF + UGFS2 + SSF
DyGF + UGFS3 + SSF

The criteria for selection of each combination will be discussed in chapter 16.6.

Table 16.4 Summary of considerations concerning MSF treatment

Issue Comment concerning MSF treatment

Quality of treated water It is a good alternative to improve the physical, chemical and
bacteriological quality of the water. In many areas and
particularly those with a less developed infrastructure, MSF is
the only feasible treatment option.
Ease of construction The relatively simple design facilitates the use of local materials
and local manpower. There is no need for special equipment.
Construction cost The construction in local materials and with local labour reduces
the cost. Usually there is no need for imported materials.
Ease of operation and After a short period of training, local operators with a minimum
maintenance of formal education can operate and maintain the system.
Cost of operation and The cost of operation and maintenance and the requirements in
maintenance electrical energy are minimum and less than required for other
systems. There is no need for chemical products for coagulation.
Reliability A low risk of mechanical problems or problems related to the
changes in the raw water quality, as these can be absorbed
without interrupting the service in the majority of cases.
Cleaning The cleaning process is simple although laborious, but almost
always involving low cost, as in many countries labour is
relatively cheap.
Requirements of surface area A conventional RSF plant in respect to storage zones,
management of chemicals etc., may require comparable areas to
an MSF system.
It is not a panacea There are levels of contamination that limit the efficiency or
interfere with the treatment.

349
Performance of multi-stage filtration systems
The number of full-scale MSF plants in the world is limited. Most comprehensive
research on functioning and performance has been carried out in Latin America.
The following observations on performance are therefore mainly from that region.
In general, performance findings are very satisfactory. Nevertheless, the performance
may be different, that is higher or lower, in other regions of the world. Much depends
on the characteristics of the raw water in terms of turbidity, suspended solids, particle
size distribution, true colour and temperature. Climatic seasonal fluctuations also
influence the performance of MSF.

Fig. 16.8. Components of multi-stage filtration systems

The characteristics of the different coarse gravel media filter units are presented in table
16.5. The gravel filters were evaluated for filtration rates of 0.30, 0.45, 0.6, 0.7 and 1.0
m/h. The research included physical, chemical and bacteriological parameters and
established the limits under which each unit could still operate. The last step in each
treatment line was an SSF operated at 0.15 m/h.

The removal efficiencies of basic water quality parameters in MSF pilot system at Puerto
Mallarino are shown in tables 16.6 and 16.7. Four different periods were evaluated, each
with a different filtration rate.

The HGF, having a larger sludge storage capacity and similar removal efficiencies, may be
an alternative for surface water high in suspended solids, even though it is more expensive.

350
 Chapter 16

Table 16.5 Characteristics of the treatment units in Puerto Mallarino Research Station, Cali, Colombia

Treatment Unit Number of units Filtration area Filter medium

in series (m2) Size (mm) Length (m)
DyGF 1 0.75 6-25 0.6
URGS3 3 3.14 25-1.6 4.55
HGF 1 1.54 25-1.6 7.2
HGFS 3 1.54 25-1.6 4.55
UGFS2 2 3.14 25-1.6 3.1
UGFL 1 3.14 25-1.6 1.55
SSF 1 3.14 Uc = 1.57 D10 = max.: 1.0
0.23 mm min.: 0.6

Uc: Uniformity coefficient.

D10: Effective diameter

The results show that the combination of two stage gravel filters (DyGF + CGF) very
much improves the performance of the SSF. Nevertheless, in cases of highly
contaminated surfaces water sources, particularly if the levels of suspended solids are
high (above 100 mg/l), a very critical selection of treatment barriers is required and has
to be in coherence with the risk level in the water source and its variation over time.
The preferred option would be to select an alternative water source. If this is not
possible detailed pilot studies are needed to ensure the viability of the solution.

Performance of full-scale MSF systems in Colombia

Seven community managed MSF systems in the Cauca Valley have been monitored for
a period of seven years. The systems receive water from catchment areas with low or
moderate levels of human intervention. Water quality of the different sources indicates mean
turbidity levels between 0.9 and 15 NTU, faecal coliform counts between 52 and 51,916
FCU/100 ml, and true colour levels between 3 and 30 PCU. The mean removal efficiencies of
basic water quality parameters in full-scale MSF plants are shown in table 16.6 and 16.7. The
wide ranges are due to different CMFs applied and the different local operations.

The composition of the systems matches the multi-barrier concept that implies that
more than one stage of treatment is needed, combined in such a way that together the
barriers have a removal efficiency that is sufficient to ensure low dose disinfection as the
final and efficient safety barrier.

All systems produce water with turbidity below 1 NTU, with a frequency between 65 and
98%, and below 5 NTU in more than 98% of the samples. Faecal coliforms were below
25 FCU/100 ml, with a frequency above 97%, and true colour below 15 TCU in more
than 98% of the samples. With these water qualities, constant dose disinfection with
chlorine as suggested by WHO (1996) becomes an effective safety barrier.

351
Table 16.6 Individual (at each treatment stage) and cumulative (up to the end of SSF stage) mean
removal efficiencies of basic water quality parameters in MSF pilot system at Puerto
Mallarino.

Filtration Period Influent mean values Individual mean efficiencies

Stage Rates Turbidity Colour Faecal Turbidity Colour Faecal
(mh-1) (NTU) (PCU) coliforms (%) (%) coliforms
(CFU/ (log.
100 ml) units)
DyGF stage
DyGF (0.9-1.4) I 109 81 41,184 32 11 0.2
(1.4-2.5) II 59 54 31,800 41 15 0.6
(1.4-2.5) III 51 35 97,779 43 14 0.8
(1.9-2.8) IV 52 57 108,796 40 16 0.6
CGF stage
UGFS 0.3 I 74 72 24,758 84 69 2.6
0.45 II 35 46 8843 77 54 2.3
0.6 III 29 30 16,823 77 53 2.4
0.75 IV 31 48 26,226 75 63 2.3
UGFL 0.3 I 74 72 24,758 70 44 1.8
0.45 II 35 46 8843 54 28 1.3
0.6 III 29 30 16,823 55 30 1.4
0.75 IV 31 48 26,226 61 46 1.3
SSF stage
SSF 1 0.1 I 12 22 65 64 73 2.8
0.1 II 8.1 21 45 75 67 2.7
0.55 III 6.6 14 64 82 57 2.1
0.15 IV 7.8 18 127 74 67 1.8
SSF 2 0.1 I 22 40 369 85 88 3.1
0.1 II 16 33 452 79 70 2.2
0.15 III 13 21 637 89 67 2.6
0.15 IV 12 26 1226 77 69 2.0
Treatment lines Period Effluent mean values Cumulative mean efficiencies
DyGF + UGFS + I 4.3 6 0.1 96 93 5.6
SSF1 II 2.0 7 0.1 97 87 5.5
III 1.2 6 0.5 98 83 5.3
IV 2.0 6 2.2 96 89 4.7
DyGF + UGFL + I 3.2 5 0.8 97 94 5.1
SSF2 II 3.3 10 2.7 94 81 4.1
III 1.4 7 1.7 97 80 4.8
IV 2.8 8 10.7 95 86 4.0

352
 Chapter 16

Table 16.7 Individual (at each treatment stage) and cumulative (up to the SSF stage) mean removal
efficiencies of basic water quality parameters in full-scale MSF plants.

Filtration Filter bed Influent mean values Individual mean efficiencies

Stage Rates length Turbity Colour Faecal Turbidity Colour Faecal
-1
(mh ) (m) (NTU) (PCU) coliforms (%) (%) coliforms
(CFU/ (log.units)
100 ml)
DyGF stage
DyGF 0.9 – 1.6 0.3 – 0.6 3.8 – 24 15 – 30 2895 – 21 – 57 10 – 24 0.2 – 0.7
51,916
CGF Stage
CGF 0.5 – 0.9 0.9 – 4.0 2.8 – 17 5 – 27 330 – 30 – 71 17 – 41 0.7 – 1.0
10,063
SSF Stage
SSF 0.08 – 1.0 – 1.2 0.8 – 4.9 4 – 16 52 – 50 – 87 25 – 75 1.7 – 3.3
0.17 2008
Treatment plant Effluent mean values Cumulative mean efficiencies
DyGF + CGF + SSF 0.4 – 0.9 3 – 6 0.7 - 79 – 96 40 – 87 2.6 – 4.7

MSF treatment can adapt itself to the type of raw water and the concentration of
contamination. The systems give higher removal efficiencies for water that is higher in
contamination. This implies that the barriers become more effective if the water to be
treated has a higher risk and still can produce a water with a low sanitary risk level. MSF
technology has a great potential to reduce the physical-chemical and bacteriological risk
associated with surface water sources. However, the MSF technology is not a panacea
and has its limitations, particularly with high levels of contamination, not always
producing water of a quality that can be properly disinfected.

16.5 Cost considerations

Some components of a filtration system have the greatest impact (about 80%) on the
construction cost. These include civil works, filter media, the excavation and the valves.
The cost efficiency increases with the size of the system. Nevertheless, for this type of
filtration systems the economy of scale is limited, which favours a relative short design
period of some ten years.

The operation and maintenance cost of MSF systems is mainly determined by labour
cost; in Colombia staff costs made up 85% of the total.

353
 16.6 Selection of MSF alternatives

Different combinations of filtration stages are identified to treat raw water types. In
general, filter bed lengths increase with the contamination levels in raw water types
while filtration rates decrease. Unsurprisingly, capital and running costs of MSF plants
increase with increasing contamination levels in their raw water types. Table 16.8 gives
a selection guide based on the parameters faecal coliforms densities, turbidity and
colour. All these MSF alternatives fulfil proposed water treatment objectives.

Table 16.8 An example of a selection guide for MSF alternatives fulfilling established
treatment objectives for removing turbidity, faecal coliform bacteria and colour,
based on experiences in the Andean Colombian Cauca Valley.

Y4 DyGF2.5 DyGF2.0 DyGF2.0 DyGF1.5 DyGF1.5

Mean < UGFS(3)0.6 UGFS(3)0.6 UGFS(3)0.6 UGFS(3)0.45 UGFS(3)0.3
15,000 SSF0.15 SSF0.15 SSF0.15 SSF0.15 SSF0.15
Max. <
45,000
Y3Mean DyGF2.5 DyGF2.0 DyGF2.0 DyGF1.5 DyGF1.5
< 5000 UGFS(2) 0.6 UGFS(2) 0.6 UGFS(3) 0.6 UGFS(3) 0.45 UGFS(3)0.3
Max. < SSF0.15 SSF0.15 SSF0.15 SSF0.15 SSF0.15
15,000
Faecal coliforms (CFU/100ml)

Y2Mean< DyGF2.5 DyGF2.0 DyGF2.0 DyGF1.5 DyGF1.5

1500 UGFL 0.6 UGFL 0.6 UGFS(3) 0.6 UGFS(3) 0.45 UGFS(3)0.3
Max.< SSF0.15 SSF0.15 SSF0.15 SSF0.15 SSF0.15
5000
Y1Mean< DyGF2.5 DyGF2.0 DyGF2.0 DyGF1.5 DyGF1.5
750 UGFL 0.75 UGFL 0.6 UGFS(3) 0.6 UGFS(3) 0.45 UGFS(3)0.3
Max.< SSF0.20 SSF0.15 SSF0.15 SSF0.15 SSF0.15
2500
Y

Mean < 5 10 16 20 25
P 95% < 15 30 50 60 70
Turbidity
(NTU)

Max. < 50 100 150 225 300

X1 X2 X3 X4 X5
X

Mean < 10 13 16 18 20
Colour
(PCU)

Max. < 30 40 50 55 60
Z1 Z2 Z3 Z4 Z5
Z

354
 Chapter 16

Explanation of the selection guide in table 16.8:

1. The number between brackets indicates the number of filtration steps in UGFS
alternatives. The sub-index means filtration rates in mh-1.
2. Raw water may be directly disinfected (without filtration) if turbidity and faecal
coliform levels are below 5 NTU and 20 CFU/100 ml in 95% of samples
respectively. These low contamination levels must be confirmed periodically with
sanitary inspections and analyses in the watershed area.
3. DyGF + SSF (without CGF stage) could be applied if turbidity and faecal coliform
levels are below 10 NTU and 20 CFU/100 ml in 95% of samples respectively. These
low contamination levels must be confirmed periodically with sanitary inspections
and analyses in the watershed area.
4. Turbidity treatment objectives (< 10 and 5 NTU in CGF and SSF effluents
respectively) should be obtained with 95 percentile (P95) turbidity values. It is
expected that maximum (peak) turbidity values can be treated thanks to the
protection capacity of the DyGF stage, combining flow reductions with higher
removal efficiencies.
5. Faecal coliform treatment objectives (< 1000 and 10 CFU/100 ml in CGF and SSF
effluents respectively) should be obtained with maximum faecal coliform levels. With
medium faecal coliform levels in raw water sources SSF effluents should have
effluents with mean values < 3 CFU/100 ml before terminal disinfection).
6. Colour treatment objective (< 15 PCU in SSF effluents) should be obtained with
maximum colour levels. This is a secondary treatment objective and should not
compromise previous treatment objectives or terminal disinfection as a safety barrier.

355
Bibliography

Amirtharajah, A. (1988). ‘Some theoretical and conceptual views of filtration’. In: Journal
American Water Works Association, vol. 80, no. 12, p. 36-46.

Baker, M.N. (1981). The quest for pure water: the history of water purification from the
earliest records to the twentieth century. Vol. I. 2nd ed. Denver, CO, USA, American
Water Works Association.

Bellamy, W.D.; Hendricks, D.W. and Logsdon, G.S. (1985). ‘Slow sand filtration: influences
of some selected process variables’. In: Journal American Water Works Association, vol.
77, no. 2, p. 62-66.

CEHE (1999). Development and integration of small scale multistage treatment for
drinking water. Draft. Guildford, UK, Centre for Environmental Health Engineering
(CEHE). P. 1-142.

Cleasby, J.L.; Hilmoe, D.J.; Dimitracopoulus, C.J. and Diaz-Bassio, L.M. (1984). ‘Slow sand
and direct in-line filtration and surface water’. In: Journal American Water Works
Association, vol. 76, 12, p. 44.

Collins, M.R. et al. (1985). ‘Removal of organic matter in water treatment’. In: Journal of
environmental engineering, vol. 111, no. 6.

Collins, M.R.; Eighmy, I.T. and Malley, J.P. (1991). ‘Evaluating modifications to slow sand
filters. In: Journal American Water Works Association, vol. 83, no. 9, p. 62-68.

Collins, M.R. and Graham, M.J.D. (1994). Slow sand filtration: an international
compilation of recent scientific and operational developments. Denver, CO, USA,
American Water Works Association.

Craun, G.F. et al. (1994). ‘Balancing chemical and microbial risk of drinking water
disinfection. Part I. Benefits and potential risks’. In: Aqua: journal of water supply
research and technology, vol. 43, no. 4, p. 192-199.

Craun, G.F. et al. (1994). ‘Balancing chemical and microbial risk of drinking water
disinfection. Part II. Managing the risks. In: Aqua: journal of water supply research and
technology, vol. 43, no. 5, p. 207-218.

Ellis, K.V. (1985). ‘Slow sand filtration’. In: CRC critical reviews in environmental control,
vol. 15, no. 4, p. 315-354.

356
 Chapter 16

Fox, K.R. et al. (1984). ‘Pilot-plant studies of slow rate filtration’. In: Journal American
Water Works Association, vol. 76, no.12, p. 62-68.

Galvis, G.; Latorre, J. and Visscher, J.T. (1998). Multi-stage filtration: an innovative water
treatment technology. (Technical paper series; no. 34-E). The Hague, The Netherlands,
IRC International Water and Sanitation Centre.

Graham, N.J.D. and Collins, M.R. (1996). Advances in slow sand and alternative biological
filtration. Chichester, UK, John Wiley.

Hudson, H.E. (1981). Water clarification processes: practical design and evaluation. New
York, NY, USA, Van Nostrand Reinhold.

Huisman, L. and Wood, W.E. (1974). Slow sand filtration. Geneva, Switzerland, World
Health Organization.

Ives, K.J. (1957). ‘Algae and water supplies: physical removal of algae’. In: Water
engineering, vol. 61, p. 432.

Joshi, N.S. et al. (1982). ‘Water quality changes during slow sand filtration’. In: Indian
journal of environmental health, vol. 24, no. 4, p. 261-266.

Logsdon, G.S. (1991). Slow sand filtration. New York, NY, USA: American Society of Civil
Engineers.

Pardón, M. (1989). Treatment of turbid surface water for small community supplies.
Report. Guildford, UK, University of Surrey. PhD Thesis.

Shulz, R.C. and Okun, D.A. (1992). Surface water treatment for communities in
developing countries. (Wiley-interscience). New York, NY, USA,Wiley; London, UK, ITDG
Publishing.

Slezak, L.A. and Sims, R.C. (1984). ‘The application and effectiveness of slow sand
filtration in the United States’. In: Journal American Water Works Association, vol. 76, no.
12, p. 38-43.

Smet, J.E.M. and Visscher, J.T. (1989). Pre-treatment methods for community water
supply: an overview of techniques and present experience. The Hague, The Netherlands,
IRC International Water and Sanitation Centre.

357
Sundaresan, B.B.; Paramasivam, R. (1982). Slow sand filtration: research and
demonstration project - India: final report. Nagpur, India, National Environmental
Engineering Research Institute.

Visscher, J.T., Paramasivam, R., Raman, A., Heijnen, H.A. (1987). Slow sand filtration for
community water supply: planning, design, construction, operation and maintenance.
(Technical paper series; no. 24). The Hague, The Netherlands, IRC International Water
and Sanitation Centre.

Wegelin, M. (1996). Surface water treatment by roughing filters: a design, construction

and operation manual. (SANDEC report; no. 2/96). St. Gallen, Switzerland, SKAT (Swiss
Centre for Development Cooperation in Technology and Management).

Wegelin, M.; Boller, M. and Schertenleib, R. (1986). ‘Particle removal by horizontal-flow

roughing filtration’. In: Aqua: journal of water supply research and technology, vol. 35,
no. 3, p. 115-125.

Yao, K.M.; Habibian, M.T. and O'Melia, C.R. (1971). ‘Water and wastewater filtration:
concepts and applications’. In: Environmental science and technology. p. 1105.

358