FILTRATION
Design and operation of high-rate filters Practical observations improve the design and operation of filtration systems. Susumu Kawamura
W
hen designing a filtration system, a design engineer can specify certain parameters: type of filter; number of filters; size, type, and thickness of the filter bed; filtration rate and filter control system; filter washing system; type of filter underdrain; available head loss; filter-to-waste (rewash); and the potential for future conversion of the regular filter media to granular activated carbon (GAC). Moreover, certain hydraulic principles and more general principles of physics must be apFiltration is the fundamental system in a water treatment process plied if design and operatrain that removes suspended solids, including microorganisms tion problems are to be such as Cryptosporidium oocysts, Giardia cysts, and parasite eggs. To avoided. (This article does constrain costs, the design of any filter should be simple, reliable, not discuss the theory of proven, easy to build, and easy to operate; and it should require filter design and hythe minimum capital, operation, and maintenance costs while draulics, which are deproviding maximum operational flexibility. flexibility. A robust filter design scribed in several refershould be based on knowledge, experience, and user feedback. ence books.1–6) This article provides practical guidance on the design of high-rate filtration systems. For executive summary, see page 182.
DECEMBER 1999
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
77
FIGURE 1
Performance of filters with four L/d ratios at 7 gpm/sq ft (17.5 m/h)
0.70
0.60
0.50
u t n 0.40 — y t i d i b r 0.30 u T
Filter 2 (L/d = 615)
Filter 1 (L/d = 790)
0.20 Filter 3 (L/d = 930)
0.10
Filter 4 (L/d = 1,100)
0.05 0
1
2
3 4 5 Time—hours after start
6
7
8
Source—Greeley and Hansen, Pilot Filt er Test, report for South Water Treatment Plant (1995)
TABLE 1
Physical characteristics of filter beds
Filter Bed
Total Depth in. (mm)
Dual-media
30 (750)
Trimedia
33 (825)
Coarse deep monomedia
Dual coarse deep bed
72 (1,800)
72 (1,800)
Type and Size of Media
20 in. (500 mm) of anthracite Effective size = 1.0 mm, uniformity coefficient = 1.4 10 in. (25 mm) of sand Effective size = 0.55 mm, uniformity coefficient = 1.4 18 in. (450 mm) of anthracite Effective size = 1.0 mm, uniformity coefficient = 1.5 12 in. (300 mm) of sand Effective size = 0.55 mm, uniformity coefficient = 1.5 3 in. (75 mm) of garnet Effective size = 0.30 mm, uniformity coefficient = 1.5 72 in. (1,800 mm) of anthracite Effective size = 1.4 mm, uniformity coefficient = 1.4 60 in. (1,500 mm) of anthracite Effective size = 1.4 mm, uniformity coefficient = 1.4 12 in. (300 mm) of sand Effective size = 0.73 mm, uniformity coefficient = 1.4
pretreatment and quality of filter influent; and site topography. Only the first three are discussed in this article. Conditions at the plant site, such as climate, the availability of qualified plant operators, and how the plant is financed are perhaps the most important determinants of filter type. If a private contractor provides design–build–operate services, the filter installed typically is a proprietary filter of an equipment manufacturer that is under the umbrella of the financing institution. Proprietary filters supplied by equipment manufacturers are generally used by small plants (< 3 mgd [0.011 X 10 6 m3/d]) because they are cost-effective, allow more rapid design and construction, and are easy to operate and maintain. Number of filters. The number and size of each filter are closely related to site configuration and the size of the water treatment plant. At least four filters are recommended for midsized plants (10–30 mgd [0.038–0.11 X 106 m3/d]); small plants processing < 10 mgd (0.038 X 106 m3/d) may use two filters if financing is limited. If a plant has few filters, then the filtration rate among the remaining filters substantially increases whenever one or two of the filters is placed off line for washing or repair. The author recommends the following formula as a guide in determining the required number of filters based on plant capacity. It is based on the analysis of about 80 treatment plants designed after 1970 in the range 1–600 mgd (0.004–2.3 X 10 6 m3/d). N = 1.2 Q0.5
Filters Type of filter. The first decision, type of filter, is commonly dictated by seven conditions: local conditions; type and amount of funding and capital; treatment plant size; design requirements established by local regulatory agencies; raw water quality; type of
78
VOLUME 91, ISSUE 12
in which N is the total number of filters and Q is the design flow rate of the plant in mgd. If a self-backwashing filter* is to be used, then each module should have four—or preferably six— *Greenleaf filter, Infilco Degremont Inc., Richmond, Va.
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA
If backwash waste is to be recycled, it should be held in a waste holding tank that can retain the waste of at least two washings and preferably three or four.
filters even in a small plant. Otherwise, during periods of low plant flow, the system may not be able to produce the volume of backwash water required to wash one filter unless supplemental water is pumped into the effluent channel. Size. Maximum filter size is dictated by three things: the need to provide uniform flow distribution of backwash water throughout the entire filter bed; the economically justifiable size of the filter backwashing system, pumps, and blowers; and the cost of the waste-wash handling facilities. In general, the obtained from pilot filter studies performed during practical maximum size of an individual filter bed is winter through summer conditions, and the L/d 2 about 1,000 sq ft (90 m ), provided that the plant is ratio—a simple parameter that is a reliable and proven not exceptionally large. However, a filter may be indicator. L/d ratio. Pilot studies should be conducted for designed as two filter beds sharing a central gullet so that one filter can have up to 2,500 sq ft (225 m 2) of at least six months to yield reliable data, but their filter bed area for extremely large plants. Filters may cost may exceed a client’s budget. Thus, the L/d ratio, be designed to wash only one side-cell at a time, thus reducing the size of the washing system but requirenerally, a standard dual-media or coarse ing twice as much time to wash the entire filter bed at deep bed may be used for high-rate once. Therefore, the filter filters. may be inoperable for almost 1 h during washing.
G
Filter media Basic gravity filters are composed of granular media of a certain size and depth. Sand, anthracite coal, and garnet (or ilmenite) are most widely used. Filter beds may be designed as monomedia, dual media, trimedia, or mixed media. The proper media and the size and depth of the filter bed will help maximize filter efficiency. Composition. In addition to the granular media listed previously, pumice or synthetic materials are also used provided there is no measurable leaching of these materials from the media, no significant attrition, and the cost is reasonable. For example, pompon-shaped synthetic fiber balls of about 1 in. (25 mm) in diameter are used in filters such as the Ishigaki filter developed during the early 1980s in Japan and in the recently marketed “fuzzy filter” of the United States. The filtration rate of these types of filters has been reported to be as high as 20 gpm/sq ft (50 m/h). Depth. Two basic kinds of information are used to select the proper depth and size of filter bed: results
DECEMBER 1999
in which L is the bed depth in millimetres and d is the effective size of the media in millimetres, may be used instead. It is based on more than 200 pilot studies and the performance data of many operational filters. The value of the L/d ratio should be > 1,000 in rapid sand filters and standard dual-media filter beds, > 1,250 in regular trimedia filter beds, > 1,300 in most coarse deep beds in which d is 1.2–1.4 mm and L is 1.8–2 m, and > 1,500 in most coarse deep beds in which d is 1.5 mm. When the diameter of the media exceeds 1.5 mm, the space between the grains becomes large compared with the void space in regular filter beds. The void space triples when the diameter of the grain is doubled. Thus, the L/d ratio should be used as an estimate only when the media is > 1.5-mm diameter. In recent years, filters have been designed to produce a filtered water turbidity 0.1 ntu. In dualmedia, multimedia, and coarse deep-bed filters, this low turbidity is achieved by preloading a small amount of polymer as a filter aid. Nonionic, highmolecular-weight polymers are generally used at a
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
79
flocculation). Filtration rates were 6, 9, and 12 gpm/sq ft (15, 22.5, and 30 m/h). Filter runs were terminated Case 1 t f Unit filter run volume Particle removal when either the head loss q12,000 2 500 2.5 s m reached 10 ft (3 m) or the / / l 3 a g 9,500 m effluent turbidity exceeded 2.0 —400 — e e 0.1 ntu (Figure 3). m m u u 7,000 The unit filter run volume l 300 1.5 l o o V V (UFRV), expressed as gal/sq ft n n 1.0 u 4,500 u 200 (m3/m2), is a product of the R R r r e e filter run length and the fil t t l i 100 i 2,000 l 0.5 F F tration rate. In this case, the t t i i n n goal of UFRV is 9,820 gal/sq 0 U 0 0.0 U L/d L/d L/d L/d L/d L/d L/d L/d L/d ft (400 m 3/m2) because this is 1,250 1,875 2,500 1,250 1,875 2,500 1,250 1,875 2,500 the condition attained by one Filtration rate = 15 m/h Filtration rate = 22.5 m/h Filtration rate = 30 m/h (6 gpm/sq ft) (9 gpm/sq ft) (12 gpm/sq ft) filter wash per day conducted at a filter rate of about 7 Initial conditions —a lum dose = 2 mg/L, cationic polymer = 0.75 mg/L, GT = 54,000; unit 3 2 gpm/sq ft (17.5 m/h). filter run volume goal —4 00 m /m (9,820 gal/sq ft); log particle removal goal —2 .0 log When the diameter of the Source —P ilot study, Symore Water Treatment Plant (1997) media remained constant at 1.2 mm, a deeper filter bed was more advantageous because it produced a higher L/d typical dosage of 15–25 µg/L. If a filter aid is not ratio. Under the tested conditions, an L/d ratio of used, then the L/d ratio will need to be increased by 2,500 achieved the desired UFRV at filter rates up to a minimum of 20 percent. Additional increase in L/d 9 gpm/sq ft (22.5 m/h). It was assumed that the filis not cost-effective because of the deeper filter structer bed may be either monomedia or dual-media with ture, the higher initial head loss, and a longer filter the same L/d ratio. This assumption was confirmed washing time that does not appreciably improve filusing either 2 or 4 mg/L of alum with cationic polyter performance. mer (Figure 3). The results not only confirmed the When GAC is used as the adsorption media, the assumption but also demonstrated that placing a sand depth of the filter bed is governed by the required layer (a barrier) at the bottom of the filter bed empty bed contact time (EBCT), and the L/d ratio is improved filter performance even if the total bed not a design indicator. However, when a GAC bed depth of the dual-media filter was shorter than that must accomplish both filtration and adsorption, both of the monomedia filter. In this experiment a high the EBCT and the L/d criteria should be used to ratio of cationic polymer to alum as coagulant determine the depth of the filter bed. improved overall filter efficiency. Pilot studies. For example, consider the relation between filtered water tur bidity and the L/d ratio of a filter bed at 7 gpm/sq ft nalyses of core samples taken (17.5 m/h) (Figure 1). from the filter bed provide substantial These data are based on a pilot study for a conveninformation about the condition tional treatment process. 7 of the filter. An L/ d ratio > 1,000 strongly affected filter performance, and the quality Settling velocity. In dual and trimedia filter beds, of filtered water improved. However, further increases in filtration rate have correlated with higher L/d and the size of the media must be carefully matched to distinctly better filter performance. provide similar settling velocities. The size of the Next, consider the results of a pilot study congrains and the specific gravity of each layer greatly ducted for a direct filtration process of lake water determine the effectiveness of filter backwashing on the northwest coast of Canada (Figures 2 and 3). because of the potential loss of filter media or the Each filter run was to be longer than 24 h, and there heavy accumulation of floc or sludge in the mixedwas to be no more than one filter wash a day. The media zone. An inappropriate combination of media coagulants were alum and two types of cationic polywill allow only a limited portion of the filter bed to be mers, and after optimization tests, mixing conditions adequately cleaned; the rest of the bed will either were set as Gt = 54,000 ( Gt represents the degree of remain dirty, or a large portion of the media will be FIGURE 2
Performance of filters using 1.2-mm anthracite media at three bed depths
A
80
VOLUME 91, ISSUE 12
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA
washed away at a specific backwash rate. Figure 4 illustrates the accumulation of a large amount of floc or sludge in the anthracite and sand mixed zone despite air-scouring backwash of the filter at this lime-softening process plant. To ensure that two types of filter media grains have the same settling velocity, the following relationship should exist: d 1
=
d 2
– 0.667
2
1 –
FIGURE 3
Performance of three filters with different bed design Case 2
t f 14,500 2 600 q m s / / l 3 a12,000 m500 g
—
e m 9,500 u l o V 7,000 n u R 4,500 r e t l i 2,000 F t i n U 0
Unit filter run volume
Particle removal 3.0 2.5
—
e m400 u l o V300 n u R200 r e t l i F100 t i n U 0
2.0 1.5 1.0 0.5
Alum 2 mg/L
Alum 4 mg/L
Dual Media/Shallow Sand Layer 2,700 mm, 1.4 mm; 300 mm, 0.7 mm L/d = 2,350
Alum 2 mg/L
Alum 4 mg/L
Alum 2 mg/L
Dual Media/Deep Sand Layer 1,800 mm, 1.4 mm; 900 mm, 0.7 mm L/d = 2,500
Alum 4 mg/L
0.0
Fine Monomedia 3,000 mm, 1.2 mm L/d = 2,500
in which d 1 and d 2 are grains of two media with specific Initial conditions —G T = 54,000; filtration rate = 15 m/h (6 gpm/sq ft); cationic gravities of 1 and 2, and is polymer —2 .5 mg/L at 2 mg/L alum; 1.6 mg/L at 4 mg/L alum; unit filter run volume 3 2 goal —4 00 m /m (9,820 gal/sq ft); log particle removal goal —2 .0 log the density of water with a specific gravity of 1.0. Source —P ilot study, Symore Water Treatment Plant (1997) The size relationship among the three kinds of media in trimedia may also be determined by the formula. Because anthracite coal specific types of filters and filter beds when one filabsorbs water and thus increases its density by many ter is off line. Most states restrict the maximum rate percentage points, the density of anthracite coal media of regular rapid sand filters to 3 gpm/sq ft (7.5 m/h) should be measured after 24 h of soaking (to simulate and of dual-media filters to 4–5 gpm/sq ft (10–12.5 working conditions). To minimize intermixing of the m/h). The West Coast restricts the maximum rate of two or three layers, the size of each successive layer, dual-media and coarse deep filters to 6 gpm/sq ft from top to bottom, should be 5–7 percent larger (15 m/h). Higher rates may be allowed if the design than the size calculated by the equation. engineer presents an extensive filter pilot study that Uniformity coefficient. The uniformity coeffi- justifies those rates to the satisfaction of a reviewing cient (uc) of granular media should be specified. authority. For example, dual-media filters have run Smaller uc values result in slightly better filtrate at 8 gpm/sq ft (20 m/h) and as high as 13 gpm/sq ft quality and longer filter run lengths (i.e., better fil(32.5 m/h) in plants fitted with a 6-ft-deep coarse ter performance), and, in the case of dual media or (effective size, 1.4 mm) deep bed monomedia filtrimedia beds, a smaller mixed-media zone after fluter. In these cases, the higher rates were justified idization backwash. Based on proven filter perfor by extensive pilot studies using raw water of good mance of numerous operational treatment plants, quality from protected sources that was pretreated a uc of < 1.40 and preferably 1.30 should be speciwith ozone. fied in the design of any filter bed. Contrary to popular belief, the quality of filtered Grain shape. The shape of grains in the filter media water produced by high-rate filters is not substantially may affect filtration. Angular grains usually perform affected by filtration rates up to 10 gpm/sq ft (25 m/h) better than do rounded or worn grains when polyprovided that the water is properly pretreated. Susmer is not used as a filter aid, because of the larger pended colloids in raw water should be coagulated by porosity ratio and possibly the availability of more a combination of metallic coagulant and a small adsorption sites on each angular grain. This phenomamount of polymer as a filter aid, and the excess susenon has been observed since the 1940s. Yet, few of pended solids should be removed during pretreatment. today’s filtration plants detect any notable increase in However, the rate at which head loss develops during filtered water turbidity because of “worn-out filter higher filtration rates is quite dramatic. media,” provided that polymer is fed as a filtration aid. Generally speaking, either a standard dual-media The physical characteristics of standard dualor coarse deep bed may be used for high-rate filters. media, trimedia, and typical coarse deep filter beds are Yet, when the water temperature falls below about given in Table 1. 45oF (8oC), filtered water quality deteriorates and filter run length decreases in both high-rate and rapid Filtration sand filters that use regular alum pretreatment. Many Rates. In the United States, all state regulatory case histories have shown that under these condiagencies restrict the maximum filtration rates of tions it is difficult for filters to produce < 0.2-ntu fil-
DECEMBER 1999
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
81
FIGURE 4
0
. n i
—
d e B f o e c a f r u S w o l e B h t p e D
Floc retention on 100 mL of filter media
l a o c e t i c a r h t n A d n d a l n a a o s C
5
10
15
0
125
Before filter wash
— 250 d 375
500
20
d n a S
25
After air-scouring wash
625
30
750
35
875 0
m m
200
300 400 500
1,000
2,000 3,000 4,000 5,000
e B f o e c a f r u S w o l e B h t p e D
10,000
Turbidity —ntu
TABLE 2
Characteristics of surface backwash systems
Type
Flow Rate gpm/sq ft (m/min)
Driving Pressure psi (kPa)
Fixed nozzles Rotating arm
3–4 (0.12–0.16) 0.5–0.7 ( 0.02–0.03)
20–25 (140–170) 70–100 ( 480–690)
Lateral Spacing ft (m)
About 3 (0.9) Maximum diameter 14 (4.3)
A self-backwashing filter,* developed in the mid-1970s, is a reliable constant-flow-type control scheme. This system eliminates the need for backwash headers, valves, pumps, and an elevated washwater storage tank. Yet the filter system does have drawbacks: the backwash rate is not changed as easily as it is in regular filters, there is no filter-to-waste system on standard designs, the filter structure is about 4 ft (1.2 m) deeper than in ordinary filters, and influent water cascades > 6.5 ft (2 m) into the filter cells during the early stages of the filtration cycle. Thus, it is recommended that a small amount of polymer be continuously added to the filter influent to strengthen the alum floc. The Metropolitan Water District of Southern California uses a self-backwashing filter in two of its five treatment plants; the total production rate of these plants is 1,080 mgd (4 X 10 6 m3/d).
tered water. One practical solution is to use ferric iron Filter washing coagulant with cationic polymer as coagulants. Ferric and underdrains Filter wash systems. Filter wash systems fall floc forms several times as fast as alum floc in cold water sources, and the polymer physically strengthens into two basic types: fluidized bed backwash with or the floc. Substituting polymerized aluminum chlowithout surface wash and air-scour with or without ride for alum may reduce this problem during the low-rate backwash (Table 2). winter. Even with modifications to the coagulant, the Fluidized-bed backwash. The surface-wash sysrecommended cold-weather safe filtration rate for tem has been widely used in the United States since high-rate filters under a normal conventional process the early 1950s and has been effective when properly is generally around 6 gpm/sq ft (15 m/h). designed and operated. However, many case histories Control of rates. Filtration rates may be conof treatment plants operating during the past 50 years trolled by constant-rate or by declining-rate schemes. demonstrate that backwash alone does not maintain The current standard is constant-rate filtration; it is granular filter beds in a reasonably clean condition, a proven method that provides better operational control of the filters, has proven performance, is better able to meet the criteria roper filter design alone will not yield set by regulatory agencies, good filtered water. and is overwhelmingly favored by plant operators. The declining-rate filtration system8–10 is simple in its design and operation even when optimal backwash rates were used. An becau se of the lack of a flow-modulating system. auxiliary scouring system is necessary whenever coagHowever, it delivers the highest filtration rate at the ulants are used in the pretreatment process. beginning of the filtration cycle even when a flowBoth of the surface-wash systems (Baylis-type restricting device such as an orifice plate is fitted in fixed-nozzle 11 and rotating-arm with jet nozzles) are the effluent pipe, and it allows plant operators little effective to depths of 4 ft (1.2 m) (Table 2). Neveroperational control. Thus, few modern filtration theless, the fixed-nozzle system is preferred because plants use this system. of the lack of moving parts and because, unlike the
P
82
VOLUME 91, ISSUE 12
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA
rotating-arm system, it reFIGURE 5 Movement of media during air-scouring mains effective even when the filter bed has lost several inches (0.15–0.2 m) of filter media. Moreover, rotatingExpanded bed level Boil arm systems may still leave filter beds dirty because of clogged jet nozzles, slow or e m nonrotating arms, and im n ) . o m n z 0 i proper distance between the g 5 0 n 2 1 i nozzles and the filter bed. The – – l i 0 ( o 5 6 o water jet angle is only 15 B 1 below the horizontal and in most cases the arm is only s e about 2 in. (5 cm) above the t l n z Ascending air e z bed. In contrast, a fixed-noz o bubbles with m s n e e l zle system uses a water jet of mushroom v n z o z e shape e o m about 25 fps (7.5 m/s) at an m n t w d u o r i e e t angle of 27–35 and thus can a d v b n e o n w e scour more than half the n M b o m w f a i e t o o n o v depth of regular filter beds. o o D n i m t o m Air-scour. Air-scour effec i t o d r d c m a r e tively cleans some filter beds. a w v d r w n a n p However, air-scouring filter o w w U C p o d wash systems vigorously agi u t w s l tate (and thus clean) only the a o F S • • top 6–10 in. (0.15–0.25 m) of the filter bed. This surface boiling may make it seem that the entire filter bed is violently Air agitated. However, some mud Strainer nozzle balls that have fallen below Sludge the boiling zone will remain deposit Plenum unbroken. Water Mushroom-shaped bub bles of compressed air emanating from the bottom strainer nozzles in an underdrain will slowly rise to the surface almost directly above each nozzle. The bubbles will persist until near the top run time before backwash, then it is worthwhile to of the bed. The higher the air-scouring rate, the more consider a surface wash system in conjunction with direct the vertical rise (i.e., the bubbles will not spread an air-scouring system that has strainer nozzles in out sideways). Between the upwelling streams of the underdrain. Applying a surface wash and then bubbles, bed media is pulled down (in accordance air-scour will break up the top 8 in. (20 cm) or so of with the mass balance law). This behavior is more compacted media. distinct when a slow backwash is applied with comAn air-scouring system complicates not only the pressed air (Figure 5). design and construction of the filter but also filter operThe two types of air-scouring are sequential and ations, as air blowers, air piping, and air and backconcurrent. Both perform well when the systems are wash sequence control are required. Air-scour requires properly designed and operated (Figure 6). In Europe, twice as much filter-washing time as the surface-wash sequential air-scouring or air-scouring without backsystem. Furthermore, air-scouring systems appear to wash typically is used. In the United States concurrent have a higher incidence of catastrophic underdrain air-scouring is more common (6–8 gpm/sq ft [15–20 upsets because of the explosive nature of compressed m/h] of slow backwash and compressed air). Pilot air. A standpipe should be placed in the backwash filter studies and core sampling of a few filters that use pipeline as a pressure relief measure for an air-scourair-scouring wash systems confirm the increased accuing wash system. Its size need be only 25 percent of the mulation of sludge deposited in the lower portion of size of the backwash pipe, and its height is normally 6 the filter bed.12 ft (1.8 m) above the top slab of the filter structure. If operational practices allow > 7 ft (2.1 m) of *Greenleaf filter, Infilco Degremont Inc., Richmond, Va. head loss to build up or more than two days of filter
DECEMBER 1999
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
83
Monitoring by operators. Consistent observation by the plant operating staff is a key component of a properly A —Recommended surface and backwash sequence functioning system. The plant 0 1 2 3 4 Surface washing —min operator must regularly Surface wash observe the filter bed, espeConcurrent Backwash— m in wash cially its top, during the filtraBackwash tion cycle and backwashing. 0 1 2 3 4 5 6 Large cracks on the top of the 1. Draw down water level to about 4 in. (100 mm) above the bed bed or a channel along the before surface wash starts. 2. Longer concurrent wash will result in greater loss of medium. sidewalls are signs of a dirty filter bed. Boiling of the media B —Normal concurrent air and water air-scouring wash for common during backwashing may be a filter bed sign of gravel bed and under20 drain failure; if the top of the 18 t f 40 t 16 f q filter bed contains concave Air purge washing (optional) h s 14 / h q / / s m spots, then underdrain failure / m 12 Pause m p — m f g 10 r (optional) has most likely occurred. Noz — c e i r — 8 20 t r a — A zles of the surface-wash sys r t e i W A a 6 tem commonly become Slow backwash W 4 60 clogged with media. Any 2 Air scour 0 0 0 unusual phenomenon must 0 5 10 15 be recorded in a log so that Washing Duration —min adjustments or repairs can be made. 1. Draw down water level to about 4 in. (100 mm) above the bed. 2. The first slow backwash must be stopped before the water Wash troughs. The relevel rises to 6 in. (150 mm) below the top waste weir. quired number of troughs and 3. Backwash in rinsing stage requires about 18 gpm/sq ft (45 m/h) the top elevation above the to purge air in a coarse-medium bed as well as restratification of dual-media bed. filter bed can be controversial. For example, some European-designed filters do not include wash troughs with an air-scouring system. The Underdrains. The filter underdrain system design criteria for wash troughs in the United States depends on the type of filter-wash system. If airwas first delineated in Water Treatment Plant Design scouring is incorporated into the filter design, then the Manuals of Engineering Practice.13 The Ten States Stanfilter underdrain system is almost always limited to dards, 14 first published in 1953 and revised several the proprietary strainer nozzles with a false bottom or times since, lists the parameters recommended for a plastic block system for air-scouring wash. In contrast, several proven types of filter underdrains are available for use with a suronsistent observation by the plant face-wash system with backwash. operating staff is a key to good filtered The most important water. characteristics of filter underdrain systems for either surface-wash or air-scouring systems are that features such as the top elevation of the trough in the system has proven satisfactory, has an adequate relation to the top of the filter bed and spacing of anchor system, and is manufactured by a company the troughs. that practices good quality control. Moreover, conTwo types of wash troughs are commonly used: struction of the underdrain system should be superthose with a wide, shallow cross section and a vised by an experienced construction supervisor. A nearly flat bottom, and those with narrow, deep plenum type of underdrain system should be set in a cross sections that have a U-shaped or V-shaped monolithically cast reinforced-concrete false floor. bottom. Troughs with wide cross sections produce Avoid the use of precast elements that must be bolted higher upflow velocity when the backwash flow onto pedestals, especially for air-scouring, because exceeds the elevation of the trough bottom. after several decades of use the cement grout will fail Troughs with narrow cross sections have greater and massive amounts of air will leak around the blocks. structural integrity that permits their fabrication FIGURE 6
Recommended filter-washing sequence
C
84
VOLUME 91, ISSUE 12
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA
from thinner materials such FIGURE 7 Filter backwash rate at water temperature of 20 oC (68oF) as fiberglass. Sequential air-scour. Un4.0 Specific gravity of mineral variety like common fluidized back30 1.2 wash systems, sequential airscouring systems do not use 1.70 4.0 1.70 wash troughs across the filter 1.65 bed. Rather, a single overflow 25 1.0 1.55 wall along a central gullet and 4.1 a single V-shaped trough along 2.65 the side wall opposite the 4 0 5 1 . 3 t . n f 20 0.8 wash-waste overflow wall o 1 b n q r i 2.60 a s / c provide the surface-sweeping m d / d t n m t e a e m p l function. The width of each a n S a g i v r — t o a c e c — filter cell must be limited (to t a 5 G e e 5 r a . t t 15 i 1 0.6 R l a a c about 13–14 ft [4–4.5 m]) for n u s i n R r a h e a r h r s t h the wash-waste to be effec G a n t i c s e A a w h t k n tively removed. The distance w c S y k a c between the top of the over B a 0.4 B10 flow wall and the top of the filter bed is about 1 ft (0.3 m) in most French designs; this feature enables the total filter 0.2 5 box to be much shal lowe r than in filters with sequential washing. Concurrent air-scour. Con0 0 current air-scouring filters 0 0.5 1.0 1.5 2.0 2.5 3.0 have wash troughs arranged 60 percent weight grain size calculated as effective size t imes uniformity coefficient in a fashion similar to troughs Source —K awamura, S. Integrated Design of Water Treatment Facilities. John Wiley & in historical filter designs, Sons, New York except for trough elevation. The top of the wash troughs should be about 6 ft (1.8 m) above the top of the bed because roughly 8 gpm/sq cell from the influent–wash-waste gullet should be ft (0.33 m/min) of slow backwash is applied for about carefully fixed. With respect to regular high-rate fil5 min in conjunction with compressed air. The airters, the top of the forebay wall should be higher scouring wash should begin with a water level of than the lip of the wash troughs and ideally higher about 4 in. (100 mm) above the bed in a filter cell. than the high-water level in the filter cell. In many Because the water above the bed during air-scouring high-rate filters in which the top of the gullet wall is contains a large number of grains of media, it should level with the wash troughs, filter media migrate not be overflowed into the wash troughs. Thus, a because the filter influent is unevenly distributed. distance of 6 ft (1.8 m) is essential. The elevation of existing wash troughs needs to be Troughless filters. In common high-rate filters assessed if the filter is renovated. In rapid sand filthat use dual-media or trimedia beds and surface ters, the top of the wash troughs is located 24–30 in. wash, the top of the wash troughs is generally (600–750 mm) above the bed. In most cases, the located about 4 ft (1.2 m) above the bed, and the troughs are manufactured of reinforced concrete. If troughs are spaced 8–10 ft (2.4–3 m) on center. Yet the bed is converted to either dual media or trimedia, the total number of troughs may be reduced witha large amount of anthracite media will be lost. out harming the effective washing of filters. The city Although relocating all wash troughs is costly, adding of Sacramento, Calif., evaluated troughless filters at baffles around the troughs is one relatively inexpenone of its two filtration plants (100 mgd [0.38 X 10 6 sive solution. A hydraulic model study can help 3 m /d]) designed in 1961. All existing wash troughs determine an effective arrangement of baffles. Spewere removed from one of the 16 filters in 1992. cially designed baffles installed on both sides of each After several years of continuous operation, the trough could reduce loss of GAC media by 70 percent troughless filter was determined to be in a condieven when a backwash rate of 21 gpm/sq ft (52 m/h) tion comparable to that of filter beds fitted with is used.16 wash troughs, provided that the washing time was Backwash rate roughly 20 percent longer. 15 Elevation of forebay wall and troughs. The eleThe proper backwash rate should be based on the vation of the forebay wall that separates the filter grain size of the filter media, its specific gravity, and
DECEMBER 1999
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
85
FIGURE 8
Turbidity of filtered water at beginning of filter run
0.5 Plant A Filter 5 (after 12-min surface wash)
0.4 0.3 0.2 0.1 0.0 0
5
10
15
20 Time —min
0.5 Plant A Filter 5 (after 6-min surface wash)
0.4 0.3 0.2 0.1 0.0 0
5
10
15
20 Time —min
0.5 Plant A Filter 1 (air-scour wash)
0.4 0.3 0.2 0.1 u t n
0.0 0
—
y t i d i b r u T
10
20
30 Time —min
40
50
60
Plant B Filter 15 (air-scour wash)
0.5 0.4 0.3 0.2 0.1 0.0 0
10
20
30 Time —min
40
50
60
1.0 Plant C Filter 11 (surface wash)
0.8 0.6
in which t is the temperature in o C, v b – t is the backwash rate at temperature t , v b – 20 is the backwash rate at 20 oC, and µ is the water viscocity in centipoise at temperature t . Figure 7 applies only to a media with a uc 1.5 and only if fluidization of the filter bed is required (fluidization air-scouring filter wash). A backwash rate of at least 18 gpm/sq ft (45 m/h) is necessary to purge air bubbles trapped in coarse deep filter beds after air-scouring. Some treatment plant operators calculate the appropriate backwash rate as that which provides 50 percent filter bed expansion during backwash. During the late 1950s Baylis and others 11,17 emphasized that 50 percent sand expansion should be achieved for rapid sand filters. This requirement is justified when the sand bed is composed of small grains (effective size = 0.40–0.45 mm), but few filters were fitted with auxiliary scouring (surface wash) systems during the late 1950s. The preferred expansion rate for high-rate filters during backwash with surface wash is about 37 percent for a typical sand bed with a porosity ratio of 0.45; the expansion rate for an anthracite bed with a porosity ratio of 0.50 is about 25 percent. These figures are based on operational data of a prototype filter and the equation for optimal expansion rate, 0.22
(0.1) – f Optimal expansion rate = 1 – (0.1) 0.22
0.6 – f = 0.4
0.4 0.2 0.0 0
10
20
30
40
Time —min 1.0
Plant C Filter 12 (surface wash)
0.8 0.6 0.4 0.2 0.0 0
10
20 Time —min
the high and low water temperature of the region (Figure 7). The adjustment of backwash water for water temperature can be calculated using the following equation: v b – t = v b – 20 X µt – t /3
86
in which f is the porosity ratio.
VOLUME 91, ISSUE 12
Filter-to-waste
A filter-to-waste system diverts inferior filtered water to the waste line. Inferior filtered water may be generated not only under abnormal conditions but also at the beginning of the filtration cycle under normal conditions. Inferior filtered water. Turbidity breakthrough that leads to inferior filtered water is commonly observed when pretreatment is inadequate and filter beds are unripened. Inferior filter water may be produced under these abnormal conditions: (1) new unripened filter media—the bed cannot effectively capture suspended solids because it has not been properly conditioned by a small amount of polymer fed as a filtration aid or by the deposition of the correct quantity
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA
of appropriately charged floc on the filter FIGURE 9 Filter-to-waste flowsheet media; (2) after disturbance of the filter bed, such as the repair or modification of the gravel bed or underdrain system; 0.3 m air gap above the water surface (no cross-connection) (3) accidental initiation of surface wash or air-scouring during the filtration cycle; Influent channel (4) liquefaction of the filter beds as the result of strong earthquakes; or (5) a wrong coagulant dosage. Although almost all filters are equipped with a To washwater holding tank drain pipe, it is too small to divert filtered Sampling tap for water flow to the waste line. particulate counter Filters Polymer. The US Environmental ProTurbidimeter tection Agency (USEPA) turbidity goal for filtered water is now 0.1 ntu. The Flow meter departments of health of some states, including California, set a more stringent guideline by advocating the wasting of filtered water with turbidity > 0.25 ntu. The California Department of Health SerFilter-to-waste P pump(s) vices endorses slow starting of filters after M M M M filter washing to circumvent initial tur bidity breakthrough. USEPA and a few states allow the addition of coagulant or polymer to the Clearwell backwash water as an alternative to filterto-waste. This practice is intended to condition the filter bed before the filtration cycle begins. Unfortunately, it is not always successful. Floc formation and Special features: sludge deposition in the clearwells have 1. No additional capacity is needed for washwater holding tank(s). 2. Slow filter starting can be achieved (3 4 5 gpm/sq ft [7.5 10 been reported when this alternative was 12.5 m/h]. used. Despite the drawbacks, polymer is 3. Filter-to-waste rate can be matched to actual seasonal filtration rate. commonly added as a coagulant because 4. Instrumentation and control can be simplified. 5. Turbidimeter in the recycle line can determine the filter-to-waste time. it is a cheaper method of meeting regu6. Filtered water of questionable quality can be discharged to waste line. latory requirements. Polymer may be added at the influent of the filter at the beginning of the filtration cycle rather than to the backwash water. A number of treatment plants have successfully adopted this pracoperated have consistently produced filtered water tice. A substantial initial turbidity breakthrough is of < 0.1 ntu and fewer than 50 particulates/mL, unusual in well-ripened filter beds. Examples of inidespite an average influent turbidity of 3–6 ntu. tial turbidity breakthrough patterns are shown in The Aqueduct Filtration Plant of Los Angeles and Figure 8. the Utah Valley Water Treatment Plant both have Each new filter should be fitted with a filter-todirect filtration processes and have successfully waste piping system to manage initial turbidity breaktreated raw water turbidity spikes as high as 30–50 through under normal—or abnormal—plant condintu lasting two to three wee ks. (To do so, the filter tions. Among the several ways to manage filterwas washed every 4–6 h.) to-waste, the system depicted in Figure 9 should be The turbidity of the filter influent and removal of considered. protozoan cysts are determined by a treatment plant’s site and the treatment processes. To achieve an approMiscellaneous issues priate influent may require a high degree of clarificaTurbidity of filter influent. One of the most tion before filtration. Under certain conditions, adecontroversial issues in pretreatment is the turbidquate filtration may follow proper coagulation– ity of the filter influent. Some regulatory agencies flocculation, thus eliminating the need for sedimenadvocate that filter influent turbidity should be < 2 tation before filtration. If the filter requires an influntu or even < 1 ntu. Certain treatment plants ent turbidity of < 1 ntu to consistently produce good require < 0.5 ntu settled water turbidity to profiltered water, then the filter is not functioning propduce filtered water with < 0.1 ntu. Conversely, erly—either coagulation of suspended solids in raw direct filtration plants that do not have a sediwater is insufficient, or filter beds are inadequately mentation process but that are well designed and conditioned, or both.
DECEMBER 1999
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
87
one filter. The choice of elevated tank or direct pump wash is influenced by the site topography, the size of each Very clean bed Not well ripened filter bed, aesthetics, and local Clean bed conditions such as winter Partially temperature. ripened Reasonably clean Backwashing filters by Ripened direct pumping has the Dirty bed advantage of easy backwash Well ripened Dirty bed with mud rate control without the Provide an effective auxiliary balls need for a holding tank or scouring in addition to backwash Well ripened a long backwash main line. Very dirty bed However, this system reMany mud balls quires a large-capacity Extremely dirty bed Needs to be replaced pump, a high-power motor, 10 20 30 40 50 100 200 300 400 500 1,000 2,000 3,000 4,000 5,000 and a large storage well. Floc Deposition on 100 mL of Filter Media After Filter Wash —ntu There is also the potential to violate the required disinfection time if the clearwell is not large enough. Water depth above filter bed. Air-binding may Backwash recycles. The practice of recycling be minimized with proper design. To obtain reason- backwash waste means that organic constituents may ably long filter runs in water of regular filter depth, be returned to the raw water if the waste is not clarthe water depth above the filter bed must be at least ified. Although wash-waste is commonly clarified in 8 ft (2.4 m). The rule of thumb is to specify a water lagoons, this practice does not remove algae, Cryptodepth equal to or higher than the available head loss sporidium oocysts,18,19 organic compounds that impart for filtration, which is usually 8 ft (2.4 m). However, bad taste and odor, or precursors of disinfection bymost patented automatic backwash filters commonly products. Sludge from sedimentation tanks should allow a water depth of only 1 ft (0.3 m) or less. On the never be mixed with backwash waste because the other hand, pressure filters generally provide a tersludge contains large amounts of undesirable subminal head loss of > 30 ft (9.3 m). Thus, the water stances such as heavy metals and other materials that depth or driving water pressure is a function of design degrade the quality of finished water. specifications. If backwash waste is to be recycled, then it should The decision to wash filters should not be solely be held in a waste holding tank that can retain the based on a fixed level of head loss during the filter waste of at least two washings and preferably three or run. Turbidity breakthrough and the concentration of four. A shortage of holding capacity prohibits filter particulates per unit volume (i.e., 50/mL) for a cer backwashing and thus becomes a bottleneck in the tain size range (i.e., 2–10 µm) should be used to entire water treatment process. A cylindrical tank determine the end of the filtration cycle. with a tangential inflow concentrates sludge and washed-out filter media on center because of a Filter piping. The most important design issue for filter piping in the pipe gallery is simplicity and cyclone effect, and thus a sludge scraper can be elimaccessibility to all control systems, including valve inated. The stored water is then pumped into a reacactuators. Moreover, the main backwash pipe must tor–clarifier located at a higher elevation than the be locat ed below the water level of the filter cell head of the plant, and the waste is treated at a and preferably below the top of the wash troughs to reduced and constant rate (no greater than 10 percent avoid entraining air in the backwash pipe through of the plant flow rate). The clarified and disinfected leaky backwash valves. Entrained air can disturb water may now be recycled to the head of the plant the gravel bed and upset the filter underdrain. Air by gravity. The concentrated sludge in the clarifier is entrainment may be minimized or even eliminated then discharged to the sludge-handling system at if a master backwash control valve is placed fixed intervals. Filter control and instrumentation. A control upstream of the filter banks. When the wash valve is opened slowly, allowing a flow of only about 8 and instrumentation system should help the plant –1 gpm/sq ft (20 m/h) for a period of 30 s min, the air operator observe filter washing and investigate any in the backwash main will purge every time a filter unusual behavior in a filter. The essential items to is backwashed. A self-backwash filter has the simcontrol and record are filtration rate; head loss across plest filter piping design. the bed; backwash rate and duration; auxiliary scourFilter backwash mode. Filters may be backing rate and duration; and filtered water quality. Conwashed either by an elevated backwash tank or by trol and instrumentation for these items should be direct pumping; self-backwash filters use filter effluprovided at each filter and in the central control room. ent from all the filters remaining on line to wash The capacity for manual control, in addition to autoFIGURE 10
88
Range of filter bed cleanliness
VOLUME 91, ISSUE 12
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA
matic control, is essential at each filter Vertical distribution of media sizes in a filter bed FIGURE 11 for calibrating the automatic system and in case the remote control system is disAnthracite Sand Mixed abled. It is not necessary to have a console for each filter in the case of local 30 control. In fact, one console may furnish control, an indication system, and record m ing for two to four filters. A portable 600 m 25 m m m m plug-in control and indication system 3 m 8 9 2 . 8 m 9 . 0 . may be used for each filter, thus elimi m 0 0 0 . m 1 nating the use of consoles. 1 . 20 1 Other aspects requiring control and . m 9 . n i m 0 instrumentation are the water level in m — — d 400 d m the elevated tank and waste wash water e e 3 B15 B 0 f . holding tank, and water elevation in the f 1 o o h h clearwells. Each filter should be furnished t t p p e e with a turbidimeter (although additional D D 10 particle sizing and counting systems are even better) for process control. 200 Filter housing. The local consoles m m m m m m m m m m should be housed in a building because 5 2 5 2 1 1 5 m 5 5 5 5 . . . . . the computer control mechanisms re 0 0 0 0 0 m 2 quire a dust-free environment and a 5 . 0 constant, cool, ambient temperature for 0 0 proper functioning. The building may Original Filter Filter Filter Filter Filter Design also shelter plant operators during Zone Numbers inside the bars are the effective size of the media. severe weather. Automatic backwash filters certainly require housing; exposing the filters to open air increases the likelihood of contaminating the filtered water. Several articles on filter design by Monk 20,21 are good references. filter beds at least once a year, depending on the Many engineers believe that the removal of 2–13 local conditions. If the filter bed is dirty, it should be µm particulates is a good surrogate measure of the washed longer than usual with a higher backwash removal of protozoic cysts and oocysts ( Giardia and rate and auxiliary scouring such as surface wash. Cryptosporidium) and parasite eggs. Unfortunately, Such washing two or three times in succession may cysts and oocysts can adhere to filter media. Several be needed to clean the filter bed. Although some recent pilot studies 18,19,22–24 report that the removal water professionals firmly believe that a “thorough” of cysts and oocysts may be 1.5–2.5 logs higher than filter wash is absolutely necessary, the author questhe removal of particulates of the same size range tions its effectiveness. The concept is correct in prinunder the same filtration rate. ciple, but the definition is vague, the process is Core sampling of the filter bed. Filter perforwasteful of wash water, and the maturation level mance is commonly evaluated on the quality of the of the filter bed is reduced; therefore, the benefits are filtered water and the length of the filter run under questionable. given operating conditions. These methods, however, Media-grain gradation analysis. A media gradado not provide a complete picture of the condition of tion analysis includes determination of uniformity the filter. Analyses of core samples taken from the coefficient and the specific gravity of the media, effec1 filter bed provide substantially more information. tive size of the media, mixed zone, and bed depth. FigThe analysis should include a filter-media grain ure 11 depicts the physical conditions of each filter gradation profile and a floc-retention profile through- bed compared with the original design conditions out the depth of the bed. Because the core sampling (specifications) for a 120-mgd (0.45 X 10 6–m 3/d) process also measures the filter bed depth, the topogplant. A frequent surprise of trimedia filter beds is raphy of the support gravel bed mounding can be the loss of practically all of the small, high-density established (provided a sufficient number of core media, generally garnet or ilmenite, after two to three samples are taken). years of regular filter use. Floc retention. The cleanliness of the filter bed Summary and effectiveness of the filter washing system and technique may be evaluated by considering floc Filtration is the final barrier to removing susretention (expressed as turbidity) for the cored sampended solids such as cysts, oocysts, and parasite eggs. ple after filter wash (Figure 10). Plant operators The common denominator of the pretreatment and should consider core-sampling and analyzing the filtration processes is coagulation; thus, proper coag-
DECEMBER 1999
© 1999 American Water Works Association, Journal AWWA December 1999
S. KAWAMURA
89
ulation and flocculation are essential to producing good filtered water. Filter performance is dictated mainly by the filter media, the filtration rate, and the effective use of polymer as a filter aid. Two proven high-rate filtration beds are the standard dual-media and the coarse deep bed. A thin (10-in. [250-mm]) sand layer at the bottom, to act as a barrier, is worth consideration. For better filter performance, the media should have a uc < 1.4. The trimedia bed commonly behaves as a dualmedia bed because it frequently loses its bottom garnet layer. Each filter must be fitted with auxiliary scouring and backwashed at the appropriate rate to maintain the filter bed in clean yet ripened condition. Standard dual-media beds expand about 25 percent (not 50 percent) during backwash. A coarse deep bed (effective grain size > 1.5 mm and depth > 4 ft or 1.2 m) must have auxiliary air-scouring. A surface-wash system may need to be added if air-scouring with an underdrain nozzle system with backwash does not prevent mud ball formation (particularly when polymer is used as coagulant). Each media in a filter bed composed of two or more types of granular media should fluidize to the same degree during backwash, to prevent intermixing and the loss of media. Under the current Surface Water Treatment Rule and the upcoming Enhanced Surface Water Treatment Rule, the goal is to produce a filtered water turbidity of 0.1 ntu. To meet this goal, the ratio of the bed depth to grain size of the media (effective size) should be at least 1,000. Plant operational staff should examine the condition of the on-line filters, especially the cleanliness of the filter beds and the media-grain gradation profile. It is thus recommended that the filter bed be core-sampled and analyzed once a year. Proper filter design alone will not yield good filtered water. The importance of thorough and consistent observations of filter bed conditions cannot be underestimated, especially at plants where filter controls are highly automated.
References 1. KAWAMURA, S. Integrated Design of Water Treatment Facilities. John Wiley & Sons, New York (1991). 2. ASCE, AWWA. Water Treatment Plant Design. McGraw-Hill, New York (3rd ed., 1998). 3. Culp, Wesner, Culp. Handbook of Public Water Systems. Van Nostrand Reinhold Co., New York (1987). 4. J.M. Montgomery Consulting Engineers. Water Treatment Principle and Design. John Wiley & Sons, New York (1985). 5. WEBER, W.J. J R. Physicochemical Process for Water Quality Control. Wiley International, New York (1972). 6. FAIR, G.M. & GEYER, J.C. Water Supply and Wastewater Disposal. John Wiley & Sons, New York (1954).
90
VOLUME 91, ISSUE 12
7. Greeley and Hansen. Pilot Filter Test Report for South Water Purification Plant (draft). Chicago (1995). 8. BAUMANN , E.R. Granular-media Deep-bed Filtration. Water Treatment Plant Design (R.I. Sanks, editor). Ann Arbor Science, Ann Arbor, Mich. (1979). 9. C LEASBY, J.L. Declining-rate Filtration. Jour. AWWA, 73:9:484 (Sept. 1981). 10. HUDSON, H.E. JR. Water Clarification Processes. Van Nostrand Reinhold, New York (1981). 11. BAYLIS, J.R. Nature and Effects of Filter Backwashing. Jour. AWWA, 51:1:126 (Jan. 1959). 12. KAWAMURA, S. Optimization of Basic Water Treatment Processes—Design and Operations: Sedimentation and Filtration. Jour. Water SRT—Aqua, 45:130 (1996). 13. ASCE. Water Treatment Plant Design Manuals of Engineering Practice. Amer. Soc. Civil Engrs., No. 19 (1940). 14. Health Education Service. Recommended Standards of Water Works—The Ten States’ Standards. Health Education Services, Albany, N.Y. (1968). 15. WILCZAK , A. ET AL. Filter Performance Using a Troughless Design at the Sacramento Plant. Conf. California–Nevada Section AWWA (Oct. 1994). 16. K AWAMURA , S. ET AL . Modifying a Backwash Trough to Reduce Media Loss. Jour. AWWA, 89:12:47 (Dec. 1997). 17. BAYLIS, J.R. Design Criteria for Rapid Sand Filters. Jour. AWWA, 51:11:1433 (Nov. 1959). 18. J.M. Montgomery & CH2M Hill. Joint Rept. South Fork Tolt River Pilot Study for Seattle Department of Water (Oct. 1992). 19. JEFFERY, J. Cryptosporidiosis and Water Supply. Jour. Water SRT—Aqua, 40:5:110 (Apr. 1991). 20. MONK, R.D.G. Improved Methods of Design of Filter Boxes. Jour. AWWA, 76:8:54 (Aug. 1984). 21. MONK, R.D.G. Design Option for Water Filtration. Jour. AWWA, 79:9:93 (Sept. 1987). 22. C UMMINGS , L. Removal Correlation of Cryptosporidium and Giardia muris With Particle Size and Turbidity for Pilot Scale. Proc. 1994 AWWA Ann. Conf., New York. 23. NIEMINSKI, E. Giardia and Cryptosporidium Removal Through Direct Filtration and Conventional Treatment. Proc. 1994 AWWA Ann. Conf., New York. 24. Montgomery Watson. Optimization of Filtration for Cyst Removal (90699). AWWA Res. Fdn., Denver (1994). About the author: Susumu Kawa-
mura is a senior vice-president and corporate technical director of Mont gomery Watson, 300 N. Lake Ave., Suite 1200, Pasadena, CA 91101. He has nearly 50 years of experience in water treatment engineering, particularly the design of new water treatment plants that include softening processes.
© 1999 American Water Works Association, Journal AWWA December 1999
JOURNAL AWWA