United States
Environmental Protection
Agency
Water Engineering Research
Laboratory
Cincinnati OH 45268
Research and Development
EPA'600/S2-85/026  May 1985
Project  Summary
Filtration  of Giardia Cysts  and
Other Substances:  Volume  2.
Slow  Sand  Filtration

William D. Bellamy, Gary P. Silverman, and David W. Hendricks
  Slow sand filtration was evaluated for
a range  of operating conditions and
simulated ambient conditions using 1 -
ft-diameter laboratory filters in two
phases of experimentation. The objec-
tive was to determine its effectiveness
as a process in drinking water treatment
for  removal of Giardia lamblia cysts,
total coliform bacteria, standard plate
count bacteria, turbidity, and particles.
  During Phase I experiments,  three
filters were operated for 16 months at
hydraulic loading rates of 0.04,  0.12,
and 0.40 m/hr using raw water from
Horsetooth Reservoir, located adjacent
to the Engineering Research Center at
Colorado State University. In Phase II
experiments,  six filters were operated
for 12 months, all at hydraulic loading
rates of  0.12 m/hr, each under a dif-
ferent operating condition (e.g.. depth
of sand, size of sand, disinfection of raw
water, nutrient addition, sand size, and
temperature).
  Phase  I results showed removals of
Giardia cysts that exceeded 99.9 per-
cent for the  three hydraulic  loading
rates used. The most important oper-
ating condition was the development of
a biopopulation within the sand bed.
Cysts removals were about 99.0 per-
cent with new sand, but as the biopopu-
lation matured (after about 40 weeks),
removals were 100 percent, qualified
by detection  limits. Removals of total
coliform bacteria  related well to the
development of the biopopulation with-
in the sand bed, showing 90  percent
removal for a  new sand bed operated at
0.40 m/hr hydraulic loading rate, and
99.99 percent removal for a mature
sand bed and established schmutzdecke
operated at 0.04 m/hr. Removal of the
schmutzdecke caused removals to de-
cline to 99.9 percent, but recovery to
99.99 percent removal occurred within
a few days.
  Removals of standard plate count
bacteria usually ranged from 88 to 91
percent. Because the sand bed compris-
ing the filter develops an internal micro-
biological  population,  organisms were
continuously sloughed from within the
sand bed, causing significant counts of
standard plate count  bacteria in the
effluent. Particle count removals in the
size range of 6.35 to 12.7 fjm ranged
from 96 to 98 percent. Also because of
the sloughing of material from the filter
bed. significant numbers of particles
occurred in the effluent. Turbidity re-
moval was usually 27 to 40 percent.
The mineral particles that made up the
turbidity within the Horsetooth Reser-
voir consisted mostly of particles 1 fan
or smaller, which passed readily through
the filters.
  Phase II testing was done using total
coliform bacteria as the primary meas-
ure of effectiveness and periodic spiking
with Giardia cysts. Removals of total
coliform bacteria ranged from 60  per-
cent for the filter  maintained with no
biological activity (e.g., chlorinated
between tests) to 99.9 percent for the
filter with nutrients added. Coliform
removal for the control filter averaged
97 percent. Using a larger sand  size
(0.62  instead of  0.29 mm) caused a
decline in  removal rates, as did using a
sand depth of 48 instead of 97  cm.
Operation at 2° instead of 17°C caused

-------
a decline in removals to 92 percent com-
pared with 99 percent for the control
filter. Removals of Giardia cysts were
100 percent for all  tests conducted
(again, qualified by detection limits).

  This Project Summary was developed
by EPA's  Water Engineering 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
  This report is  the  second of three
volumes describing the research con-
ducted under  EPA-CSU  Cooperative
Agreement No. CR808650-02. The first is
entitled, "Filtration of Giardia Cysts and
Other Substances: Volume 1. Diatoma-
ceous Earth Filtration," EPA/600/2-84/
114, June 1984, and the third is entitled,
"Filtration of Giardia  Cysts  and Other
Substances: Volume 3. Rapid Rate Filtra-
tion," EPA/600/2-85/027, April 1985.

Objective
  This report describes the results  of
experimental research to  evaluate the
effectiveness of slow  sand filtration for
removal of Giardia lamblia cysts. Other
variables studied were turbidity, particles,
total coliform bacteria, and standard plate
count bacteria. These dependent variables
were evaluated  with respect to the influ-
ence of the independent variables—
design  and operating conditions.  The
research had two phases.  Phase I oper-
ating conditions included hydraulic load-
ing  rate,  cyst  concentration,  bacteria
concentration, biological maturity of sand
in the filter bed, age of schmutzdecke, and
temperature.  Phase II operating condi-
tions included depth of sand, size of sand,
temperature, disinfection of  raw water,
and nutrient addition to raw water.

Design and Operation of
Slow Sand Filters
  Slow sand filtration is a  passive filtra-
tion process—that is, it is subject to very
little control by  an operator. The process
involves no chemical  addition or back-
wash. For recommended design, effective
sand size ranges from 0.15 to 0.35 mm,
with a uniformity coefficient of less than
2. The sand bed depth ranges from 60 to
120 cm and is supported by graded gravel
30 to 50 cm deep. Drain tiles are placed at
the bottom of the gravel support to collect
the filtered water. Hydraulic loading rates
range from 0.04 to 0.40 m/hr.
  During operation of the slow sand filter,
biological growth occurs within the sand
bed and the gravel support. Also, a layer
of inert deposits and biological material
called  the  schmutzdecke forms on the
surface of the sand bed. In the literature,
the effectiveness of slow sand filtration is
attributed  mostly to the role of the
schmutzdecke, but this  research  has
found that  both the schmutzdecke and
the biological growth within the sand bed
have important roles in the effectiveness
of slow sand filtration. The  latter may
require weeks or months to develop,
depending  on  the nutrients within the
raw water.
  Operation of a slow sand filter requires
two periodic tasks: (1) cleaning by removal
of the schmutzdecke, and (2) replacing of
the sand. Schmutzdecke is removed when
the headloss exceeds the designed value,
which  may range from 1 to 2 m. After
draining the filter, the schmutzdecke is
removed by scraping about 2 cm from the
surface of  the  sand bed. The removal
interval depends on the rate of accumula-
tion of material, which in turn is related to
the contaminants present in the  raw
water  and  the hydraulic loading rate.
Since operating expenses are affected by
the frequency of schmutzdecke removal,
pilot testing is advisable to determine this
important operating parameter. Replacing
sand is necessary after repetititve scrap-
ings  have reduced the  sand  bed in the
filter to its lowest acceptable depth.
  A number of slow sand filtration treat-
ment plants have been built in the United
States, but most were completed in the
first  decades of the century. Today the
technology  is well established in Europe,
though it never gained a firm foothold in
the United  States. The process seems
well  suited to small communities that
need a technology  less  complex than
rapid sand filtration.


Experimental Apparatus and
Methods

Apparatus
  The Phase I experimentation was con-
ducted using  three 1-ft-diameter pilot
plant units operated in parallel. The filters
were packed with 96 cm of sand (dio =
0.28mm, d8o = 0.41 mm) supported by 46
cm of coarse sand and gravel. The effluent
was  routed through  a  241 -mm, 5-jum-
pore-size membrane filter for  Giardia
sampling, or it could flow directly to the
constant head discharge device.  For
temperature  control,  the filters were
equipped with cooling coils in the heads,
and  the filter feed tank had a built-in
temperature control. Temperatures were
maintained constant throughout the sys-
tem within the range 3° to 17°C.
  Sixslow sand filter columns were used
in Phase II arranged in a circular configu-
ration about an operating platform. The
constant head tank in the center dis-
tributed equal flow to each filter by means
of orifices. The six columns were operated
continuously for a 12-month period.
  The effluent flow could  be directed
through the constant head outflow device
or through a 142-mm membrane filter
used for  Giardia sampling.  A cooling
element  was  used for two filters to
maintain temperatures at  5° and 2°C.
The  feed  water  to the six filters  was
maintained at 17°C by means of cooling
and heating elements located in a 1200-L
feed tank.
Operation
  The three slow sand filters in Phase I
were operated continuously from August
1981 to Decemoer 1982 at  hydraulic
loading rates of 0.04,  0.12,  and 0.40
m/hr filters  designated  1,  2, and  3,
respectively.  The  common  feed tank
delivered the same influent to each of the
slow sand filters,  thus  allowing for the
evaluation of process response to differ-
ent hydraulic loading rates.
  The other operating variables studied
were(1) temperature, (2) influent bacteria
concentration and cyst concentration, (3)
age of the schmutzdecke, and (4) biologi-
cal maturity of the sand in the filter bed.
These variables were changed system-
atically to  determine  their effect  on
removals of Giardia cysts, bacteria, tur-
bidity, and particles.
  Temperature effects were examined by
operating the system at 5° and 15°C. The
highest temperature permitted during
Giardia testing was 15°C. This upper limit
was based on observation that the cysts
deteriorate at higher temperatures.
  Giardia cyst concentrations were varied
between 50 and 5,075 cysts/L. Because
the filtration  removal processes  were
highly effective, high influent cyst con-
centrations were  necessary to ensure
passage of a few cysts through the filter.
The high cyst  concentration also permit-
ted discernment of possible functional
relationships and encompassed the high-
est expected ambient concentration,
which was estimated as 500 cysts/L.
  Total  coliform bacteria  ranged from
almost 0 to about 300,000/100 ml. These
latter levels were the result of adding
primary settled sewage to the filter feed

-------
tank and from fecal residue accompanying
Giardia cyst addition. No attempt was
made to .change these concentrations
systematically. The bacteria were added
to challenge the filter, since the raw
water  bacteria counts  were normally
quite low.
  The effect of the schmutzdecke was
determined by testing after it had been
allowed to develop and immediately after
scraping. A developed schmutzdecke is
defined as one that has had at  least 2
weeks to establish itself.
  The biological maturity of the sand bed
indicates the degree of microbiological
development throughout its depth. This
condition  is not measurable, but  is a
function of the number of  weeks  of
undisturbed filter operation. To determine
the influence of microbiological maturity,
testing was done for three filter condi-
tions: (1) new sand bed  and new gravel
support (which simulated start  up of a
new filter); (2) new sand bed with micro-
biologically mature gravel support (which
simulated a filter that  had just  had its
sand totally replaced); and (3) sand bed
and gravel support that are both microbio-
logically mature (which simulated steady-
state operation). Testing under the third
condition was done at various filter ages,
ranging from 26 to 80 weeks. The age of
the filter  can  be used  as an index  of
microbiological maturity for given raw
water  conditions.  The  most pertinent
conditions that affect the length of time to
bed maturity are nutrient availability and
temperature.
  A test run consisted of filling the batch
feed tank with lake water and then spiking
the water  in  the tank with a  known
concentration  of Giardia  cysts.  When
additional conforms were desired, the
tank was also spiked with primary settled
sewage. The feed tank was then sampled
for Giardia cysts, total coliform bacteria,
standard plate count bacteria, particles,
and turbidity.  The same sampling and
analyses were performed on the three
filter effluents the following day to allow
for the needed volume  displacement
within the filter column. This procedure
(i.e., spiking, sampling of the feed tank,
and sampling of the filter effluents) was
continued daily for 3 to 11 days, depend-
ing on the particular test run.
  The six pilot filters in Phase II were
operated in parallel with a common raw
water source. Filter No. 1  was operated as
the control, providing a basis for compari-
son with the other filters. Each of the
other columns was operated with one of
the process variables having a different
magnitude than did the control.
  With  the  six filters, three  levels of
biological activity were studied along with
three other  variables—sand bed depth,
sand size, and temperature. Filter 1, the
control, had the  amount of biological
development that would typically occur
with the  Horsetooth Reservoir  water.
Filter 3 was subjected to 5 mg/L residual
chlorine between runs to minimize bio-
logical growth. Sterile synthetic sewage
was added to Filter 4 to promote additional
biological  growth within  the sand bed.
Filter 2 had a 48-cm  sand bed depth
(instead of the 97 cm of the other filters).
Filter 5 was  packed with sand having d10
size 0.615 mm (instead of the 0.287-mm
size in the other filters). Filters 5 and 6
were operated at 5°C continuously (in-
stead of at the 17°C of the other filters).
  To evaluate the effects  of process
variables, the filters were spiked with a
laboratory culture  of total coliform bac-
teria. Filter No. 3, which was disinfected
by a sodium hypochlorate solution, was
purged of disinfectant with sodium thio-
sulfate before such tests. Effluent sam-
ples from  the six filters were  obtained
once each day during the test period. This
series of measurements, together  with
the spiking,  constituted a test run. Such
test runs were conducted at various times,
usually weekly, throughout the 11 -month
period of continuous filter operation. In
addition to  the total coliform testing,
removals of  turbidity and standard plate
count bacteria were monitored routinely.
Tests with Giardia cysts, concentrated
from dog  feces, were  conducted on a
more limited basis to minimize the fouling
of the sand surface caused by debris and
fats present in the Giardia cyst concen-
trate.

Results

Phase I Removals
  Table 1 summarizes the removals from
Phase I averaged for all data over the
period August 1981  to December 1982
for the three filter columns  operated at
0.04,0.12, and 0.40 m/hr. The number of
samples obtained for each variable and
the range of each are included.
  The  data  showed  uniformly high re-
movals for all dependent variables except
turbidity, which ranged from 27 to 39
percent for raw water turbidities ranging
from  2.7 to 11  NTU. These, turbidity
removals are not as high as  reported by
others (e.g.,  the Kassler plant in Denver)
because of the small clay  particles that
make up the suspended matter in the raw
water  source,  Horsetooth  Reservoir.
About  30 percent of the turbidity in this
water will pass through a 0.45-fjm mem-
brane filter.
  Removals of Giardia cysts, total coli-
forms, and fecal coliforms were all high.
At optimum conditions of  operation, ef-
fluent concentrations of each  approached
their respective detection limits.

Hydraulic Loading Rate
  Well-defined relationships  can be seen
from the data  in Table  1, in which
removals of coliform bacteria,  standard
plate count bacteria, Giardia cysts,  and
turbidity decline with increasing hydraulic
loading rate. For example, average re-
 Table 1.    A verage Percent Removals for Dependent Variables in Slow Sand Filter Columns

                                                  Percent Removal of Parameter
Dependent
Variable
Giardia cysts
Total coliforms
Fecal coliforms
Standard plate
count
Total
Number of
Analyses
222
243
81
351
Range of
Variable in
Raw Water
50-5,075
cysts/liter
0-290.0OO
coliforms/ 100 ml
0-35,OOO
coliforms/ '100 ml
10-1.010.OOO
organisms/ml
Filter 1
v = 0.04
m/hr
99.991
99.96
99.84
91.40
Filter 2
v = 0.12
m/hr
99.994
99.67
98.45
89.47
Filter3
v - 0.40
m/hr
99.981
98.98
98.65
87.99
Turbidity
Particles
(6.35-1 2.7 urn)
891
39
2.7-11 NTU
62-40,506
panicles/ 10 ml
39.18
96.81
32.14
98.50
27.24
98.02

-------
movals of total coliform bacteria declined
from 99.991  percent at 0.04  m/hr to
99.981 percent at 0.12 m/hr. Though
hydraulic loading  rate has an influence
on filtration efficiency, the effect is not
great enough to warrant establishing a
firm design criterion for this parameter.
Rather, the  concern should  be with
respect to economic aspects. For example,
the advantage of reduced construction
costs for a design with a high hydraulic
loading rate  must be weighed against
increased operating costs caused by the
need for more  frequent schmutzdecke
removals.  Performance would be only
slightly poorer  at the higher hydraulic
loading rate.
Microbiological Conditions

The biological conditions governing the
process effectiveness of the filter are: (1)
the degree of schmutzdecke formation;
and (2) the microbiological maturity of the
sand bed. Figure 1 illustrates how these
conditions affect coliform effluent  con-
centrations (i.e., the percent remaining at
hydraulic loading rates of 0.04,0.12, and
0.40 m/hr). Also, each of the bars shows
effluent  coliform concentrations calcu-
lated from a hypothetical influent density
of 1 million coliforms per 100 ml. These
figures  are derived  from  the  percent
remaining data and permit a more tangi-
ble means for comparing results in terms
of whole numbers.
  To evaluate the respective roles of the
schmutzdecke and the  maturity of the
sand bed, it is useful to examine first a
filter  column with a  new sand  bed,
including a new, graded gravel  support.
This simulates a newly constructed filter
during startup when there is no biological
development in the  sand  bed and  no
schmutzdecke. For this condition of new
sand (as indicated in Figure 1  for Run
118), 15.4 percent coliforms remained, or
154,000 coliforms/100 ml remained from
a hypothetical 1 million coliforms/100 ml
in the influent. In other words, filtration
through the new sand will cause an order
of magnitude reduction.
  In contrast to the  initial startup of a
filter is the filter that has been in operation
for a  period of time and has a mature
biological population and an established
schmutzdecke. Such a case is represented
by runs 104, 105, and 106, which show
that a mature filter will reduce  the coli-
form concentration by 2.5  to 4 logs, or
from 1  million coliforms/100 ml to 40,
1000, and 4000 coliforms/100 ml, re-
spectively.
                                   4
700-
1 I 10~
,° IT: 1
V = 0.04M/H VS
o g> o/-
K .g
|| 0.07-
fc «
a. .
1




xH
18
o «~
^
S
	



^
|8
^
3
	 ^--
o >~


106 1O9 112 113
    Simulated Operating
    Condition During Test

Condition of Schmutzdecke

      Condition of Sand

Condition of Support Layer
 Estab.   _.     . Cleaned/ Replaced  Filter
Operation Cleaned Disturbed   Sand ' Start-Up

  Estab.    None   None    None   None

 Mature   Mature n^turbed  New   New

 Mature   Mature Mature  Mature  New
Figure 1.   Effect of schmutzdecke and sand bed conditions on percent of remaining total
           coliforms for three hydraulic loading rates.
  Schmutzdecke removal will  result in
approximately a 1 -log decrease in treat-
ment efficiency  when compared with
operation  under  established conditions.
This result can  be demonstrated  by
comparing runs  107, 108,  and 109 (at
300, 10,000, and 28,000 coliforms/100
ml, respectively) with runs 104,105, and
106  (at 40,  1000, and 4000 coliforms/
100  ml, respectively).
  Replacing sand will result in almost a
2-log decrease  in treatment efficiency.
                . Run 116 shows 70,000 coliforms/100 ml
                 remaining,  compared with  only  1000
                 coliforms/100 ml  for  the established
                 condition represented by run 105.
                   One additional condition tested  was
                 the effect of removing the schmutzdecke
                 and then disturbing the sand bed, as
                 illustrated by runs 110, 111, and  112.
                 This experiment was intended to simulate
                 the effects of a full-scale filter operation
                 in which the filter is drained and the sand
                 bed is disturbed by the movement of men

-------
and  equipment over the filter  surface
during cleaning. The experimental dis-
turbance was accomplished for each filter
by draining the filter for a 2-day period,
removing the schmutzdecke, mixing the
top 10 cm of sand, and pounding on the
sand surface. This experiment caused an
additional 0.5-  to 1-log decrease  in
treatment efficiency compared with the
filter cleaning procedure when no disrup-
tion  occurred.
  The test results as shown in Figure 1
confirm the importance of the role  of
microbiological conditions in the  treat-
ment effectiveness of slow sand filtration.
The best treatment can be expected from
a filter that has been in operation for an
extended period of time. This filter will
have a mature biopopulation within the
filter bed and will have an established
schmutzdecke. The treatment efficiency
will deteriorate markedly as greater por-
tions of the biological  community are
disrupted as shown in Figure 1.
  The data in Table 2 show the effects of
the various filter operating conditions on
removals of Giardia cysts. The first two
rows compare removals of Giardia cysts
between a control filter and one that has
new sand and new gravel support. No
cysts were detected in the effluent of the
control filter, and only 17 cysts/L were
found in the effluent of the  new media
filter. This result  demonstrates that a
filter with a mature biological population
can  remove cysts to the detectable limit,
and  that even a new filter removes 99
percent of the influent cysts. Both filters
were subjected to an influent cyst con-
centration of 2000 cysts/L.
  Results for a similar experiment with a
new sand bed and a mature gravel support
(Run 116) showed zero cysts/L in the
effluent, compared with an influent cyst
concentration of 3692 cysts/L. This result
indicates that  even a modest amount of
microbiological growth in the sand bed, or
indeed in the gravel support, can provide
the  marginal  effect needed to cause
removal of influent cysts to levels below
the detection limit.
  The third secton of Table 2 presents the
results of 15 Giardia removal test runs on
filters with-freshry scraped sand surfaces
(i.e.,  no schmutzdecke  development).
These test runs were arranged chrono-
logically according to continuous filter
operation, ranging from 26 to 70 weeks.
Table 2 shows that removal of  Giardia
cysts to below the detection limit was
achieved in all but four of these test runs.
The  key difference between those tests
 hat  achieved  nearly complete removal
and those that did not was the degree of
microbiological maturity within the sand
bed. All four of the tests in which cysts
were passed occurred during the first 41
weeks of filter operation, indicating that
when the microbiological population has
developed to maturity, complete removal
of Giardia  cysts can be expected. As
demonstrated in Table 2, this result
occurs independently of hydraulic loading
rate, influent Giardia cyst concentration,
and presence of a schmutzdecke.
  The same improvement in Giardia cyst
removal with time is demonstrated by
results shown in the fourth part of Table
2, where the test data are summarized in
chronological order for 24 Giardia tests
with filters having established schmutz-
deckes. These results show that the
removal of cysts improved steadily with
time and was independent of schumtz-
decke age. Cysts were passed through
filters with  12-week-old schmutzdeckes,
whereas they were removed below the
detectable  limit  with 4- to 5-week-old
schmutzdeckes when the microbiological
population  within the filter was given a
longer time to mature. In fact, after 49
weeks of operation, cyst removal below
detection limit was achieved in all cases,
even when influent cyst concentrations
as high as  5075 cysts/L. These results
demonstrate that the age of the schmutz-
decke is not as important for Giardia cyst
removal as the maturity of the microbio-
logical population throughout the sand
bed and gravel support.

Phase II Removals

Design—
  Phase II testing was designed to ascer-
tain the effects of sand size, sand bed
depth, and sustained low temperatures
on removals of Giardia cysts and other
parameters. No experiments to determine
removals of Giardia cysts were done with
the chlorinated filter and the nutrients-
added filter, to avoid the influence of such
testing  on  their performance (e.g., the
influence of nutrients and debris).
  The results of the Phase II  Giardia
removal tests demonstrated that removal
was not affected by increasing the effec-
tive sand size to 0.615 mm, by continuous
operation at 5°C, or by reducing the sand
bed depth to 0.48m. Each of the filters (1,
2, 5, and  6)  had a mature  biological
population  and the same influent water,
and each  was operated  at  hydraulic
loading  rates of 0.12 m/hr. Cysts were
not detected in any of the effluent sam-
ples.
  To induce cyst breakthrough, another
filter  column  was newly packed  with
0.615 mm sand, operated at 0.47 m/hr,
and then challenged during the first 2
days of operation with 2770 cysts/L. This
test resulted in passage of 26 cysts/L
through the filter.  Even under  these
extreme conditions, removal  was  99
percent. Testing with new sand was also
carried out during rapid sand experimen-
tation, which was the subject of Volume 3
of this study. Testing with 0.43 mm sand
at hydraulic loading rates in excess of 14
m/hr and without  chemical addition
resulted in four of eight test runs with
removals of less than 50 percent.

Effects of Process Variables on
Filter Performance in Phase II—
  The effects of process variables on filter
performance were evaluated by using the
percent removals of total coliform bacteria
as the measure of efficiency. Removals of
standard plate count bacteria and turbidity
were determined also, but they were not
suitable for this purpose because hetero-
trophic bacteria were shed by the filter as
a result of internal growth, and turbidity
removal from Horsetooth Reservoir water
had little relation to operation because of
its  unique and  nonrepresentative be-
havior, as reported in the Phase I results.
  Biological Community—Phase II inves-
tigations were designed to study the
effects of low, natural, and accelerated
biological activity on filter performance,
as  represented by Filters 3,  1, and 4,
respectively. For low biological  activity,
growth was prevented in Filter 3 by main-
taining a 5-mg/L chlorine residual be-
tween test runs and dechlorinating with
sodium thiosulfate a test run. Augmented
biological activity was created in Filter 4
by continuously adding sterile synthetic
sewage to the filter. Filter 1 was a control
filter that used raw water from Horsetooth
Reservoir with no alteration; this  filter
represented the natural condition.
  The results of the Phase II testing
demonstrated that as the activity of the
biological community increased from
minimal  biological  community  for the
chlorinated filter to augmented activity
for  the nutrients-added filter,  the re-
movals of conforms, standard plate count
bacteria, and turbidity increased signifi-
cantly. For Filters 3, 1, and 4, removals
were 60, 98, and 99.9 percent for total
coliform bacteria; -89, -41, and 58 per-
cent for standard plate count bacteria;
and 5, 15, and 52 percent for turbidity.
These results  demonstrate the unmis-
takable influence of biological activity on
filter performance.
  Temperature—Decreasing the temper-
ature from 17° to 5° or 2°C decreased

-------
Table 2.    Effect of Operating Conditions on Giardia Cyst Removal by Slow Sand Filtration
           Condition of
Test       Sand Bed and
Objective   Gravel Suport
                Age of   Length of                       Influent      Effluent
               Schmutz-   Time of            Filtration      Cyst         Cyst
                 decks   Operation   Run      Rate       Cone.        Cone.
                (Weeks)   (Weeks)   Number   (m/hr)    (cyst/liter)    (cyst/liter)
                                                                 Percent
                                                                Removal
                                                                   Detection
                                                                     Limit
                                                                   (cyst/liter)
                                                                      Effluent
                                                                      Volume
                                                                      Sampled
                                                                       (literr
Effect
of New
Sand Bed
and New
Gravel
Support
New Sand Bed/
New Gravel
Support

Control Filter:
(Mature Sand
Bed/Mature
Gravel Support)

New Sand Bed/
Mature Gravel
Support
Effect of
New Sand  Control Filter:
Bed        (Mature Sand
           Bed/Mature
           Gravel Support)
10
         0/0
80
                                      0/67
                            67
         118
119
                   116
                   117
                             0.40
                             0.40
                                                           0.12
                             0.12
                     2000
2000
                              3692
                              3692
             17.05
0.0
                                  0.0
                                  0.0
            99.15
                                                        1OO
                                                                  100
                                                                                    100
0.046
0.049
                                                           0.039
                                                                               0.40
610
770
                                                   497
                                                   566
0
0
0
0
0
'fleet of Mature 0
'chmutz- Sand Bed/ 0

-------
removals of coliform bacteria and stan-
dard plate count bacteria from about 99
percent nominally to 90 percent nominally
for each. The filtration efficiency was not
reduced as sharply as expected. The liter-
ature has reported sharp reductions in
percent  removals  as  a result of lower
temperatures.
  Sand Bad Depth—The removals of total
coliform bacteria were 97 percent at a
sand bed depth of 1 m and 95 percent at
0.5 m. This result indicates that bacterial
removal is not overly sensitive to sand
bed depths above 0.5 m. In practice, this
result means that a series of schmutz-
decke removals with the resulting attrition
of the sand bed from 1  m to 0.5m will not
seriously impair  the  efficiency  of  the
filtration process.
  Sand Size—To discern better the role of
effective sand  size,  three filters were
packed with sand with effective sizes of
0.62, 0.28, and 0.13 mm. Eighteen test
runs using pure cultures of total coliform
bacteria were then conducted parallel
with each filter. Each of these filters had a
mature biological population. The coliform
removal improved from 96.0 to 98.6 to
99.4 percent for effective sand sizes of
0.615,0.278, and 0128 mm, respectively.

Summary of Results
  Findings from the experimental pro-
gram are summarized first in terms of the
overall removal effectiveness of slow sand
filtration for the parameters tested,  and
second in terms of the effects of operating
conditions. The effectiveness of slow sand
filtration for  removing the parameters
tested is summarized as follows:
  1.  Giardia cyst removal exceeded 98
     percent for all operating conditions
     tested. Once a microbiological popu-
     lation is established within the sand
     bed, removal will be virtually 100
     percent.
  2.  Coliform  removals  exceeded 99
     percent on the average over all
     operating conditions.  Even with
     new sand, coliform removals were
     85 percent.
  3.  Removals of standard plate count
     bacteria and particles range from
     88 to 91 percent and from 96 to 98
     percent, respectively.
  4.  Turbidity removals  averaged from
     27 to 39 percent. This low removal
     was caused by the fine clay turbidity
     particles characteristic of the lake
     water used in the testing program.
Operating conditions affected removals
of Giardia  cysts, total  conforms,  and
standard plate count  bacteria  in  the
following ways:
  1.  Hydraulic loading rate. Removals of
     Giardia cysts,  coliform  bacteria,
     standard plate count bacteria, and
     turbidity  declined with increasing
     hydraulic loading  rate.  However,
     even  at  0.40 m/hr, removals of
     Giardia cysts and coliform bacteria
     were high—99.98  percent  and
     99.01 percent, respectively.
  2.  Temperature—The Phase II experi-
     ments for mature filters showed.
     that Giardia removals were uniform-
     ly 100 percent for continuous oper-
     ation at both  17°  and at 5°C.
     However, removals of total coliform
     bacteria declined from 97 percent
     at 17°C to  87 percent at 5°C.
     Effluent concentrations of standard
     plate count bacteria were 100 times
     higher at 2°C than at 5°C.
  3.  Influent concentration of bacteria
     and Giardia cysts. Effluent concen-
     trations of coliform  bacteria  and
     standard plate  count bacteria in-
     creased  with increasing influent
     concentrations. At the same time,
     removals increased. A similar rela-
     tion would be expected for removals
     of Giardia cysts, but data were not
     sufficient to establish it. Though
     this information may be of academic
     interest, removals of the above are
     influenced more  strongly by  the
     microbiological maturity of the sand
     bed than by influent concentrations.

  4.  Conditions of the sand bed. A new
     sand  bed removed 85 percent of
     influent coliform bacteria and 98
     percent of influent Giardia cysts. As
     the sand bed mature biologically,
     removals improved to greater than
     99 percent for coliform bacteria and
     virtually  100 percent for Giardia
     cysts. Disturbance of the sand bed
     caused reduced coliform removals,
     but it  had no effect on Giardia cyst
     removals.  Development of  the
     schmutzdecke  further  improved
     removals of coliform bacteria by an
     order  of magnitude. The presence
     or absence of a schmutzdecke has
     essentially no influence on Giardia
     cyst removal efficiency. The micro-
     biological maturity of the sand bed
     is the most  important variable in
     removal of Giardia cysts and coli-
     form bacteria. This mature condition
     develops over a  matter of weeks or
     months, depending on raw water
     conditions.
  5.  Sand Bed Depth—Coliform remov-
      als  averaged 97 percent  for the
      control filter with a bed depth of 1.0
      m and declined only to 95  percent
      for the filter with a bed depth of 0.5
      m. These results demonstrate that
      the bed depth can be reduced to 0.5
      m by repeated schmutzdecke re-
      movals without significant impair-
      ment of filtration removal efficiency.

  6.  Sand Size—Removals of  Giardia
      cysts were 100 percent for all sand
      sizes tested. Removals  of total
      coliform bacteria declined from 99.4
      percent for 0.128- m m sa nd to 96.0
      percent for 0.615-mm sand. Though
      results showed that sand size had a
      functional  influence on  bacteria
      removals, the removals were high
      even with the 0.615-mm sand. So
      the argument for using smaller sand
      is not strong from the standpoint of
      removal effectiveness. Instead, the
      argument for  using  an effective
      sand size  of about 0.35  mm is
      economic. The schmutzdecke will
      penetrate to a greater depth with
      larger sand, necessitating removal
      of more sand during schmutzdecke
      removal  and  resulting  in  higher
      operating costs.  Thus  using the
      smaller  sand  size  is preferable
      when the choice is economically
      favorable and  if  the trade off in
      higher headloss is not appreciable.
  7.  Biological Activity—Phase II results
      showed that the average coliform
      removals for Filter 3, which was
      chlorinated between test runs and
      had  no biological community, were
      only 60  percent. For the  control
      filter, the average coliform removal
      was 98 percent. Filter 4, which had
      nutrients added, showed an average
      coliform  removal of 99.9 percent.
      These  results  augment  those of
      Phase I and establish unequivocally
      the  importance of the biological
      community and its leyel of  activity
      within the sand bed.

Conclusions
  Slow sand filtration is an effective water
treatment technology as  determined by
removals  of total coliform bacteria  and
Giardia cysts. Furthermore, the process is
passive in nature, requiring little action
on  the part of the operator. This tech-
nology should be considered as an alter-
native when water treatment systems are
being selected. Pilot plant testing should
be  done,  however,  to  determine  the
technical feasibility of each alternative.
                                                         ICt: 1MB- 339-111/10845

-------
 The selection should be made by eco-
 nomic analysis and judgments of how
 appropriate the technologies are in terms
 of community attitudes toward operation.

   The full report was submitted in fulfill-
 ment of EPA Cooperative Agreement No.
 CR808650-02 by Colorado State Univer-
 sity under the  sponsorship of the U.S.
 Environmental Protection Agency.
        William D. Bellamy is presently with CH2M-HHI Consulting Engineers, Newport
          Beach, CA  92660; the EPA  author  Gary P. Silverman is with the U.S.
          Environmental Protection Agency, San Francisco, CA 94105; and David W.
          Hendricks is with Colorado State University, Fort Collins, CO 80523.
        Gary S. Logsdon is the EPA Project Officer (see below).
        The complete report,  entitled "Filtration of Giardia Cysts and Other Substances:
          Volume 2. Slow Sand Filtration," (Order No. PB 85-191 633/AS; Cost: $26.50.
          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:
                Water Engineering Research Laboratory
                U.S. Environmental Protection Agency
                Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
     BULK RATE
POSTAGE & FEES P
        EPA
  PERMIT No. G-35
Official Business
Penalty for Private Use $300
                              0169Q64
                              US  EPA  REGION
                              LIBRARY
                              £JO  S  DEARBORN
                              CHICAGO
                  ST.
                     it
                                                             60604

-------