United States
Environmental Protection
Agency
Water Engineering Research
Laboratory
Cincinnati OH 45268
'/i
Research and Development
EPA/600/S2-85/052 June 1985
Project Summary
Slow Sand Filter and
Package Treatment Plant
Evaluation: Operating
Costs and Removal of
Bacteria, Giardia, and
Trihalomethanes
Gordon R. Pyper
A study was conducted to evaluate
two simple methods of water filtration
for small water systems: slow sand
filtration and pressure diatomaceous
earth (DE) filtration. The study address-
es the concerns of small water systems
with regard to Giardia cysts, bacteria,
trihalomethanes (THM's) and operating
costs. Objectives are (1) to determine
effectiveness of the two filtration sys-
tems for removing bacteria, turbidity,
and Giardia cysts under various loading
conditions, (2) to spike bacteria and
Giardia cysts into the raw water under
various loadings to determine the break-
through point, (3) to determine the level
of technical expertise needed to operate
the systems, (4) to obtain operating and
maintenance data and costs, and (5) to
evaluate the potential for the formation
of THM's with the two filtration sys-
tems.
The study was conducted at Mclndoe
Falls, Vermont. Raw water was of high
quality with respect to most parameters,
but the source was a highly organic
impoundment site. Raw water values
were generally low for the principal
parameters studied (total coliforms,
standard plate count, Giardia cysts,
particles, and water temperature).
Slow sand filtration provided depend-
able water treatment with little attention
required, but capital cost was high.
Bacteria and Giardia cysts were re-
moved very dependably at warm water
temperatures (above 5° to 10°C) but
less efficiently at lower temperatures.
Turbidity was below 1 IMTU 99.19% of
the time.
Pressure DE filtration also reduced
Giardia cysts and bacteria dependably,
but the system required full-time, skilled
operation when running and careful
attention to every detail.
Both systems failed to reduce THM
precursors significantly, and both sys-
tems incurred comparable costs for
producing small quantities of water.
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 study compared two simple meth-
ods of water filtration that could be used
for small water systems. Many small
water utilities in traditionally cold, clear
water areas of the country have continu-
ously used such water without filtration
or other treatment, and some do not even
chlorinate. Chlorination might be consid-
ered a minimum requirementfor bacterial
control, but with the increasing occur-
rence of Giardia cysts in surface water
supplies, chlorination adequate to inactive
-------
coliform bacteria may not provide ade-
quate protection. Filtration is needed also.
This study addresses the concerns of
the small water system with particular
emphasis on turbidity, bacteria, and
Giardia cysts. The principal objectives of
this study were:
1. To determine the effectiveness and
efficiency of slow sand filtration
and pressure diatomaceous earth
(DE) filtration on removal of bacteria,
turbidity, and Giardia cysts under
various loading conditions.
2. To spike bacteria and Giardia artifi-
cially into the raw water under
various loading conditions to de-
termine when the loadings might
produce a breakthrough in the
systems and contribute bacteria
and Giardia to the effluent.
3. To observe, record, and evaluate
the level of technical expertise
required to operate the systems
(observations were to be based
primarily on ambient operating
conditions).
4. To obtain operating data in terms of
hours and costs associated with
operating and maintaining the sys-
tem (costs related to chemical addi-
tions, cleaning and restoring the
systems to operation, and similar
well-defined operational require-
ments).
5. To evaluate the potential for and
formation of trihalomethanes
(THM's) in the untreated water and
compare the results with those
from the effluent of a slow sand
filter and a DE package treatment
plant treating the same source of
supply.
This study was conducted at Mclndoe
Falls, Vermont, at the site of the slow
sand filtration plant. The water source
was a small brook fed from a marsh. The
raw water was of high quality with respect
to most parameters, but the source was a
highly organic impoundment site. The
water could be described as low in color,
turbidity, chloride, manganese, calcium,
hardness, alkalinity, nitrate, and sus-
pended solids; pH was neutral. The water
had moderate to heavy levels of iron (0.1
to 1.5 mg/L), total Kjeldahl nitrogen (17
to 56 mg/L), ammonia nitrogen (7 to 25
mg/L), and sodium (1 to 155 mg/L). No
primary standard chemicals were found
to be above MCL limits.
The principal parameters considered
during the study were turbidity, total
coliforms, standard plate count, Giardia
cysts, particles (7- to 12-/um range), and
water temperature. Raw water values for
these parameters were generally low.
Raw water turbidity averaged 1 .4 NTU
during the study (83% of the samples
were 2.0 NTU or less), but some high
spikes occurred during storms or during
road work in the impoundment area.
Concentrations of total coliform and
standard plate count bacteria were in-
fluenced by rain and snow storms. Levels
averaged 296/1 00 mL for total coliforms
and 185/mL for standard plate count
bacteria. Fifty percent of the total coliform
samples have values of 1 00/1 00 mL, and
the standard plate count had values of
80/mL or fewer. The average particle
concentration (7- to 12-//m range) was
1 2,780 per mL, and 50% of the samples
contained 5300 per mL or fewer. Water
temperature tended to be cold most of the
time except in the middle of the summer.
The average temperature was 9.7°C, but
38% of the readings were 2°C or below
from about April 4 to December 1 5.
Slow Sand Filtration
The slow sand filter was operated under
normal ambient conditions and under
special biological stress conditions. The
rate of filtration was maintained at a
constant value of 0.08 m/hr (2 million
gallons per acre per day [mgad]) through-
out the study because summer flows
from the source were not dependable
above this rate.
Total Coliforms and Standard
Plate Counts
The total coliform and standard plate
count results for ambient conditions are
summarized in Figures 1 and 2 and Table
1. Under ambient loading conditions,
reductions averaged 80% for total coli-
forms and 90% for standard plate count
bacteria. However, 90% of the total
coliform and standard plate count samples
showed reductions of 80% or more. Also,
60% of the effluent samples contained
total coliform concentrations of 1/100
mL or fewer, and standard plate count
values were 2/mL or fewer. Eighty per-
cent of the effluent samples showed total
coliform values of 7/100 mL or fewer and
standard plate count values of 4/mL or
fewer. These results showed relatively
dependable bacterial quality in the efflu-
ent with rather variable raw water quality.
The slow sand filter did not show an
immediate response to sudden improve-
ments in raw water quality. When the
bacterial concentration rapidly declined
in the raw water, the concentration of
bacteria in the filtered water may have
been close to or greater than the concen-
tration in the raw water for a day or so.
This circumstance would cause very low
removal percentages or negative removal
percentages (increases).
Recovery of Filter After Scraping
Normal recovery after filter scraping
was an important consideration for filter
operation. Bacterial quality of the water
did suffer immediately after filter clean-
ing, particularly during cold water condi-
tions. About 2 days after cleaning,
reductions decreased from approximately
95% to 20% for total coliforms, and from
about 90% to -300% for standard plate
count bacteria. In warmer water situa-
tions, the reduction in total coliforms
dropped to about 55% in about 7 days and
to approximately 93% for the standard
plate count in the same time period. This
disruption of the treatment capability was
700
I
80
o? 40
§ 20
o Total Coliforms - Influent
• Total Coliforms - Effluent
I ill
JO 100
Total Coliform No./100 mL
WOO
Figure 1.
Total coliforms in influent and effluent of slow sand filter during
normal operation, filtration rate of 0.08 m/hr.
-------
I
I
CO
700
80
60
40
20
o Standard Plate Count - Influent
• Standard Plate Count - Effluent
\ \ \
\ \ \
\
10 100
Standard Plate Count Bacteria/mL
1000
Figure 2. Standard plate count bacteria in influent and effluent of slow sand
filter during normal operation, filtration rate of 0.08 m/hr.
Table 1. Percent Reduction of Bacteria by Slow Sand Filtration*
Parameter
Total Coliforms/ 100 mL
Standard plate count
bacteria
Number of
Samples
67
67
Mean
79%
89%
Maximum
99.99%
99.99%
Minimum
-60%
-200%
*Ambient operation for all periods from 6/18/82 to 5/4/84.
much less severe. These results demon-
strated that temperature must be con-
sidered when evaluating any biological
impact on slow sand filtration.
Spiking the Filter After Cleaning—
During the summer months when the
water was warm (22°C), the filter was
spiked with bacteria immediately after it
was cleaned. The filter showed very little
disruption of bacterial treatment capabil-
ity after being cleaned, even with heavy
spikes of total coliform and standard plate
count organisms. The filter effluent aver-
aged 7/100 mL for total coliform 8 to 10
days after cleaning and 2 to 3/100 mL 15
to 20 days after cleaning. The standard
plate count results were similar. The
effluent averaged 2/mL for standard plate
count bacteria 8 to 10 days after cleaning,
and it decreased to 1/mL 20 days after
cleaning.
Turbidity After Filter Cleaning—
Turbidity reductions after filter cleaning
were similar to bacterial results. Under
warmer ambient water conditions, turbid-
ity reductions tended to remain at 92% to
95% removal throughout the recovery
period. I n cold water conditions, removals
dropped from 92% to 95% to about 50% in
1 to 5 days. Recovery during cold weather
tended to take 10 to 20 days and was
much more erratic than warm weather
results. However, under the worst of the
cleaning conditions, the filtered water
never exceeded 0.9 NTU except for the
start-up condition when the filter had just
been cleaned. The erratic response would
be expected because the filter had not
been used for several years and thus
represented a biologically immature sand
bed. The results for all values, including
turbidity, were very erratic and irregular
for about 100 days after this initial
cleaning, a result completely different
from subsequent cleaning and normal
operating results.
Particle Count After
Filter Cleaning—
Information on reductions of particles
(7- to 12-//m range) proved to be very
erratic for this water. Reductions were at
times in the 90% to 95% range; but at
other times they were -100% to -200%,
with no particular pattern to their changes.
For normal ambient operation, the aver-
age particle reduction was about 45%,
and during recovery from cleaning, aver-
ages were about the same. Recovery from
cleaning with bacterial spiking demon-
strated severe impacts, with an average
reduction of -8%. Little correlation was
apparent with bacteria, turbidity, or
Giardia removal. Slime growths in the
water were believed to contribute to this
erratic behavior, as such organisms
occasionally clogged the particle counter
and did continuously clog the automatic
turbidimeters, often within a day after
cleaning.
Bacterial Spiking
Bacterial spiking was also evaluated
under normal operating conditions. The
temperature influence was again demon-
strated during cold water conditions (1 °C).
Effluent bacterial reductions deteriorated
steadily during spiking and showed signs
of breakdown in treatment about 10 days
after the spike started. During warmer
water conditions(9°C), the effluent hardly
showed any disruption for either total
coliforms or standrd plate count, with
bacterial loads of 1000 to 10,000/100
mL for total coliforms and 100 to 1000/mL
for standard plate count organisms.
Cyst Reduction
The removal of Giardia lamblia cysts
was an important consideration in this
research. Eight spikes of fluorescent-
tagged cysts were applied to the slow
sand filter during this study, including a
series of 5 spikes applied at 1-month
intervals. After each spiking, 8% of the
filter effluent was sampled each day on a
continuous basis. The sampling periods
ranged in length from 6 days to 5 months;
the 5-month sampling period was done
during the series of 5 cyst spikes. Cyst
removal was excellent, even though some
cysts did appear in the effluent. Cyst
reduction through the filter is summarized
in Figure 3 and Table 2. Removals tended
to be best (99.9%) during warm water
conditions and less effective (99.5%)
during cold water conditions, except for
one result in cold water that yielded only a
93.7% reduction. This somewhat lower
reduction occurred during sewage spiking
(water temperature 0.5°C), and it appear-
ed that the biological capabilities of the
filter were stressed almost to the limit
since removals of both bacteria and
Giardia cysts showed degradation in their
removal patterns. This hypothesis should
be investigated further because it has
important ramifications when considering
cold water situations that might involve
sewer breaks or other contamination with
Giardia cysts. During the study, water
temperatures were 2°C or below for 3.5
months or more per year.
-------
Turbidity
Turbidity reductions under normal am-
bient conditions were uniformly good.
The results are summarized in Table 3.
The average slow sand filter effluent
turbidity value wasO.22 NTU, and 90% of
the effluent samples showed 0.5 NTU or
less.
Trihalomethane Precursors
THM precursor reduction was studied
for both warm and cold water conditions.
Precursor reduction was evaluated by
generating THM's under conditions of
excess free chlorine residuals. No appre-
ciable reduction in precursors occurred
when passing through the slow sand
filter.
Diatomaceous Earth Filtration
Diatomaceous earth filtration was
evaluated using pressure (0.74 and 0.93
m2, or 8 and 1 0 ft2 septum area filters).
Filtration ratesaveraged2.4and4.3 m/hr
using Celite 503®.* At low ambient
bacterial loads (5/100 mL of total coli-
forms and 30/mL or fewer standard plate
count bacteria), the reduction in bacteria
appeared to be low.
Total Coliforms and Standard
Plate Counts
The average reduction was 87% for
total coliform and 89% for standard plate
count bacteria. Of these runs, 50%
showed 92% reduction in total coliforms
and 90% or more reduction in standard
plate count bacteria. Filtration at 4. 3 m/hr
demonstrated slightly lower reductions in
bacteria compared with the 2.4 m/hr
rate: the average reduction was 77% for
total coliforms and 87% for standard plate
count bacteria at the high rate of filtration
compared with the 92% and 88%, respec-
tively, at the low rate; but this difference
did not constitute a significant variation.
Bacterial Spiking
Under bacterial spiking conditions, the
average reduction was 98% for total
coliforms and 93% for standard plate
count bacteria. Of these runs, 50% re-
duced total coliforms 98% or more and
standard plate count, 94% or more.
• jf
99.80 - /
/
/
| 99.60 -J
^
03
*
I 99.40 -
£ •
^
39.20 -
»-
1
1
0 5 10 15 20
Temperature ° C
Figure 3. Slow sand filter reduction of
f filtration rate of 0.08 m/hr).
Giardia cysts at
various temperatures
Table 2. G iardia Cyst Removal with the Slow Sand Filter*
Days Since Number of
Most Recent 24-Hour
Filter Samples Cysts
Scraping Collected Recovered
34 6 4032
88 26 3214
117 26 3503
50 38 4090
144 28 485
174 32 51
82 7 8
35 5 42
Spike Date Sample Dates
2/28/83 3/ 1 / 83-3/6/83
1/16/84 1/17/84-2/14/84
2/14/84 2/15/84-3/12/84
12/8/83 12/9/83-1/16/84
3/12/84 3/13/84-4/9/84
4/9/84 4/10/84-5/11/84
5/ 1 6/83 5/ 1 7/83-5/23/83
8/8/83 8/9/83-8/12/83
'Filtration rate is 0.08 m/hr.
Number of
Samples With
No Cysts
3
9
10
19
2
19
6
2
Temperature
f°C)
0.5°
0.5°
0.5°
0.75°
0.75°
7.5°
11°
21°
Cyst
Removal t%)
_ _ —
93.7
99.62
99.46
99.36
99.91
99.99
99.98
99.98
Cysts Applied
2.1 x10e
2.55 x 107
2.31 x 107
2.3 x 107
2.55 x 107
2.31 x 10r
2.1 xW6
a
8 x 10e
Turbidity
Turbidity reduction was fairly consis-
tent during the DE runs. The average
reduction was 71'
effluent of 0.5 NTU.
with an average
'Mention of trade names or commercial products
does not constitute endorsement or recommenda-
tion for use.
Giardia Cyst Spiking
One Giardia spike of 8 x 106 cysts
produced 99.97% removal—a result con-
sistent with the DE results previously
reported by other researchers. No other
Giardia spike applications were conducted
because cyst supplies needed to be
conserved and because other work had
shown similar results. The results of this
study are shown in Table 4. The reduction
-------
Table 3. Turbidity Removal with the Slow Sand Filter
Turbidity (NTU)
Sample
Point
Influent
Effluent
Number of
Observations
674
701
Average
1.4
0.22
Maximum
59
8.0
Minimum
0.2
0.05
Table 4. G iardia Cyst Removal by the Diatomaceous Earth Filter with Celite 503®
Filtration
Rate m/hr
Temperature
°C
Cysts
Applied
Actual
Cysts
Recovered
Portion of
Effluent
Sampled
Cyst
Removal
3.8
23°
8x10"
48
5.3%
99.97%
of 99.97% was excellent at the 4.3 m/hr
filtration rate.
THM Precursors
The DE filter did not affect the THM
precursor reduction, and the results were
similar to the values for the slow sand
filter.
For this water body, feed rates of 3 to 4
mg/L appeared to produce the best results
when related to pressure buildup. The
higher filtration rate (4.4 m/hr) appeared
to have a slight advantage over the lower
rate (2.4 m/hr) filtration.
Grade of Diatomaceous Earth
DE grade effects were not studied
extensively. From very limited studies, it
did appear that a finer grade might have
been advantageous for this water in that
body feed rates on the order of 28 mg/L
produced much longer operation time
than the 3 mg/L at the same rate of
filtration. The longest operation times
were provided by the coarse grade at body
feed values of about 3 to 4 mg/L and by
the fine grade at about 28 mg/L or
possibly more.
Water Plant Operation
Operation and cost information was
accumulated during the course of this
study and analyzed to evaluate slow sand
filter needs and compare them with DE
operation.
Cleaning requirements for the slow
sand filter can be expressed by the
relationship:
Y = 1.6 + 3.5x±1.0
where Y = person hours to clean, and x =
removed sand volume (m3). A considerable
amount of variation existed in a single
determination. The type of installation
and operating conditions affect the clean-
ing results.
The length of the slow sand filter runs
could not be extrapolated from head loss
information. For this full-scale filter and
the particular water source, filter runs
could be expected to range from 100 to
250 days, but plots of head loss versus
time tended to be very flat for many
months and then suddenly increase ex-
ponentially to limiting head loss values.
Many more studies over 5 to 10 years
would be required to provide sufficient
data for predicting a pattern, if a pattern is
possible.
Operation time data were recorded
during the study to determine the time
required to obtain and record turbidity,
temperature, and chlorine residuals and
also to sample for bacteria and make
chlorine solutions. The mean time re-
quirements are as follows: 1.46 hr for
reading, testing, and recording turbidity
and temperature; 1.54 hr for bacteria,
turbidity, and temperature; 0.38 hr for
chlorine residual; and 0.20 hr for chlorine
preparation. Results will vary consider-
ably depending on facilities and person-
nel.
Production costs for water were eval-
uated. If the slow sand filter used in the
study were constructed new in 1984 and
operated at full capacity (0.08 m/hr),
water would cost $4.60/1000 gal. A
similar-capacity DE pressure filter might
produce water at a comparable cost. The
DE studies did not provide sufficient
operating data to permit extrapolating DE
costs in a meaningful manner. The re-
search included too many operational
variables to permit development of infor-
mation that would be comparable to that
from a functioning treatment plant. The
DE filtration research operating data were
not comparable to the slow sand filter
operating data.
The stated cost values are produced
costs, not delivered costs. These costs are
high. For small systems, however, they
are comparable to costs that could be
incurred when individuals drill wells and
install private water systems that can
produce high quality water equalling or
exceeding the Safe Drinking Water Act's
quality requirements.
Conclusions
Slow Sand Filtration
1. Slow sand filtration provided de-
pendable water treatment with a
minimum of attention, but capital
cost was high.
2. Turbidity was below 1 NTU 99.19%
of the time. After the first 100 days
of operation, the effluent turbidity
values were below 1 NTU 99.68%
of the time. Turbidity values were
0.2 NTU, or less, 72% of the time.
3. Slow sand filtration reduced total
coliforms to 10/100 mL, or fewer,
86% of the time under ambient load
conditions.
4. The standard plate count bacteria
were reduced to 10/mL, or fewer,
94% of thetime under ambient load
conditions.
5. Massive spikes of total coliform and
standard plate count bacteria were
removed from raw water at tem-
perature conditions above 5° to
10°C.
6. Slow sand filtration was not as
efficient in removing bacteria at
temperatures below 5°C, particu-
larly around 0° to 1°C.
7. Giardia cysts were removed very
dependably; 99.98% removals or
better were achieved under warm
temperature conditions.
8. Giardia cysts were not as complete-
ly removed at low temperatures; at
temperatures below 7°C, removals
were 99.36 to 99.91%.
9. Heavy applications of bacteria and
Giardia cysts to the filter at the
same time under cold conditions
produced signs of competition for
the biological treatment capability.
Giardia cyst removal was reduced
to 93.7%, and reduction of total
coliforms and standard plate count
bacteria dropped to 43% and to 79%
to 82%, respectively.
-------
10. Slow sand filtration did not produce
any significant reduction of THM
precursors.
11. Erratic particle reduction in the 7-to
12-/ym range did not compare with
the Giardia cyst removal results.
12. Particle reduction did not provide a
dependable method of predicting
Giardia cyst removal in this full-
scale operating filter experiment
with this particular water.
13. The mature filter recovered from
cleaning within 2 weeks to provide
dependable bacteria and turbidity
removal. Limited data showed that
at times under warm weather con-
ditions the effluent water contained
satisfactory bacteria and turbidity
concentrations before 2 weeks had
elapsed.
14. A minimum of 1.5 hr of operation
were required each day to run the
system properly and meet monitor-
ing requirements.
Pressure Diatomaceous Earth
Filtration
1. Pressure DE filtration removed
Giardia cysts dependably using
Celite 503® with 99.97% reduction.
2. Total coliforms were reduced 86%
or more in 70% of the samples, and
standard plate count bacteria were
reduced 80% or more in 70% of the
samples.
3. Eighty-six percent of the average
run values for total coliforms did not
exceed 8/100 ml, and 82% of the
average run values for standard
plate counts did not exceed 12/mL.
4. The average bacterial content in the
effluent under ambient conditions
was 38/100 mL for total coliforms
and 6/mL for standard plate count
bacteria.
5. Under spiking conditions, the aver-
age reduction was 97.6% for total
coliforms and 92.7% for standard
plate count bacteria. Eighty percent
of the average run values showed
total coliform reductions of 95.8%
or more and standard plate count
reductions of 87.5% or more.
6. Under spiking conditions, effluent
total coliforms averaged 122/100
mL (107/100 mL or fewer for 77%
of the runs), and standard plate
counts averaged 47/mL (7/mL or
fewer for 77% of the runs).
7. Pressure DE filtration provided rapid
cycle time and flexible filter water
production capability.
8. The system required full-time oper-
ation when running and careful
attention to every detail of opera-
tion.
9. Highly skilled operators are needed
for dependable production of the
most satisfactory water the treat-
ment process can produce.
10. The costs for producing small quan-
tities of water are comparable with
those for the slow sand filter.
11. The process is labor-, energy- and
materials-intensive, as compared
with that of slow sand filtration.
12. Particle reductions in the 7- to 12-
pim range were erratic for this
water. Slime organisms may have
contributed to the erratic results.
The full report was submitted in fulfill-
ment of Cooperative Agreement No. CR-
809284010 with Mclndoe Falls Fire
District #3, under the sponsorship of the
U.S. Environmental Protection Agency.
Gordon R. Pyper is with Dufresne-Henry, Inc., North Springfield, VT 05150.
Gary S. Logsdon is the EPA Project Officer (see below).
The complete report, entitled Slow Sand Filter and Package Treatment Plant
Evaluation: Operating Costs and Removal of Bacteria, Giardia. and Trihalo-
methanes,"(Order No. PB 85-197 051 /AS; Cost: $17.50. 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
t U.S. GOVERNMENT PRINTING OFFICE 1«W - 559-111/10862
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
OCOC329 PS
U 3 ENVIR PROTECTION AGENCY
REGION 5 LIBRARY
230 S DEAPBCRN STREET
CHICAGO It- 6Q<$0<
------- |