Filtration, as it applies to water treatment, is the passage of water through a porous medium to remove suspended solids. According to Baker (1948), the earliest written records of water treatment, dating from about 4000 B.C., mention filtration of water through charcoal or sand and gravel. Although a number of modifications have been made in the manner of application, filtration remains one of the fundamental technologies associated with water treatment.
Filtration is needed for most surface waters, to provide a second barrier against the
transmission of waterborne diseases. Although disinfection is today the primary defense, filtration can assist significantly by reducing the load on the disinfection process, increasing disinfection efficiency, and aiding in the removal of precursors to disinfection by-product (DBP) formation. The Surface Water Treatment Rule (SWTR) and Enhanced Surface Water Treatment Rules (ESWTR) recognize three categories of granular filtration techniques:

• Rapid sand
• Slow sand
• Diatomaceous earth

This chapter covers the design of the first category of filters. However, in this instance the term rapid sand includes not only sand, but also other types of filter media such as crushed anthracite coal and granular activated carbon (GAC). Chapter 9 covers the other two categories of granular filtration techniques. Chapter 14 further discusses activated carbon processes, including GAC filters/ adsorbers.

Removing suspended solids by high-rate granular media filtration is a complex process
involving a number of phenomena. Attempts to develop theories that quantitatively predict solids removal performance with sufficient precision and versatility to be of use in practical filter design have met with relatively little success. Consequently, filter media selection is often an empirical process. Pilot investigations are common tools for assessing the performance of a particular filter design (see Chapter 28).
In current high-rate granular media filtration techniques, solids removal occurs primarily as a two-step process (Cleasby, 1972). During the initial transport step, particles are moved to the surfaces of media grains or previously captured floc. Transport is believed to be caused largely by hydrodynamic forces, with contact occurring as stream lines converge in pore restrictions. The second step is particles’ attachment to either grain or floc surfaces. Electrokinetic and molecular forces are probably responsible for the adherence of particles on surfaces within the bed (O’Melia and Crapps, 1964; Craft, 1966; O’Melia and Stumm, 1967). Physical straining through the surface layer of solids and biological growth (schmutzdecke) is the principal filtration mechanism of a slow sand filter, but it is generally a minor means of solids removal in high-rate granular media filters.

A number of interrelated components are involved in the overall design of a high-rate
Granular media filtration system:
• Regulatory requirements
• Pretreatment systems
• Filter media
• Filtration rates
• Depth of the filter box
• Mode of operational control
• Filter washing system
• Filter arrangements
• Underdrain system
• Filter performance monitoring
• Auxiliaries
These components are discussed in detail in the following sections.

Filter Media
Although the selection of filter media type and characteristics is the heart of any filtration
system, selection is usually based on arbitrary decisions, tradition, or a standard approach. Pilot plant studies using alternative filter media and filtration rates can determine the most effective and efficient media for a particular water.
In drinking water applications in North America, the most commonly used filter media are natural silica sand, garnet sand or ilmenite, crushed anthracite coal, and GAC. Selecting appropriate filter media involves a number of design decisions concerning source water quality, pretreatment, and desired filtered water quality. Filter media cleaning requirements and underdrain system options depend on the filter configuration and filter media selected. Media variables the designer can control include bed composition, bed depth, grain size distribution, and, to a lesser extent, specific gravity. In addition to media design characteristics, media quality can be controlled to some extent through specifications covering,
where applicable, hardness or abrasion resistance, grain shapes, acid solubility, impurities, moisture, adsorptive capacity, manner of shipment, and other such factors.


Suggested criteria and a discussion of the applicability of these parameters can be found in the AWWA Standard for Filtering Material and Standard Granular Activated Carbon .
In the United States, granular media have been traditionally described in terms of effective size (ES) and uniformity coefficient (UC). The ES is that dimension exceeded by all but the finest 10% (by weight) of the representative sample. It is also referred to as the “10% finer” size. The UC is the ratio of the “60% finer” size to the ES. Common practice in Europe is to express media sizes as the upper and lower limits of a range.
These limits may be expressed either as linear dimensions or as passing and retaining sieve sizes (that is, 1.0 to 2.0 mm or -10 + 18 mesh).
The filtration process also affects the selection of the filter bed because of the special requirements of each type of process. The direct and in-line filtration processes must have filter beds with a large floc holding capacity. A reverse-graded filter bed, such as a dualmedia or coarse deep bed, satisfies this requirement. In two-stage filtration, the filter bed of the first stage acts as a roughing filter and carries out the flocculation process. Data obtained from pilot filter tests and actual installations using the two-stage filtration process indicate that the first-stage filter bed may be designed in the same fashion as an ordinary filter
Rapid sand filtration, with filtration rates ranging from 2 to 3 gpm/ft 2 (5 to 7.5 m/h), usually uses medium-sized sand (0.5-mm ES). High-rate filters of 5 to 10 gpm/ft 2 (12.5 to 25 m/h) always consist of a reverse-graded filter bed or a deep, large-sized monomedia. Filter beds may be classified as graded fine-to-coarse, ungraded, graded coarse-to fine, or uniformly graded, depending on the distribution of grain sizes within the bed during filtration. Transition from the ungraded media of a slow sand filter to the fine-to-coarse high-rate granular media filter resulted from dissatisfaction with the low loading rates and laborious cleaning procedure characteristic of slow sand filters. Filters with uniformly graded or coarse-to-fine beds are now operated at higher filtration rates and for longer run times than are feasible with conventional rapid sand filters.

Ungraded Media. The slow sand filter is a primary example of an ungraded bed. Because slow sand filters are not backwashed, no hydraulic grading of the media occurs. Distribution of the various grain sizes in the bed is essentially random. Typical slow sand filter beds contain 2 to 4 ft (0.6 to 1.2 m) of sand with an ES of 0.15 to 0.35 mm and a UC not exceeding 3.0. Refer to Chapter 9 for further information on slow sand filtration design.

Fine-to-Coarse Media. Fluidization and expansion of rapid sand filter beds during backwashing result in accumulating fine sand grains at the top of the bed and coarse grains at the bottom. Consequently, filtration occurs predominantly in the top few inches, and head loss increases relatively rapidly during operation. This sand medium typically has an ES of 0.35 to 0.60 mm (generally 0.5 ram) and a UC of 1.3 to 1.8. Grains passing a no. 50 sieve (0.3 ram) or captured on a no. 16 sieve (1.18 mm) are normally limited by specifications to very small portions of the medium. Bed depths are typically 24, 30, or 36 in. (0.6, 0.75, or 0.9 m), respectively.
Single-medium anthracite beds have been used in the same basic configuration as rapid rate beds. Because anthracite is more angular than sand, the porosity of an anthracite bed is higher than that of a sand bed containing media with the same ES. The porosity of a sand bed is generally 40% to 45% whereas a typical anthracite bed has a porosity of 50% to 55%. Consequently, anthracite does not perform in exactly the same manner as sand of equivalent size. Because of the lower specific gravity, anthracite beds are also easier to fluidize and expand than sand beds.

Coarse-to-Fine Media. In a coarse-to-fine bed, both small and large grains contribute to the filtering process. The presence of fine media in a filter is desirable because of the relatively large surface area per unit volume that fine media provide for particle adhesion.
Fine media are when placed before fine media in the filtering sequence, decrease the rate of head loss buildup and increase available storage capacity in the bed.
Dual-media beds normally contain silica sand and crushed anthracite coal and are a
very common filter media design. Triple-media beds contain an additional layer of garnet or ilmenite sand. Specific gravities of materials used in filtration are roughly as follows:
• Silica sand, 2.55 to 2.65
• Anthracite coal, 1.5 to 1.75
• Garnet, 4.0 to 4.3
• Ilmenite, 4.5
A typical dual-media bed contains 6 to 12 in. (0.15 to 0.3 m) of silica sand (ES 0.45 to 0.55 ram) overlaid by 18 to 30 in. (0.46 to 0.76 m) of anthracite (ES 0.8 to 1.2 mm). A typical mixed-media filter bed contains 3 to 4 in. (5 to 10 cm) of garnet (ES 0.15 to 0.35 ram), 6 to 9 in. (0.15 to 0.3 m) of silica sand (ES 0.35 to 0.5 ram), and 18 to 24 in. (0.5 to 0.6 m) of anthracite (ES 0.8 to 1.2 ram).
The degree to which media layers are intermixed in the bed depends on the sizes and shapes of the media used, the nature of the backwashing procedure, and the specific gravities of the different media. Disagreement exists over whether distinct layers or intermixed layers are most desirable. If layers mixed completely, the purpose of using more than one medium would be defeated. If no mixing occurs, individual fine-to-coarse layers would result, and the possibility of rapid clogging at interfaces would be raised.
Proponents contend that in a properly designed mixed-media filter, a gradual decline in pore sizes from top to bottom of the bed is established after backwashing. The original argument can be traced to Conley and Pitman (1960), Conley (1961), and Camp (t961,1964) in the early 1960s. Brosman and Malina (1972) concluded that a slightly mixed bed was superior to a distinctly layered bed in terms of head loss development, filter run time, and filtered water turbidity. Cleasby and Sejkora (1975), however, disagree that superior performance can be attributed to interfacial intermixing in and of itself; rather, it is a result of differences in the media sizes required to construct mixed and separated beds. They found that to provide a relatively sharp interface in a dual-media bed, fairly coarse sand was required. The resulting bed would not provide the same filtered water quality as a bed using finer sand that mixed more readily with the coal.
The anthracite coal and silica sand used in dual-media filters inevitably result in some intermixing of layers. In a triple-media bed, intermixing of silica sand and garnet sand normally occurs more readily than mixing of silica sand and coal. Cleasby and Woods (1975) suggest that, as a rule of thumb, the ratio of the average particle size of coarse silica grains to the size of coarse garnet grains should not exceed 1.5, to ensure that some garnet remains at the bottom of the bed. They also suggest that a ratio of coarse coal grain size to a fine silica sand grain size of about 3 results in a reasonable degree of mixing in dual- or mixed-media beds. Brosman and Malina (1972) found that anthracite sand filter media with a size ratio at the interface of less than 3:1 exhibits little mixing and that the zone of mixing increases linearly as the size ratio increases above 3:1.
In a number of U.S. installations, taste and odor removal and filtration have been combined in a single unit using GAC (Hager, 1969; Hansen, 1972; Blanck and Sulick, 1975; McCreary and Snoeyink, 1977). GAC is sometimes added to existing rapid sand units from which some sand has been removed. GAC depths of 12 to 48 in. (0.3 to 1.2 m) over silica sand layers of 6 to 18 in. (0.14 to 0.5 m) have been reported. Typically, GAC with an ES of 0.5 to 0.65 mm has been used. This technique is usually applicable only where taste and odor, and not turbidity, are of primary concern. If turbidity levels are high, GACe instrumental in achieving the best-quality filtered water. pores become rapidly plugged, and carbon life is quickly reduced. If turbidity, taste, and odor are all significant problems, GAC beds should be preceded by conventional granular media filtration. If carbon adsorption is desired to remove organics, the depth of GAC that can be provided in a converted gravity filter is likely to be too shallow to provide adequate contact time; however, this may offer some additional carbon removal after conventional

Uniform Media. Uniformly graded deep-bed filters use relatively coarse media, ranging from 0.5 mm to as much as 6.0 mm. The UC is typically 1.2 to 1.3, but values as high as 1.5 may be found. Greater media depth is substituted for the lack of fine media in the bed. Such a substitution requires more vigilant operation of pretreatment systems to avoid breakthrough. Depths of 4 to 6 ft (l.2 to 1.8 m) are common, and in some cases media depths reach 8 ft (2.4 m). Filters of this type are not expanded during backwash, and stratification of grain sizes does not occur. These filters are generally designed to use air or air/water backwash. There are many possible combinations of filter media size d and depth L. J. M. Montgomery
(1985) presents a methodology for determining the optimum relationship between
these two variables. The relationship between L and ES de (10% finer) of many high-rate filters is shown in Figure 8.1. In this figure, the average ES for a dual- or mixed-mediaCoarse filter was computed as a weighted average. Data in the figure indicate that as the media used become coarser, the required depth is increased, and as the media become finer, the depth required is reduced. Kawamura (1999) uses a length-to-diameter L/d ratio to select the proper depth and size of filter beds (with both L and d measured in millimeters). A value of the L/d ratio should be > 1,000 in rapid sand filters, > 1,250 in trimedia filters, and > 1,300 in most coarse deep beds where d is 1.2 to 1.4 mm and > 1,500 is most coarse deep beds where d > 1.5 mm.


Slow sand filters, designed for filtration rates of 3 to 6 mgd/acre at a rate of 0.05 to 0.10 gpm/ft 2 (0.1 to 0.2 m/h), were initially replaced by rapid sand filters that operated at rates of 1 to 2 gpm/ft 2 (2.4 to 5.0 m/h). The 2 gprn/ft 2 (4.9 m/h) rate became widely accepted as an upper limit in U.S. water supply practice for many years. For the past 30 years, it has been demonstrated that dual-media and mixed-media, as well as single-medium (sand or anthracite), filters can be successfully operated at much higher rates.
A number of investigators found dual- and mixed-media filters to operate successfully at rates from 3 to 8 gpm/ft 2 (8 to 20 m/h) in a variety of locations (Conley, 1961, 1965; Robeck, Dostal, and Woodward, 1964; Dostal and Robeck, 1966; Laughlin and Duvall, 1968; Tuepker and Buescher, 1968; Rimer, 1968; Westerhoff, 1971; Kirchman and Jones, 1972). The quantity of evidence of the practicality of high-rate filtration was such that in 1972 the AWWA Committee on Filtration Problems concluded that it had been amply demonstrated that filters could be designed and operated to produce water of acceptable quality at flows substantially higher than the rate of 2 gpm/ft 2 (5 m/h), once considered the maximum. Over the last 30 years, a number of pilot-scale and full-scale deep-bed uniformly graded anthracite filters have been operated reliably at rates of 10 to 15 gpm/ft 2 (24 to 37 m/h).
Average filtration rates of roughly 2 to 7 gpm/ft 2 (5 to 17 m/h) are reported for the upflow, biflow, and deep-bed filters discussed previously (Hamann and McKinney, 1968; Jung and Savage, 1974). Logsdon et al. (1993) demonstrated that deep-bed monomedia direct filters could be operated at filtration rates up to 9 gpm/ft 2 on higher-turbidity waters (up to 60 ntu) with proper chemical pretreatment and polymer filter aid selection. Filtration rates are impacted by water temperature. Generally, when water temperatures drop below 45 ° F (8 ° C), water quality and filter run length deteriorate in high-rate filters (Kawamura, 1999). Many regulatory agencies will not approve rates in excess of 4 to 5 gpm/ft 2 (10 m/h) without successful pilot-scale testing. The designer should make every effort to obtain approvals for operation at higher rates. The quality of the raw water and extent of pretreatment will play a large role in the acceptable filtration rate.