United States
                       Environmental Protection
Municipal Environmental
Research Laboratory
Cincinnati OH 45268
                       Research and Development
EPA-600/S2-84-083 May 1984
&ERA           Project   Summary

                       Water  Filtration  at   Duluth
                       Frank X.  Schleppenbach
                         A wide variety of studies were con-
                       ducted at a new water filtration plant in
                       Duluth, Minnesota, and in the distribu-
                       tion system. The plant was originally
                       constructed for removing asbestos fi-
                       bers from drinking water. Studies were
                       undertaken after the treatment plant
                       began operations. Work included moni-
                       toring the quality of raw, filtered, and
                       distribution system water;  evaluating
                       monitoring  methods;   studying  the
                       chemical conditioning of raw water be-
                       fore filtration; evaluating filtered water
                       quality  during  rate changes  (filter
                       stress); evaluating bacteria removal by
                       filtration; studying sludge dewatering by
                       natural freezing; and modifying the cor-
                       rosive nature of the water caused by
                       alum coagulation.
                         Results showed  that the  filtration
                       plant  can remove  asbestiform fibers
                       from drinking water on a  day-to-day
                       basis. Filtered water turbidity was also
                       consistently reduced below 0.1 ntu. The
                       various filter rate changes studied did
                       not have a continuous  detrimental ef-
                       fect on the effluent quality at the filter
                       plant. Rusty water problems were alle-
                       viated by use  of zinc orthophosphate
                       and sodium tripolyphosphate. Coliform
                       bacteria were removed by filtration in all
                       but one case, but numbers  of bacteria
                       in the influent were low to begin with.
                       Alum sludge was effectively disposed of
                       by natural freezing in lagoons. Static
                       freezing in deep  lagoons was not ef-
                         This Project Summary was developed
                       by EPA's Municipal Environmental 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
  On October 16, 1975, the City of Duluth
received a demonstration grant to construct
a water filtration  plant for removal  of
asbestos fibers from  drinking water. The
grant provided partial  support for the plant
construction  and full support for a wide
variety of studies to be carried out at the new
filtration plant and in the Duluth distribution
system. This report describes the studies that
were undertaken after the treatment plant
began operation. [Work funded at Duluth in-
cluded monitoring raw, filtered, and distribu-
tion system  water quality; evaluating  a
number of monitoring methods; studying
chemical conditioning of raw water before
filtration;  evaluating filtered water quality
during rate changes (filter stress); evaluating
bacteria removal by filtration;  studying
sludge dewatering by natural freezing; and
modifying the more corrosive nature of the
water caused by alum coagulation.

Water Quality Monitoring
  The Lakewood Filtration Plant was built
to remove asbestos  fibers from drinking
water. When the plant was built, the only
method for quantifying amphibole fibers in
water was the transmission electron micro-
scope (TEM)  method. Arrangements were
made to perform TEM asbestos fiber count-
ing at Duluth  because this reduced the time
needed for fiber counting; data could be ob-
tained from the electron microscope labo-
ratory as quickly as from the X-ray diffraction
laboratory. Nonetheless, the method was
slow and expensive,  and other means of
monitoring water quality had to be evalu-
ated. Effective control of the filtration pro-
cess requires monitoring techniques with
rapid response time. Methods  tested  in-
cluded X-ray  diffraction, particle counting,
and turbidimetry.
  X-ray diffraction was used during the ini-
tial Duluth pilot plant research to measure

amphibole mass and thus give an indication
of amphibole fiber  removal.  The method
does  not give actual fiber  counts,  and
estimating fiber numbers on the  basis of
mass is very uncertain, especially when fiber
diameters are variable. In addition, as the
filtered water quality improved, most  am-
phibole mass determinations were below the
detectable limit (BDL). BDL values could not
be mathematically correlated to other water
quality data, so amphibole  mass measure-
ment  ceased.
  Particle counting was done at Duluth with
an HIAC* particle counter and with a special-
ly fabricated optical detector that could dif-
ferentiate between rods (amphiboles)  and
spheres (clays). This detector was developed
and tested by the  University of Minnesota-
Duluth and described in "Optical Detection
of Fiber  Particles  in  Water,"  (EPA-600/
2-79-127,  U.S. Environmental  Protection
Agency,  Cincinnati, Ohio 45268).
  The HIAC optical particle detector,  pur-
chased by the City of Duluth Water Division,
has been used as an in-line detection device.
Typical particle counts in the 1- to GCKm size
range would be 1000 particles/L when the
plant is operating properly and in-line count-
ing is used. When grab samples are  collected
and taken to the counter for analysis, counts
are higher. This result suggests that some
atmospheric or sample container contamina-
tion may occur in  the grab sample  pro-
cedure.  Particle  counting  is a  sensitive
monitoring method even though  it is not
specific for asbestos fibers.
  Turbidity has been monitored with a Hach
2100A bench model for grab samples and
Hach  1720 continuous flow turbidimeters for
on-line monitoring. An in-line Sigrist Photo-
meter (Model  UP52B2) was  also used to
measure turbidity of treated water from one
dual-media filter and one mixed-media filter.
  The plant operation is very closely moni-
tored  and carefully controlled at Duluth. As
a result, turbidity is  routinely reduced from
the typical 1 nephefometric  turbidity  unit
(ntu)  raw water value to 0.04 or 0.05 ntu in
filtered water. This  high level of treatment
can be attained even when the raw water
temperature is near freezing, a situation that
usually makes effective coagulation  more dif-
  Monitoring of amphibole asbestos fiber
counts was done using the TEM  method.
Asbestos fibers in water are  so small  that
most  of them can be seen only with the aid
of an electron microscope. The procedure
is slow and expensive (a few hundred dollars
per sample), so it is suitable only for confirm-
'Mention of trade names or commercial products
 does not constitute endorsement or recommenda-
 tion for use.
ing the efficacy of treatment as indicated by
cheaper, faster monitoring methods. More
than 500 asbestos samples were submitted
for analysis during  this project. Some of
these were routine monitoring samples,
whereas others were collected during special '
  Results of routine monitoring showed that
the filtration plant could dependably, day
after day, reduce the amphibole asbestos
fibers from counts on the order of 10 million
to 1 billion fibers/L in the raw water to levels
below 0.10  million fibers/L. Many filtered
water samples contained fewer than 0.01
million fibers/L.
  Duluth has about 3 days of storage capac-
ity in its  distribution system. As the treat-
ment plant was being completed, the Water
Division  conducted  a  program of  main
flushing and reservoir cleaning to remove the
debris that had accumulated during decades
of unfiltered water usage. After a year and
a half of filter plant operation, the amphibole
fiber counts in the distribution system were
similar to those  in filtered water  at the
Lakewood Filtration Plant.

Chemical Conditioning and
Corrosion Control
  Chemical  conditioning of water  before
filtration  is very important for all rapid-rate,
granular  media filters, and it is essential for
asbestos fiber removal. Chemical condition-
ing was  studied  at  Duluth  using both the
10-gpm pilot plant and the full-scale  plant.
Early in the operation of the Lakewood Filtra-
tion  Plant, pilot- and later full-scale filtration
results showed that anionic polymers were
not  needed for  effective removal  of am-
phibole fibers. Monitoring data showed no
chrysotile in raw water, so  use of anionic
polymer  was discontinued.
   For effective filtration of  Lake Superior
water, careful  control of pH was shown to
be necessary. The  proper operating pH
varies with temperature, ranging from pH 6.8
at 50 F to  pH 7.35 at 32 F.
   Other  chemical  conditioning  research
showed  that fluoride should not be added
before filtration  when aluminum  sulfate
(alum) is the coagulant. Fluoride can form
a soluble complex with aluminum, and pilot
plant results showed that when 1 mg/L of
fluoride was added before filtration, an extra
4 mg/L  of alum (14 versus 10 mg/L) was
needed for  effective filtration. This  effect
should be considered when fluoridation is
   Pilot plant studies of the treatment of tur-
bid water (4 to 10 ntu) were conducted by
injecting a native clay slurry into the pilot
plant influent water. Alum doses of 15 to 25
mg/L were adequate to control turbidity and
amphibole fiber count  in the filter effluent.
  Soon after the filtration plant began op-
erating, the City of Duluth Water Division
began receiving  complaints about  rusty
water. Problems existed with galvanized iron
or steel plumbing. In response to this prob-
lem, zinc orthophosphate was fed to the
distribution system. A zinc concentration of
0.13 mg/L as Zn seems to be adequate for
control  of  corrosion  in  the distribution
system. To reduce problems associated with
corrosion of plumbing,-a sequestering agent
(sodium hexametaphosphate) was added at
a dosage of about 1  mg/L. This treatment
seems to have been satisfactory. After more
research, sodium tripolyphosphate was de-
termined to be a more economical sequester-
ing  agent, and  it is now  being used.

Bacteria  Removal by Filtration
  For a  3-week period  at  Duluth,  pre-
chlorination  was discontinued and chlorine
was added  to  the  filter effluent channel
downstream from the bacterial sampling
points. Bacterial counts were made on raw
and filtered water. Results showed that coli-
form bacteria were successfully removed in
all but one case. The numbers of bacteria
naturally present in the raw water were low,
however, so the filtration plant was not
challenged by large numbers of microor-
ganisms  that might be  present at  other

Effects of Filter Rate  Changes
on Water Quality
  When granular-media filters are operated
at a constant rate,  shear forces slowly  in-
crease within the filter bed as coagulated par-
ticles are removed  from  the  water and
gradually fill the pore spaces between the
media grains. The result of this gradual clog-
ging is seen as slowly increasing head loss.
When filtration  rates change quickly, the
shear forces, or stresses, within the bed also
change  quickly.  Rapid rate changes  can
cause deterioration of filtered water quality.
  At Duluth, three rate change or stress con-
ditions  and two filtration rates were  in-
vestigated. The plant is operated at 20 or 30
mgd (3.2 or 4.5 gpm/ft2), so changes to or
from these rates were studied. The first
stress was caused by stopping plant opera-
tions during a  filter run  and then starting
plant operations again approximately 6 hr
later. This procedure is referred to as filter
restart. Since filters had not reached the ter-
minal headloss or suffered turbidity break-
through, the filters were restarted at the end
of this shutdown period in a partially used
condition. The filtration rate changed from
0 to 3.2 gpm/ft2 or from 0 to 4.9 gpm/ft2 at
the higher  plant rates in approximately 15

  The second stress was caused by remov-
ing one of the four filters for backwashing.
The  filtration rate of the  three remaining
filters was increased from 3.2 to 4.3 gpm/ft2
or from 4.9 to 6.5 gpm/ft2 within 60 sec. A
slight  buffering  of  the  increase occurs
because the water level over the filters in-
  The third stress was caused by placing a
clean filter back into service. The filtering rate
increased from 0 to 3.2 gpm/ft2 and from 0
to 4.9 gpm/ft2. No predeposited material ex-
isted to be sloughed off during this filter rate
increase. During the setup or filter ripening
time of approximately 1 hr, the effluent tur-
bidity dropped  from its startup high (less
than 0.10 ntu) to the normal operating rate
of less than 0.04 ntu.
   Studies of the effect of filter stress or rate
change were made  using both  the  dual-
media filter and  one of the three mixed-
media filters installed at the Lakewood Filtra-
tion  Plant. Water quality was assessed both
by asbestos fiber analysis and turbidity mea-
surements in a  total  of 44 filter runs.
   Results generally showed that when the
coagulation  process  is  effective,   rate
changes of the magnitude experienced  at
Duluth did not cause substantial deteriora-
tion of filtered water quality. In two tests in-
volving restart of dirty filters, dual-media and
mixed-media filters were started when they
were sufficiently clogged to have 9 ft of head
loss.  In these  instances,  amphibole fiber
count in the filtered water  ranged from 2 to
4 million fibers/L during the first 10 min  of
the run. Fiber count dropped rapidly after
this time. Fiber counts in the range of 1  to
2 million fibers/L were observed twice with
dual-media filters when clean filters were
returned to production. Once during a back-
wash-induced rate increase, the dual-media
filter had 5 to 15 million fibers/L in the ef-
fluent. In most of the runs tested, asbestos
fiber concentrations did not exceed 1 million
fibers/L even during the first 5 min after the
stress or rate change had occurred.

Sludge Handling and Disposal
   Disposal of alum coagulation sludge like
that produced at Duluth is generally difficult.
Because  Duluth's  water filtration  plant
sludge contained asbestos fibers, the Min-
nesota  Pollution  Control  Agency imposed
stringent ultimate disposal requirements on
the Water Division. To minimize the amount
of sludge requiring ultimate disposal, a study
of sludge freezing was undertaken.
   When the Lakewood Filtration Plant was
built, three sludge lagoons were provided
(342-ft long, 90-ft wide, 8 ft deep, the maxi-
mum sludge depth).  The original plan for
using the lagoons was to permit sludge in
one lagoon to  freeze for  an entire winter
season while newly produced sludge was
added to one or both of the other lagoons.
This freezing concept is referred to as static
freezing. Experience showed that the static
freezing concept was not the most effective
method available.
  When natural freezing of sludge is prac-
ticed to dewater the sludge and condition it
for disposal, the static freezing process has
some important limitations. Because the at-
mosphere is below freezing but the soil at
the base of the lagoon is not, freezing pro-
ceeds from top to bottom. Thus, snow cover
cannot be allowed to remain on the partially
frozen sludge. To promote faster  freezing,
snow cover must be removed from the la-
goon  as soon as it can be safely done. At
Duluth, static  freezing  with  good snow
removal practice yielded ice formation at a
reasonable rate until ice was about 2 ft thick.
After  this point, the rate declined  substan-
  Because sludge freezes faster when it is
close to subfreezing air, sludge pumping was
studied. Unfrozen sludge was added to the
lagoon on top of previously frozen sludge.
Newly produced sludge or unfrozen sludge
pumped out from under the ice was spread
on  the  lagoon surface. Adding  unfrozen
sludge to the top of the ice helped maintain
higher freezing rates. A comparison of freez-
ing rates observed at Duluth appears in Table
  Natural sludge freezing has been very ef-
fective at Duluth.  After the sludge in the
lagoon freezes completely and thaws in the
spring, the clear supernatant water can be
decanted. The thin layer of remaining solids
has the appearance of instant coffee, and the
sludge volume is reduced so much that the
residue  is about an inch deep. Sludge vol-
ume reduction  has been so  complete that
after 5 years, the lagoons  have not  had to
be cleaned.


Filter  Plant Operation
  The  filtration  plant  has successfully
demonstrated that asbestiform fibers can be
removed from drinking water on a day-to-
day basis. Asbestos fiber counts have been
consistently below 0.1 million fibers/L since
the plant started operation.
  With proper chemical treatment, filtered
water turbidity can consistently be reduced
to below 0.1 ntu, generally to  0.04 or 0.05
  The various rate changes studied did not
have a continuous detrimental effect on the
effluent quality of the filter plant. Changes
occurred, but  these  were  generally
  As a result of these studies, filters are no
longer shut down with a high headloss. If a
filter has not reached the run termination
headloss of 8.0 ft prior to plant shutdown,
it is backwashed if the headloss is 7.5 ft or
greater. This procedure lessens the chance
of a major breakthrough of flocculated ma-
terials when the plant is  restarted.
  Alum dose varies with suspended solids
and ranges from 10 to 20 mg/L. Asbestiform
fibers are the last solids to be removed in the
process. The alum dose must  satisfy all  of
the coagulant needs of the other incoming
suspended solids before  fibers will  be
  Optimum pH for fiber removal depends on
raw water temperature and varies from 6.85
when the raw water is relatively warm to 7.35
when it is colder.
  The chlorine injection  point was moved
from the  rapid mix effluent to  rapid mix
chamber #1 to eliminate a change  in pH after
the coagulant is added.
  Sodium hydroxide application was moved
to rapid mix #1 for pH control.
  Because fluoride complexes  alum, the
point of addition was moved from the filter
influent flume to the filter effluent flume.

Water Quality Monitoring
  Continuous-flow  turbidimeters can  be
used to monitor filter behavior. No direct cor-
relation exists between the results from the
HIAC particle counter and the results from
the electron microscope analysis, but the
particle counter can be used as an indicator
of water quality. Ninety percent of the
samples  with particle counts  of less than
1000 particles/100 mL contained less than
Table 1.    Effect of Lagoon Operation of Sludge Freezing
                       Type of Operation
                                                           Freeze Degree Days*
                                                         to Freeze One Inch of Ice
   Static freezing, 66% snow cover
   Static freezing, 50% snow cover
   Static freezing, good snow removal
     tup to 25 in, of ice formed)
   Layer freezing on top of ice
   Layer freezing, good snow removal

                     28 to 34
 *The number of freeze degree days for one day is defined as that day's average temperature subtracted
from 32F. Total freeze degree days equals the sum of the freeze degree days for all days involved.

   20,000 amphibole fibers/L. Since the parti-
   cle counter gives a direct measurement of
   participates present in the water, it is more
   useful than turbidity as a measurement of
   water quality.

   Corrosion Control
     Because  Lake  Superior raw water is
   saturated with air and is a  very soft water,
   rusty water problems did occur and were ap-
   parently aggravated by introduction of alum
   coagulation. Two  corrosion control chem-
   icals, zinc orthophosphate and sodium tri-
   polyphosphate, have alleviated most of the
   rusty water problems in domestic galvanized
   iron piping.

   Bacteria Removal
     Coliform  bacteria  were successfully
   removed by the filter plant in all but one case.
   Note, however, that the filter plant was not
   challenged  by large numbers of microor-
   ganisms,  as might be the case at other

   Backwash Sludge
     The data gathered in this project indicate
   that with appropriate control, the  natural
   freezing of alum sludge is highly successful
   in reducing sludge volume. Static freezing
   in deep lagoons is not efficient for  several
   reasons. Freeze  degree days can be deter-
   mined from existing climatological data and
   can be used as a guideline to design lagoons
   for sludge volume reduction by natural freez-
   ing. During freezing weather, sludge should
   be delivered by spraying onto already formed
     At Duluth, sludge  is distributed to the
   lagoons by spraying from the hydrants with
   nozzles. The lagoon inlet structures are not
   used. This  method of oepration solved a
   sludge distribution problem by giving even
       distribution of sludge and a smooth surface
       for winter freeze operation and  snow
         After 5 years of operation, no sludge solids
       have had to be removed from the lagoons
       because of the very effective dewatering at-
tained in the sludge freezing process.
  The full report was submitted in fulfillment
of Grant No. S-804221 by the Duluth Depart-
ment of Water and Gas under the sponsor-
ship of the U.S. Environmental Protection
          Frank X. Schleppenbach is with the Duluth Department of Water and Gas, Duluth,
             MN 55802.
          Gary S. Logsdon is the EPA Project Officer (see below).
          The complete report, entitled "Water Filtration at Duluth," (Order No. PB 84-177
             807;  Cost: $14.50. 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:
                  Municipal Environmental Research Laboratory
                  U.S. Environmental Protection Agency
                  Cincinnati, OH 45268
                                               US GOVERNMENT PRINTING OFFICE 1984  759-015/7709
United States
Environmental Protection
Center for Environmental Research
Cincinnati OH 45268
                          BULK RATE
                     POSTAGE & FEES PAI
                       PERMIT No  G-35
Official Business
Penalty for Private Use $300
       LOU  W  ULLEY
       RtGION  V  EPA
       CHICAGO  IL  60&04