Surface Water Treatment by Roughing Filters - A Design, Construction and Operation Manual (SANDEC - SKAT, 1996, 180 p.)

Surface Water Treatment by Roughing Filters - A Design, Construction and Operation Manual (SANDEC - SKAT, 1996, 180 p.) Part 1: General aspects of roughing filter application 1. Historical development and experience with water treatment 2. Water treatment concept 3. Raw water quality 3.1 Raw water characterisation 3.2 Catchment area 3.3 Water quality analysis 4. Solid matter separation (introduction...) 4.1 Sedimentation 4.2 Roughing filtration 5. Bacteriological water quality improvement (introduction...) 5.1 Slow sand filtration 5.2 Chlorination 6. Layout of a water supply scheme 6.1 General considerations 6.2 Hydraulic profile 6.3 Treatment steps 6.4 Water distribution 7. Roughing filtration application 7.1 Historic use 7.2 Development of roughing filters

Surface Water Treatment by Roughing Filters - A Design, Construction and Operation Manual (SANDEC - SKAT, 1996, 180 p.)

Part 1: General aspects of roughing filter application

1. Historical development and experience with water treatment

" In the earliest days of the human race, water was taken as found. It might be pure and abundant, plentiful but muddy, scarce but good, or both scarce and bad. To get more or better water, man moved to other sources rather than transport better water to his own location or to try to improve the quality of water at hand". This cited text marks the beginning of Baker's epilogue in "The Quest for Pure Water" [1], a reference book he started compiling at the beginning of this century and which was finalised in the 1940s. Baker continues by saying " Man's earliest standards of quality were few: freedom from mud, taste and odour". However, an increase in man-made water pollution, the development of technical and public health science, as well as the consumers' greater need for clean water contributed to the development of the water purification technology.

At the beginning of the 19th century, the first water treatment plants for public water supplies were constructed in Britain and France. They generally comprised settling basins followed by gravel and sand filters. In the course of time, slow sand filters were developed as an efficient water treatment process, and used by many water authorities at the end of last century. By this time however, the Industrial Revolution came up with the "mechanical" filters as rapid sand filters were initially called. The growing water demand and the subsequent discovery of chlorine to disinfect the water enhanced the use of rapid sand filters. In 1940, there were about 2,275 rapid filter plants in the United States as opposed to about 100 slow sand filter plants. Another outstanding feature with regard to the water treatment technology was the use of aluminium and iron salts as coagulants in water treatment. Since the beginning of this century, coagulation and flocculation combined with sedimentation, rapid filtration, and final chlorination are now commonly used in water treatment. This treatment combination is now usually regarded as conventional.

Water treatment plants are either built in situ, usually as reinforced concrete structures, or installed as package plants manufactured by the water industry. Fig. 1 illustrates the extensive use of chemicals in conventional water treatment. Colloidal matter has to be destabilised by coagulants, such as aluminium sulphate or ferrous sulphate, possibly in combination with lime dosage for pH adjustment and polymers or polyelectrolytes to improve flocculation. As rapid filters do not significantly improve the microbiological water quality, chlorine has to be used as final treatment step to produce water which is safe for consumption. Finally, the numerous chemicals added may also have changed the chemical water characteristics. The treated water, which may either be corrosive or deposit-forming, could greatly harm the distribution system. Consequently, the treated water often has to be stabilised with a final dose of lime.

Conventional water treatment also requires a substantial input of energy and mechanical equipment. Frequently, the raw water has to be pumped through the different treatment stages. Flocculation requires energy inputs for hydraulic or mechanical mixing, sludge removal in sedimentation tanks is often carried out with mechanical scrapers, and rapid sand filters are backwashed for filter cleaning. Dosing pumps are necessary for adequate chemical application. In brief, conventional treatment calls for an extensive use of power-driven, mechanical and often sophisticated equipment.

Fig. 1 Operational Problems in Conventional Water Treatment Plants

A reliable and efficient operation of a conventional water treatment plant is a demanding task. A continuous supply of different chemicals must be guaranteed, spare parts of mechanical equipment must be stocked or easily available, and the treatment plant operated by well-trained and skilled personnel. The local infrastructure should support maintenance and repair of treatment plant components. However, these criteria are hardly ever met by local conditions prevailing in rural areas of developing countries.

Wagner states in the preface of the manual "Upgrading Water Treatment Plants" [2], which is the result of a WHO working group on operation and maintenance established in the 1990s: "In the majority of plants, especially in the less developed countries, much of the expensive equipment does not operate properly due to lack of understanding or disregard of maintenance and operation recommendations". Only a few plants are designed on the basis of bench and pilot plant testing. The need for careful design is most urgent in countries with the least resources. However, design studies are in fact considered a luxury rather than a necessity in these countries. The most widely encountered deficiency in water treatment is the application of coagulants to raw water. Incorrect dilution of the solution, inadequate doses and inappropriate dosing are the most common mistakes. Difficulties are also experienced with the flocculation step. Uncontrolled energy inputs result in small floes of low settleability. Sedimentation tanks are often not well-designed; short circuiting and incorrect water abstraction lead to poor clarification and overloading of the subsequent filters. These in turn cannot be backwashed properly and produce filtered water of high turbidity. Finally, poorly or inoperative chlorination equipment is commonplace in rural water treatment plants in developing countries, as the equipment usually originates from industrialised countries and, hence, foreign exchange is required to purchase these installations and spares. The described difficulties encountered with conventional water treatment will result in the production of water of erratic quality which is often unsafe for consumption.

Objectiveness demands that earlier experienced operational difficulties with slow sand filters have to be mentioned at this point. Initially, slow sand filters were developed to combat the cholera and typhus epidemics in Europe last century. On account of its simplicity and low-cost, the slow sand filter concept was then indiscriminately exported to developing countries in the early days of technical cooperation. Slow sand filters operate perfectly well with raw water of low turbidity as generally encountered in European surface waters. However, raw water quality in tropical climates can vary considerably, especially as regards turbidity and solid matter load. Therefore, this direct transfer of technology has proved inadequate. The inability of slow sand filters to sustain adequate filter runs when subject to high turbidity loads became obvious. Worldwide practical experience revealed that the slow sand filter design concept was often misunderstood, the use of pretreatment processes, such as plain sedimentation or flocculation and sedimentation, were either inefficient or unreliable as well as inappropriate, and that operation and maintenance deficiencies contributed to the poor performance of slow sand filters. In the early 1960s in Brazil, for example, the communities were not adequately trained in slow sand filter operation, thus causing a high failure rate of the slow sand filters [3]. In Cameroon on the other hand, slow sand filters were operated adequately twenty years ago. However, due to the raw water quality deterioration caused by progressive deforestation of the catchment areas, these filters faced increasing operational difficulties which required treatment plant rehabilitation [4]. Finally, an evaluation of the performance of four slow sand filter plants carried out in India in 1993 revealed that its current design, construction and operation, including source protection, is far from being satisfactory and often leads to poor filter performance [5].

Successful projects call for a multidisciplinary approach requiring various types of inputs. Sociocultural, institutional, and natural conditions must be considered along with financial and technical aspects. The synthesis of all these inputs lead to appropriate and sustainable solutions. This manual focuses mainly on technical aspects and gives answers to perhaps the least difficult problems. From the technical viewpoint, development of the roughing filter technology has contributed towards an efficient and reliable slow sand filter operation appropriate for rural water supply schemes in developing countries.

Photo 1 Compact Plant An Example of Conventional Water Treatment

Photo 2 Roughing and Slow Sand Filter - An Alternative Treatment Option

2. Water treatment concept

Water Treatment is usually a complex process which is often bound to fail if the objectives are not defined, the raw water properties not closeIy examined and the treatment methods inadequate. With a clear treatment concept, including a reasonable appreciation of the raw water characteristics and seasonal variations of the water quality, logically combined with the most appropriate treatment processes, failures can be avoided.

Fig. 2 Solid Matter Content in Surface Water

Fig. 3 Multiple Barrier Water Treatment Concept

A bucket filled with turbid river water, as illustrated in Fig. 2, often contains floating matter, such as debris of wood, leaves and grass, fine and coarse sand, which has settled at the bottom, and some fine suspended matter in the form of silt and clay particles or algae. However, harmful microorganisms, carriers of so many infectious diseases transmitted by consumption or contact with polluted water, cannot be detected with the naked eye. The size of such organisms, such as protozoa, bacteria and viruses, range within a few micrometers (1 mm is a thousandth of a millimetre) or even less. Removal or inactivation of these pathogenic organisms should, however, be given first priority in any water treatment concept. A difficult task, considering their small size and possibly low concentration in such a large volume of water. Slow sand filtration and chlorination are thus the two most widely used treatment processes, as they are capable of improving, in particular, the microbiological water quality.

The efficiency of chlorination and slow sand filtration is strongly influenced by the level of turbidity of the water to be treated. Turbidity mainly reflects the amount of fine suspended solids present in the water. A large number of microorganisms, tired of swimming around, attach themselves like "boat people" to the surface of these solids. The solids protect the microorganisms from the deadly chlorine. In slow sand filters, the pathogens will triumphantly observe how the fine particles block the sand surface. Hence, an efficient use of chlorine and slow sand filters is only possible with a low water turbidity virtually exempt from sol id matter.

As illustrated in Fig. 3, water has to undergo a step-by-step treatment, especially if it contains differently sized impurities. The first and easiest step in sound water treatment schemes is coarse solids separation. Finer particles are separated in a second pretreatment step and, finally, water treatment will end with the removal or destruction of small solids and microorganisms. These different pretreatment steps will contribute to reducing the pathogenic microorganisms. The '´boat-people" or pathogens attached to the surface of suspended solids will get stranded when the solids are separated. Some of the microorganisms floating in the water might also get pushed to the surface of the treatment installations and adhere to biological films. Solid matter and microorganisms, therefore, face a multitude of treatment barriers. Since treatment efficiency of each barrier increases in the direction of flow, it becomes increasingly difficult for the impurities to pass through each subsequent treatment barrier.

Surface water treatment thus requires generally at least two treatment steps as shown in Fig. 4. The first step, also called pretreatment, concentrates mainly on the removal of solids. Screens, grit chambers, sedimentation tanks, gravel and coarse sand filters are typically used as pretreatment units. The second step, commonly considered as main treatment, is used especially to remove or destroy the remaining microorganisms and the last traces of solid matter. Slow sand filtration and chlorination are the most commonly applied treatment processes in this second step.

Fig. 4 Surface Water Treatment in Two Stages

3. Raw water quality

3.1 Raw water characterisation

Surface water must generally be treated before it is used as drinking water as it is highly exposed to natural and man-made pollution. The extent of treatment depends, however, on the degree of water pollution to be assessed prior to designing any treatment facility. The design of a rural water treatment plant is based mainly on the following important water quality parameters:

· turbidity
· true colour
· solids concentration
· degree of faecal pollution

Quite often, however, hardly any information is available on the surface water quality of a raw water source meant for a rural water supply system. In such a case, the following preliminary surface water quality assessment steps can be used:

· sanitary inspection of the catchment area
· water quality analysis of the raw water

Reference [6] contains a detailed description of these two main rural water quality assessment steps. The information obtained through a sanitary inspection is more of a qualitative or descriptive nature and reflects the long-term situation of an assessed water course. The results of a water analysis present a quantitative assessment of the examined water source, and might only reflect the actual water quality at the time of sampling. Both methods complement each other, however, a thorough sanitary inspection of the catchment area often provides a more reliable and practical method of risk identification and general water quality assessment. Several water analyses have to be carried out to determine extent, duration and frequencies of water quality fluctuations. However, such information is rarely available prior to treatment facility design. Water quality analysis is often performed at a later stage to monitor only the performance of constructed treatment plants.

Detailed information on raw water quality will ease filter design. Nevertheless, accurate prediction of filter performance is hardly possible due to the complexity of filter processes.

3.2 Catchment area

An overall characterisation of the catchment area and its hydrology, along with a sanitary inspection of the area, can provide relevant information on the raw water quality. The specific characteristics of the catchment area, such as climate, hydrogeology, topography, vegetation, as well as human and animal activities greatly influence the qualitative and quantitative levels, as well as the surface water variations. Total rainfall and its annual distribution, together with soil conditions and topography, are significant criteria influencing the natural characteristics of a flowing surface water. Human activities, (deforestation, agriculture and settlements) in the catchment area will induce qualitative and quantitative changes in the natural regime of the surface water.

Turbidity level and suspended solids concentration are often correlated with the seasonal fluctuations of a river discharge. The size of the catchment area usually influences the period of high discharges; short heavy storms normally affect the discharge of small highland rivers to a greater extent than of large lowland streams. Inspection of the river bed and its embankments will certainly provide first-hand information on flow characteristics of the river. Closer inspection of the bed sediments and embankments will supply some details on the type of solids carried by the river at different periods of discharge. Information provided by the locals will focus more on frequency and length of turbidity peaks rather than on absolute turbidity levels, which can only be determined with measuring equipment.

Faecal pollution is not visible in a water sample. Even clear and pleasant water may carry harmful and disease-causing microorganisms. Population density, wastewater disposal practice and general public health condition will influence the bacteriological quality of a surface water. This quality varies widely, e.g. a highland river draining a well-protected, unpopulated area has probably a low public health risk level when used as drinking water, whereas a surface water draining wastewater from a slum area without proper sanitation facilities will certainly have an extremely high public health risk level even when used as washwater. Points of surface water pollution have to be detected by a sanitary inspection of the catchment area. Source protection is the first step in water treatment. Hence, remedial actions must be taken when such pollution points are identified. A survey of the public health condition is necessary to assess the presence of endemic diseases. Such a survey might also determine the need to improve the situation with the construction of a water supply system and, particularly, with the installation of water treatment facilities. Nevertheless, surface water remains unprotected and is, therefore, permanently exposed to human and animal faecal contamination and other man-made pollution. As a result it will generally have to be treated before it is used as drinking water.

3.3 Water quality analysis

In rural areas, the main surface water treatment objective is to improve its bacteriological quality. Drinking water should not contain any pathogenic organisms, which are often difficult to detect analytically. Therefore, the bacteriological water quality is analysed for faecal indicators. The bacteria used for such analysis are faecal coliforms, Escherichia coli and faecal streptococci present in large concentrations (10 - 1,000 million conform bacteria are found in 1 gram of faeces) in the faeces of humans and warm-blooded animals. If waters contain faecal indicators, pathogenic microorganisms are also considered to be present.

Faecal conform analyses are performed either by the membrane filtration technique or by the multiple tube method. Field test kits (e.g. manufactured by DelAgua Ltd. [7] ) are available and generally use the membrane filtration technique. The multiple tube method is often applied in central laboratories. The use of field test kits requires some basic training in test procedures, initial supervision of field analysis and, at a later stage, correct and careful handling during routine work. To obtain reliable data, the analysis of faecal coliforms should be carried out by specially trained people.

Type and amount of solid matter is the second most important aspect in surface water characterisation. Expensive and very sensitive laboratory equipment has been developed for the analysis of size, shape and concentration of solid particles. However, such equipment is hardly available nor necessary for the design of treatment facilities. Even the standard routine method of determination of the suspended solids concentration is often not possible as it requires a highly accurate scale, a vacuum pump and a drying furnace installed in an air-conditioned room. Such equipment is often unavailable or has fallen into disrepair. Hence, determination of the physical characteristics of the solids, to be separated by adequate treatment processes, requires sturdy and simple field test methods.

The physical characteristics of the solid matter can be assessed by different simple analytical methods easily applied by any treatment plant operators. These simple tests are described in Annex 1 and include the following:

- turbidity test by means of a simple tube
- determination of the settleable solids volume with a test cone
- determination of the filterability by means of a filter paper
- suspension stability test using a vessel and turbidity readings
- solid classification test using a common bottle
- particle size characteristics by sequential membrane filtration

Chemical water quality parameters should be determined on a case by case basis if water pollution levels caused by hydrogeological conditions, agriculture or industry are likely to occur. Simple field test equipment, as described in [8, 9], could be used for preliminary chemical water quality assessment. Especially, manganese, true colour and water aggressivity are important parameters which need to be examined. Furthermore, the amount of dissolved organic matter should be determined as it will greatly influence the extent of biological activity and oxygen demand in the filters.

4. Solid matter separation

Let us now examine the first treatment step; i.e., the separation of solid matter. We might be confronted with a great variety of solids as observed in our bucket filled with turbid river water. The variety, illustrated in Fig. 5, is greatly dependent on the type of surface water and whether natural purification processes can separate part of the solids or possibly generate undesirable particulate matter by organic growth. Natural purification should largely be integrated into the treatment design when determining surface water source and intake location.

Fig. 5 Solid Matter Content for Separation

Sedimentation and filtration processes are mainly applied for solid matter separation. These shall be discussed in detail in the next two sections.

Yet, let us focus first on the peculiarities of the various types of surface water and their impact on the different solids in the raw water:

· The properties of the drained catchment area and the characteristics of the surface water influence the type and concentration of solid matter in the raw water. Flow velocity and rate of erosion determine the amount of settleable solids carried by the water. Flowing and still surface waters greatly differ with respect to the encountered type of solid matter. The turbulent flow of a water course may carry coarse settleable solids, which settle in gently flowing or impounded surface water. Algae found in ponds and lakes contribute to the suspended solids concentration of the water.

· Flowing surface water is often subjected to drastic quantitative and qualitative changes. The annual rainfall distribution influences the seasonal surface water fluctuation mainly with regard to turbidity and solids concentration. Flowing surface water will usually carry settleable solids at varying concentrations during different periods of time. During the dry season, small upland rivers are generally low in turbidity, however, they can exhibit high short-term turbidity peaks during heavy rainfalls. Larger lowland rivers may be of moderate turbidity throughout the year but with relatively long periods of increased turbidity levels.

· In still surface water, amount and type of solid matter change only gradually in the course of a year. In fact, the large volume of stored water in lakes, reservoirs and ponds preconditions the water quality. Coarse inorganic particles settle at the bottom of the receiving water body, light organic solid debris tend to float on the water surface, and dissolved organic matter may be transformed by photosynthetic processes to algae and plankton. Hence, each still water source acts as a first pretreatment step since the incoming and stored water is exposed to natural purification. As a result, impounded water is generally characterised by smaller water quality fluctuations. This higher stability of the raw water quality usually facilitates treatment plant operation.

· Flowing surface water carries solids of different sizes, such as coarse sand and silt to fine clay. Due to the irregular flow conditions, the solids are unevenly distributed over the cross section of a river bend. Coarse solids drift towards the outer side of the bend whereas the fine solids are washed to the inner side of a river bend and form a silting zone. Selecting an appropriate location for the intake structure contributes to reducing the levels of fine particles which are difficult to remove in treatment processes. The intake should, therefore, be placed at the outer or erosion side of a river bend in order to reduce the abstraction of fine matter and to avoid the silting of intake works.

· Surface water can also carry coarse floating matter which may block or even damage part of the water supply installations. The undesirable material is thus retained right from the beginning either by screens or by a scumboard. Fixed screens (e.g. a coarse screen followed by a finer one) are most commonly used to avoid blockage and excessive headlosses.

In short, if surface water is used as raw water source in a water supply scheme, preference should be given to still water provided excess amounts of algae or colour do not create special treatment problems. Natural purification processes reduce in particular the solid matter concentration by sedimentation, and smaller water quality variations often decrease and simplify the required degree of treatment. Flowing surface water frequently exhibits rapid water quality changes which render water treatment more difficult.

4.1 Sedimentation

Small pebbles or sand particles will undoubtedly settle in still water. This process, called sedimentation, is dependent on the physical properties of the solid matter and water. The settling velocity is influenced by density, size and shape of the particle, as well as by viscosity and hydraulic conditions of the water. Stilling basins and sedimentation tanks are quite efficient in removing relatively heavy and coarse solids, such as sand and silt particles. Inorganic matter larger than about 20 mm (0.02 mm) can usually be removed by plain sedimentation and without the use of chemicals.

Stilling basins can often be installed in small rivers. As shown in Fig. 6, a small weir is placed in the water course to raise the water depth and to reduce the flow velocity. Easily settleable matter can now be separated in the backwater of the weir equipped with a small gate to ease periodic removal of the settled material. The intake of the water supply scheme may be integrated into the sidewall of the weir, in a zone with sufficient water current to achieve removal of floating matter retained by the scum-board.

Sedimentation tanks are either rectangular, square, or circular in shape. The tanks are operated on a continuous or intermittent basis. In continuously operated tanks, the flow direction is either horizontal or vertical. In circular tanks, the flow pattern is complex, and the conditions are unstable in vertically operated tanks. Therefore, rectangular tanks operated on a horizontal flow and continuous basis are recommended for rural water supply schemes.

Fig. 6 Layout and Design of a Stilling Basin

Fig. 7 Layout and Design of a Sedimentation Tank

Sedimentation tanks separate finer solids, such as silt, clay and part of the suspended solids. The raw and turbid water enters on one side of the tank and is evenly distributed over the entire tank cross section. The solids then settle under laminar flow conditions to the tank bottom, and the clarified water is abstracted uniformly over the full width on the opposite side of the tank. In order for the particles to separate, each solid particle has to overcome a settling distance equal to the tank's depth, e.g. around 1 to 3 m. The accumulated sludge is periodically removed from the tank bottom. The solids removal efficiency of a sedimentation tank depends mainly on the hydraulic surface load, tank depth, and retention time. Some general design values fore sedimentation tank are given in Fig. 7, however, they should be chosen according to the settling characteristics of the solids. These can be determined in a sedimentation test using a transparent test tube; for additional information consult Annex 1. The recorded time necessary to attain a certain clarification level in the test has to be multiplied by a factor three to allow for unfavourable flow conditions in a full-sized tank. Low surface loads should be applied with raw water of poor settling properties, and in small plants with variable operating conditions.

Even properly designed and operated sedimentation tanks will separate only part of the suspended solids. With the help of coagulants, such as alum or iron salts, suspensions can be destabilised. The small particles lose their repulsive force, cluster together and coalesce to larger floes of improved settling characteristics. Coagulants are extensively used in conventional water treatment systems. However, the flocculation/sedimentation process is already an advanced treatment technique requiring qualified personnel and well-equipped facilities; both scarce in rural areas of developing countries. Chemicals often have to be imported from abroad and paid for in foreign currency. Since transport problems are pertinent to many developing countries, the adequate and reliable supply of chemicals to remote treatment plants is yet another stumbling-block. Dosage is also an art in itself, as it must be adapted to the varying raw water quality and thus requires professional water quality monitoring. Accurate and sensitive dosing instruments are attacked by the corrosive action of the chemicals. Chemical water treatment calls for skilled personnel trained in water quality monitoring, dosage adjustment, as well as in maintenance and repairs of dosing equipment. Finally, use of chemicals often greatly increases operating costs. In practice, rural water supplies often face considerable problems with chemical water treatment. A reliable and successful application of chemical flocculation is, therefore, rather unrealistic for many small water supply schemes. The chemical coagulation and sedimentation process applied in conventional water treatment schemes for separation of suspended solids and colloidal matter will generally fail in rural water supply schemes and is therefore not recommended.

To conclude, it can be said that stilling basins and sedimentation tanks are quite efficient in removing coarse and easily settleable solids. They are used as preliminary treatment step, especially to treat raw water drawn from running water courses containing high solids concentrations. In rural water supply schemes, use of chemicals to enhance sedimentation by flocculation is difficult and, therefore, quite unreliable.

4.2 Roughing filtration

The water quality of contaminated surface water can be improved significantly when filtered through gravel and sand layers. Therefore, favourable hydrogeological conditions allow polluted and turbid river water to be drawn as clear and safe groundwater from a shallow well located next to a river. However, local soils are quite often impervious for lack of gravel and sand layers. Nevertheless, why should nature's excellent purification capacity be ignored just because of unfavourable hydrogeological conditions at the site of a new water supply scheme? Let us then copy nature and construct an artificial aquifer by filling a sedimentation tank with gravel. As illustrated in Fig. 8, the solids removal efficiency of such a tank will drastically increase due to the greatly reduced settling distance in the gravel material. In other words, the fine solids crossing an ordinary sedimentation tank have to overcome a vertical settling distance of 1 to 3 m before coming into contact with the tank bottom. The same solids flowing through a filter will fortunately touch the gravel surface already after a few millimetres. Thus, filtration becomes a more effective process for solids removal since the settling distance is drastically reduced by the filter material. Presence of a small pore system and large internal filter surface area enhances sedimentation and adsorption, as well as chemical and biological activities.

Design and application of prefilters vary considerably. The different filter types are classified according to their location within the water supply scheme, their main application and flow direction. Intake and dynamic filters, which often form part of the water intake structure, differ from actual roughing filters which are generally located at the water treatment plant. As illustrated in Fig. 9, roughing filters are further subdivided into down, up and horizontal-flow filters. Finally, vertical-flow filters can be classified according to the manner in which the gravel layers are installed. The different gravel fractions of roughing filters "in series" are installed in separate compartments, while those of roughing filters "in layers" are placed on top of each other in the same compartment.

Fig. 8 Particle Removal in a Sedimentation Tank and a Roughing Filter

Roughing filters usually consist of differently sized filter material decreasing successively in size in the direction of flow. The bulk of the solids is separated by the coarse filter medium located next to the filter inlet. The subsequent medium and fine filter media further reduce the suspended solids concentration. The filter medium of a roughing filter is composed of relatively coarse (rough) material ranging from about 25 to 4 mm in size. Gravel is generally used as filter material. Significant solids removal efficiencies are only achieved under laminar flow conditions since sedimentation is the predominant process in roughing filtration. Therefore, roughing filters are operated at small hydraulic loads, which have been defined as flow rate Q divided by the filter area A perpendicular to the direction of flow. Filtration velocity, synonymous with hydraulic load, usually ranges between 0.3 and 1.5 m/h. The coarse filter material and the small hydraulic load limit filter resistance to a few centimetres.

Fig. 9 Classification of Prefilters

Filter cleaning is carried out manually and hydraulically depending on the pattern of solids accumulated in the filter. Intake and dynamic filters separate the solids mainly at the inlet zone of the filter and, thus, act as surface filters. These filters are therefore manually cleaned by scouring the top of the filter bed with a shovel or rake. Compared to intake and dynamic filters, roughing filters act as space filters on account of the deep penetration of the solids into the filter medium. The accumulated solid matter is periodically flushed out of roughing filters by hydraulic filter cleaning. If necessary, these filters can be cleaned manually by excavating the filter material from the filter compartment, washing and refilling it into the filter boxes.

Sedimentation is the main process in roughing filtration. It is responsible for solids separation from the water as observed in laboratory tests conducted with roughing filters [10, 11, 12, 13,14]. The filter acts as multi-storage sedimentation basin since it provides a large surface area to accumulate the settled matter. As shown in Fig. 10 and illustrated in Photo 3, the deposits are retained on top of the collectors where they grow to dome-shaped aggregates with advanced filtration time. Part of the small heaps drift to the filter bottom when the loosely accumulated aggregates become unstable. In horizontal-flow roughing filters, this drift regenerates filter efficiency of the upper gravel layers and allows accumulation of a considerable amount of retained material at the filter bottom. Depending on the organic characteristics of the raw water, other processes such as biological oxidation or adsorption of solid matter at the slimy filter surface may occur. Under these circumstances, enhanced aggregation and consolidation of deposits have been reported [12]. This poses inherent difficulties during hydraulic cleaning and filter regeneration.

Fig. 10 Solid Removal in a Horizontal-flow Roughing Filter

Filter regeneration can be enhanced by filter drainage. The loosely accumulated aggregates collapse and are washed down to the filter bottom if the water table in the filter is lowered. Part of the accumulated solids can be flushed out of the filter with high filter drainage rates and adequate installations.

Photo 3 Accumulation of Kaolin in a Horizontal-flow Roughing Filter after 24h, 100h, and 300h of Filter Operation

5. Bacteriological water quality improvement

The water in our bucket is now clear but still unsafe for consumption. The turbid river water has changed its appearance as the solid matter has been separated by the pretreatment processes discussed in the foregoing chapter. The water has lost its brownish tinge and turned into a clear and pleasantly looking liquid. However, the water is still not as pleasant and safe as it looks. As schematically shown in Fig. 11, disease-causing pathogenic microorganisms are usually not visible to the naked eye of the consumer who could get a severe attack of diarrhoea a few hours after drinking this water. Hence, the pretreated water still needs further treatment for final removal or inactivation of pathogens. Slow sand filtration and chlorination are the two most commonly applied treatment processes for bacteriological water quality improvement.

Fig. 11 Microorganisms for Separation

5.1 Slow sand filtration

Slow sand filtration plays a key role in rural water treatment. Design and application of this treatment process is well-documented in the available literature [15, 16, 17]. Since slow sand filters reduce the number of microorganisms present in the water, they improve the bacteriological water quality. In addition, fine organic and inorganic matter is separated, and the organic compounds dissolved in the water are oxidised. Since the effluent of a well-designed and operated slow sand filter is virtually free from pathogenic microorganisms, water treated by such a slow sand filter is safe for consumption. Furthermore, a comparative evaluation [18] of slow sand and rapid sand filter efficiencies revealed that slow sand filters are more efficient in the removal of several commonly occurring pesticides. In contrast, they were found to be poorer than coagulant-assisted rapid sand filters for the removal of dissolved organic carbon and organic colour. However, slow sand filtration is one of the most efficient processes for the production of hygienically safe drinking water with a possibly small bacterial regrowth.

The slow sand filter technology copies nature. The sand layers of aquifers convert unsafe surface water into good quality drinking water. Especially the harmful bacteria, viruses, protozoa, eggs, and worms are most effectively removed by physical and biochemical processes to a level which no longer endangers human health. These natural purification processes are also used by the slow sand filters - a technology which was introduced last century. At that time, Europe was struck by cholera epidemics, which forced the waterworks to take quick action. The advantages of slow sand filtration were then discovered. This water treatment technique proved to be efficient against water-borne diseases and, in combination with other public health improvements, these epidemics were eradicated in Europe. Numerous water supplies in industrialised countries are still using slow sand filters. Thames Water supplies for instance two thirds of London's population with slow sand filter treated surface water drawn from the River Thames which carries a very high percentage of sewage effluent from upstream settlements during drought years. This is a tribute to the efficiency and reliability of the slow sand filter technology.

The layout of slow sand filters is simple and straightforward. As shown in Fig. 12, a slow sand filter contains an open box filled with a sand layer of a depth of about 0.8 to 1.0 meter. The upper part of the filter box is filled with water flowing by gravity through the sand bed. The filtered water is then collected by an underdrain system and conveyed to the clear water tank. The well-graded sand of the filter bed is relatively fine; i.e., its effective size ranges between 0.15 and 0.30 millimetre, but recent field experience revealed that also somewhat coarser sand can be used [4].

Fig. 12 Layout and Design of a Slow Sand Filter

Slow sand filter operation is easy and reliable. Slow sand filters are usually operated at 0.1 to 0.2 m/h filtration rates. Consequently, an area of 1 m² sand produces about 2.5 to 5 m³ of water per day. The flow rate is preferably controlled at the filter inlet, and the water level is maintained at a minimum level above the sand bed by means of a weir or effluent pipe located at the filter outlet. Effective biological treatment can only be achieved if a reasonably steady throughput is maintained. Therefore, a 24-hour operation is recommended as it makes maximum use of the available filter installations. The initial filter resistance of a clean sand filter ranges between 0.20 and 0.30 meter. The headloss gradually increases with progressive filtration time. The sand filter has to be cleaned when filter resistance amounts to about 1 meter.

Slow sand filters act mainly as surface filters. The water quality changes at the surface of the sand bed, in the so-called "Schmutzdecke" and, to a lesser extent, in the first 20 to 30 centimetres of the send bed. A thin layer on the surface of the sand bed, formed by retained organic and inorganic matter, and a large variety of biologically active microorganisms, are responsible for the physical, chemical, and biological improvement of the water. This thin biological layer must first develop in a new slow sand filter. The initial ripening period normally requires two to four weeks. Cleaned filters will regain their full biological activity within two to four days, provided shut down time for filter cleaning is short; i.e., not more than 6 - 12 hours.

Filter cleaning must be carried out once the supernatant water has reached its maximum permissible level; i.e., when maximum filter resistance of about 1 meter is attained for the designed filtration rate. Filter cleaning starts with drainage of the supernatant water and dewatering of the top part of the sand bed. Subsequently, the biological skin and 1 to 2 centimetres of sand are removed from the sand bed as shown in Photo 4. Resanding is possibly performed after removal of the top sand layer. Thereupon, filter operation is immediately restarted to avoid disrupting biological filter activity more than is necessary. The filter bed is refilled with water introduced via the under-drainage system. This drives the air out of the pores of the sand and completely saturates the filter bed. Normal operation is then reassumed by opening the inlet valve and adjusting the filtration rate.

Well-operated slow sand filters should at least achieve more than 1 to 3 months of filter runs. The term "filter run" is defined as the time between two subsequent filter cleanings. In order to realise such long filter runs, slow sand filters have to be supplied with relatively clear water. Reasonable filter operation can only be expected with inlet water turbidities below 20 to 30 NTU. Higher turbidities, with consequently higher solids concentrations, will rapidly clog the sand surface and interfere with the biological processes. Hence, it is strongly recommended that surface water is pretreated prior to slow sand filtration.

Photo 4 Cleaning of a Slow Sand Filter

Design deficiencies will cause problems to slow sand filters. In the past, several slow sand filter plants in developing countries have faced operational problems or had to be closed down. Serious design faults, inadequate operation and poor water quality supplied to the slow sand filters are the main reasons for the problems and failures experienced. As illustrated in Fig. 13, a lack of flow control equipment, inadequate pipe installations, a soiled and poorly graded sand which does not conform to the recommended size, or missing water level control systems, are the most common design errors encountered. Random filter operation under variable and often too high filtration rates by insufficiently trained caretakers, are generally the causes of inadequate filter efficiency.

Poor quality raw water, inadequately pretreated, also contributes to poor slow sand filter performance. Frequently, slow sand filters are directly fed with raw water or are often combined with inefficient or inappropriate pretreatment processes. Slow sand filters usually face serious operational problems when chemical flocculation and sedimentation are used as pretreatment. The local caretaker might not be able to control flocculation as it is an unstable pretreatment process difficult to operate. Light floes often get washed onto the slow sand filters, or a lack of chemicals greatly reduces the solid removal efficiency of the sedimentation tank. Premature, rapid filter clogging and frequent filter cleaning are the resulting consequences. Therefore, efficient pretreatment of the surface water, such as for instance by roughing filters, is necessary to avoid serious operational difficulties with slow sand filters. Small slow sand filter units receiving raw water of moderate turbidity can also be protected by layers of non-woven synthetic filter fabrics [19,20] or by a layer of gravel [21 ] installed on top of the sand bed.

Fig. 13 Common Design Faults of Slow Sand Filters

In summary, slow sand filtration can thus be regarded as a safe, stable, simple and reliable treatment process. Filter construction makes extensive use of local material and skills. Filter operation neither requires sophisticated mechanical parts nor the use of chemicals. Construction, operation and maintenance of the filters are easy and require only limited skills. However, adequate filter operation is only possible with raw water of low turbidity; i.e., virtually free of solid matter. Pretreatment of surface water is therefore necessary. In combination with adequate pretreatment methods, slow sand filtration is considered a most appropriate water treatment technology for developing countries.

5.2 Chlorination

Chlorination aims at destroying or, at least, inactivating harmful microorganisms, such as pathogenic bacteria, viruses, and cysts present in the water. Chlorine is a strong oxidant, which not only reacts with the enzymes vital to the metabolic processes of living cells, but is also responsible for other chemical reactions. Dissolved organic matter, for instance, depletes by fast chemical reaction the available chlorine that will then be unavailable for water disinfection. Or chlorine reacts with nitrogen to form the more stable chloramines often purposely generated by the addition of ammonia to the water so as to cope with any type of pollution problems in the distribution system.

The advantages why chlorination is widely used in water treatment practice are the following:

- Chlorine is a strong disinfectant when applied to low water turbidity with a small dissolved organic content.

- Residual chlorine content is extremely simple to determine by calorimeters, which is not the case for other disinfection processes such as ozone or UV radiation.

- Since chlorination installations are relatively small, they do not require large civil engineering structures and their investment costs are relatively low.

- Chlorine is often applied as a safeguard (especially in the form of stable chloramines) against secondary water pollution. Although small quantities of chlorine may deal with minor contaminations resulting from incorrect water handling at household level, they will never be able to combat heavily contaminated water caused by cross-connections or wastewater infiltration in intermittently operated water supply systems.

Numerous disadvantages of chlorination, however, question the application of this water treatment process in rural water supply schemes. Chlorination is associated with the following problems:

- Chlorination requires a reliable water treatment system. It is neither applicable to turbid water nor to water of high organic content.

- With inadequately pretreated water, chlorine forms by-products (e.g. trihalomethanes) that are considered carcinogenic.

- Chlorine is usually an unstable and corrosive chemical that loses its disinfecting power during storage, and attacks the delicate installations in the dosing room.

- Dosing equipment and chemicals must often be imported, which leads to a foreign currency demand and high operating and maintenance costs.

- Consumers frequently refuse to drink chlorinated water for reasons of taste and odour.

Accurate chlorine dosage is essential to attain efficient disinfection. Only partial disinfection is achieved with chlorine dosages lower than the chlorine demand of the water. Water containing a too high chlorine concentration might not be accepted by the consumers, as chlorinated water has a distinct odour. A strong smell develops when chlorine reacts with ammonia to form chloramines. People often reject chlorinated water even when chlorine is carefully handled and dosed at low concentrations.

Chlorine is available in gaseous, solid, and liquid form. Chlorine gas is extremely toxic, difficult to handle and, therefore, usually inappropriate for rural water supplies. Chlorinated lime, commonly known as "bleaching powder", calcium hypochlorite powder, or sodium hypochlorite solution, also called "Javel water", are used as chlorine derivatives. Since a chlorine solution is preferably added to the water, chlorinated lime and calcium hypochlorite should be dissolved in the water to a stock solution containing usually 1 to 3 percent active chlorine. Chlorine solutions require careful preparation; i.e., it is extremely dangerous to spill water on dry hypochlorite. Fig. 14 summarises the different chlorine applications. Please note that adequate disinfection is attained not only with a sufficient chlorine concentration C (mg/l), but also requires an appropriate contact time T (min) as the inactivation of mircroorganisms is dependent on the product C x X.

A constant rate of chlorine solution is added to the water by dosing devices. The relatively small doses of chlorine call for accurate dosing equipment. These are, however, exposed to the corrosive action of the chlorine and often get damaged. The dose has to be adjusted to the chlorine demand of the water to be disinfected. In practice, a limited chlorine dosage adjustment is possible, e.g. on a day-to-day basis. Reliable water treatment prior to chlorination is consequently necessary. Adequate water disinfection is generally feasible only with water virtually free of solid and organic matter.

Fig. 14 Application of Chlorine

A reliable supply of chlorine is often difficult to obtain. Chlorine must be purchased on a regular basis as its unstable nature does not allow lengthy storage. Chlorine generally has to be imported and, thus, requires foreign currency; an often scarce amenity in developing countries. In addition, these countries usually face other difficulties, such as communication and transportation problems. Finally, chemical water treatment requires skilled personnel often unavailable in rural areas. All these aspects are a stumbling block to a reliable and efficient chlorine application and, more generally, to the use of chemicals in rural water supply schemes in developing countries. This observation is endorsed by numerous cases of malfunctioning or abandoned chemical water treatment installations.

Since conventional disinfection methods are generally unsuccessful in small rural water supply schemes, simple, robust and easily maintained low-cost, reliable methods of disinfection are thus necessary.

New water disinfection techniques have already been developed and field-tested [22]. The use of iodine instead of chlorine bound onto resins housed in a cartridge is but one alternative. By placing this cartridge in the water, the microorganisms are rendered non-viable by oxidative reaction with iodine. Compared to chlorine, iodine does not react so easily with organic compounds in the water. However, this disinfection method requires further development before it can be used on a larger scale, especially with regard to fixing the iodine on an adequate supporting material. Furthermore, dosing of iodine must be well-controlled - at high dosage, it may pose a health hazard, particularly to pregnant women.

The use of an electrolytic cell which produces an oxidising gas when an electrical current is passed through a saturated sodium-chloride solution is a second water disinfection alternative. The Moggod method ("mixed oxidant gas generated on demand") requires salt, water and electric current to produce a strong oxidising gas. This method, however, is sensitive to the use of normal salt as it creates substantial problems when associated with calcium and magnesium deposits on the membrane. Further investigations on the nature of the gas produced and on operational aspects regarding the use of low-quality salt, are necessary before this disinfection method can be recommended for wider use.

The described methods suggest different processes rather than real disinfection alternatives to replace or produce chlorine at the site. Other processes (e.g. the MIOX method) are being developed and field-tested. A comprehensive description of chlorination and alternative disinfection methods is beyond the scope of this manual, however, reference is made to the relevant literature [23, 24, 25].

To conclude it can be said that an efficient and reliable disinfection with chlorine requires pretreated water virtually free of solid and organic matter. The use of chlorine in rural water supply schemes often creates enormous problems and is, therefore, frequently bound to fail as documented by numerous treatment plants. Furthermore, the rural population often rejects chlorinated water. Thorough technical, institutional and sociocultural investigations are necessary before chlorination is introduced in rural water treatment.

6. Layout of a water supply scheme

6.1 General considerations

From the technical point of view, the following three main questions have to be answered during the planning phase of a water supply scheme:

· which raw water source should be used for the water supply scheme?
· if treatment is necessary, what type of treatment scheme should be favoured?
· how much water should be distributed to the consumers, and at what service level?

Source selection is a very basic decision entailing numerous consequences for the future water supply scheme. The different local water sources have to be evaluated with respect to their quantity, quality and accessibility. The future water demand must be covered by the selected source with the best possible water quality, and located as close as possible to the supply area.

Fig. 15 Layout Possibilities of water supply Schemes Using Surface Water

Since water treatment is usually the most difficult element in any water supply scheme, it should be avoided whenever possible. The general statement that no treatment is the best treatment especially applies to rural water supply schemes which generally exhibit a poor infrastructural and institutional framework to adequately maintain water treatment facilities. The use of better water quality sources is, therefore, an alternative which will always have to be taken into serious consideration. If no other alternative is available, rural water treatment must concentrate on improving the bacteriological water quality by locally sustainable treatment processes.

Water distribution systems depend on the type of water source used, on the topography, and on the provided supply service level. Individual water supplies, e.g. rainwater harvesting and shallow groundwater wells equipped with hand pumps usually do not need piped supply systems. Treated surface water, however, is normally distributed by a piped system. A suitable topography often allows the installation of a gravity system which will improve reliability and supply continuity. Since pumped water supply schemes depend on the reliable supply of energy and spare parts, they are very susceptible to temporary standstills. Finally, the service level of water supply strongly governs water demand. Water usage increases drastically with the provided service level, e.g. public standpost, yard connection, multiple tap house connection. Water supply is always interlinked with wastewater disposal. The health situation of a community supplied with treated water does not necessarily improve, especially if public health and wastewater disposal issues are neglected. The main components necessary to significantly improve the public health situation of a community are therefore a reliable and safe water supply, an adequate waste disposal system and a comprehensive hygiene education programme.

As schematised in Fig. 15, surface water has to be collected, treated and stored before it reaches the consumer. These activities can be met by different water supply layout options. Figs. 15 and 16 only illustrate some arrangement examples.

6.2 Hydraulic profile

Selection of the hydraulic profile is a basic criteria when planning a water supply scheme. First choice must be given to gravity supply systems since they guarantee reliable operation at low running costs. Schemes, which integrate the use of handpumps, are given second choice. The installation of mechanically driven pumps should be chosen as last option and only applied in special cases where a reliable and affordable energy supply is guaranteed, including the infrastructure for pump maintenance and repair work. Hydraulic rams making use of the potential energy of a large water volume to pump a small fraction of this water volume to a higher level [26] may be an appropriate option where surface water gravity is available and water volume abundant. Under special local conditions, collection and pretreatment of the raw water may be combined in a single installation such as infiltration galleries.

Fig. 16 General Layout of Pumped Water Supply Schemes

Water treatment plants should, whenever possible, be operated by gravity and with a free water table to minimise water pressure on the structures. The total headloss through the treatment plant will amount to 2 or 3 m. In general, any type of water lifting, except through handpumps, should be avoided as the supply of energy and sophisticated spare parts is generally unreliable. If water lifting is absolutely necessary for topographical reasons, the number of pumping steps must be limited. As illustrated in Fig. 16, a one-stage pumping scheme should be chosen for raw water to be pumped to an elevated site where the treatment plant and reservoir are located. Such a one-stage pumping scheme has greater advantages over a two-stage scheme as it increases its reliability by a factor of 2. Moreover, the risk of flooding in lowland areas can often not be excluded entirely. Protecting a high-lift pumping station against floods is easier than a full-sized treatment plant. However, a two-stage pumping system is unavoidable for a piped supply on a flat area devoid of natural elevation and in case of serious raw water quality fluctuations, e.g. heavy sediment loads during the monsoon. In such a situation, installation of a low lift raw water pump is recommended. It may consist of an irrigation unit of low efficiency but of simple repair to limit high lift pumping for treated water and protect impellers and seals from damage. Hence, high lift pumps should be used for treated water or raw water pumped from infiltration galleries or similar intake systems.

Fig. 17 Treatment of Surface Water

6.3 Treatment steps

As discussed in Chapter 2, surface water has to undergo a step-by-step treatment. Coarse solids and impurities are first removed by pretreatment, whereas the remaining small particles and microorganisms are separated by the ultimate treatment step. Under special local conditions, raw water collection and pretreatment may be combined in a single installation, such as intake or dynamic filters or, alternatively, by infiltration galleries. Fig. 17 illustrates different schemes for surface water treatment. The required water treatment scheme is mainly dependent on the degree of faecal pollution, characteristics of the raw water turbidity and on the available tvpe of surface water.

6.3.1 Removal of Coarse Material

Separation of coarse solids from the water is preferably carried out by a high-load sedimentation tank (grit chamber) or by a plain sedimentation tank, since sludge removal from such tanks is less troublesome than from roughing filters. Simple sedimentation tanks can be designed according to the layout and guidelines given in Fig. 7, or constructed as earth basins as illustrated in Fig. 18.

Use of one sedimentation tank should be sufficient for a small-scale water supply scheme. The accumulated sludge can be removed during periods of low silt load. A bypass is required to maintain operation of the treatment plant during cleaning periods. In order not to interfere too much with normal operation of larger water treatment plants, two or more sedimentation tanks operating in parallel should be provided to allow cleaning, maintenance and repair of one tank.

6.3.2 Aeration

The water's dissolved oxygen content plays a key role in the biology of the slow sand filtration process. The activity of the aerobic biomass decreases considerably if the oxygen concentration of the water falls below 0.5 mg/l. Furthermore, nitrification of ammonia is associated with a significant consumption of oxygen, e.g. 1 mg NH₄-N/I requires 4.5 mg O₂/l. Hence, an adequate oxygen content in the water to be filtered is of prime importance. Physical processes are the main mechanisms in roughing filtration. However, biochemical reactions might also occur in the prefilters, especially if the raw water contains high organic loads.

Fig. 18 Design of an Earth Basin as Sedimentation Tank

Fig. 19 Layout and Design of an Aeration Cascade

Since turbulent surface waters are generally well oxygenated, they do not require additional aeration. Still water, however, can exhibit low oxygen contents, especially when drawn from the bottom of polluted surface water reservoirs. Multi-level drawoffs are recommended as intake structures for stratified water bodies to allow abstraction of best raw water quality. However, stagnant raw surface waters are preferably aerated.

Cascades are simple but efficient aeration devices. A submerged cascade aerator, as illustrated in Fig. 19, should be installed in gravity systems with sufficient hydraulic head. The cascade should preferably precede filters to meet the possible oxygen demand. The different weirs, used for flow control, are an additional source of oxygen supply.

6.3.3 Roughing Filtration as Pretreatment

Roughing filtration mainly separates the fine solids which are not retained by the preceding sedimentation tank. The effluent of roughing filters should not contain more than 2-5 mg/l solid matter to comply with the requirements of the raw water quality for slow sand filters.

Coarse gravel filters mainly improve the physical water quality as they remove suspended solids and reduce turbidity. However, a bacteriological water improvement can also be expected as bacteria and viruses are solids too, ranging in size between about 10 - 0.2 mm and 0.4 - 0.002 mm respectively. Furthermore, according to the specific literature [27], these organisms get frequently attached by electrostatic force to the surface of other solids in the water. Hence, a removal of the solids also means a reduction of pathogens (disease-causing microorganisms). The efficiency of roughing filtration in microorganism reduction may be in the same order of magnitude as that for suspended solids, e.g. an inlet concentration of 10 - 100 mg/l can be reduced by a roughing filter to about 1 - 3 mg/l. The bacteriological water quality improvement could amount to about 60 - 99%, or the microorganisms are reduced to about 1 - 2 log. Larger sized pathogens (eggs, worms) are removed to an even greater extent.

Roughing filters are used as pretreatment step prior to slow sand filters. Slow sand filtration may not be necessary if the bacteriological contamination of the water to be treated is absent or small, particularly in surface waters draining an unpopulated catchment area, or where controlled sanitation prevents water contamination by human waste. However, physical improvement of the water may be required with permanent or periodic high silt loads in the surface water. Excessive amounts of solids in the water lead to the silting up of pipes and reservoirs. For technical reasons, roughing filtration may therefore be used without slow sand filtration if the raw water originates from a well-protected catchment area and if it is of bacteriologically minor contamination; i.e., in the order of less than 20 50 E. coli/100 ml.

For operational reasons, at least two roughing filter units are generally required in a treatment plant. Since manual cleaning and maintenance may take some time, the remaining roughing filtration unit(s) will have to operate at higher hydraulic loads. A single prefilter unit may be appropriate in small water supply schemes treating water of periodically low turbidity.

6.3.4 Slow Sand Filtration as Main Treatment

The substantial reduction of bacteria, cysts and viruses by the slow sand filters is important for public health. Slow sand filters also remove the finest impurities found in the water. For this reason they are placed at the end of the treatment line. The filters act as strainers, since the small suspended solids are retained at the top of the filter. However, the biological activities of the slow sand filter are more important than the physical processes. Dissolved and unstable solid organic matter, causing oxygen depletion or even turning to fouling processes during the absence of oxygen, is oxidised by the filter biology to stable inorganic products. The biological layer on top of the filter bed, the so-called "Schmutzdecke", is responsible for oxidation of the organics and for the removal of the pathogens. A slow sand filter will produce hygienically safe water once this layer is developed.

Unlike roughing filters, the time for slow sand filter cleaning is determined by maximum available headloss level, and not by deterioration of effluent quality. This offers some advantages as recording of a hydraulic criteria is easier than measuring water quality parameters.

Further information on slow sand filtration is summarised in Annex 3, and detailed information on design and construction of slow sand filters is provided by different technical manuals [15, 16,17] and proceedings [28, 29, 30].

6.3.5 Water Disinfection

Water from a slow sand filter with a well-developed biological layer is hygienic and safe for consumption. Any further treatment, such as disinfection is, therefore, not necessary. As documented by numerous examples in many developing countries, provision of a reliable chlorine disinfection system in small rural water supply schemes is often not practicable. A regular supply of mostly imported chemicals, and accurate dosage of the disinfectant, are the two main practical problems encountered.

However, as regards disinfection, one has to differentiate between small (rural) and large (urban) water supply schemes. Large distribution systems with often illegal connections present a risk of recontamination, especially if the supply of water is intermittent. In large urban water supply schemes, final water chlorination is recommended as a safeguard. However, residual chlorine will be too low and contact too short to deal with serious contamination introduced by infiltration of highly contaminated shallow groundwater in intermittently operated water supply systems. In rural water supply system, implementation of a general health education programme with special emphasis on correct water handling is a more effective measure than preventive d is infection.

An example of a water treatment plant operating without any foreign chemicals or energy inputs is illustrated in Fig. 20. The pipe layout of this 60 m³/d capacity plant provides the necessary flexibility to run the plant uninterruptedly also during the required cleaning and maintenance activities.

6.4 Water distribution

6.4.1 Water Storage

To make full use of the treatment capacity and to avoid interference of the treatment process by intermittent operation, water treatment installations should preferably be operated uninterruptedly on a 24-hour basis. Particularly slow sand filters should be operated continuously to provide the biological layer with a permanent supply of nutrients and oxygen. Roughing filters are less sensitive to operational interruptions, although careful restarting of filtration should be observed in order not to resuspend the solids accumulated in the filler. Water supply schemes, operated entirely by gravity, can easily handle a 24-hour operation. However, pump operation is often reduced to 6 -16 hours a day in water supply systems requiring raw water lifting. In pumped schemes, construction of a raw water tank may offer an economically and technically sound option since it enables continuous operation of the treatment plant and also acts as presedimendation tank. Fig. 21 illustrates possible installations for a controlled and constant raw water supply of the treatment plant.

Water storage capacity must be provided to compensate for daily water demand fluctuations. In rural water supply schemes, daily water consumption occurs more or less in the morning and evening hours. Therefore, a storage volume of at least 30 to 50% of the daily treatment capacity should be provided to compensate for the uneven daily water demand distribution.

6.4.2 Distribution System

Water accessibility and not so much water quality is the most important criteria for the consumer as his main concern is the walking distance between his home and the water point. Consequently, treated or better quality water has to be brought nearer to the homes than the traditional water sources. Treated river water as a new water source is likely for instance to be more readily accepted if the original walking distance to the river can be reduced substantially by the installation of a water supply system.

A water distribution system will therefore have to be constructed. The service level of a piped system is dependent on the economic situation construction costs of a distribution system normally amount to 50 - 70% of the total investment costs of a water supply scheme, including a water treatment plant. Gravity schemes should be installed whenever possible. In many instances, however, topography is unfavourable and differences in altitude must be overcome by water lifting. Pumps require, however, relatively high investment and operating costs, spare parts and, particularly, energy, an aspect which will, in future, gain increased importance. In rural water supply schemes, pumped systems should therefore be introduced only after careful consideration and in exceptional cases.

Fig. 20 Example of a Water Treatment Layout

Fig. 15 on page Vl-1 illustrates different hydraulic layout possibilities. On the raw water side, the water flows by gravity directly to the treatment plant or, if pumped, preferably first to a raw water balancing tank. After passing through the treatment plant it is stored in a reservoir and later distributed to the consumers by a piped gravity scheme close to the houses. In a semi-piped scheme, the water flows by gravity through the treatment plant into the reservoir equipped with handpumps, or, as an extended alternative, the reservoir is connected to a system of cistern located between treatment plant and village. Treated water is now supplied by gravity to these cisterns equipped with handpumps. Each cistern acts as reservoir and water point.

Such distribution systems may increase sustainability and reliability of a water supply as the energy supplied by the consumers when operating the handpump keeps the water supply system running at low operating costs and at village maintenance level. The proposed system of storage tanks equipped with handpumps can best control excess water usage, prevent contamination and avoid wastewater disposal problems.

However, the consumer may require higher service levels than the aforementioned "handpump option". On the one hand, higher service levels run parallel with increased water consumption and wastewater disposal problems, on the other, collection of water charges may become easier if the distribution level is shifted from public to individual supply.

Concerning the different service levels, the following per capita daily water demand values are generally used:

The effective q values for the supply with public handpumps or standpipes are greatly influenced by transport distance, ranging from a few dozen to 300 and more metres. For yard and house connections, water use will be influenced by the level and manner in which the water charges are levied (e.g. as a monthly lump sum or on an effectively used water volume basis recorded by water metres). Furthermore, use of drinking water for backyard garden irrigation leads to an enormous water demand and should therefore be prohibited.

Fig. 21 Raw Water Supply and Flow Control

7. Roughing filtration application

7.1 Historic use

The natural water treatment potential was adopted long before chemical water treatment methods, such as chlorination and flocculation, were discovered and applied. Gravel and sand used as filter media are key components in natural treatment processes. Although sand was able to maintain its important role since the development of the first slow sand filters at the beginning of the last century, the use of roughing filters was successively replaced by chemical water treatment processes. A comprehensive review of gravel filter application is far beyond the scope of this manual. However, a few examples presented hereafter will document that the roughing filter technology is an old water treatment process used in the past and rediscovered in recent years.

Numerous castles and forts were constructed in Europe during the Middle Ages. They were often located at strategically important points, difficult to conquer and also to supply with water. Ingenious water supply installations were therefore constructed. A good example is the former castle of Hohentrins located on top of a steep rocky reef in the Swiss Alpine valley of the river Rhine. During periods of war, the people who sought protection in this castle depended on rainwater collected in the yard and stored in a cistern. In this extensively used area, it was, however, not possible to avoid water pollution caused by man and animal. Therefore, in order to treat the water, a gravel pack was installed around the inlet of the cistern. This is probably one of the first roughing filters used to treat surface water [31].

In 1804, John Gibb constructed the first water filtration plant for a public water supply at Paisley in Scotland. ln order to pretreat the muddy river water, John Gibb designed and constructed an intake filter described as follows:

"Water from the River Cart flowed to a pump well through a roughing filter about 75 feet long, composed of "chipped" freestone, of smaller size near the well than at the upper end. This stone was placed in a trench about eight feet wide and four feet deep, covered with '´Russian mats" over which the ground was levelled." (cited from [1]).

The pretreated raw water was then lifted by a steam engine-driven pump to a place 16 feet higher than the river from where it flowed by gravity to the water treatment plant. This installation consisted of three concentric rings each six feet wide and arranged around a central clear water tank measuring 23.5 feet in diameter. The water flowed in horizontal direction from the outer ring, which was used as settling basin, through the two other rings towards the centre into the clear water tank. The two inner rings contained coarse and very fine 9 ravel or sand as filter material respectively. John Gibb applied, already then, the multi-stage treatment approach; i.e., the intake filter, the settling basin and the gravel filter were used as pretreatment processes prior to sand filtration. Many other water treatment plants in England followed the example of Paisley and applied coarse gravel and slow sand filtration. In the last century, the general water treatment practice in Great Britain comprised the use of multiple filtration in form of roughing filters placed in front of slow sand filters. It was only in 1925 that rapid sand filters were slowly introduced to increase the capacity of slow sand filters. In the US, however, the clay content in the raw water prohibing adequate slow sand filter operation was one reason for developing rapid sand filters at the turn of the century.

Puech-Chabal filters, constructed in France in 1899 to treat part of the water supplied to the city of Paris, are another example of roughing filter application. The treatment scheme consisted of a series of filters and cascades to treat turbid surface water. The water flowed through four downflow roughing filters and one so-called prefilter before being treated by a finishing filter. Cascades were used to aerate the water in between the different filter stages. The filter material decreased successively in size, and the filtration rate was also reduced from filter to filter. The Puech-Chabal treatment system was used extensively in Europe. By 1935,125 plants were built in France, nearly 20 in Italy and some in other European countries [1].

After some time, the roughing filters were virtually converted into rapid or mechanical filters. Coagulation, combined with sedimentation, was introduced as a pretreatment method and, more recently, direct filtration (coagulation, flocculation and solids removal are carried out in filter units only) replaced the prefilter technology. In recent years, however, the roughing filter technology has been revived in Europe through its use in artificial groundwater recharging plants. In the early 1960s, the waterworks of Dortmund, Germany, constructed horizontal-flow roughing filters of 50-70 m filter length which are operated at about 10 m/h filtration rate [32]. The raw water falls over an aeration cascade, crosses a sedimentation trough before entering the roughing filter at the top of the gravel bed. The filter inlet zone is progressively impounded with increasing running time, and the entrance area of the water thus slowly shifts in direction of the filter outlet. After prefiltration, the water falls over a second cascade, percolates through the sand filter bed and finally reaches the aquifer. Other waterworks in Europe (e.g. in Switzerland and Austria) followed the example of Dortmund with modified horizontal-flow roughing filter designs as shown in Fig. 22.

European rivers usually exhibit low turbidity, however, filter operation is stopped during the short periods of high turbidity. A continuous supply of water to the consumers is guaranteed by the use of the aquifer's water storage capacity. In contrast to filter plants in moderate climates, roughing filters in tropical countries usually have to handle raw water of permanent or seasonable high turbidity. Since aquifers are often unavailable due to unfavourable hydrogeological conditions, the water supplies have to draw the water directly from surface water, treat it and supply it to the consumer throughout the year and even during periods of extremely poor raw water quality. Reliable operation is especially required during the rainy season, at the beginning of the wet period when the risk of epidemic outbreaks of diarrhoeal diseases increases as a result of rain washing poorly disposed faecal material into surface waters, and later on to cope with heavy sediment loads when the faecal pollution may be reduced by high dilution. Efficient and reliable water treatment is nevertheless also required in the dry season when surface waters in arid areas may discharge poorly diluted wastewater. The need for reliable and simple water treatment processes initiated the development of roughing gravel filtration which received considerable attention in recent years. Studies on design and performance of prefilters functioning under tropical water quality conditions have been, and are still being, conducted by various research groups.

Fig. 22 Different Layouts of Horizontal-flow Roughing Filters

7.2 Development of roughing filters

Motivated by the simplicity of horizontal-flow roughing filters, different institutions em barked on laboratory and field studies in order to assess the potential of horizontal-flow roughing filters in reducing the solid matter concentration of highly turbid surface water. In 1977, the Asian Institute of Technology (AIT) in Bangkok, Thailand, conducted laboratory tests with a prefilter composed of seven gravel layers [33]. Three full-scale water treatment plants applying the AIT prefilter design were later constructed in combination with slow sand filter units. The treatment plants, monitored for about half a year, revealed a good performance of the prefilters and enabled slow sand filter runs of several months [34]. These investigations were, however, discontinued and, therefore, marked the end of the project in Thailand. Since 1979, the Pan American Centre for Sanitary Engineering (CEPIS/PAHO) conducted an experimental programme that concluded in a comprehensive review and design manual introducing/he roughing filter technology [35].

The University of Dar es Salaam, Tanzania, embarked on laboratory filtration tests in 1980. Initially, investigations on vertical-flow roughing filters revealed short filter runs of a few days only. Subsequently, the horizontal-flow roughing filter concept was developed and the design tested with a 15-m long open channel filled with three gravel fractions ranging in size from 16-32, 8-16 and 4-8 mm. The laboratory tests clearly indicated that significant solids removal is achieved only under laminar flow conditions, as sedimentation is the predominant process in roughing filtration [36]. Field tests were then conducted to assess the applicability of the horizontal-flow roughing and slow sand filter treatment combination. The pilot plant investigations compared the developed filter resistance of different slow sand filters fed either with untreated or with prefiltered turbid river water. A significant increase in slow sand filter runs was achieved with prefiltration. The field tests revealed that horizontal-flow roughing filtration combined with slow sand filtration could be a viable system for turbid surface water treatment [37].

From 1982 to 1984, extensive filtration tests were conducted by the Department of Water and Sanitation in Developing Countries (SANDEC), formerly IRCWD, at the laboratories of the Swiss Federal Institute for Environmental Science and Technology (EAWAG) in Duebendorf, Switzerland. A model suspension of kaolin was used to investigate the mechanisms of horizontal-flow roughing filtration. Two important laboratory test results established that filter efficiency is hardly influenced by the surface properties of the filter medium, and that filter regeneration can be enhanced by drainage. The results of the research are summarised in a scientific paper [38], and the more practical aspects on implementation of horizontal-flow roughing filtration are compiled in a design, construction and operation manual [39]. In a collaborative effort, the University of Surrey, the DelAgua Organisation and CEPIS/PAHO developed and implemented vertical roughing filtration in Peru in 1985. Implementation and evaluation of horizontal roughing filters [12] were extensively supported by SANDEC in subsequent years.

Financially supported by the Swiss Development Cooperation (SDC), which already cofinanced SANDEC's laboratory tests, promotion and dissemination of the horizontal-flow roughing filter technology started in 1986. Under the technical assistance of SANDEC, engineers of local institutions designed full-scale demonstration plants in order to study this technology and gain practical experience with the treatment process. Frequently, horizontal-f low roughing filters were constructed in order to rehabilitate deficient slow sand filter plants. In the past ten years, the promoted filter technology has spread to more than 20 countries and, according to SANDEC's knowledge, over 80 horizontal roughing filter plants have been constructed during this period [40]. Fig. 23 indicates the countries where these filters have been constructed. Basic information on roughing filtration, as well as new approaches and designs developed by local engineers and practical field experience with the filter technology are presented in the following chapters of this publication.

Furthermore, several institutions conducted additional studies, usually in the form of postgraduate research work [41, 42, 43, 44, 45, 11, 12], on the horizontal-flow roughing filter process. The University of Dares Salaam, Tanzania, the Tampere University of Technology in Finland, the University of Surrey in Guildford, England, the International Institute for Hydraulic and Environmental Engineering in Delft, the Delft University of Technology in the Netherlands, and the University of Newcastle upon Tyne in England, as well as the University of New Hampshire in Durham, USA conducted, among other institutions, laboratory or field tests with roughing filters. Furthermore, laboratory tests on pebble bed filtration were carried out at the Imperial College in London, England [46].

Different pretreatment methods, including horizontal-flow roughing filtration, are currently field tested on a comparative basis by an extensive research programme in Cali, Colombia, where the Instituto de Investigaci�n y Desarrollo en Agua Potable, Saneamiento B�sico y Conservaci�n del Recurso H�drico (CINARA) investigates, in collaboration with the International Water and Sanitation Centre (IRC) in The Hague, The Netherlands, and different other international technical institutions and supporting agencies, the potential to optimise and simplify pretreatment processes [47].

Fig. 23 Geographical Distribution of Horizontal-flow Roughing Filter Use

Surface Water Treatment by Roughing Filters - A Design, Construction and Operation Manual (SANDEC - SKAT, 1996, 180 p.) Part 2: Design, construction and operation of roughing filters 8. Classification of roughing filters 9. General aspects of roughing filter design 9.1 Main features 9.2 Basic filtration theory 9.3 Design variables and guidelines 9.4 Flow and headloss control 9.5 Filter drainage system 9.6 General design aspects 10. Detailed filter design 10.1 Intake Filters 10.2 Dynamic filters 10.3 Vertical-flow roughing filters 10.4 Horizontal-flow roughing filters 11. Roughing filter efficiency 11.1 Practical experience 11.2 Pilot plant tests 12. Selection criteria for roughing filters (introduction...) 12.1 Raw water quality as selection criteria 12.2 Layout and operational aspects as selection criteria 13. Construction of roughing filters (introduction...) 13.1 Filter box 13.2 Filter material 13.3 Inlet and outlet structures 13.4 Drainage system 13.5 Gravel and sand washing facilities 14. Operation and maintenance of roughing filters (introduction...) 14.1 Caretaker training 14.2 Treatment plant commissioning 14.3 Flow control 14.4 Water quality control 14.5 Filter cleaning 14.6 Filter maintenance 15. Economic aspects (introduction...) 15.1 Construction costs 15.2 Operating and maintenance costs 15.3 Overall costs of water supply schemes 16. Design examples (introduction...) 16.1. Treatment of an upland river 16.2 Treatment of a lowland stream 16.3 Treatment of reservoir water 16.4 Rehabilitation of a slow sand filter plant 16.5 Standard designs for compact water treatment plants 17. Final remarks (introduction...) References Abbreviations

Surface Water Treatment by Roughing Filters - A Design, Construction and Operation Manual (SANDEC - SKAT, 1996, 180 p.)

Part 2: Design, construction and operation of roughing filters

8. Classification of roughing filters

As shown in Table 1, filters can be classified according to filter material size and filtration rate into the following categories: rock filters, roughing filters, rapid sand filters and slow sand filters. Roughing filters, using mainly gravel as filter medium, are operated without chemicals and do not require sophisticated mechanical equipment for operation and maintenance. Nevertheless, their design and application vary considerably. The different roughing filter types are classified according to:

- location within the water supply
- scheme main application purpose
- flow direction
- filter design
- filter cleaning technique

Construction of infiltration galleries in the river bed or next to the river embankment is an old technique used to draw surface water and pretreat it at the same time. Infiltration galleries basically consist of an excavated trench filled with gravel and sand layers surrounding a perforated pipe. However, construction work in water bearing aquifers might prove difficult unless extensive alluvial deposits and high seasonal flow variations prevail as may be the case in South-East Asia. There, the infiltration gallery could be the inlet technology of choice as it is almost maintenance free and may easily be installed during the dry season, provided the aquifer characteristics prevent clogging as well as breakthrough of fine material. Maintenance and cleaning of infiltration galleries are hardly possible unless installed in the river banks or in the bed of irrigation canals where the flow can be regulated or even interrupted. Such flow regulations allow a controlled operation, protect the installed gravel layers from being washed away, and enable maintenance work, e.g. cleaning or replacement of the filter layer.

Table 1 Filter Classification

However, due to the limited application and briefly mentioned operational inconveniences, infiltration galleries will not be further presented in this manual. Intake and dynamic filters are usually the first components of a treatment scheme. Similar to infiltration galleries, their structure often forms part of the water intake installation. Intake filters are used as first treatment step, mainly for separation of solids. The reduced solids concentration in the pretreated water allows a more economical layout and operation of the subsequent filter units. Dynamic filters are applied to safeguard the treatment plant from sudden solids concentration peaks. Hence, they are usually used not so much for water quality improvement, but to protect the treatment plant from heavy silt loads and cumbersome filter cleaning work.

Roughing filters are generally located at the treatment plant and used as last pretreatment process prior to slow sand filtration. These filters can be operated either as upflow, downflow or horizontal-flow filters. The different gravel fractions of roughing filters are installed either in separate compartments and hence operated in series, or the differently sized gravel is placed in succeeding layers in the same compartment.

Filter cleaning, which is carried out manually or hydraulically, is dependent on the pattern of the retained solids in the filter. Intake and dynamic filters separate the solids usually within the inlet zone of the filter and thus act as surface filters. The relatively fine gravel of these filters is cleaned manually by scouring the top of the filter with a shovel or rake, and flushing the resuspended solids from the filter bed. Roughing filters containing differently sized filter material act as deep bed filters and allow deep penetration of the solids into the filter medium. Removal of the accumulated solids is carried out by periodic filter flushing. Roughing filters might gradually get silted up if the retained solids are not completely removed by hydraulic filter cleaning. Such undesirable filter clogging calls for tedious manual cleaning and should thus be avoided whenever possible by regular and efficient filter drainages. Construction of shallow beds in upflow roughing filters minimises construction and maintenance work and, finally, also manual cleaning. However, such shallow upflow roughing filters should be used only with raw water of moderate turbidity.

A general layout of different prefilters is given in Fig. 24, the main differences in use and configurations of the prefilters are summarised in Table 2, and the detailed filter design and operation explained in Chapters 10 and 14 respectively.

Fig. 24 Classification of Prefilters

Table 2 Use and Layout of Prefilters

9. General aspects of roughing filter design

9.1 Main features

Different installations are required for controlled and adequate filter operation and maintenance. However, the main part of the filter is the section containing the filter material. A filter comprises the following six elements as schematically illustrated in Fig. 25:

- inlet flow control
- raw water distribution
- actual filter
- treated water collection
- outlet flow control
- drainage system

The inflow to a filter has to be reduced to a given flow rate and maintained thereafter at this rate as constant flow conditions are essential for efficient filter operation. In order to simplify flow control during operation, the adjusted constant flow rate can remain unchanged even during filter cleaning. However, intake filters require a controlled increase of the flow rate in order to provide sufficient washwater to flush the resuspended solids out of the filter surface.

The raw water distribution on a filter should be homogeneous to achieve uniform flow conditions in the filter bed. Therefore, the flow emerging from a pipe or a channel ought to be evenly distributed over the entire filter surface. Submerged filter beds, inlet weirs covering the full filter width, or perforated walls supplying the entire filter cross section are used for this purpose. To avoid scouring effects of the filter material, the hydraulic energy of fast flowing water has to be reduced by baffles positioned in the inlet zone. For the same purpose, concrete slabs or large flat stones should be placed on top of the filter material next to overfalls.

The actual filter consists of a watertight structure containing filter material. The shape of the filter box is normally rectangular and the walls vertical. However, depending on the local construction techniques, circular tanks and inclined walls may also be built. Round river bed gravel or broken stones with sharper edges are generally used as filter material, although any type of inert material resistant to mechanical forces, insoluble, and not impairing the water quality with respect to odour or colour, can be used as filter material.

Collection of the treated water also has to be uniform over the entire filter bed. Uneven water abstraction would reduce the overall filter efficiency and create undesirable hydraulic short circuits. Provision of a free water table on top of the filter bed is the best option to achieve even collection of the treated water for upflow filters, or the construction of a false filter bottom (see Fig. 47) for downflow filters. A second but less favourable option is the installation of perforated pipes in downflow filters. For horizontal-flow filters, construction of a perforated wall in the outlet chamber is necessary for even abstraction of the treated water.

Fig. 25 Main Features of a Filter

The outlet flow control prevents the filter bed from drying out. Hydraulic cleaning of a dried up roughing filter filled with accumulated solids is a very difficult if not impossible task. Therefore, all roughing filters must be operated under saturated conditions. A weir or a raised and aerated effluent pipe maintains the water above the filter bed level. Furthermore, a V-notch weir might be installed to allow flow rate measurements at the filter outlet.

The drainage system of roughing filters serves two purposes: it is used for hydraulic filter cleaning and allows complete during maintenance or repair work. Hydraulic filter bed cleaning calls for high discharge rates and, therefore, requires rather large pipes and fittings. For complete water removal, additional but smaller drains in inlet and outlet compartments can be installed.

Fig. 26 Solid Separation Mechanisms in Roughing Filters

9.2 Basic filtration theory

The following explanations aim at providing some information about the filtration mechanisms and at elucidating the process in more details. Removal of suspended solids by roughing filters is a rather complex process that includes sedimentation, adsorption and biological as well as biochemical activities. Basically, as illustrated in Fig. 26, solid particles have to be transported to a surface and remain attached to that surface before they are possibly transformed by biological and biochemical processes. The latter are also important for the removal of dissolved impurities.

Let us now follow the path of a small 4 µm (0. 004 mm) clay particle through a roughing filter. Please note that the following described journey of our clay particle through a roughing filter is not a science fiction story but a popular description of particle removal mechanisms taking place in roughing filters. Annex 4 provides additional analytical details of the processes described more scientifically in [38]. The small clay particle is exposed to different transportation, attachment and transformation mechanisms.

Transportation mechanisms

Screening removes particles larger than the pores of the filter bed. The smallest pore sizes are roughly one sixth of the gravel size.

Since our clay particle is travelling unhindered through the large pores of the coarse, medium and even fine filter gravel, as shown in Fig. 27, it will never be retained by screening mechanisms.

Sedimentation separates settleable solids by gravity. The settling velocity is influenced by mass density, size and shape of the particle, as well as by viscosity and hydraulic conditions of the water.

Let us now assume that our clay particle has reached the finer gravel fraction of our roughing filter operated at 0.5 m/h filtration rate. Even at this low filtration rate, the time of flow through a pore of 4 mm length and 1.25 mm height is only 10 seconds whereas the settling velocity of the clay particle amounts to 0.01 mm/s. Hence, our clay particle would need 125 seconds to overcome only half of the pore height and will, therefore, hardly touch the surface of a gravel grain but continue to drift deeper into the filter bed as shown in Fig. 28.

Interception is described as the process which enhances particle removal through gradual reduction of the pore size caused by accumulated material.

Our floating clay particle wants to settle on the filter medium and desperately calls his friends already resting on the gravel surface for help. He knows that hundreds of million of colleagues have already entered the roughing filter before he started his hopeless journey, and that the filter load (weight of accumulated solids per unit filter volume) has reached a value of 5 g/l. However, the initial filter bed porosity of 35% can only be reduced by 0.25% if his resting clay particle colleagues are packed like sardines. Fortunately, they are building up on the gravel grains and form structured pyramids, thereby increasing the occupied filter volume by a factor 10. As illustrated in Fig. 29, they are thus able to reduce porosity by 2.5%. However, our desperate clay particle is missing his colleagues since the settling distance is still far too big.

Fig. 27 Screening

Fig. 28 Sedimentation

Fig. 29 Interception

Fig. 30 Hydrodynamic Forces

Hydrodynamic forces are responsible for the water in the filter to flow continuously through the pore system and not to turn stagnant. The water has to surround each single gravel grain on its way through the filter. The flow lines are, therefore, not straight but curved around the gravel grains. The water has to change even its velocity, since restrictions require flow accelerations, and large pore volumes even force the water to take a short rest.

As illustrated in Fig. 30, our old clay particle is also exposed to this flow pattern and hydraulic shear forces which drive him on a twisted trail. He gets thrown off track by these hydrodynamic forces that lead him into a filter compartment with stagnant water where he has time to settle and join his waiting colleagues.

Hence, our clay particle had to be grateful to the flow pattern and hydrodynamic forces which transported him closer to the filter grains or to a quiescent zone where he could settle on the filter material. On his way through the filter he noticed that some very tiny particles, known as colloids, had slightly changed their direction when compared to the bulk of the solids, and had diffused due to molecular forces (Brownian movement) in some other directions. However, these forces did not affect him in the least.

Attachment mechanisms

Our clay particle, glad to have escaped the flow is still exposed to the water current which tries to drag him away. In this delicate situation, the clay particle can count on the help of his colleagues and on the support of the grain surface.

Mass attraction and electrostatic force - a combination of these two forces is frequently called adsorption, enable the particles to keep in contact with other solids and the filter material. Mass particle attraction (van der Waals force) and the attraction between opposite electrically charged particles (double layer forces) very much decrease with increasing distance between the particles. In roughing filters, these forces are important only to hold the settled particles together on the grain surface.

Biological activity will develop in the filter when particles of organic origin are deposited on the filter material. Bacteria and other microorganisms will form a sticky and slimy layer around the gravel or may build a large chain of organic material floating in the pores of the filter material.

This biological microcosm is alive. A forest composed of microbes is inhabited by monsters in the form of larvae and smaller microorganisms such as bacteria. The forest is subjected to constant changes, the micro-inhabitants are eaten up by their macro-residents, thus making a prediction of the behaviour pattern almost impossible. Particles readily adhere to this organic material and are retained in the filter.

Electrostatic and mass attraction as well as biological activity allow particles to remain on the deposited material.

Transformation mechanisms

As time goes by, new particles settle on top of our small clay particle and slowly turn it into a firm structure of accumulated material. However, he is no longer alone with his clay particle colleagues since other material of organic origin and biological matter have started to invade the pyramid-like structure. He also notices the change in water quality within the structure as he is no longer exposed to fresh water flowing on the surface of the accumulated material.

Biochemical oxidation starts to convert organic matter into smaller aggregates and finally into water, carbon dioxide and inorganic salts. Also part of the dissolved matter is subjected to these chemical and biochemical reactions. Turbidity and colour also undergo changes, while iron and manganese traces are precipitated and removed.

Microbiological activity also play an important part in roughing filters. The back of our old clay particle started to itch and he realised that three tiny microorganisms had attached themselves to his surface. He then remembered that some faecal coliforms, which were tired of swimming around, had asked him for a lift before entering the roughing filter He /lad agreed to give them a ride and they therefore started their journey through the filter like boat-people on the back of the clay particle. They also remained together as the particle settled on the filter material. But as time went by, these faecal coliforms started to starve and were attacked by other microorganisms. It was their last twitch just before they passed away which disturbed our clay particle colleague.

Hence, biologically active roughing filters are not only efficient in removing solid matter but also in improving significantly the chemical and microbiological water quality.

The "1/3 - 2/3 Filter Theory"

Our old clay particle still did not feel comfortable even after the death of the three faecal coliforms as he was embedded in a large clay deposit. Conversation with his colleagues became boring and he started to ponder on how to change his present unsatisfactory situation. He remembered seeing a lot of clay fellows sitting cheerfully on the larger grains while he had to travel through the filter uncomfortable and squeezed in. Thanks to his quick mind he developed the "1/3 - 2/3 filter theory".

He knew by experience that a particle can bypass a gravel grain either on the left or on the right or settle on its surface. Hence, the chances to fall on the grain is 1/3. However, the game continues as there is a second, third and many other gravel grains to settle on. At this point our clay particle started making some calculations. He assumed that if about 300 clay particles enter the filter 100 clay particles would settle on the first layer of grains and 200 clay particles would have to continue their journey to reach the second layer of grains. Here again, 1/3 or 67 particles would attach themselves to the second line, and 2/3 or 133 particles remain in the water flow.. The next line of particles would be split in 44 to 89 particles. He continued his mental arithmetic and was glad that someone was writing it down in Annex 4.

Proud of himself, the clay particle evaluated his calculation and came to the conclusion that 90% of the 300 particles which enter a filter are removed already after the 5th or 6th gravel layer. The remaining 10% have to travel through another five to six gravel layers in order to achieve a 99% particle separation. Hence, compared to an efficiency of only 1.5% per layer in the second filter section, the first part of the filter is apparently more efficient in particle removal, since every layer of this filter section retains about 16% of the particles. The following filter sections are obviously less efficient in particle removal. However, our clay particle found it hard to believe that a gravel grain in the inner part of the filter with the exact same size and shape as a gravel grain located in the filter inlet should be less efficient in particle removal. He then remembered that filter efficiency is dependent on particle concentration: the higher the concentration of impurities, the greater the apparent efficiency.

Nevertheless, the clay particle was not yet satisfied with his filtration theory as it would mean that the number of particles found on the gravel would continuously decline on his journey through the filter. Nevertheless, he remembered that the has encountered a sudden increase of settled particles at some specific points in the filter. He quickly concluded that these places were identical with the changes in the gravel fractions. Furthermore, he recalled that his finer clay particle fellows did not have the same settling pattern; i.e., they penetrated deeper into the roughing filter.

And finally he had the feeling that the settling conditions were not always the same. There was a crowd at the filter inlet and it was hard to find a free space to rest. However, the inner filter part provided more space, and settlement on the gravel grains was even supported by a sticky layer built on the gravel surface. Hence, filter efficiency not only depends on the size of the filter medium and particles, but also on the actual filter load and biological filter activity.

Our clay particle was pleased that his filter theory and conclusions were endorsed by some researchers carrying out filtration tests. Their results and developed correlations between the different parameters on horizontal roughing filtration are summarised in Annex 4. However, based on his own experience and reflections, our small filtration expert was convinced that this filter theory is basically also valid for upflow and downflow roughing filters. Hydrodynamic forces are present in all roughing filters where quiescent conditions allow particles to settle. Nevertheless, he was frustrated that his "1/3 - 2/3 filter theory" did not fit this ratio. Recorded filter efficiencies are much smaller on account of the numerous flow lines curving around the gravel grains.

Sudden liberation

Our puzzled clay particle became aware that his environment was deteriorating during his reflections on filtration. Together with his colleagues he was densely packed in a structure made of decomposing organic matter and hungry microorganisms. The water around them was also foul-smelling. However, this liquid hardly mixed with the water that was flowing gently at a very constant rate and under laminar flow conditions through the filter. The clay particle had the strong urge to dive into this fresh water, or even to swim away.

All of a sudden, the quiescent and dull conditions stopped. The filter was filled with tremendous noise which felt worse than a giant earthquake. The water was shooting downwards through the gravel to the filter bottom, and all the clay particles were anxiously holding each other. A side of his structure was flushed away. Our terrified particle then saw a clump of particles roll down like an avalanche. The dragging forces grew stronger, his structure collapsed and then everything went very fast. Under very turbulent flow conditions, he was flushed to the filter bottom, pressed through a pipe where a valve nearly broke his neck, and was finally discharged into a lagoon. Our old fellow gained his liberty and felt like a new-born. However, he was not used to the bright sunshine and decided to settle again, but this time under different conditions.

Closing remark: since the complex mechanisms of hydraulic filter cleaning are not yet fully explored, our clay particle had no time for further philosophical contemplation on this matter. This is where our excursion through the basic filter theory comes to an end.

9.3 Design variables and guidelines

The main objective of roughing filters is the reduction of solid matter in the raw water from a specific, in many cases however unknown concentration, to a level which allows a sound slow sand filter operation. A turbidity value of about 10 20 NTU, or a suspended solids concentration of 2 - 5 mg/l, is generally considered an adequate pretreated water standard for slow sand filtration. Furthermore, since the roughing filters have to treat a certain volume of water per day, a reasonable operational period is necessary between two filter cleanings. Explicitly, roughing filters have to meet the following three design targets:

- reduce turbidity and suspended solids concentration by DC (mg/l) to a level required for adequate slow sand filter operation

- produce a specific daily output Q (m³/d)

- allow adequate operation during a determined filter running period T_r (days or weeks).

Filter design has to meet these targets and is defined by the following six design variables which can be selected within a certain range:

- filtration rate or filter velocity v_F (m/h)
- average size dg_i (mm) of each filter medium
- individual length I_i(m) of each specific filter medium
- number n_i of filter fractions
- height H (m) and width W (m) of filter bed area A (m²).

Filtration rate V_F generally amounts to 0.3 -1 m/h. Filters are occasionally operated at a filtration rate of up to 1.5-2 m/h or even 9 m/h as in intake and dynamic filters. However, the applied filtration rate significantly influences filter performance although removal efficiency does not seem very much affected in between varying filtration rates of 0.3 and 0.6 m/h [48]. Filtration rate or filter velocity expressed in (m/h) is defined as the hydraulic load (m³/h) applied to the filter and divided by the area (m²) of the filter bed perpendicular to the flow direction.

Table 3 Gravel Fraction Sizes for Roughing Filters

Size dg_i of the filter material usually ranges between 20 and 4 mm. The gravel should be rather uniform to achieve large porosity. The uniformity coefficient U, defined here as quotient between the largest and smallest size of a filter fraction (U = dg_{i max}/dg_{i min}), should be in the order of 2 or less. Filter medium fractions as listed in Table 3 would meet the recommended uniformity.

Since length I_i of the filter material is dependent on the filter type, it may vary greatly. Dynamic and intake filters acting as surface filters require a smaller filter depth of about 40 - 60 cm, compared to roughing filters operated as deep bed filters. The depth of upflow and downflow roughing filters is limited by structural constraints, however, it is generally between 80 and 120 cm. The length of horizontal flow roughing filters is, in this respect, not limited. However, overall length normally lies within 5 and 7 m.

Number n_i of filter fractions is also dependent on filter type. Surface filters might only need one fraction whereas roughing filters are usually composed of three gravel fractions. The required overall filter length can substantially be reduced with the use of differently graded filter fractions as illustrated in Fig. 31. The bulk of the solid matter is removed by the coarse filter fraction, the medium sized gravel has a polishing effect, and the finest gravel ought to remove only the remaining traces of solid matter. Therefore, individual filter length l_i of roughing filters are often designed in a 3:2:1 ratio.

Fig. 31 Turbidity Reduction along a Roughing Filter

Height H (m) and width W (m) are dependent on structural and operational aspects. Shallow structures of about 1 - 2 m are recommended to avoid problems with respect to water tightness. In view of a possible manual filter cleaning, 1-m deep structures are even recommended for easy removal of filter material. The width of the filters should also be limited to allow efficient hydraulic cleaning and avoid washwater disposal problems. Therefore, filter width should generally not exceed 4 - 5 m, and filter surface area A for vertical flow filters should not be larger than 25 - 30 m² or 4 - 6 m² (cross section area) for horizontal-flow roughing filters. Since these recommended maximum sizes limit hydraulic filter capacity, several filter units operated in parallel are required to meet the requested treatment plant output.

Construction of at least two parallel filter units is anyhow advisable to allow continuous treatment plant operation even during major maintenance and repair work.

9.4 Flow and headloss control

The hydraulic conditions in roughing filters are determined by the filtration rate V_F (m/h), calculated as flow rate Q (m³/h) divided by the active cross-sectional filter area A (m²), e.g.:

V_F (m/h) = Q (m³/h) / A (m²)

For adequate filter performance, flow control is essential and must meet the following targets:

- maximum flow through the treatment plant should be limited in general, and through the filter units in particular

- total flow should be distributed evenly over the parallel running filter units

- controlled water levels should be maintained within the filter units.

Weirs, overflow pipes and valves are used to control the flow through the treatment plant and the different filter units. Maximum flow through the treatment plant is limited by an overflow generally located at the intake. At the treatment plant, the flow is equally divided by a distributor box or channel into the different filter units. The simplest flow control device is a V-notch weir. Finally, maximum flow through the filter unit is limited by an overflow located upstream of the V-notch weir.

The outlet structure controls the water level in roughing filters. Installation of a V-notch weir to maintain a fixed water level is the simplest flow control option. Even a normal effluent pipe can keep this water table at a constant level. However, such an effluent pipe, without weir but connected to adjacent pipe installations, does not allow discharge measurements necessary for detection of possible leaks in the filter structure.

Filter resistance increases with progressive filter operation. Final headloss in a roughing filter is usually small; i.e., 10 to 20 cm, or 30 cm at the most. Headloss variation in the filter can be recorded by the water level in the inlet filter compartment. A general layout of inlet controlled filter is illustrated in Fig. 32. More details on discharge measurements are contained in Annex 2.

A variable water level on the effluent side is achieved with the installation of a manually operated valve, a self-regulating floating weir or a constant flow device as suggested in [49]. However, since final headlosses for horizontal-flow roughing filters are relatively small, use of a variable effluent level is not recommended.

Fig. 33 contains the main details of self-regulating flow rate devices. Such installations are useful to maintain a constant flow throughout the treatment plant in pumped raw water supply schemes, particularly at night, provided a raw water balancing tank is installed for continuous raw water supply.

Fig. 32 Layout of an Inlet-Controlled Roughing Filter

9.5 Filter drainage system

Accumulation of large volumes of solids in the filter media decreases filter porosity and ultimately also filter efficiency and increases filter resistance. To maintain adequate filter performance and limit filter headloss, periodic removal of the accumulated solids from the filter media is essential.

Fig. 33 Layout of Flow Rate Controlling Devices

Roughing filters are cleaned either manually or hydraulically. Manual filter cleaning (excavation, washing and refilling of the filter media) is labour-intensive and cumbersome. Therefore, hydraulic filter cleaning plays a key role in long-term and efficient roughing filter operation.

Hydraulic filter cleaning entails fast filter drainage of the accumulated solids, which are flushed down to the filter bottom, dragged to the drainage system and washed out of the filter. The following are most important design variables for hydraulic filter cleaning:

- filter drainage velocity v_d (m/h)
- inlet area A_d of the drainage system
- horizontal distance L_d (m) between the drains or openings in the filter bottom
- washwater volume V_w
- cleaning frequency 1/T_r or filter running period T_r.

Filter drainage velocity is identical with the dropping rate of the water table in the filter. High initial filter drainage velocity v_d is recommended for efficient cleaning. Turbulent flow conditions are absolutely necessary for resuspension and transport of the accumulated solids through the filter. Therefore, a drainage velocity of at least 30 m/h, or preferably 60 - 90 m/h, is required for efficient hydraulic cleaning.

Maximum drainage rate velocity is very much influenced by minimum cross section area avail able for washwater flow. The cross section of the drainage pipes constitutes a bottleneck, the other limiting factor is the overall inlet area A_d of the drainage system, which should be designed as large as possible. Perforated false filter bottom systems provide a larger inlet area than a perforated drainage pipe system.

After having been flushed to the filter bottom, the resuspended solids have to be transported to the inlet of the drainage system. The horizontal distance L_d between the openings of the drainage system should be as small as possible to prevent gradual accumulation of sludge at the filter bottom. Here again, the installation of a false filter bottom is recommended since the washwater can be collected more evenly than in a perforated drainage system, which should be installed with a small horizontal distance L_d of maximum 1 - 2 m between the drains.

Hydraulic filter cleaning is carried out with the washwater volume V_w stored in the roughing filter. Normal filter operation is interrupted and the drains opened. Hence, compared to rapid sand filters, hydraulic roughing filter cleaning does not require additional equipment, such as backwash pumps or even air compressors. To prevent loss of washwater, fast opening valves and gates are necessary to make best use of the washwater stored in the filter bed. Since such devices of relatively large diameter (about 150 - 250 mm) are rather expensive, alternative installations, as presented in Fig. 34, have been developed by local institutions. Fast opening devices must be simple in design, sturdy and easy to operate as well as watertight. Furthermore, they should be fitted with a closing device to save washwater during drainage. Locally manufactured devices need to be carefully field-tested prior to their use in full-scale filter units.

Fig. 34 Layout of Fast Opening Devices for Filter Drainage

Cleaning frequency or filter running period T_r of roughing filters is dependent on solid matter load and biological activity in the filter. General recommendations are not possible since each filter operates under specific local conditions. Nevertheless, periodic hydraulic cleaning is advisable to prevent gradual accumulation of solids in the filter that filter drainage cannot remove due to agglomeration and consolidation. Cleaning frequencies could amount to once every one to two weeks during the rainy season, and once every one to two months during the dry season. However, very high solid loads with turbidities > 1000 NTU call for daily hydraulic flushes. Furthermore, excessive biological activities could hinder efficient hydraulic cleaning or affect taste and odour of the water. Such conditions would also require frequent hydraulic cleaning. Research in the laboratory [11] with biologically ripened roughing filters suggests that a drying period will have a positive impact on hydraulic filter cleaning. However, this observation is in contrast with the general recommendation on keeping roughing filters always wet.

Safe disposal of the washwater is important. Filter flushing generates relatively large washwater volumes (up to 10 m³) within a short time (about 1 - 2 minutes). To prevent erosion in steep regions, intermediate storage in a small, separately constructed pond may be necessary. Such an installation would allow gradual and controlled discharge of the washwater in a water course or its agricultural use. Furthermore, the solids washed out of the filter and settled in the pond are a valuable soil conditioner and fertiliser.

The drainage system used for hydraulic filter cleaning might not be designed to drain the entire filter structure. However, complete drainage is required during maintenance and repair of the filter. Additional small drainage installations are thus necessary for complete removal of the water stored in the filter. For this purpose, small drainage pipes equipped with taps or plugs can be used in large roughing filters. Small structures can, however, be dewatered with buckets or a tube used as siphon.

9.6 General design aspects

Treatment facilities have to be dimensioned for extreme loads; i.e., in terms of solids removal for maximum solids concentration in raw waters. However, it is preferable to pretreat the raw water in a sequence of different treatment units. Gradual reduction of suspended solids, turbidity or pathogenic microorganisms by a sequence of different pretreatment units probably offers the most economic option with respect to investment and operating costs. Small pretreatment units, such as intake filters or sedimentation tanks, can significantly reduce solid matter concentrations or turbidity peaks. Furthermore, cleaning of these installations is generally easier than roughing filters. Hence, roughing filters should preferably not be designed to handle maximum water peaks, but pre-conditioned water that has already been subjected to pretreatment.

Filter length and permissible filter running period are correlated. Horizontal-flow roughing filters in particular were originally designed to provide a large silt storage capacity at low headloss, as filter cleaning was carried out manually. Relatively important filter lengths of 9 to 12 m were the consequence of this original design approach, permitting filter runs of several months, similar to those of well-operated slow sand filters. However, importance and benefits of hydraulic cleaning have meanwhile been recognised.

Current design practice tends to reduce filter lengths and incorporate efficient hydraulic cleaning facilities. Regeneration of filter efficiency through frequent hydraulic cleanings has to counterbalance shorter filter lengths.

The use of smaller filter material can improve filter efficiency. However, besides efficiency in suspended solids separation, other criteria such as terminal headloss, filter running time and filter cleaning aspects have to be taken into consideration. Use of only a uniform and fine filter material allows sufficient pretreatment of the raw water, but at the expense of high head losses, short filter runs and filter cleaning difficulties. The roughing filter technology requires the use of coarse filter material sizes between 20 - 4 mm graded in different fractions. However, the use of filter material coarser than 20 mm with lower removal efficiencies is not advisable as it would require longer filters to achieve the same treatment efficiency. Furthermore, the filter material should not be smaller than about 4 mm to facilitate hydraulic filter cleaning. These recommendations are not applicable to intake and dynamic filters as these operate differently. Since intake and dynamic filters are basically surface filters, they require small filter material sizes between 2 - 8 mm. These filters act as surface filters, and their filter depth therefore has no great influence on the overall efficiency.

Filtration rate greatly influences filter efficiency. Sedimentation is the main solids separation process in roughing filters. Therefore, roughing filters must be operated under laminar flow conditions to achieve adequate solids removal efficiencies. Flow conditions are described by the Reynolds Number. At a value of less than 10, laminar flow can be expected (see also Fig. 35). As the Reynolds Number is directly proportional to the filter material size, maximum allowable filtration rate for laminar flow conditions will be determined by the coarsest gravel fraction in a roughing filter. Hence, for optimum filter use, coarser filter material requires smaller filtration rates. However, filtration rate can only partly be increased by applying smaller filter material, as particle size distribution of the solids and suspension stability also determine the fillers' solids separation efficiency.

Fig. 35 Roughing Filter Efficiency in Correlation to Flow Conditions

10. Detailed filter design

10.1 Intake Filters

Intake filters are combined with water abstraction structures and installed next to small and narrow river beds as illustrated in Photo 5 and Fig. 36. Intake filters are often used as first pretreatment unit in a water treatment scheme. A small weir regulates the water level of the surface water and channels part of the flow into an adjacent filter compartment. This filter box is filled with two gravel layers. The top layer consists of relatively fine gravel of less than six millimetres in diameter.

Photo 5 Example of an Intake Filter

Fig. 36 Layout of Intake and Dynamic Filters

The lower coarser gravel layer acts as filter support and allows an even abstraction of the prefiltered water through perforated drainage pipes. The abstracted raw water, after being distributed evenly by a small weir over the entire width of the filter box, flows gently over the gravel bed surface. Part of this water percolates through the gravel layers and the remaining water is discharged over an outlet weir back to the river. Intake filters are constructed along rivers and not directly in the river bed, as the filter material would be washed out during periods of high river discharge. Construction of a separation wall between river bed and filter box is recommended to prevent the filter from being washed out.

Intake filters can also be installed in the bed of small canals. Upland rivers with a steep river bed and a suitable topography might allow the accommodation of a small diverting canal. The filter bed, comprising different gravel layers, is installed over a small stretch in the canal. Part of the canal water is filtered through the series of fine to coarse gravel layers, while the remaining water is returned to the river. The prefiltered water is collected by perforated drainage pipes laid at the bottom of the coarse gravel layer, and the discharge rate regulated by a valve placed at the filter control box. The flow velocity in the canal must be regulated by the canal's intake structures to protect the filter bed from being washed out during periods of high river discharge. The flow velocity in the canal should actually range between 0.10 and 0.30 m/s to prevent fine matter from settling and remaining on top of the gravel bed, and also to avoid fine filter material from being washed out. This layout may also be applied to irrigation canals, provided they are continuously supplied and regulated through out the year. However, construction of intake filters along rivers is strongly recommended as these filters allow a more reliable operation than intake filters installed in canal beds.

Finally, "intake" filters may be located directly at the treatment plant site and function as pretreatment facility. This particular location is recommended in gravity water supply schemes with a raw water intake located in a remote area of difficuIt access. Such a layout will facilitate monitoring and regular filter cleaning.

Filtration rates of intake filters range between 0.3 and 2 m/h. However, significant solids removal rates can be expected only at filter velocities smaller than 1 m/h, Design of the hydraulic structures should be based on maximum filter resistance of 20 to 40 cm. This figure will not be exceeded if regular filter cleaning, (e.g. once a week), is observed.

Relatively small filter material of less than 6 mm is used in intake filters which act as surface filters as the solids mainly accumulate on top of the filter bed. Since filter cleaning is carried out manually, the different gravel layers might be disturbed and mixed up if filter material of different sizes is used in intake filters. A filter cloth is sometimes placed in-between the different gravel layers to avoid mixing of the filter fractions and possibly reducing filter porosity and efficiency. However, coarser gravel hardly contributes to solids removal, but allows an even abstraction of the pretreated water. Nevertheless, regular filtered water collection is also possible with a single filter bed layer, moderate filtration rates, medium-sized filter structures and reasonable layout of the perforated drainage pipes. Design guidelines are summarised in Fig. 37. Use of a single filter layer and false filter bottom offers a favourable design alternative for intake filters, as filter material mixing is no longer possible, and even abstraction of the prefiltered water is guaranteed.

Fig. 37 Design of Intake and Dynamic Filters

10.2 Dynamic filters

Dynamic filters protect the treatment plant units from high turbidity peaks. Highly turbid surface water can quickly clog filters, especially slow sand filters. Therefore, during periods of extremely high raw water turbidity, the flow may be interrupted to reduce cumbersome filter cleaning. Separation of solids is only of secondary importance in dynamic filters.

Dynamic filter performance is, as described by its adjective, dynamic. The water quality between filter inlet and outlet hardly changes during periods of low raw water turbidity. During raw water turbidity peaks, however, the quantitative change is drastic as no water is available in the filter outlet! Dynamic filters act like turbidity meters connected to an open-close valve; i.e., they rapidly get clogged when raw water of high turbidity passes through the filter.

Dynamic filters are similar in layout to intake filters, but differ in filter material size and filtration rate. Especially the gravel size of the top filter layer is smaller; i.e., less than 6 millimetres in diameter, while filtration rate is usually more than 5 m/h, Maximum available headloss is still limited and ranges between 20 and 40 cm in spite of finer filter material and greater filter velocity. Finally, the horizontal flow velocity over the filter bed surface should be small or non-existent; i.e., less than 0.05 m/s or nil, to prevent removal of accumulated silt during turbidity peaks.

Dynamic filters are cleaned after each raw water turbidity peak. Cleaning is also carried out manually and in the same way as in intake filters. During periods of dynamic filter interruptions, treatment plant operation is reduced (e.g. by declining filtration rate operation in slow sand filters). Nevertheless, dynamic filters should only be used with raw water experiencing short turbidity peaks; i.e., from a few hours to maximum half a day. Dynamic filters are preferably located at the site of treatment plants to facilitate monitoring and cleaning by the caretaker.

Russian sanitary engineers introduced the idea of dynamic filters to Argentina [50], where about 50 filters were installed and operated in the late 1970s. However, filter design differs from the one presented here. There raw water flows into a dissipation chamber and from there over an inlet weir on top of a sand filter. The filter is operated in such a way that a water layer of a few millimetres flows over the sand surface. This flow washes the solid matter deposited on top of the filter bed into a sand recovery chamber installed at the end of the filter. Pilot plant studies conducted by Watertek with Water Research Commission funding to test and demonstrate the use of this filter design [51 ] are in progress in South Africa.

10.3 Vertical-flow roughing filters

Roughing filters can be considered a major pretreatment process for turbid surface water since they efficiently separate fine solid particles over prolonged periods. They are therefore placed at the treatment plant site and operated in combination with other pretreatment units such as dynamic filters or sedimentation tanks. Roughing filters precede final treatment processes, such as slow sand filtration and chlorination.

Vertical-flow roughing filters usually consist of three filter units arranged in series as shown in Fig. 38. The water to be treated flows in sequence through the three filter compartments filled with coarse, medium and fine filter material. The size of the three distinct filter material fractions is generally between 20 and 4 mm, and graded, for example, into fractions of 12-18 mm, 8-12 mm and 4-8 mm.

Vertical-flow roughing filters operate either as downflow or upflow filters. They are hence either supplied by inflowing water at the filter top or at the filter bottom. The filter material of vertical-flow roughing filters is completely submerged. A water volume of about 10 cm depth usually covers the gravel. The top should be covered by a layer of coarse stones to shade the supernatant water and thus prevent algal growth often experienced in pretreated water exposed to the sun. Drainage facilities, consisting in perforated pipes or a false filter bottom system, are installed on the floor of the filter boxes. Finally, pipes or special inlet and outlet compartments are required to convey the water through the subsequent three filter units.

Fig. 38 Layout and Design of Vertical-flow Roughing Filters

Vertical-flow roughing filters are usually operated at 0.3 to 1.0 m/h filtration rates. Vertical-flow roughing filters may be sensitive to hydraulic fluctuations, especially if loaded with large amounts of solids. Settled matter might be resuspended at increased filtration rates, causing solids to break through the filter. Filter operation at constant flow rates is, therefore, recommended. Raw water containing colloidal matter and a high suspension stability should be treated at low filtration rates and preferably with fine filter material. Filter resistance is usually less than 20 cm per filter unit and, hence, not a decisive operational criteria for properly designed and operated roughing filters.

Due to structural constraints, vertical-flow roughing filters have a relatively small filter depth of about 1 m. Total filter depth of the three filter units used in series is thus 3 m. This total available filter dentin limits vertical-flow roughig filter application. It can generally and efficiently handle moderate raw water turbidities of 50 to 150 NTU. Raw water pretreatment by intake filters, reduction of filtration rate or provision of additional filter boxes would be required to treat raw water of higher turbidities.

In vertical roughing filters in layers, where all three gravel fractions are installed in one filter box with a total filter distance of about 1 m, low turbidity raw water can be pretreated. However, due to cleaning aspects this filter design can only be used for upflow operated roughing filters. In such filters, the coarse filter material is placed at the bottom and the finest material at the top of the filter. The separated solids, which accumulate mainly in the coarse filter fraction next to the filter bottom, can be easily flushed out with the water stored in the filter. Therefore, the use of upflow roughing filters in layers is recommended. Down flow roughing filters in layers face considerable problems with hydraulic filter cleaning. The bulk of solids accumulated in the coarse filter material on top of such filters would have to be flushed through the finer rather clean filter material and would thus soil the entire filter bed.

Adequate and efficient washwater collection is important for reliable roughing filter operation. Perforated pipes or a false filter bottom can be installed in vertical-flow roughing filters. Perforated pipes, which should be laid in a coarse gravel pack to support an even washwater abstraction, would require the installation of additional filter material in vertical-flow roughing filters. Preference is given to false filter bottoms as they allow an even washwater abstraction and do not require additional gravel layers. Although special perforated concrete slabs will be necessary, they may be readily produced locally.

A comparison of downflow with upflow roughing filters reveals the following:

· Direction of flow and sedimentation are obviously the first differences which might interfere or support solids settling on the filter material. Solid removal efficiency should consequently vary in the two filter types. Theoretically, downflow filters should have a better performance than upflow filters as the solid particles are more likely to settle on top of the gravel surface in the direction of flow than under countercurrent conditions. However, practical field experience has shown a similar efficiency for both filters. In dead filter zones, where the water flow is reduced to a minimum, solids settle regardless whether the roughing filter is operated in upflow or downflow direction. Hence, filter efficiency is similar in both filter types.

· The accumulation pattern of retained solids is another difference between downflow and upflow filters. The bulk of the solids is deposited at the inlet of the filter; i.e., for downflow filters in the upper part of the filter, and for upflow filters in the filter medium located next to the filter bottom. This, however, has a tremendous impact on hydraulic filter cleaning. In downflow roughing filters, the buIk of accumulated solids has to be flushed with a relatively small washwater volume from the soiled filter top through the lower and cleaner filter part to the filter bottom. The opposite is true for upflow roughing filters. The bulk of retained solids is accumulated next to the drainage system and a relatively large washwater volume, accommodated in the upper filter part, is available to flush the solids out of the filter. Owing to the important filter cleaning aspect, use of upflow roughing filters rather than downflow filters is recommended.

10.4 Horizontal-flow roughing filters

Unlimited filter length and simple layout are the main advantages of horizontal-flow roughing filters. Generally, the shallow structure does not create structural problems, and the filter length is not limited to a few metres. Furthermore, its simple layout does not require additional hydraulic structures and installations as in vertical-flow roughing filters. The raw water runs in horizontal direction from the inlet compartment, through a series of differently graded filter material separated by perforated walls, to the filter outlet as illustrated in Fig. 39. Filter material also ranges between 20 and 4 mm in size, and is usually distributed as coarse, medium and fine fraction in three subsequent filter compartments. To prevent algal growth in the filter, the water level is kept below the surface of the filter material by a weir or an effluent pipe placed at the filter outlet.

Fig. 39 Layout and Design of a Horizontal-flow Roughing

Filtration rate in horizontal-flow roughing filters ranges between 0.3 and 1.5 m/h. It has been defined here as hydraulic load (m³/h) per unit of vertical cross section area (m²) of the filter. Filter length is dependent on raw water turbidity and usually lies within 5 to 7 m. Due to the comparatively long filter length, horizontal-flow roughing filters can handle short turbidity peaks of 500 to 1,000 NTU.

Drainage facilities, such as perforated pipes, troughs or culverts, allow hydraulic filter bed cleaning. These drainage systems are placed at the filter bottom perpendicular to the direction of flow. Drainage facilities in flow direction must be avoided as they could create short-circuits during normal filter operation. Hence, false filter bottom systems cannot be installed in horizontal-flow roughing filters. Since most of the solids accumulate at the inlet of each filter medium, drainage facilities should be placed at the inlet of each filter compartment to enhance hydraulic cleaning efficiency. Installation of troughs complicates construction of the filter box floor. Furthermore, since the horizontal distance L_d between the troughs is usually large, an even abstraction of the sludge is correspondingly difficult. Therefore, use of perforated pipes is the best drainage system for horizontal-flow roughing filters, as it allows easy installation of a dispersed system. Although prefabricated culverts may allow a more even solids removal, connection to the washwater effluent pipes is more complicated.

Horizontal-flow roughing filters have a large silt storage capacity. Solids settle on top of the filter medium surface and grow to small heaps of loose aggregates with progressive filtration time. Part of the small heaps will drift towards the filter bottom as soon as they become unstable. This drift regenerates filter efficiency at the top, and slowly silts the filter from bottom to top. Horizontal-flow roughing filters also react less sensitively to filtration rate changes, as clusters of resuspended solids will drift towards the filter bottom or be retained by the subsequent filter layers. Horizontal-flow roughing filters are thus less susceptible than vertical-flow filters to solid breakthroughs caused by flow rate changes. However, they may react more sensitively to short circuits induced by a variable raw wafer temperature.

Periodic cleaning is also essential for horizontal-flow roughing filters. Hydraulic cleaning is carried out by fast drainage of the water stored in the filter. During filter drainage, the small unstable heaps of accumulated solids collapse and are flushed towards the filter bottom. The solid matter stored in the filter material is washed out of the filter box through the drainage system. Drainage velocities of 60 to 90 m/h are necessary to achieve a good hydraulic cleaning efficiency. Drainage pipes of adequate size are required to achieve the recommended velocity which drains the filter within 1 to 2 minutes. Depending on the solids concentration in the raw water, regular hydraulic filter cleaning, at intervals of every few weeks, is required to avoid deterioration of filter efficiency and development of excessive filter resistance. However, filter resistance will not exceed 20 cm if normal filter operation and regular cleaning are observed. Frequent and efficient filter drainages will also defer the need for manual filter cleaning, which nonetheless becomes unavoidable after some years of filter operation.

Photo 6 Inside View of a Roughing Filter Bed during Hydraulic Cleaning

11. Roughing filter efficiency

11.1 Practical experience

Treatment efficiency is dependent on raw water characteristics, layout and operation of roughing filters. On the one hand, size, concentration, type of particles and suspension stability are the most important water quality parameters influencing suspended solids removal efficiency. On the other hand, filter material size and filter length, applied filtration rate and cleaning frequency are the key factors determining filter efficiency. Hence, roughing filters with identical layout and operation may vary in filter performance with different raw water sources. Even a specific filter will most probably not have a constant filter efficiency with the same raw water source: high particle removal rates will be recorded during periods of high raw water turbidity where a slower rates will be experienced during periods of moderate raw water turbidity. Therefore, an exact indication of filter efficiencies is generally quite impossible.

Treatment efficiencies of different roughing filters have been studied extensively by CINARA [47] at a pilot plant in Puerto Mallarino, Cali, Colombia (reported also in [48]). These field tests are considered the most comprehensive pilot plant studies for roughing filter development. The pilot plant consists of a first pretreatment step using intake/dynamic filters to precondition the raw water drawn from the Cauca River. The flow is then split into five lines where the filter performance of different roughing filters is tested in combination with identical slow sand filters used as reference. The following types of roughing filters are installed at Puerto Mallarino:

All roughing filters have similar gravel fractions but differ in filter length. Total filter length of URFS, MHRF and DRFS, amounts to 4.40 m. Total filter length of the HRF unitis 7.10 m and 1.60 m for the URFL unit. The slow sand filter units are circular in shape, 2.00 m in diameter and 2.00 m in height. They were filled with a 1 -m deep sand layer, which was gradually reduced due to subsequent sand cleanings, but never fell below 0.60 m. The sand has an effective diameter of 0.2 mm and a uniformity coefficient of 1.57.

Fig. 40 summarises filter efficiencies of the different roughing filters with respect to turbidity removal at different filtration rates and two distinct raw water turbidity levels. The graphs show higher removal rates of generally 85 - 90% or more for periods of high turbidity (150 - 500 NTU). Filter efficiency is reduced to about 80 85% or less during periods of moderate turbidity (30 - 50 NTU), and is hence in accordance with the general filter theory. The different but small filtration rates had no significant influence on the turbidity removal efficiencies of the filters, as laminar flow prevailed in all gravel fractions also at the highest filter velocity of 0.60 m/h, The upflow roughing filter in series and the horizontal-flow roughing filter unit exhibited best performance throughout all test conditions. The smallest turbidity removal efficiency was achieved by the downflow roughing filter unit.

Fig. 41 elucidates the efficiency of the different pilot plant treatment steps with respect to suspended solids and faecal conform reduction. The intake/dynamic roughing filters reduced the average suspended solids concentration by 55% from about 200 mg/l to 90 mg/l. This concentration was further reduced to less than 5 mg/l by the roughing filters and the filtrate of the slow sand filters had an average suspended solids concentration of 0.2 - 0.3 mg/l. The relatively high suspended solids concentration of the upflow roughing filter in layer might be an indication of a comparatively low process stability of this filter.

Average faecal coliform concentration of the raw water of about 40,000 CFU/100 ml was subsequently reduced to about 24,000, 400 and to less than 1 CFU/100 ml by the treatment scheme consisting of intake/dynamic filters, roughing filters and slow sand filters. The modified horizontal-flow roughing filter had the smallest faecal coliform removal efficiency of 96.5%. This rate also influenced performance of the subsequent slow sand filter, which produced an average effluent of 2.6 CFU/100 ml. All the other slow sand filters had average faecal coliform concentrations of less than 1 CFU/100 ml in their effluents.

Fig. 40 Turbidity Removal by Different roughing filters

Fig. 41 documents the high treatment efficiency of the pilot plant. The two pretreatment steps and the slow sand filters were able to reduce the suspended solids concentration from about 200 mg/l to about 0.2 mg/l, or by 3 log, whereas the faecal coliform concentration was reduced from about 40,000 CFU/100 ml to generally less than 1 CFU/100 ml, which corresponds to a 4 - 5 log reduction.

Full-scale treatment plants are not so extensively monitored and controlled as pilot plants. Nevertheless, Fig. 42 documents the development of biological filter activities in the filters of a community water supply [52]. The raw water of La Javeriana's treatment plant originates from the Pance River, a highland river of moderate turbidity. The water is treated by an intake filter, two horizontal-flow roughing filters, and two slow sand filters operated at 1.3, 0.6 and 0.08 m/h filtration rates. Faecal coliform concentration ranging between about 1,000 and 10,000 CFU/100 ml is proof of relatively high faecal contamination of the raw water. Turbidity of about 20 NTU is relatively low during the dry periods, but increases to short turbidity peaks after periods of precipitation. The apparent colour averaging about 100 CU/I follows a similar pattern as turbidity. Mean faecal coliform concentration in the pretreated water amounted to about 200 CFU/100 ml and did not decline during the monitoring period of half a year. Although the treated water had initially somewhat elevated faecal coliform concentrations of more than 10 CFU/100 ml, the effluent concentration levelled out to about 1 CFU/100 ml after three weeks of operation. This corresponds to the period of maturation of the slow sand filter. The overall turbidity and apparent colour reduction, however, indicated a distinct improvement within the first six months of operation. With progressive filter operation, the respective treatment efficiencies also increased in the roughing filter, most probably on account of the gradual development of biological processes in this filter.

Table 4 summarises treatment efficiencies of roughing filters operated at different flow directions. The filter material in the downflow and horizontal-flow roughing filter is rather coarse compared to the one used in the upflow roughing filter operated, however, at more than double the normal filter velocity. Nevertheless, in all three treatment plants, turbidity reduction by the roughing filters amounts to about 70 - 90%. The bacteriological water quality improvement was about of the same order for these three treatment plants.

Treatment efficiency of roughing filters is also limited as they are not capable of treating any type of water as illustrated by the following example [56]. Construction of a water supply scheme was one component of the Laka Laka multi-purpose project. The project team decided to draw raw water from the newly constructed water irrigation reservoir to supply their scheme. High raw water turbidity lead to extremely short filter runs of the two slow sand filters. Two horizontal-flow roughing filters were therefore designed on the basis of the available literature data in order to improve operation of the slow sand filters. The 18-m long roughing filters were operated at 2.5 m/h filtration rate. Since filter efficiency was very poor at this rate, filter velocity was gradually reduced to 0.5 m/h, but without achieving a substantial treatment efficiency improvement. These problems were encountered mainly because roughing and slow sand filters have never been used in this area before. However, they could have been avoided by pilot plant tests (which are strongly recommended in such a situation) conducted during the project design phase.

Fig. 41 Suspended Solids and Faecal Coliform Reduction by Roughing and Slow Sand Filtration

11.2 Pilot plant tests

As described in the previous chapter, the worldwide experience with roughing and slow sand filters documents the significant potential of this treatment concept in producing potable water from polluted turbid surface water. There is no doubt about the general treatment efficiency of slow sand filters, since a biologically mature filter will consistently reduce the concentration of microorganisms by 2 to 4 log (99 to 99.99% reduction). It is more a question of specific treatment efficiency of roughing and slow sand filters when fed with a local raw water source. Hence, treatability of a particular raw water is of major concern to design engineers, especially if local practical experience with the considered treatment process and raw water source is not available.

Fig. 42 Turbidity, Apparent Colour and Faecal Coliform Reduction at the Treatment Plant La Javeriana, Colombia

Table 4 Examples and Practical Experience with Roughing Filters