'/
<•#
United States
Environmental Protection
Agency
Municipal Environmental
Research Laboratory
Cincinnati OH 45268
Research and Development
EPA-600/S2-84-088 May 1984
Project Summary
Effective Filtration Methods for
Small Water Supplies
John L Cleasby, David J. Hilmoe, Constantine Dimitracopoulos, and
Luis M. Diaz-Bossio
A 2-year study was conducted of
various simple water filtration systems
potentially appropriate for high-quality
surface waters serving small systems. A
slow sand filter without coagulant and a
direct, rapid filter with coagulant were
operated in parallel. Direct filtration
with and without flocculation were
compared in parallel in one phase of the
study; declining- and constant-rate
filtration were compared in parallel in
another phase. The study was designed
to emphasize simple treatment systems
for small supplies where operational
skill and attention may be lacking. The
systems were compared while
monitoring turbidity, particle count,
and coliform bacteria in the influent and
filtered water.
Slow sand filtration was the most
effective for particle removal, but filter
runs were as short as 9 days during algal
blooms. If the raw water is consistently
high in quality and land is available, the
slow sand filter would be the simple
system of choice. All three direct filtra-
tion systems studied were capable of
meeting the 1 -nephelometric-turbidity-
unit (NTU) maximum contaminant level
(MCL), except during the first hour of
the filter cycle. Flocculation was bene-
ficial to the filtrate quality and head loss
in direct filtration, but it was
detrimental to the terminal break-
though. Declining-rate filtration did not
improve the filtrate compared with
constant-rate filtration.
This Project Summary was developed
by EPA's Municipal Environmental
Research Laboratory. Cincinnati, OH,
to announce key findings of the research
project that is fully documented in a
separate report of the same title (see
Project Report ordering information at
back).
Introduction
The increasing number of outbreaks of
intestinal disease caused by Giardia
lamblia in upland , high-quality surface
waters has focused attention on the
water treatment deficiencies related to
such outbreaks. Because the .cysts of
Giardia lamblia resist conventional disin-
fection procedures, effective filtration
must serve as an additional barrier to
prevent such disease transmission.
transmission .
Some communities (especially small
ones) served by upland surface water
supplies presently provide no treatment
except for disinfection. Such supplies
may exceed the EPA MCL for turbidity (1
NTU) in public water supplies during
some seasons of the year, and they may
contain the cysts of Giardia lamblia.
Such communities are faced with the
need to construct and operate some form
of treatment system to produce water
that will consistently protect the public
health and meet the drinking water
standards. Small communities need
simplified treatment systems that can
operate effectively with a minimum level
of operator skill to ensure acceptable
levels of treated water quality.
This report presents the results of a 2-
year pilot study of simplified filtration
techniques potentially applicable to small
public water supplies treating high-
quality surface waters. The results can be
divided into four main parts:
1. Results for a slow sand filter oper-
ated for 14 months without
chemical pretreatment of any sort.
-------
2. Results for a rapid, dual-media,
constant-rate filter operated for 1
year in the direct, in-line filtration
mode using alum or cationic
polymer as a sole coagulant.
3. An evaluation of the impact of floc-
culation on direct filtration obtained
by a 3-month parallel operation of
two constant-rate filters with the
same chemical pretreatment. One
filter was operated with flocculation
before filtration and one was
without flocculation.
4. A parallel comparison of constant-
rate filtration with declining-rate
filtration while both systems were
operated in the direct, in-line
filtration mode.
In all portions of the study, raw and
filtered water were monitored for
turbidity, particle count, coliform
bacteria, and head loss development.
Special attention was focused on the
initial improvement period of the filter
runs and the terminal breakthrough
period if it occurred. The raw water was a
high-quality surface water in an Iowa
gravel pit. The study period covered a full
range of seasonal extremes, with water
temperatures ranging from 2° C in the
winter under ice cover to 25° C in the
summer. One summer season included
several intense algal blooms.
Results
Slow Sand Filtration Study
The slow sand filter had a 0.76-m
inside diameter with an initial sand depth
of 0.94 m. The sand had an effective size
of 0.32 mm and a uniformity coefficient of
1.44. The filter was operated at a
constant rate of 0.12 m/h by influent flow
splitting. Runs were terminated at the
overflow head loss of 1.35 m of water.
Eleven filter runs were completed over
the 14-month period of operation. The
performance of the filter was as follows:
1. The filtrate quality was somewhat
inferior for 1 to 2 days at the
beginning of each filter run when
compared with the quality for the
remainder of the run. This result
was more evident if the filter was
idle for several days between runs.
2. The filtrate quality was well below 1
NTU during all filter runs, even
during the first 2 days of each. After
the first four filter runs, the average
turbidity of the filtrate (excluding the
initial 2-day period) was consist-
ently near 0.1 NTU. Typical data for
four filter runs spread over the 14-
month period appear in Table 1.
Table 1. Slow Sand Filter Operating Data
lor Selected Filter Runs Spanning
the 14 Months of Operation
Average Turbidity. NTU
Table 2.
Effluent
Run
A
C
F
J
Length,
Days
34
22
9
41
Influent
4.4
6.9
4.9
3.9
First
2 Days
0.42
0.24
0.14
0.13
Remainder
of Run
0.39
0.24
0.10
0.07
3. A gradual improvement occurred in
the filter performance over the
series of filter runs as evidenced by
all four parameters (turbidity,
particle count, total coliform
bacteria, and chlorophyll-a).
Excluding the first 2 days of the
filter runs, the removal efficiency for
turbidity, particle count, and total
coliform bacteria was always at
least 90% (one log) and often about
99% (two log).
After the first four filter runs
spanning an 8-month period, the
performance in each subsequent
run was excellent, as follows (based
on the data after the first 2 days of
the run):
--average turbidity removal for
each run was 97.8% or better;
--7- to 12-fjm particle removal for
each run was 96.8% or better;
-1- to 60- m particle removal
was 98.1% or better, except in
one run with 92.8% removal;
--coliform bacteria removal was
99.4% or greater, reaching
100% in one filter run; and
--average chlorophyll-a removal
was 95% or better, even after
the second filter run.
Lower percentage removals were
typically associated with low
influent values. Typical results for.
G/'a/tf/a-cyst-sized particles (7- to
12 fjm size range) for selected runs
appear in Table 2 along with total
particles in the 1- to 60-um size
range.
4. Filter run length was generally rath-
er short—41 days or less—in 9 out of
A verage Particles per mL in Influ-
ent and Effluent of Slow Sand
Filter for Selected Filter Runs
Particles per mL
Effluent
Run
Influent
First Remainder
2 Days of Run
7- to 12-/jm
Particles:
A
C
F
J
1 - to 60-um
Particles:
A
C
F
J
2,242
3,745
10.3O5
753
70
59
48
34
13
95
18
5
50.156 5.740 2.384
96.252 1,030 879
42,400 866 298
33,355 1.351 346
10 complete runs, all of which were
terminated by a steeply accelerating
head loss curve. A long filter run of
123 days was achieved only under
winter conditions, when algal
populations were reduced. During
serious algal blooms, filter runs
were as short as 9 days. Increasing
the available head loss would not
have increased these run lengths
appreciably because of the
exponentially increasing head loss
curves.
Turbidity alone was not an adequate
predictor of the probable filter run
length to be expected. Algal
population was a dominant factor
affecting filter run length.
Chlorophyll-a levels of less than 5
mg/m3 were associated with run
lengths of more than 30 days.
During these runs, the mean
turbidity of the raw water was 4 to 5
NTU, with short-term peaks as high
as 16 NTU.
. No evidence showed that the filter
was clogging to any substantial
depth as indicated by initial head
loss observations and by scanning
electron microscope examination of
the sand at the end of the last filter
run.
-------
Direct, In-Line Filtration Study
A dual-media filter was housed in a
Plexiglas* tube that was 0.10 m in inside
diameter and 2.88 m high. The media
were 0.41 m of anthracite (with an effec-
tive size of 1.54 mm and a uniformity
coefficient of 1.18) over 0.30 m of sand
(with an effective size 0.43 mm and a
uniformity coefficient of 1.53). The filter
was operated as a constant-rate filter by
influent flow splitting.
Since the emphasis of this research
was on small treatment systems, the
primary goal was to evaluate the simplest
systems for high-quality surface waters.
For that reason, only a single coagulant
was used-either alum or a cationic poly-
mer. In some filter runs using alum as a
coagulant, pH was lowered with sulfuric
acid to about 6.8 in hopes of achieving
better results. The acid was needed
because of the relatively high alkalinity of
the quarry water (150 to 200 mg/L as
CaCO3), which buffered the pH above 7.5
even after alum addition. Most upland
waters of low alkalinity would have the
pH reduced suffifiently by the alum alone
so that adding acid to reduce pH would
not be necessary. An alternative
approach would be to use cationic
polymer as the sole coagulant.
Also, in view of the small system
emphasis, the range of filtration rates
studied was limited to 6.6 to 16.1 m/h.
Higher rates were considered inappropri-
ate for small systems.
Rapid mixing of the chemicals with the
filter influent water was achieved by
static mixers. No flocculation time was
provided, but some detention after rapid
mixing did exist in the influent hoses and
in the water above the filter media.
Because of the clarity of the raw water
and the low doses of chemicals used, no
visible floe particles were evident in the
water above the filter media. Neverthe-
less, the evidence of destabilization was
dramatized by the quality of the filtrate
and by the abrupt loss of quality if the
chemical feed was terminated, either
intentionally or accidentally.
Performances during the direct, in-line,
rapid filtration studies using alum or
cationic polymer as a coagulant are
described as follows:
1. An initial period of poorer filtrate
quality existed in all filter runs, as
evidenced by turbidity, 7- to 12-//m
particle count data. 1- to 60-//m
particle count data, and total
coliform data. Peak turbidity during
this period often exceeded 1 NTU.
The period of initial improvement
lasted several hours in some cases,
although the worst effects were
over in 1 h. Thus, a filtering-to-
waste period would be appropriate,
especially where Giardia cysts are
concerned.
2. When serious algal blooms were not
in progress, alum dosages between
5 and 10 mg/L (as AI2(S04)3 • 18
H2O) or cationic polymer (Cat-Floe T)
dosages between 0.09 and 1.49
mg/L could treat raw waters with
average turbidities of 8 NTU and
peak turbidities as high as 16 NTU
and produce (a) acceptable filtrate
with average turbidities well below
1 NTU before breakthrough, and (b)
reasonable filter run length. Turbid-
ity data for typical filter runs appear
in Table 3.
3. During a period of heavy blue-green
algal population with chlorophyll-a
level of 130 mg/m3 and with an
average turbidity of 20 in the raw
water, prechlorination was essen-
tial to the reasonable success of the
direct, in-line filtration process.
Alum dosages up tp 20 mg/L were
used with filter cycles as short as 12
hours at 7.3 m/h. Without
prechlorination, filtrate quality of
less than 1 NTU could not be
assured. Even with prechlorination,
the 1-NTU limit was sometimes
exceeded.
4. With low algae (chlorophyll-a less
than 5 mg/m3), the mean solids load
for the filter media of this study was
1.9 Kg suspended solids applied per
square meter of filter area per meter
of head loss increase (Kg/m2/m)
when using alum and 2.5 Kg/m2/m
when using cationic polymer. With
moderate algae, the value dropped
Table 3. Turbidity Data for Rapid, Constant-Rate Filter Runs
Run
Number
Water Raw-Turbidity, NTU
Temp, Filtrate
°C pH High* Average^ Low§
Filtrate Turbidity
NTU
High* Average^ Low§
Alum Runs at 6 to 8 m/h:
pH Controlled:
A-1 14
B-11 2
H-2d 28
pH Uncontrolled:
J-6 20
J-7 20
Alum Runs at 11 to 16 m/h:
pH Controlled:
A-4 13
B-1 7
B-10 2
pH Uncontrolled:
C-3 7
J-1 24
Cat-Floe Runs at 6 to 8 m/h
B-4 4
G-2 23
J-9 17
6.8
6.8
7.0
7.8
7.8
6.8
6.8
6.8
7.6
7.8
8.6
8.4
8.4
6.6
7.6
20.0
4.4
4.8
7.8
11.1
4
7.0
16.0
9.1
9.5
4.6
5.2
5.1
18.1
3.2
3.3
7.0
7.9
5.7
8.2
2.9
3.O
1.7
4.7
4.0
16.1
2.4
2.2
6.1
6.2
5
5.2
1.9
2.0
1.4
0.9
0.90 0.18
1.74 0.23
16.0 1.68
0.44 0.21
0.44 0.20
1.50
2.05
1.05
1.60
0.81
1.05
1.04
0.73
0.21
0.28
0.33
0.35
0.27
0.21
0.51
0.29
0.15
0.17
1.10
0.18
0.17
0.14
0.11
0.19
0.22
0.20
0.16
0.42
0.21
Cat-Floe Runs at 11 to 16 m/h:
B—'
B-7
l-6c
J-3
4
3
24
23
8.6
8.3
8.4
8.5
5.8
1.0
6.0
3.7
5.2
0.35
2.7
2.3
4.5
0.3
1.6
1.6
2.46
0.60
1.28
0.76
0.27
0.13
0.55
0.34
0.19
0.09
0.38
0.27
'Mention of trade names or commercial products
does not imply endorsement or recommendation for
use.
* Highest value at beginning of filter run.
t Average for entire run up to time of breakthrough.
% Lowest value of run.
-------
to 1.T Kg/m2/m when using alum
and 1.8 Kg/m2/m when using
cationic polymer.
Based on these values, the follow-
ing limits of average raw water
turbidity were calculated to achieve
24-hr cycles at 7.5 m/h filtration
rate with 2 m of head loss increase
available (above initial clean filter
system head loss).
Average
Average Suspended
Turbidity Solids,
NTU mg/L
During low algae
Using alum 12 21
Using cationic
polymer 1 6 28
During moderate algae
Using alum 7 12
Using cationic
polymer 11 20
Table 5. Mean Total Coliform
Season
1981:
Fall
1982:
Winter (ice
covered
Snow melt
Spring (ice
gone)
Summer
Fall
Run Dates
10/20 to 12/15
1/4 to 2/22
2/24 to 3/13
3/29 to 4/21
6/1 to 6/30
9/2 to 10/2
Removal by Rapid Filter
Mean
Chemical Used
Alum
Cat-Floe T
Cat-Floe T
Alum
Cat-Floe T
Alum
Alum
Cat-Floe T &T-1
Alum
Cat-Floe T
Influent.
No./
100 mL
1300
8200
1500
1600
640
350
90
50
550
170
Percent Coliform Removal
First Hour
%
90.5
88
77.7
93
72
79
80
81.5
86.5
70.5
No."
4
2
3
1
1
3
1
2
2
2
Remainder
%
91
96.5
89.7
96
89
91.3
86
86
89
86.5
No.*
3
2
3
1
1
3
1
2
2
2
Higher values for short periods
during the filter run can be tolerated
providing the average is not
violated.
The percent removal of 7- to 12-^m
particles after the first hour of the
cycle (Table 4) was above 85% in all
cases, exceeded 90% in 8 of 11
cases, and exceeded 95% in 6 of 11
cases.
"Number of mean filter run values used to calculate the mean percent removal value.
6. The percent removal of total
coliform bacteria after the first hour
of the cycle (Table 5) was greater
than 86% in all cases,-greater than
90% in 4 of 10 cases, and greater
than 95% in 2 of 10 cases.
7. The percent removal of 7- to 12-//m
particles generally exceeded the
Table 4. Mean Particle Removal by Rapid Filter in 7- to 12-pm Size Range
Mean % Particle Removal
Mean
Influent.
First Hour
Remainder
Season
1981:
Fall
1982:
Winter (ice
covered)
Snow melt
Spring (ice
gone)
Summer
Fall
Run Dates Chemical Used
10/20 to 12/15 Alum
Cat-Floe T
1/4 to 2/22 Cat-Floe T
2/24 to 3/1 3 Alum
3/29 to 4/21 Cat-Floe T
Alum
6/1 to 8/1 8 Alum
Cat-Floe T & T-1
Alum & C/2
9/2 to 10/2 Alum
Cat-Floe T
No/mL
2320
1170
370
2190
1620
2860
13040
1350
2730
1640
340
%
97.6
91.9
68.7
97.0
97.0
92.0
85.0
89.0
86.0
94.0
87.0
No."
5
2
3
1
1
3
1
2
3
2
2
%
98.8
96.7
87.0
99.0
98.0
94.0
99.0
85.5
92.0
96.5
87.5
No."
5
2
3
1
1
3
1
2
3
2
2
"Number of mean filter run values used to calculate the mean percent removal value.
percent removal of total coliform
bacteria.
8. The performance of direct, in-line
filtration was not impaired by cold
water as low as 2° C. In fact, when
the best raw water was treated
during the winter ice cover,
excellent filtrate and long filter runs
were obtained.
9. The cationic polymer produced
substantially longer filter cycles
than alum but a slightly inferior
filtrate, as judged by all three
parameters. Run-length data are
summarized in Table 6. Run length
and filtrate quality comparisons for
various coagulants are clouded by
the fact that the comparison runs
were sequential rather than parallel.
10. Selecting the optimum coagulant
dose for direct, in-line filtration was
difficult because of the variability of
raw water quality. Overdosing with
alum caused excessive head loss
and early breakthrough. Overdosing
or underdosing with cationic
polymer resulted in poorer filtrate
quality throughout the run.
11. Selecting the optimum dosage of
cationic polymer was more difficult
than selecting the optimum dosage
of alum. The proper dosage of alum
was easier to select because it was
much less sensitive to raw water
-------
able 6. Mean Run Lengths for Rapid Filter Comparing Cat-Floe T and Alum
10.0
Mean Run Length, h
Season
Run Dates
Chemical
At 7.3 m/h' At 12,2 m/h*
981:
all
982:
finter (ice
10/20 to 12/15 Alum
Cat-Floe T
54
95
28
26
covered}
now melt
oring (no ice)
ummer
ill
1/4 to 2/22
2/24 to 3/1 3
3/29 to 4/21
6/1 to 8/18
9/2 to 10/2
Cat-Floe T
Alum
Cat-Floe T
Alum
Alum
Cat-Floe T&T1
Alum + C/2
Alum
Cat-Floe T
52
22
120
29
4
48
21
26
109
76St
6.5
No data
6
No data
No data
17
10
31
lominal rates; actual rates are somewhat higher or lower.
Midwinter with extremely good raw water (Runs B-7 and B-8).
quality than the cationic polymer
dosage. At a particular dosage of
alum between 5 and 10 mg/L, raw
water turbidity changes from 2 to 20
NTU had practically no impacron the
filtrate quality.
2. Selecting the optimum dose of
cationic polymer was assisted by
briefly halting the polymer feed
(about 10 to 20 min) and observing
the turbidity response. If the earlier
dosage was too high, the filtrate
improved briefly (as the dosage
residual in the filter diminished) and
then deteriorated as the residual
• disappeared. If the earlier dosage
was too low, the filtrate began to
deteriorate immediately upon
cessation of polymer feed.
ipact of Flocculation on
irect Filtration
Two identical constant-rate filters were
erated in parallel during this phase of
e study. Filters, filtration rates, and
emical pretreatments were identical to
ose described in the earlier phase on in-
ie filtration. Filter media were slightly
ferent, consisting of 0.46 m of
thracite (with an effective size of 1.40
n and a uniformity coefficient of 1.36)
er 0.30 m of sand (with an effective size
0.52 mm and a uniformity coefficient of
10).
Dne of the two filters received
cculated water, and the other operated
an in-line filter without flocculation.
icculatiofi was provided in a 1.52- x
30- x 0.30-m tank divided by baffles
o four cells in a series. Each cell was
equipped with a variable-speed, 3-bladed
turbine paddle. Paddles were operated at
a constant speed of 60 rpm, which
provided root mean square velocity gradi-
ents ranging from 48 s-' at 7°C to 59 s-' at
22° C. Detention time in the flocculation
tank was held constant at 14 min.
Providing flocculation as described
above had the following impacts on the
direct filtration performance (observa-
tions are based on turbidity and particle
count data alone; bacterial results had to
be rejected as a result of experimental
difficulties):
1. The filter receiving flocculated
water had a shorter initial
improvement period, as evidenced
by lower average effluent turbidity
and particle count data during the
first hour of the run for the filter with
flocculation. A typical run appears in
Fig. 1. This result was less clearly
demonstrated with cationic polymer
and with the 7- to 12-//m particle
data. Both filters occasionally
exceeded 1 NTU average turbidity
during the first hour of the run.
2. The average quality of the filtrate
during the remainder of the filter
run (after the first hour and before
terminal breakthrough) was
superior for the filter with floccula-
tion, as evidenced by turbidity and
particle count data. Again, this
result was less clearly demonstra-
ted for all parameters when cationic
polymer was used. Both filters were
well below 1 NTU average turbidity
0.01
Figure 1.
Filter Effluents
* #2. Flocculated
#3, Unflocculated
i i i i i i
Turbidity for run L -1 at 7.0 m/h
(2.86 gpm/ft*J using alum.
during this remainder-of-run
period.
3. Providing flocculation reduced the
rate of head loss buildup when
either alum or cationic polymer was
the coagulant. But in many alum
runs, with or without pH adjust-
ment, flocculation caused earlier
breakthrough of turbidity.
4. When terminal breakthrough was a
problem, as it was in many alum
coagulated filter runs, the lower
head loss of the filter receiving
flocculated water was of no benefit
to the run length because the
effective run length was controlled
by breakthrough rather than
available head loss.
Declining- Versus Constant-
Rate Filtration
A bank of four declining-rate filters was
operated in parallel with a single constant-
rate filter. The four declining-rate filters
were placed in Plexiglas housings that
were 0.15 m inside diameter and 3.28 m
high. Both systems operated at the same
mean filtration rate and received the same
chemically pretreated water. Chemical
pretreatment and mean filtration rates
were the same as in the previously
described in-line study. Identical dual
media were installed in all five filters and
consisted of 0.35 m of anthracite (with an
effective size of 1.40 mm and a uniformity
coefficient of 1.36) over 0.25 m of sand
(with an effective size of 0.52 mm and a
uniformity coefficient of 1.40). Both filter
systems were operated in the direct, in-
-------
line filtration mode without flocculation
for a period of about 4 months in the
summer of 1982. Various problems
occurred during the first 3 months of
operation, so the following observations
are based on the final month of operation
in Septemer 1 982:
1. No water quality advantage
occurred for the declining-rate
operation in turbidity, particle count,
or total coliform removal compared
with constant-rate operation. This
conclusion contrasts with an earlier
study in which a significant
qualitative advantage for declining-
rate operation was reported while
filtering water from a lime-softening
plant.
2. Rate of head loss increase was the
same for the constant- and
declining-rate operation at either
7.70 or 13.35 m/h.
3. The highest flow rate in the bank of
declining-rate filters always occurred
in the cleanest filter just after
backwash.
4. The effluent turbidity, particle
count, and total coliform counts
were higher at the beginning of the
run during the initial improvement
period for both the declining- and
constant-rate filters. No substantial
decrease occurred in average
effluent turbidity when a filter-to-
waste period was used-a period
that consisted of wasting all effluent
at the beginning of the run until the
turbidity dropped to 0.5 NTU.
Conclusions
The following general conclusions are
drawn from the results and from the
operational experience of the study:
1. The slow sand filter system studied
in this research outperformed the
direct, rapid filtration system oper-
ating with alum or cationic polymer
as a primary coagulant. This
conclusion was substantiated by
turbidity, particle count, and total
coliform bacterial data.
2. Where simple operation is impor-
tant (as in small water supply
systems), a slow sand filter system
is superior to a direct, rapid filtration
system, but the raw water must be
of consistent high quality and low in
algae to avoid excessively short
cycles for the slow sand filter.
Turbidity alone was not an adequate
predictor of the probable run length.
Algal enumeration or a surrogate
measure of algal population, such
as chlorophyll, are essential
parameters for judging the accept-
ability of a raw water for slow sand
filtration. Chlorophyll-a concentra-
tions of less than 5 mg/m3 along
with turbidities of 5 NTU or less are
suggested as upper limits for slow
sand filter application.
3. For waters of somewhat poorer
quality, direct, rapid filtration can be
used, but it requires substantially
more operational skill and attention
and poses a greater potential risk if
improperly operated. Other
alternatives such as diatomaceous
earth filtration also should be
considered.
4. The collection of raw water data on
turbidity, suspended solids, and
chlorophyll-a over a period of at
least 1 year and including all
seasonal extremes would be
essential to make rational decisions
among filtration alternatives.
5. Both slow sand filtration and direct,
rapid filtration exhibited a period of
poorer filtrate quality at the
beginning of the filter runs. Thus
both systems require a filtering-to-
waste period where Giardia cysts
are of concern. Minimum wasting
periods of 2 days for slow sand
filtration and 1 h for direct, rapid
filtration are suggested from the
results of this study.
Because of a need for a filter-to-
waste period, at least two filters are
mandatory, even for the smallest
system. Two filters will also allow
for periodic filter maintenance and
for slow sand filter draining and
scraping after each cycle.
6. The influent flow-splitting system
used in the pilot plant of this study is
an ideally simple system that would
be appropriate to both rapid or slow
sand filter plants for small installa-
tions. This arrangement (a)
eliminates the possibility of sudden
rate changes, (b) eliminates the pos-
sibility of negative head and
consequent air binding, (c) elimi-
nates the need for rate control
equipment or head loss equipment,
and (d) can be easily made fail-safe
with a high water overflow to waste
and a turbidity monitoring and auto-
matic shut-down capability.
7. A good parallelism was evident fo
for the three parameters of filtrau
quality used in this study (turbidity,
particle count, and total coliforrr
count). Thus a good job o
continuous turbidity monitoring car
give a good indication of particular
removal and should be an essentia
minimum of instrumentation for al
plants, large or small, when a higt
degree of particle removal t;
essential on a continuous bash
(e.g., when Giardia lamblia may b<
present in the raw water).
The following conclusions apply t<
direct, rapid filtration systems as appliei
to small water treatment system!
treating high quality surface waters:
1. Declining-rate filtration did no
produce better filtrate thai
constant-rate filtration in thi
application. Thus declining-rat
filtration (which is more difficult t
understand) should not b<
recommended for small system;
Influent flow-splitting would be
superior system of operation.
2. A short period of flocculation o
about 10 min should be provided ii
direct, rapid filtration. Thi
flocculator should be provided witl
three or four compartments ii
series, a complete bypass to thi
filters, and bypasses at eacl
compartment to allow flexibility ii
flocculator detention.
3. Chemical coagulants should b<
applied in direct, rapid filtratior
systems even when the raw water
are below the MCL of 1 NTU
Substantial numbers of particle:
can still be removed during suet
periods.
4. The research reported hen
demonstrated that the best direc
filtration operation occurred durinj
midwinter uner ice cover with wate
temperatures of 2° C and witf
stable raw water quality. Cold wate
is therefore not an impairment t<
direct filtration.
5. Many existing conventional plant
in northern climates could benefi
by operating in the direct filtratioi
-------
mode during seasons of better raw
quality (e.g., in the winter during ice
cover).
The full report was submitted in
jlfillment of Cooperative Agreement No.
IR 808837-01-0 by Iowa State
niversity, under the sponsorship of the
.S. Environmental Protection Agency.
John Cleasby, David Hilmoe. Constantino Dimitracopoulos. and Luis Diaz-Bossio
are with Iowa State University, Ames, I A 50011.
Gary S. Logsdon is the EPA Project Officer (see below).
The complete report, entitled "Effective Filtration Methods for Small Water
Supplies," (Order No. PB 84-187 9O5; Cost: $22. OO, subject to change) will be
available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Municipal Environmental Research Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
-------
United States Center for Environmental Research BULK RATE
Environmental Protection Information POSTAGE & FEES P
Agency Cincinnati OH 45268 EPA
PERMIT No. G-3£
Official Business
Penalty for Private Use $300
MtRL0063240
LUU W TILLEY
RtGiON \l EPA
UBKARIAN
2iO S OtARBQRN ST
CHICAGO IL 60604
U.S. GOVERNMENT PRINTING OFFICE: 1984-759-102/96
------- |