War. Res. Vol. 21, No. 4, pp. 403-410, 1987 0043-1354/87 $3.00+0.

00
Printed in Great Britain. All rights reserved Copyright 1987 PergamonJournals Ltd

SLOW S A N D FILTRATION AS A TECHNIQUE

FOR THE TERTIARY TREATMENT OF
MUNICIPAL SEWAGES
K. V. ELLIS
Department of Civil Engineering, University of Technology, Loughborough, Leicestershire, U.K.

(Received May 1985)

Abstract--The investigation was designed to demonstrate the viability, or otherwise, of slow sand filtration
as a means of tertiary treatment for secondary effluents derived from conventional aerobic, biological
treatment processes operating with municipal wastewaters. Secondary effluents derived from both an
activated-sludge plant and from a percolating filtration plant were employed.
The basic slow sand filtration unit used consisted of a 140 mm i.d. perspex cylinder, 2.65 m in height
containing a 950 mm depth of fine sand. Treatment rates were either 3.5 or 7.0 m d -~ and the sand used
was of an effective size initially of 0.3 mm and then later of 0.6 mm.
This investigation has demonstrated that a laboratory-scale slow sand filtration unit is capable of
consistently removing at least 90% of the suspended solids, more than 65% of the remaining BOD and
over 95% of the coliform organisms from the settled effluent from an operational percolating filter plant.
The length of operational run averaged 20 days at 3.5 m d -~ and 13 days at 7.0 m d -t. Slightly inferior
results were achieved when using the settled effluent from an operational activated sludge unit.
Further investigation employing a horizontal-flow gravel pre-filter demonstrated that at flows of 2 m h -
with a contact time of 33 rain up to 82% of the suspended solids in the secondary effluent could be removed
prior even to slow sand filtration.

Key words--tertiary treatment, slow sand filters, percolating filters, activated-sludge, secondary effluent,
coliform bacteria, BOD5 removal, gravel filtration

INTRODUCTION otherwise acceptable effluent. Upward-flow, deep-bed

filtration is a popular and effective form of polishing
Tertiary treatment (frequently referred to as effluent process. Slow sand filtration is rarely employed and
polishing) can mean different things on opposite sides this represents an enigma to those aware of its
of the Atlantic, but in Britain, tertiary treatment is effectiveness in the potable water industry.
usually held to refer to those processes which primar-
ily reduce the suspended solids content of a secondary SLOW SAND FILTERS
effluent and, by doing so, also reduce the level of the
BOD. This is not a comprehensive definition as some Slow sand filters were the first of the modern
processes may be held to reduce the BOD, to some treatment techniques devised for the purification of
extent, by biological activity and one specific process potable water. In the potable water industry slow
(nitrifying filtration) employs biological activity not sand filtration is still extensively employed and slow
to remove material from the water but merely to sand filters are known for their ability to produce
change its form. consistently a high class filtrate with the m i n i m u m of
Generally the techniques of tertiary wastewater control. They are also noted for being able to reduce
treatment can be listed as: microstrainers, grass plots, the bacterial count in water by up to 99.9%.
lagoons, sand filtration (slow filtration, rapid gravity Published results of the operation of slow filters in
filtration, upward flow filtration), upward flow the wastewater industry, however, (Truesdale et al.,
clarifiers, and nitrifying filters. 1964; HMSO, 1963; Black, 1967; Pullen, 1976; Ker-
Of these, microstrainers and rapid gravity filtration shaw, 1976) suggest only a moderate removal of
go right back to the inception of tertiary sewage suspended solids through the filter of 6 0 - 6 5 % with a
treatment at the East Hyde treatment works, Luton, limited 35-55% BOD removal. Similarly, reports of
in the early 1950s. They are still effective and still the percentage removal of coli-aerogenes bacteria are
popular. Grass plots and lagoons are also widely far less at 3 8 - 6 2 % (Truesdale et al., 1964) than might
employed and have definite advantages for specific be expected by comparison with the operation of
situations. Upward-flow clarifiers are widely em- potable water slow filters.
ployed, with varying success, usually on the effluents Because of this paradox between the known abili-
from small and remote works. Nitrifying filters oper- ties of slow sand filters in the potable water industry
ate merely to reduce the a m m o n i a content in an and their apparent lack of success with the waste-

403
404 K . V . ELLIS

w a t e r i n d u s t r y it w a s d e c i d e d to r u n a series o f sumptive coliform organisms in The Bacteriological Exam-

i n v e s t i g a t i o n s in a n a t t e m p t to d i s c o v e r t h e t r u e ination of Water Supplies (HMSO, 1977). The nitrate deter-
minations were by the 2,4-Xylenol method.
c a p a b i l i t i e s o f slow s a n d filtration as a n effluent
p o l i s h i n g t e c h n i q u e . T o this e n d t w o l a b o r a t o r y - s c a l e First stage investigation
slow s a n d filters were o p e r a t e d o v e r a p r o l o n g e d The principal objective of the first stage investigation was
p e r i o d t r e a t i n g t h e settled effluents f r o m initially a to compare efficiencies of two different grades of sand, one
p e r c o l a t i n g filtration p l a n t a n d t h e n f r o m a n set in each of the two filters employed. In one was a 0.95 m
depth of 0.3 m m effective size (UC 2.0)* sand while the
a c t i v a t e d - s l u d g e unit. T h e r e s u l t s o b t a i n e d were q u i t e
second filter contained a similar 0.95 m depth of a coarser
revealing. 0.6 m m effective size (UC 1.2) sand.
The filters were located at the Loughborough Wastewater
Recovery Works. Settled, secondary effluent from the
work's percolating filter plant was withdrawn continuously
EXPERIMENTAL from the channel by means of a 340 W Stuart Turner p u m p
This reported investigation into the slow sand filtration of and fed to a header tank (capacity 3841. to the overflow).
secondary sewage works effluent was carried out in four As a result of the height of the header tank above the
stages at the Loughborough and at the Wanlip Wastewater effluent channel there was only a slow overflow from this
Recovery Works of the Soar Division, Severn Trent Water tank. In order to maintain all the solids present in sus-
Authority. A subsequent, limited, investigation into the pension this tank was slowly and continuously stirred at the
pre-filtration of secondary effluents prior to slow sand rate of about 6 0 r p m using a 33.5 5 m m stirrer. From this
filtration was also carried out at the Loughborough works. tank the secondary effluent flowed along screw-clip con-
trolled plastic tubes to the two filters at the rates required.
Apparatus Both the header tank and the filters were fitted with overflow
The principal apparatus employed in these investigations devices to allow constant, m a x i m u m heads to be maintained
consisted of a 140 m m i.d. vertical perspex tube 2.62 m in at all times.
height with a flanged joint 1.07 m from the bottom. The During this first stage both the filters were operated at the
lower section contained a 120 m m depth of graded gravel rate o f 3.5 m d-~ (m 3 m - 2 d - t) and were cleaned when, with
held 20 m m clear of the bottom by a coarse plastic screen. the m a x i m u m hydrostatic head over the filter sand, it was
On top of the gravel was placed the 0.9 m depth bed of sand. no longer possible to maintain the desired filtration rate.
The depth of the sand bed was so arranged that its top Cleaning was accomplished by decanting the water above
surface was at a level with the flanged joint in the filter tube. the sand bed to within about 100 m m of the sand surface
A few millimeters below the top sand surface an annular and then draining down the remainder through the sand
incision was made into the wall of the perspex tube in an bed. The top 25 m m or so of the sand was then carefully
attempt to reduce the effect of water short-circuiting down removed with a spatula and replaced from a reserve stock
the filter walls. of identical sand. The sand bed was then filled initially with
Piezometer tubes were set, at various levels, into the filter fresh water from the bottom until the water level was about
wall in order to allow the head-loss through the system to 200 m m above the sand, and then from the top in the normal
be gauged. An overflow weir was attached to the top o f the manner.
filter column. At the bottom of the column a T-joint This stage of the investigation continued for 4 months
connection was inserted, through which the filtered water although samples were not taken for analysis until the end
was withdrawn and through which water was added to refill of the first m o n t h in order to allow the filters to mature. In
the column after cleaning. Outside the filter column the all, samples were taken for analysis on 12 occasions. Details
filtered-water tube was led in an inverted U-tube to a of the analytical results, removal efficiencies and run lengths
position just above the top sand level, both to prevent the are shown in Table 1.
development of negative head and to prevent the sand from During all the investigations spot samples only were taken
accidentally drying out. In this manner the U-tube fulfilled for analysis. The large capacity of the stirred feed-tank
the same function as the effluent weir in a conventional slow protected the sand filters against any sudden variation in
sand filter. quality of the secondary effluent applied and ensured that
For the additional stage of the investigation a horizontal the feed to the filters and, more particularly, the filtrate from
flow pebble filter was used with overall dimensions of 2.2 m them could only vary extremely slowly. There was therefore
long, 0.35 m high and 0.15 m wide. The interior was divided little advantage to be gained by employing a periodic
into five separate pockets. First, there was a small pocket sampling device as the filtrate sampled at any time could be
0 . 1 5 m in length containing coarse 1 4 - 2 0 m m rounded realistically assumed to have emanated from a secondary
gravel and then the main gravel pocket 1.5 m long contain- effluent feed nearly identical in quality to that sampled at the
ing smaller 5.0-6.3 m m rounded gravel. Then following the same time as the filtrates.
main mass of gravel there were three further consecutive
pockets of 0.I, 0.1 and 0.15 m length containing 6.3-10 m m Second stage
gravel, 10-14 m m gravel and 14-20 m m gravel. Each pocket Having established that a good quality filtrate could be
was divided from the next by a coarse plastic screen set into obtained using a 0.6 m m effective size slow sand filter it was
vertical slides. The effluent was removed from the equipment necessary to determine whether or not this relatively coarse-
by a perforated pipe set across the bottom of the filter box sand filter could be operated at a higher rate of flow. For
at the end of the final gravel pocket. the second stage of the investigation two identical 0.6 m m
effective size, sand filters were employed but with one
Analyses operating at 3.5m d -* ( 3 7 . 4 m l m i n -*) and the other at
The methods of analysis employed during this in- 7 . 0 m d -I (74.8 ml min *). During the second stage the two
vestigation were those recommended in The Analysis of filters were again positioned at the Loughborough treatment
Raw, Potable and Waste Waters (HMSO, 1972) with the works.
exception of the coliform determination which was carried The filter which had been operated as the coarse medium
out according to the multiple-tube technique for pre- filter during the first stage of the investigation was now
continued in operation at the original rate of 3.5 m d -~ and
was now referred to as the slower filter. Consequently it did
*UC = Uniformity coefficient. not require an extended period of maturation to bring it to
 Slow sand filtration o f secondary sewage 405

Table 1 This second-stage of the investigation ran for 3 m o n t h s

Mean during which samples were taken for analysis on 20 occa-
inflow % sions. Details of the mean analytical results together with
quality Removal Maximum Minimum percentage removals and run lengths are given in Table 1.
1st Stage Fine filter(3.5md i)
22 69 82 53 Third stage
BODs
Suspended By the end o f the second stage an appreciable a m o u n t of
solids 24 88 97 71 information had been gathered concerned with the slow
COD 106 54 79 12 sand filtration of the effluent from a percolating filter unit.
Coliforms 1,366,000 97 98 78 The results of operation obtained indicated a high degree of
Nitrate both B O D and of suspended solids removal. It was now
(as N) 18
decided to continue the investigation by operating the filters
Average length of filter run 7.1 days (max. II, min. 7)
with the effluent from an activated-sludge plant in order to
Coarse filter (3.5 m d t) discover whether this would entail any appreciable
BOD s 22 76 87 65 differences in the efflciencies o f the filters. The activated-
Suspended sludge plant selected was that at the Wanlip works. Wanlip
solids 24 88 98 74
COD 106 47 79 11 was selected not only because it was convenient but also
Coliforms 1,366,000 97 99 75 because the effluent obtained is consistently of a high quality
Average length of filter run 19.7 days (max. 36, min. 14) with regard to the BOD and suspended solids content. As
2rid Stage Faster filter (7.0 m d- ~) there was some interest during this stage of the investigation
BODs 16 65 80 45 in the potential o f slow sand filters to nitrify a m m o n i a the
Suspended effluent used was taken after the secondary settlement but
solids 16 92 97 91 before the tertiary nitrifying filter stage.
COD 110 37 53 28 In this third stage two similar filters--both containing
Coliforms 442,000 96 99.9 88 0.6 m m effective size s a n d - - w e r e again operated. The "fast"
Nitrate filter operated at a rate of 7.0 m d - t (74.8 ml min - ~) and the
(as N) 18 22 53 18 "slow" filter operated at 3.5 m d-m (37.4 ml min-~). This
Average length of run 12.8days (max. 15, rain. 11)
comparison of slow sand filter operation at different rates
Slower filter (3.5 m d -t) for the filtration o f a good-class activated-sludge effluent
BOD s 16 76 88 3I continued for three-and-a-half m o n t h s during which sam-
Suspended ples were taken for analysis on 16 occasions. Details of the
solids 16 93 98 91
COD 110 50 68 33 analytical results and of filter performances are shown in
Coliforms 442,000 99 99.9 90 Table 1.
Nitrate
(as N) 18 41 66 8.3 Fourth stage
Average run length 20 days (max. 23, min. 14) T h r o u g h o u t the third stage of the investigation, and
3rd Stage Faster filter (7.0 m d- ~) indeed throughout all the stages, little or no further
BOD5 18 69 73 44 nitrification of the effluent was recorded as it passed through
Suspended the sand filters. This was contrary to what had been
solids 16 58 89 38 expected but was, at this stage, t h o u g h t to be due to the low
COD 110 34 46 12 levels of dissolved-oxygen existing in the effluent in the
Coliforms 548,000 91 96 53 water reservoir. Largely in an attempt to discover whether
Nitrate or not nitrification could be induced through tertiary treat-
(as N) 2.4 42 74 -- ment slow sand filters it was decided, as stage 4, to continue
Average length of filter run 13.4days (max. 20, min. 8)
the operation o f the slower filter ( 3 . 5 m d -~) but with an
Slower filter (3.5 m d i) air-diffuser installed halfway down the effluent reservoir.
BODs 18 67 77 54 The immediate effect o f this was to increase the dissolved-
Suspended oxygen level above the sand from approx. 1.5 to 8.0 m g 1-t.
solids 16 65 86 52
COD I 10 29 76 5 This stage was operated for more than 4 m o n t h s during
Coliforms 548,000 91 97 63 which 11 samples were taken for analysis (Table 1). The
Nitrate temperature of the system was consistently between 12 and
(as N) 2.4 39 75 -- 15C.
Average length of filter run 25.6days (max. 33, min. 17)
#th Stage Additional stage
BODs 18 79 93 68 In the potable water industry the use of slow sand filters
Suspended has been limited traditionally by the turbidity of the feed
solids 16 71 90 53 water. Normal suggested limits vary from 10 to 50 T U (Cox,
COD 67 61 73 33 1969; H u i s m a n and Wood, 1974; T h a n h and Hettiaratchi,
Coliforms 388,000 99 99.5 74 1982; Paramasivan et al., 1981). Recently a great deal o f
Nitrate investigation has been carried out (Thanh and Ouano, 1977;
(as N) 4.3 42 88 1.4
Average length of filter run 24days (max. 35, rain. 13) University of Dar es Salaam, 1980, 1982; Wegelin, 1983;
Trueb, 1982; Symons and Pardoe, 1984; I R C W D , 1984;
Analytical results are given in terms of mg I ~ with the exception Boller, 1982) into possible methods of reducing initial high
of the coliform count which is stated in terms of a number per
100 ml samples. turbidities prior to slow sand filtration. One of the most
successful methods employed to reduce turbidity has been
the horizontal-flow pebble filter and it was considered that
full efficiency. The other filter which had contained the finer the successful application of this principle prior to a tertiary
medium for the first stage investigation had been emptied treatment slow filter might appreciably extend the period
and refilled with the same coarser medium as the first filter o f the filter runs and consequently greatly reduce the cost of
and operated during the second stage at the faster rate o f filter operations. To this end the additional stage o f the
7.0 m d-~. This was referred to as the faster filter and as it practical research was dedicated to an investigation of
had been refilled with clean sand it had to undergo a period the potential of such a pre-filter to reduce significantly the
of maturation before full efficiency could be expected. suspended solids content of the secondary effluent.
406 K.V. ELLIS

Table 2
Additional stage Horizontal flow pebble filter
Nominal Mean secondary effluent quality
retention
Flow rate period Suspended Turbidity
(m h ~) (min) BOD 5 solids COD (NTU)
1.2 60 10 9,4 50
2.4 30 22 14 45
2.0 36 18 13
4.0 18 14.6 l0
Mean % removals
Nominal Suspended
Flow rate retention BOD 5 solids COD Turbidity
1.2 60 78 75 17
2.4 30 79 74
2.0 36 75 48
4.0 18 60 30
Coliform removal
99.5 from 200,000/100ml (one sample only) at 1.2 m h
86 from 170,000/100 ml (one sample only) at 4.0 m h i
Analytical results are given in mg I ~with the exceptions of the coliform count which
is given as a number per 100 ml sample and of the turbidity which is recorded
as nephelometric turbidity units.

The horizontal-flow gravel filter employed (described was the possible flocculating action of the stirring
above) was operated using the settled effluent from a mechanisms in the header tank. The stirrer was
percolating filter plant at rates of up to 4 m h-t. The unit essential to maintain all the suspended solids in
was not operational continuously but merely started-up
about 24 h before samples were taken. Eleven runs were suspension. These two factors might have been ex-
made and 11 sets of samples removed for analysis, the pected largely to neutralize one another. No obser-
results of which, together with removal efficiencies, are vations were made on the possible action of the
shown in Table 2. delivery pump although it was considered that the
limited contact between impeller and suspended sol-
DISCUSSION ids would minimise any disintegrating action. As to
the potential flocculating actions of the stirrer mech-
Two conclusions were immediately apparent from anism, it is possible to be more definite. Initially when
the results of the first stage investigation. The im- the unit was first set up the pump delivered directly
provement in effluent quality was consistently far to the sand filters with overflow systems available to
superior to the results published in the literature and take off the excess secondary effluent. This arrange-
the quality of the filtrate from the finer filter (effective ment was soon abandoned as it made collection of
size 0.3 ram) was not substantially better than that representative input samples difficult. However, no
from the coarser filter (ES 0.6 mm). difference was noticed in the efficiency of sand filters
The 88% removal of suspended solids, 76% re- whether being fed directly with secondary effluent or
moval of BOD5 and 97% removal of coliform or- via a stirred header tank.
ganisms were all remarkably superior to the 35-45% As a result of doubling the rate of filtration in one
removal of BOD5 and 60% removal of suspended filter the mean rates of BODs, suspended solids and
solids suggested in published results (Truesdale et al., COD removal dropped to 65, 92 and 37% as com-
1964; HMSO, 1963; Black, 1967) of slow sand filter pared with the 76, 93 and 50% in the reference filter
operation. The only marked difference between the which operated at the original slower rate of
results of the operation of the finer filter and that of 3.5 m d 1 Rather strangely, however, the removal of
the coarser filter was in the run length which for the total coliforms improved from 96% in the slower
latter was, on average, more than twice the period of filter to 99% in the faster filter. The length of filter
the former. This suggested that rate of treatment run, as expected, was appreciably less for the faster
might be a more important parameter than sand filter with a mean of 12.8 days as compared with the
grain size (within limits) and hence there was little 20 days for the slower filter. This was a decrease of
difficulty in the decision to proceed with the in- only 36% resulting from a 100% increase in filtration
vestigation employing only the coarser of the two rate.
sands in both slow filters and doubling the flow rate The change from filtering the effluent from a
in one. percolating filtration plant to filtering that from an
Two additional factors may have influenced the activated-sludge unit (stage 3) brought about an
ability of the slow sand filters to remove secondary appreciable but not spectacular reduction in the
solids from the effluents. The one was the disin- overall efficiency of filter operation. Although the
tegrating action of the impeller in the small pump mean percentage removal of BOD5 increased slightly
used to lift the effluents to the header tank. The other at the faster rate (7.0md-~), the mean percentage
 Slow sand filtration of secondary sewage 407

removals of suspended solids, COD and coliforms in an activated-sludge effluent but at 25-30 and
together with the BOD 5 from the slower filter all 6-7.5 gm. Unfortunately for this theory they also
decreased appreciably. The percentage removal of reported a similar particle size distribution for the
nitrate was generally high and comparable with the effluent from a percolating filtration plant. More
percentage removal from the faster filter in stage 2, investigation is obviously required as to the particle
but the amounts of nitrate to be reduced were far less sizes passing into and out of the slow filters.
(Table 1). Filtrate samples were also taken, during the first
Certainly the removal of suspended solids from and second stages, from sample points immediately
activated-sludge effluent by slow sand filtration was below the surface sand of the slow filters (Table 3).
far less than from a percolating filter effluent. This Comparing the quality of these samples with those of
reduction in the percentage of suspended solids re- the samples taken at the filter bottom it was obvious
moved might possibly be the result of the two very that, generally, all the removable solids had been
differently sized fractions of activated-sludge solids taken out at the surface layer but that the removal of
discovered by Tchobanoglous and Eliassen (1970). biodegradable organic material continued to a sub-
The larger fraction (80-90#m) would be readily stantial extent down the whole depth of the filter. The
trapped within the filter while the smaller-sized frac- removal of the coliforms organisms also was achieved
tion (3-5 pm) might pass through. West et al. (1979) principally in the surface layer but this removal also
in their work for the Thames Water Authority dis- continued, sometimes substantially, through the
covered similar peaks in the particle size distribution whole sand bed. In addition, approx. 50% of the

Table 3. Results from top and bottom sample points

Stage l - - F i n e filter
Top sample point Bottom sample point
Mean
inflow % %
quality Removal Maximum Minimum Removal Maximum Minimum
BOD 5 22 66 78 50 69 82 53
Suspended
solids 24 88 95 79 88 97 71
COD 106 40 54 8 54 79 12
Coliform 1,366,000 88 96 85 97 99.7 78
Nitrate

Stage 1--Coarse filter

Top sample point Bottom sample point
Mean
inflow % %
quality Removal Maximum Minimum Removal Maximum Minimum
BOD 5 22 67 78 61 76 87 65
Suspended
solids 24 88 94 83 88 98 74
COD 106 36 52 31 47 79 II
Coliform 1,366,000 95 98 93 97 99 75
Nitrate

Stage 2--Slower filter

Top sample point Bottom sample point
Mean
inflow % %
quality Removal Maximum Minimum Removal Maximum Minimum
BOD s 16 71 78 41 76 88 31
Suspended
solids 16 90 97 85 93 98 91
COD ll0 40 50 23 50 68 33
Coliform 442,000 94 99.9 88 99 99.9 90
Nitrate 18 23 30 7.5 41 66 83

Stage 2 - - F a s t e r filter
Top sample point Bottom sample point
Mean
inflow % %
quality Removal Maximum Minimum Removal Maximum Minimum
BOD s 16 42 78 19 65 80 45
Suspended
solids 16 80 90 73 92 97 90
COD 1 l0 32 45 14 37 53 28
Coliform 442,000 91 99.9 83 96 99.9 88
Nitrate 18 Il 51 7.4 22 53 18
408 K.V. ELLIS

denitrification achieved was accounted for in the was not to be and, in fact, it was the process of
surface layer and the remainder through the total denitrification which was amongst the most promi-
depth of the sand. Only in the faster filter of the nent features. During the second stage the faster filter
second stage was there any appreciable improvement (7.0 m d-~) removed on average 22% of the applied
in the removal of suspended solids after the surface nitrate (mean value in secondary effluent 18 mg 1-~)
layer and this was reflected in the ~ 30%, increase in while the slower filter (3.5 m d -j) managed to remove
BOD 5 removal. 41% of the nitrate. The nitrate content of the
Overwhelmingly then most of the purification oc- activated-sludge effluent applied during the third
curred at or about the surface sand layer in the stage was much lower than that of the percolating
mixture of humus, sand, algae, protozoa and meta- filter effluent investigated earlier with an average
zoa which in a potable water filter would be referred value of only 2.4 mg l -L, of which the faster filter
to as the filter-skin schmutzdecke. Whether or not removed 42% and the slower filter 38%. It had been
this should still be referred to as the filter-skin in a thought that this denitrification was the result of the
wastewater filter is questionable as the material low-dissolved-oxygen content of the secondary
skimmed-off during cleaning was most unlike the effluents (about 1.5 mg I-t) so in the fourth stage this
schmutzdecke of potable water treatment in that it was artificially increased to about 8.0 m g l - t by pos-
possessed the consistency of black mud and the itioning an air-diffuser half-way down the column of
colour of digested sludge. secondary effluent above the sand. This too was
Further consideration, however, of the quality of unsuccessful in inducing additional nitrification and,
the samples (Table 3) taken revealed more definitely in fact, denitrification continued at an unreduced rate
that the purification achieved was not purely the (42% removal from a mean input of 4.3 mg l-J). This
result of a straining action at the surface. During the effect must be indicative of the intensity of biological
first stage operation the ratio of BOD5 removed to activity on and within the sand.
that of suspended solids removed through the whole The proven ability of slow sand filters to remove
depth of the filter was 0.72 for the finer filter and 0.79 coliform organisms during the treatment of potable
for the coarser filter. This ratio increased slightly waters suggested that worthwhile results might also
through the second stage to 0.8 (faster filter) and 0.85 be achieved with a slow filtration of secondary waste-
(slower filter). For the results of microstraining, water effluents. This proved to be so. A remarkably
however, in which the mechanism is purely mechan- consistent percentage removal of more than 90 was
ical straining, the ratio of BOD5 removed to sus- achieved throughout all four stages of the in-
pended solids removed over a 6-yr period at the vestigation despite a varying count in the secondary
Basingstoke (Axtell, 1976) wastewater treatment effluent input. During stage 1 both the coarse and fine
plant was within the range of 0.6-0.43 with a mean filters (3.5 m d -~) removed, on average, 97% of the
of only 0.45. Similarly, the results published for the total coliforms from a mean inlet count of
Harpenden works for a 4-month period in 1962 1,366,000/100ml. This percentage only dropped
(Truesdale et al., 1964) give a ratio, for the operation slightly during the second stage to 96% with the
of the microstrainers, of 0.5. These figures compare faster filter but increased slightly to 99% with the
favourably with the ratio of BOD5 to suspended slower filter. In stage 3 (activated-sludge effluent) the
solids suggested by Mara (1976) of 0.54. The higher removal percentages for both slower and faster filters
ratios of BOD5 removed to suspended solids removed was 91% from a mean input count of 548,000/100 ml,
obtained from the operations of a slow sand filter to but this rose to 99% in the fourth stage (filtration rate
those obtained from the operations of microstrainers 3.5 m d - j ) from a mean count of 388,000/100 ml. It
must be the result of appreciable biological activity was possibly the improved aerobic conditions during
within the sand bed. stage 4 that were responsible for the increase from a
Interestingly the ratios of BOD 5 removed to sus- 91% removal to that of 99%. This could be consis-
pended solids removed increased substantially during tent with the findings by the National Environmental
the filtration of an activated-sludge effluent in stage Engineering Research Institute (NEERI) (Para-
3 and stage 4 to 1.17 (slower filter), 1.24 and 1.25. The masivam et al., 1980) in India that high percentage
higher ratios reveal that a greater proportion of the coliform removals occur only under aerobic condi-
purification achieved was as the result of biological tions.
activity--this biological activity being appreciably The additional stage to the investigation was in-
greater than even that observed during the filtration cluded in an attempt to discover a technique which
of the effluent from a percolating filter plant in stage would, by removing a certain proportion of sus-
1 and stage 2. This possibly indicates a greater pended solids from the secondary effluent prior to the
availability of a readily biodegradable organic mate- slow sand filters, increase the run length of the filters
rial in the effluent from the activated-sludge unit than and, as a result, make them in practice more eco-
in the percolating-filter effluent. nomically viable. The Banks pebble-bed clarifier
Prior to the commencement of the investigation it (Banks, 1964, 1965) had already demonstrated the
had been expected that continuing nitrification would ability of packed gravel to reduce both the content of
be a feature of the slow sand filtration process. This suspended solids (40-60% removal) and that of the
 Slow sand filtration of secondary sewage 409

remaining biodegradable organic material (20-40% percentage BOD removal fell somewhat with the
removal of BODs) from settled secondary effluents percolating filter effluent but, if anything, improved
(Truesdale and Birkbeck, 1967, HMSO, 1973) but the slightly with the activated-sludge effluent.
results achieved from this horizontal-flow gravel were (2) When filtering percolating filter effluent at the
appreciably superior and the unit was demonstrated normal rate (3.5 m d -~) the slow sand filters removed
as being an effective tertiary treatment process in its about 88-93% of the suspended solids.
own right. (3) Suspended solids removal fell to 60-65% when
Most interest was with the removal of suspended settled activated-sludge effluent was filtered.
solids but measurements were also made of the (4) The slow sand filters were most effective at
BODs, COD and occasionally the turbidity and the reducing the coliform count in the secondary effluent.
total coliform count. Seventy-five percent of the The lowest mean percentage reductions were
suspended solids were removed from a secondary achieved with the activated-sludge effluent at approx.
effluent concentration of 9.4 mg 1- ~ at a flow rate of 91% but using percolating filter effluent these rose to
1.2 m h -~ (60 min nominal retention time), 75% also 96% and occasionally to over 99%.
from an inflow content of 18mgl -l at 2 . 8 m h -~ (5) No nitrification was observed during the
(36min retention) and 74% from 14mgl - l at filtration processes, even when the dissolved oxygen
2.4m h -~ (30 min retention). These percentage re- content of the secondary effluent was artificially
movals of suspended solids dropped only to 60 when enhanced, but up to 40% denitrification was a con-
the flow rate was increased to 4 . 0 m h -~ (18min stant feature of the normal slow sand filtration
retention). These are fairly remarkable results to be operations.
achieved from such a simple device. BOD5 removals, (6) Slow sand filtration using a sand of effective
when recorded, were also high at 78 and 79% at 1.2 size 0.3 mm gave no advantages over slow filtration
and 2.4 m h - ~, but the COD removals were, as would using an effective size sand of 0.6 mm. Although the
be expected, relatively low. Coliform removals were degree of purification achieved was similar with both
again remarkably high--99.5% at 1 . 2 m h -~ and filters the over-frequent blocking of the finer sand
86% even with the flow rate increased to filter (every 7.1 days as opposed to 19.7 days) would
4.0 m h - ~ - - b u t turbidity removals were only mod- make it unacceptable for full-scale operation.
erate. Overall it was demonstrated that an horizontal- (7) Appreciable biological purification through the
flow gravel filter could be either an effective device for slow sand filters was indicated by the relatively high
reducing the mass of suspended material reaching the BOD5 to suspended solids removal ratios, by the
slow filters and hence for extending the filter runs or continuing denitrification recorded and by the appre-
that the horizontal-flow filter could act very ade- ciable drop in COD values between the top sample
quately on its own as an efficient effluent polishing point (immediately under the sand surface) and the
system. filtrate.
Cleaning a horizontal-flow gravel filter might (8) No dramatic decrease in filtrate quality was
present problems. Experience in the potable water recorded when the filtration rate was doubled from
industry suggests that this can only be achieved by 3.5 to 7 . 0 m d -~ but the length of the filter run
periodically removing all the gravel bed, washing it dropped to nearly half.
and replacing but that the "run" between such clean- (9) In brief, slow sand filtration of settled second-
ings would be considerable as a result of the high ary effluents using 0.6mm effective sand size was
storage capacity for the deposited solids within the shown to be most effective at a filtration rate of
gravel-bed. Recent work carried out in Switzerland 3.5 m d -~ although the efficiencies decreased appre-
( I R C W D News, 1984) had indicated that most solids ciably when settled activated-sludge effluent was em-
are removed by gravity settling and that they form ployed in place of percolating filter effluent.
loose agglomerates on top of the individual gravel (10) In addition, it was demonstrated that a 2 m
pieces. Periodic draining of the bed creates a down- long horizontal-flow gravel pre-filter containing prin-
ward movement of this accumulated material, re- cipally 5.0-6.3 mm gravel could reduce the suspended
moves much of the sediment and goes a long way to solids content of a percolating filter effluent by be-
restoring the full removal capacity of the bed. tween 60 and 80% at flow rates varying from 4.0 to
1.2 m h - ~ (nominal retention periods of 18-60 min).

CONCLUSIONS Acknowledgements--The author wishes to thank the Di-

visional Manager, Soar Division, Severn Trent Water
The conclusions of this series of investigations can Authority and his staff at the Loughborough and Leicester
be summarised as follows: Water Reclamation Works for their assistance in carrying
through the investigation reported in this paper.
(i) BOD 5 removal through a laboratory scale slow
sand filter operating at a rate of 3 . 5 m d -~ was
generally 70-75% when treating a settled percolating REFERENCES
filter effluent and 65-70% with a settled activated- Axtell R. J. (1976) Basingstoke sewage treatment works,
sludge effluent. At the higher rate of 7.0 m d-~ the 1961-75. Inst. pubL Hlth Engrs 4, 4-14, 22.
410 K.V. ELLIS

Banks D. H. (1964) Upward-flow clarifier for use in treating aspects of gravel pre-filtration. Poster paper, AquabacT
sewage effluent. Survr munic. Cty Engr 123, 21-23. 1984, University of Birmingham, U.K.
Banks D. H. (1965) Upward-flow clarifier for treating Tchobanoglous G. and Eliassen R. (1970) Filtration of
sewage effluents. Suror munic. Cty Engr 125, 45-46. treated sewage effluent. J. sanit. Engng Div. Am. Soc. cir.
Black S. A. (1967) An evaluation of effluent polishing Engrs 96, 243-265.
process installations. Ont. Wat. Resour. Comm. Div. Res. Thanh N. C. and Ouano E. A. R. (1977) Horizontal-flow
Publication No. 20. coarse-material pre-filtration. Asian Institute of Tech-
Boiler M. (1982) Optimisation of design variables for ter- nology, Bangkok, Thailand.
tiary contact filtration, IWSA Congress, Zurich, Switzer- Thanh N. C. and Hettiaratchi J. P. A. (1982) Surface water
land. filtration for rural areas: guidelines for design, construc-
Cox C. R. (1969) Operation and Control o f Water Treatment tion, operation and maintenance. Environmental Sani-
Processes, Chap. 7. World Health Organisation, Geneva. tation Information Centre, Bangkok, Thailand.
HMSO (1963) Polishing of sewage works effluents. Notes Trueb E. (1982) Horizontal-flow gravel filters for pre-
on Water Pollution, No. 22, Department of Scientific and liminary purification of surface water especially for use in
Industrial Research. developing countries. 3R Int. Switz. 21, 30.
HMSO (1973) Developments in the treatment of sewage Truesdale G. A., Birkbeck A. E. and Shaw D. (1964) A
from small communities. Notes on Water Pollution, No. critical examination of some methods for further treat-
60, Department of the Environment. ment of effluents from percolating filters. J. Inst. Sewage
Huisman L. and Wood W. E. (1974) Slow Sand Filtration, Purif. 63, 81-101.
Chap. 3. World Health Organisation, Geneva. Truesdale G. H. and Birbeck A. E. (1967) Tertiary treat-
1RCWD News (1984) An appropriate pre-treatment for ment processes for sewage works effluents. J. Wat. Pollut.
slow sand filters in developing countries. WHO Inter- Control 66, 371-385.
national Reference Centre for Wastes Disposal, Switzer- University of Dar-es-Salaam (1980) Slow sand filter research
land. project, report 2. Research Report CWS 82.2, University
Kershaw M. A. (1976) Tertiary treatment of sewage of Dar-es-Salaam.
effluents. Process Biochem. I1, 21-35. University of Dar-es-Salaam (1982) Slow sand filter research
Mara D. D. (1976) Sewage Treatment in Hot Climates. project, report 3. Research Report CWS 82.3, University
Wiley, London. of Dar-es-Salaam.
Paramasivam R., Mhaisalkar V. A. and Berthouex P. M. Wegelin M. (1983) Roughing Filters as pre-treatment for
(1981) Slow sand filter design and construction in devel- slow sand filtration. Wat. Supply 1, 67.
oping countries. J. Am. Wat. Wks Ass. 73, 178. West J., Rachwal A. J. and Cox G. C. (1979) Experience
Pullen K. G. (1976) Methods of tertiary treatment (2) slow with high-rate tertiary treatment filtration in the Thames
gravity sand filters. Pollut. Monitor 1, 14-16. Water Authority. J. Inst. Wat. Engrs Sci. 33, 45-63.
Symons C. and Pardoe M. (1984) The bacteriological