United States
Environmental Protection
KAgency
Water Engineering
Research Laboratory
Cincinnati OH 45268
..V
Research and Development
EPA/600/S2-85/056 Aug. 1985
Project Summary
Slow Sand Filter Maintenance:
Costs and Effects on Water
Quality
Raymond D. Letterman and Thomas R. Cullen, Jr.
A study was conducted to determine
the effects of scraping on slow sand fil-
ter efficiency and to quantify the labor
required to operate and maintain a
slow sand fitter. The data were ob-
tained by monitoring scraping and
other maintenance operations at a
number of full-sized slow sand filtration
plants in Central New York.
Ripening periods (the time required
for filtrate quality to improve after filter
scraping) were evident in the slow sand
filtration plants visited. Ten mainte-
nance operations were monitored in six
filtration plants. In four of the ten oper-
ations, there was some evidence of a
ripening period. This evidence included
filtrate turbidity and/or HIAC particle
counts that were greater for a recently
scraped filter than for an on-line control
filter. The length of the ripening period
ranged from 6 hr to 2 wk. The data also
suggest that a recently scraped filter is
less efficient than a control filter in at-
tenuating a spike input of lower-quality
raw water. Factors such as the use of
prechlorination, water temperature,
scraping methodology, and frequency
of filter maintenance did not seem to be
related to the presence or absence of a
ripening period. However, the nature of
the'particulate matter in the raw water
apparently has an important effect on
filtrate quality, and a pilot plant study
should always be conducted before a
slow sand filtration plant is con-
structed. Continuous monitoring of the
turbidity of each filter effluent may be
required to ensure that slow sand filter
maintenance operations do not have a
detrimental effect on treated water
quality; the capability to waste individ-
ual filter effluent for a period of time
may be necessary in some cases to pre-
vent quality deterioration.
Typical labor requirements for fitter
scraping are approximately 5 man-
hours/1000 ft2 of fitter surface. The re-
sanding operation requires approxi-
mately 50 man-hours/1000 ft2. No clear
relationship was observed between the
frequency of scraping and the raw
water quality or maintenance proce-
dures. Operational convenience ap-
pears to be a controlling factor in the
plants visited.
This Project Summary was devel-
oped by EPA's Water Engineering Re-
search 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
A large proportion of the public sur-
face water supplies in the United States
are small and unfiltered. Many of these
systems have experienced difficulty in
meeting the 1 nephelometric turbidity
unit (NTU) maximum contaminant level
(MCL) in the U.S. Environmental Protec-
tion Agency (EPA) Drinking Water Regu-
lations. Some of these communities
have failed to meet the MCL for coliform
group bacteria. The slow sand filtration
process may be an appropriate treat-
ment alternative for many of these
small systems.
When slow sand filters are used by
water utilities, the raw water is typically
given no pretreatment. Uncoagulated
water is applied and slowly passed
through the sand filter. As the run pro-
gresses, a layer of soil particles and bio-
-------
logical matter (the schmutzdecke) accu-
mulates on the top of the sand bed and
the head loss increases. When the ter-
minal head loss is reached, the water
level is drawn down to 10 cm or more
below the surface of the sand, and the
schmutzdecke and a thin layer of sand
are removed.
The primary goal of this research was
to determine the effects of slow sand
filter scraping on water quality and op-
eration and maintenance costs. The
major objectives were as follows:
1. To evaluate filter water quality be-
fore and after slow sand filters are
scraped and to compare it with the
quality of raw water and control fil-
ter effluent to determine how filter
efficiency is affected by scraping.
2. To quantify the labor required to
operate slow sand filter plants and
to compare the labor needed for
routine operation and monitoring
with that needed for scraping fil-
ters.
3. To determine the frequency of fil-
ter scraping (length of run or vol-
ume of water filtered per run) and
relate this information to raw
water quality, water treatment be-
fore filtration (if any), filtration
rate, sand size, and other relevant
design factors. A related objective
was to determine whether and to
what extent the frequency of filter
scraping varies with the depth of
sand removed during the scraping
operation.
Seven treatment plants were studied
in New York State: Auburn, Geneva,
Hamilton, Ilion, Newark, Ogdensburg,
and Waverly. The typical study visit in-
volved traveling to the plant site 1 or 2
days before a filter was to be scraped.
The plant was toured, and the plant
records were examined to determine fil-
ter run lengths and historical water
quality. The effluent from the filter to be
scraped was sampled, along with the
raw water.
The manpower, techniques, and
equipment used in scraping (or resand-
ing) the filters were determined by ob-
servation and interview and recorded.
Approximately 50 samples were
taken during each plant visit. When
water flow through the filter was started
after scraping, grab samples were col-
lected for a period of at least 24 to 48 hr.
Samples were withdrawn from the
scraped filter effluent, a control filter ef-
fluent, and the raw water. The control
was a filter that had been on-line for at
least 1 month.
The water temperature and turbidity
were measured immediately after the
sample was drawn. Standard plate
count and total coliform bacteria analy-
ses were started within 0 to 4 hr after
sampling.
The samples were transported to
Syracuse University for particle count
and size analysis on an HIAC particle
size analyzer.
Samples of the filter sand were sieved
to determine the size distribution, and a
sand dissolution test was conducted
using the procedure given in AWWA
Standard B100-80.
Results
The average operating flow rate for
the sites visited ranged from approxi-
mately 0.3 MGD at Hamilton to 6.0 MGD
at Auburn (Table 1). The average raw
water turbidity was less than 3.0 NTU
for every site except Waverly, where the
average was approximately 8 NTU. Fil-
ters are covered at all of the sites but
Hamilton, and two filters at Ilion are un-
covered.
The average operating filtration rate
is the average operating flow rate for
the slow sand filters divided by the total
filter plan area. Filtration rates ranged
from 0.04 to 0.19 m/hr and had an aver-
age value of 0.15 m/hr.
Three of the plants visited (Ilion,
Newark, and Waverly) practice prechlo-
rination. At Newark, prechlorination is
used to control biological growth in the
transmission line between the lake and
the treatment plant. Waverly uses
prechlorination to oxidize iron and man-
ganese and to decrease the filtrate tur-
bidity. The purpose of prechlorination
at Ilion was not stated by plant person-
nel.
Table 1. Characteristics of the Slow Sand Filtration Plants Visited
Location
Auburn
Geneva
Hamilton
Ilion
Average
Operating Flow
Rate for Slow
Sand Filtration
(MGD)
6.0
2.5
-0.3
1.5
Raw Water
Source
Owasco Lake
Seneca Lake
Woodman's
Pond
Several small
streams feeding
reservoir
Average Raw
Water Turbidity
from Plant
Records (NTU)
1.5-2.0
1.0
1.0-1.5
3.0
Total Slow
Sand Filter
Plant Area
(ft2)
74,100
30,492
12,724
19,526
Design
Filtration
Rate
(m/hr)
0.11
0.19
Average
Operating
Filtration
Rate
(m/hr)
0.14
0.19
0.04
0. 16-0. 18
Prechlorination
NO
NO
NO
YES
Covered
Filters
YES
YES
NO
2 uncovered,
3 covered, and
1 not used
Newark
Ogdensburg
Waverly
2.0
3.6
1.2
Canadaigua Lake 3.0 {summer)
1.0 (winter)
St. Lawrence
River
1.0-1.4
Surface runoff to 7.0-9.0 (may
reservoirs be as high as
20-40 during
high runoff
periods)
21,684
33,600
12,000
0.16
0.20
0.16
0.16
0.18
0.16
YES
NO
YES
YES
YES
YES
-------
The efficiency of filtration in a slow
sand filter is at least partly determined
by the presence of viable microorgan-
isms within the filter bed; thus the use
of prechlorination in these systems
would be detrimental to filter perfor-
mance. The effluent average,turbidity
(for the control filter) was compared
with the influent average turbidity at
each site and for each monitoring pe-
riod in which a control filter was sam-
pled. The values were averaged for the
entire length of each sampling period
(using weighted averages based on
flow volume) and used to calculate the
percent turbidity remaining in the efflu-
ent. For the three cases in which pre-
chlorination was used (4 sets of data),
the average and the standard deviation
of the percent turbidity remaining were
17% and 7.9%, respectively. For the
three cases in which there was no pre-
chlorination (6 sets of data), the average
and standard deviation of the percent
turbidity remaining were 21% and 7.6%,
respectively. Though other factors may
have obscured the true significance of
adding chlorine before slow sand filtra-
tion, these results do not clearly indicate
that prechlorination is detrimental to
performance. In fact, it may have had a
slightly positive effect on turbidity re-
moval in the plants sampled.
The effective size of the filter sand
ranged from 0.15 mm atWaverlytoO.45
mm at Auburn. The average effective
size for all sites was 0.33 mm. The uni-
formity coefficient averaged 2.1 and
ranged from 1.7 at Newark and Ogdens-
burg to 2.4 at Auburn, Hamilton, and
Waverly.
Standard B100-80 of the American
Water Works Association states that a
high-quality filter sand should not lose
more than 5% of its weight when it is
treated in a prescribed way with 1:1 HCI
solution. At two of the seven sites, the
sand meets this requirement. When the
sand dissolution tests were conducted,
significant effervescence was noted in
most of the treated samples, suggesting
that these sands contain significant
amounts of CaC03. The significance of
this in terms of filter performance and
operation is not known.
Table 2 summarizes the results that
pertain to the filter scraping operation.
The water production per filter run
ranged from approximately 3000 gal/ft2
at Ogdensburg to 16,000 gal/ft2 at
Geneva and Ilion. The average fre-
quency of filter scraping ranged from
approximately twice a year at Geneva,
Hamilton, and Ilion to 12 times a year at
Ogdensburg. Twice a year at Auburn
(usually during the colder months), the
filters are raked and no sand is re-
moved. According to Auburn person-
nel, raking effectively reduces the head-
loss across the bed without having an
adverse effect on filtrate quality. The
frequency of 4.3 times per year listed in
Table 2 for Auburn includes scraping
(i.e., sand removal) and raking.
The water production (3200 gal/ft2)
and scraping frequency (9.7 times/year)
listed for Waverly in Table 2 are based
on an estimated future filter run length
of 900 hr. This estimate is based on data
obtained in a 9-mo study in which
Waverly personnel developed an opera-
tional strategy for effectively dealing
with the high raw-water turbidity and
the high iron and manganese concen-
trations that frequently occur in their
reservior supply. In the past, Waverly
operators experienced filter run lengths
as short as 2 days. If in the future the
raw water is high in turbidity (>12.5
NTU) and/or high in total iron (>3.0 mg
Fe/L) and manganese (>1.0 mg Mn/L),
the New York State Department of
Health will require Waverly to take the
slow sand filtration plant off-line and to
use their well water supply exclusively.
The last three columns in Table 2
summarize the methods used and man-
power required for filter scraping at the
sites visited. Most of the sites remove
approximately 1 in. of sand from the fil-
ter surface with broad shovels.
The man-hours required for scraping
depend on the depth of sand scraped. In
the cases where 0.5 to 1.0 in. was re-
moved, the labor requirement ranged
from 2 to 9 man-hours/1000 ft2. At Ilion,
where 3 to 4 in. were removed, the labor
requirement was significantly greater
(23 to 42 man-hours/1000 ft2).
The method used to convey the dirty
sand from the filter area also affects the
labor requirement. For example, the
lowest labor requirement was at
Newark (2 man-hours/1000 ft2), where
Table 2. Summary of Filter Scraping Data
Location
Average Filter Run
Water Production
(gal/ft2)
Ilion
Newark
75,487
10,122
Average Frequency
of Filter Scraping
Operations (Number
per year)
1.8
3.3
Amount of Sand
Removed in
Scraping Operation
(in.)
Method(s) Used in
Removing Sand from
Filter Surface
3-4
1.0
Shovels, hydraulic
Shovels, motorized
buggy
Time Required to
Scrape Filters
(man-hours/1000 ft2)
Auburn
Geneva
Hamilton
6,844
15,718
4,302
4.3'
2.0
2.0
0.5
1.0
1.0
Shovels, hydraulic
Shovels, motorized
buggy
Shovels, 50 gal drums,
backhoe
4
4-5
8-9
23-42
2
Ogdensburg
Waverly
2,978
3,200*
12.0
9.7 f
1.0
1.0
Shovels, hydraulics
Shovels, wheelbarrows
4-5
5
"At Auburn, two scraping operations per year are actually occasions when the filters are raked and no sand is removed.
*Water production and scraping frequency estimated by the Waverly personnel for the future using data from a 9-month operations study.
Waverly has had runs as short as 2 days.
-------
Table 3. Estimated Operation and Maintenance Costs for Slow Sand Filters*
Location
Auburn
Geneva
Hamilton
Ilion
Newark
Ogdensburg
Waverly
Average Operational
Flow (MGD)
6.0
2.5
0.3
1.5
2.0
3.6
1.2
Labor for
Scraping
(man-hours/year)
1007
374
224
905
143
8736
582
Labor for
Resanding
(man-hours/year)
618
218
NA
563
226
t
420
Labor for
Day-to-Day
Activities
(man-hours/year)
365
365
365
365
365
365
365
Total Labor
Costs ($/year>
10,597
7,390
5,890
18,331
7,640
23,811
13,670
Total Operation
and Maintenance
Unit Cost
((/WOO gal)
0.5
1.1
5.3
3.3
1.1
2.0
3.7
"All cost figures are based on a $10/hr wage rate except at Auburn, where $3/hr was used because the workers are usually summer
students.
tOgdensburg scrapes and resands simultaneously.
an efficient, motorized buggy was used
to haul the dirty sand from the filter. The
greatest labor requirement for the
plants that scrape 0.5 to 1 in. of sand
was at Hamilton (8 to 9 hr/1000 ft2),
where the dirty sand removal process
involved filling 55-gal drums and haul-
ing them away with a tractor.
Under typical conditions (i.e., re-
moval of about 1 in. of dirty sand with
shovels and conveyance of this sand
from the filter hydraulically), the labor
requirement was approximately 5 man-
hours/1000 ft2 of filter surface.
Table 3 compares the estimated oper-
ating costs for slow sand filters the
treatment plants visited. Day-to-day ac-
tivities devoted exclusively to the filters
(collecting samples, checking the filters,
etc.) were assumed to require 1 man-
hour/day. The labor requirement for
scraping is based on the scraping fre-
quency listed in Table 2. Resanding was
assumed to require 50 man-hours/1000
ft2, based on data from Auburn.
The estimated operational unit costs
ranged from 0.50/1000 gal at Auburn to
5.30/1000 gal at Hamilton. The mean
value for all plants was 2.40/1000 gal.
The exceptionally low value at Auburn
was partly because of their using low
wage summer help ($3/hr) for most
scraping and resanding operations.
Table 4 summarizes the results for fil-
ters that exhibit a ripening period. The
ripening period is the interval immedi-
ately following filter scraping and/or re-
sanding during which the turbidity or
particle count is significantly greater
than that for a control filter.
Ripening periods were observed at
Auburn, Ilion, Newark, and Waverly. At
Auburn, one out of the three scraping
operations monitored exhibited a short
ripening period. For a period of about
6 hr, the filtrate turbidity and particle
count data for the scraped filter ex-
ceeded the corresponding values for
the control filter by a factor of about 2.
However, the turbidity values were al-
ways below the MCL of 1 NTU.
The measurements made at Ilion are
difficult to interpret with respect to a
ripening period. The scraped and con-
trol filters yielded very similar turbidi-
ties after scraping, but approximately
6 hr after the scraped filter was brought
back on line, the particle counts for the
scraped filter began to exceed the
values of the control filter by a factor of
about 2. The time required for this dis-
parity to disappear was about 12 hr.
Two operations were monitored at
Newark. One was a typical scraping op-
eration and the other involved resand-
ing the bed. No ripening period was ob-
served when the scraping operation
was monitored, but a ripening period
was clearly evident in the resanding
case. During the ripening period, the fil-
trate turbidity of the scraped filter ex-
ceeded that of the control by a factor of
about 3. The effluent turbidity of the
control and scraped filters never ex-
ceeded 0.5 NTU; however, the particle
count values were always less than
1000/mLfor both filters.
Ripening periods are a routine occur-
rence at Waverly. Operating personnel
are not surprised if 2 weeks elapse be-
fore the scraped filter turbidity de-
creases to values approaching those of
the control filter. During this study,
ripening was most apparent in the tur-
bidity results: particle counts for the
scraped and control filters appeared to
coincide after about 30 hr, whereas the
turbidity values converged after about
10 days.
The reason for the Waverly's prob-
lems is not clear. The raw water appears
to contain submicron-sized particles
that scatter light and increase the tur-
bidity but are not efficiently removed by
slow sand filtration. According to the
particle count data, Waverly removes
particles larger than 2 |xm as efficiently
as the other plants visited.
Conclusions
1. Four of the ten scraping and resand-
ing maintenance operations moni-
tored produced some evidence of a
ripening period. This evidence in-
cluded filtrate turbidity and/or HIAC
particle count values that were
greater for a filter that was main-
tained than for a control filter that
had been on line for a significant pe-
riod of time.
2. The length of the ripening periods
observed ranged from 6 hr to 2 wk.
The factor that seemed to have the
most significant effect on filtrate
quality was the nature of the particu-
late matter in the raw water. The
presence or absence of a ripening
period does not seem to be related to
prechlorination, water temperature,
scraping method, or frequency of fil-
ter maintenance.
3. The results suggest that a recently
scraped filter is less efficient than a
-------
Table 4. Summary of Filter Ripening Data
Location
Auburn
Auburn
Auburn
Geneva
Hamilton
///on
Type of
Operation
During Visit
11)
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
BULK RATE
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
Official Business
Penalty for Private Use $300
EPA/600/S2-85/056
0063240 MERL
U)U W TILIEY
REGION
ST
It- 60604
------- |