DESALINATION

ELSEVIER Desalination 130 (2000) 255-264 www.elsevier.com/locate/desal

Removal of iron and manganese from groundwater by oxidation

and microfiltration
• ° a*
Donald Ellis a, Christian Bouchard , Gaetan Lantagne b
aCivil Engineering Department, LarvalUniversity, Sainte-Foy (Quebec), Canada G1K 7P4
Tel. +1 (418) 656-2868; Fax +1 (418) 656-2928; email: cbouchar@,gci.ulaval.ca
bLTEE, Hydro-Quebec, 600, avenue de la Montagne, Shawinigan (Quebec), Canada G9N 7N5

Received 18 April 2000; accepted2 August 2000

Abstract
Iron and manganese can be removed from groundwater by a process which combines oxidation and microfiltration
(MF), especially when the concentrations of these metals are high and variable. The present experimental work focused
on the MF of iron and manganese oxide suspensions in order to study the effects of the operating variables (tangential
flow rate, pressure, metal feed concentrations) on permeate quality and permeation flux decline, i.e., membrane fouling.
Artificial and natural groundwaters in which iron and manganese were previously oxidized were used to perform
laboratory-scale MF experiments. The results show that the oxide particles, with sizes ranging from 1.5 to 5~m, were
efficiently microfiltered even at high concentrations. Within the experimental limits, the effect of operating pressure
appeared to be much more significant than the those of tangential flow rate and feed concentration. A relatively high and
constant permeation rate of 0.5 m/h was obtained at a pressure below 10 kPa, whereas rapid permeation flux declines
were observed at higher pressure. Both artificial and natural groundwaters always exhibited very similar behavior.

Keywords: Microfiltration; Oxidation; Iron; Manganese; Groundwater; Membrane fouling

1. Introduction domestic needs. Generally speaking, groundwater

is characterized by a low temperature (7-10°C),
Groundwater is an important source of
a low redox potential (absence of oxygen), a high
drinking water. In Canada, for example, 26% of
carbon dioxide concentration, a high mineral
the population, mostly living in small rural
content (high alkalinity and hardness) and a very
communities, depend on groundwater for their
low suspended solid content. Iron and manga-
nese, which are usually present in groundwater as
*Corresponding author. divalent ions (Fe 2+ and Mn2+), are considered as

0011-9164/00/$- See front matter © 2000 Elsevier Science B.V. All rights reserved
PII: S0011-9164(00)00090-4
256 D. Ellis et al. / Desalination 130 (2000,) 255-264

contaminants mainly because of their organo- manganese from groundwater by oxidation and
leptic properties. In Canada the maximum microfiltration (MF). This treatment is similar to
recommended levels of Fe and Mn in drinking the conventional one except that depth filtration
water are 0.3 mg/L and 0.05 mg/L, respectively is replaced by MF. The expected advantage of
[I], whereas the international standards are this treatment is to have a compact separation
0.3 mg/L for Fe and 0.1 mg/L for Mn [2]. unit which produces high quality water from a
Conventional treatment for iron and manga- wide range of raw water quality.
nese removal from groundwater consists of In the present study the MF of iron and
oxidation and depth filtration. Oxygen or stronger manganese oxide suspensions is examined. More
oxidants, such as chlorine and potassium specifically, the effects of the operating variables
permanganate (KMnO4), are generally used for (tangential flow rate, pressure, feed concen-
Fe 2÷ and Mn 2+ oxidation. The solid products of tration) on permeate quality and permeation flux
oxidation (FeOOH.H20 and MnO2) are then decline, i.e., membrane fouling, are studied using
filtered through a granular bed, commonly green artificial and natural groundwaters.
sand [3]. An aeration step may precede oxidation
for oxygenation and removal of carbon dioxide
(pH increase), sulfurs and other volatile sub- 2. Materials and methods
stances. The conventional process works fairly
well for a total concentration of Fe and Mn below 2.1. Natural and artificial groundwaters
5 mg/L and in the absence of dissolved organic A natural groundwater containing high
matter. For higher concentrations, excessive amounts of iron and manganese was chosen as
amounts of solids tend to shorten filtration cycles the model water for the study (see Table 1). This
and render the overall process unprofitable. water has been supplying the municipality of
Moreover, the filterability of manganese oxides Sainte-Marie (Quebec, Canada) for years. It
is poor and process control may become difficult contains 7-15 mg/L of Fe and 1.8-2 mg/L of Mn.
with raw water variations. No iron or manganese solids were found in this
Alternative processes have been proposed in raw groundwater because there was no signifi-
order to facilitate the operation and to allow the cant difference between the metal concentrations
removal of high amounts of iron and manganese of filtered (0.45 lam) and unfiltered raw waters. A
in the presence, or absence, of dissolved organic container, which prevents contact between air
matter. These alternatives are enhanced conven- and water (no air gap), was used to transport the
tional treatment [4], biological treatment [5] and groundwater samples from field to laboratory.
membrane treatment. Two approaches are The analytical method are referenced in
possible for the latter option: direct membrane Table 1. All the iron and manganese analyses
separation or combination of oxidation and were performed using an atomic absorption
membrane separation. Lapointe and Barrette [6] spectrophotometer (Perkin Elmer, model 3110).
reported results for reverse osmosis of a All the samples were quickly acidified (pH<2)
groundwater whereas Middleton and Hendershaw with nitric acid prior to atomic absorption
[7] presented full-scale results for nanofiltration spectrophotometry. Prelim inary tests showed that
of a colored groundwater. In both cases, a pH in our case high temperature digestion did not
adjustment was necessary to maintain iron in the bring any significant difference in metal concen-
dissolved state to avoid membrane fouling. tration measurement, and so it was decided to
Mourato and Smith [8] and later C6t6 et al. [9] avoid it. The organic carbon content was
presented a case study on the removal of iron and measured with a TOC-5000 apparatus from
 D. Ellis et al. /Desalination 130 (2000) 255-264 257

Table 1
Groundwater qualities

Parameter Natural groundwater Artificialgroundwater Analytical method"

pH 6.9-7.2 7.0 4500-H÷B

.Alkalinity, mg (CaCOJL) 190-230 210 2340B
Hardness, mg (CaCOJL) 190-230 180 From Caz+and Mg2+
Temperature, °C 7.0-7.2 7.0 2550B
Turbidity, NTU <0.2 <0.2 2130 B D
Dissolved 02, mg/L <0.2 <0.2 4500-0 G
Fe, mg/L 7-15 5 or 10 31lIB
Mn, mg/L 1.8-2.0 1.0 or 2.0 3111B
Ca, mg/L 69-75 72 3111B
Organic carbon, mg (C/L) <2 0 5310B
Na, mg/L 38-42 60 3111B
SO~z, mg/L 23-27 10-21 4110B
Sulfurs, mg (S/L) <0.1 0 2340B
Silicates, mg (SiO2/L) 23-25 0 4500-Si D

"AmericanPublic Health Association et al. [17].

Shimadzu. Prior to TOC measurement, the H20), sodium bicarbonate (NaHCO3), calcium
samples were acidified and bubbled with air, chloride (CaCI2) and sodium sulfate (Na~SO4).
which does not contain any CO2, in order to Also a gas mixture of carbon dioxide (1.75%
remove the inorganic carbon. But as the ground- CO2) and nitrogen (98.25% N2) was bubbled
water contains a large amount of inorganic through the artificial water prior to iron and
carbon (see Table 1), it was difficult to measure manganese additions. This gas composition
accurately the organic carbon content of the allowed reaching the conditions prevailing in the
water with this method. Other analysis reveals natural aquifer. As shown in Table 1, the
that the UV254 nm absorbance of the ground- artificial water properties were very similar to
water is close to 0.03 and that this groundwater is those of the natural water. All the following
totally colorless. This indicates that this experiments were conducted both with the
groundwater does not contains any significant artificial and natural groundwaters.
amount o f humic substances which can form
complexes with iron and manganese.
2.2. Oxidation procedure
In order to simulate the behavior of the natural
water and to run experiments with a lesser A two-step oxidation procedure was chosen.
amount of iron and manganese, an artificial The first step was the oxidation o f ferrous iron
groundwater was prepared (see Table 1) as per with oxygen. The water samples were oxyge-
the work ofKnocke et al. [4]. The preparation of nated by air bubbling and the pH was then
this artificial water is described in detail adjusted to 8.14- 0.1 with NaOH in order to fasten
elsewhere [10]. Deionized water and the the oxidation of iron for which 15 rain were
following salts were used: iron sulfate allowed to ensure complete iron oxidation. The
(FeSO4o7H20), manganese sulfate (MnSO4 ° second step was the oxidation of manganous ion
258 D. Ellis et al. / Desalination 130 (2000) 255-264

with K M n O 4 for which 15 min were also allowed the membrane fouling. The pressure inside the
to ensure complete manganese oxidation. This MF cell is controlled by a needle valve located
procedure was chosen because the oxidation of downstream to the cell. Inlet and outlet cell
the manganous ion with oxygen is far too slow pressures are measured in order to determine the
and because, from a practical point of view, prior average pressure difference across the membrane
treatment with oxygen reduces the consumption (APM). The tangential flow channel is a rec-
of KMnO4,which is a fairly expensive oxidant. tangular slit (5.3 cm × 0.2 cm) and the membrane
All the MF experiments were conducted with surface in contact with water is 78 cmL In order
water containing totally oxidized iron and to maintain the feed concentration as constant as
manganese. possible, the permeate and the concentrate are
The particles formed after oxidation were circulated back to the feed tank.
characterized with a particle size analyzer which A polymeric polyethersulfone MF membrane
measures laser beam diffraction by particles was used (SUPOR200 ® from Gelman Sciences).
ranging from 0.05 to 500~tm (Mastersizer This membrane was chosen for its resistance to
Microplus ®, Malvern Instrument Ltd.). The strong acidic conditions (nitric acid 10% w/w), to
particle size measurements were conducted under alkaline conditions (sodium hydroxide 10% w/w)
a strong agitation (2000 rpm) to prevent floccu- and oxidizing conditions (sodium hypochlorite
lation and to simulate the conditions prevailing in 5.25% w/w). Its given absolute porosity is 0.2 pm
a MF system where the feed water is recirculated. and its given pure water permeability is 15.7 m 3
m-2h-~bar-I at 20°C. Based on 153 measure-
2.3. Microfiltration ments, the average pure water membrane
permeability appeared to be 15.9 m 3m -~h-1bar-
The MF set-up, shown in Fig. 1, is a usual with a standard deviation of 1.9m3m-2h -I bar-L
membrane laboratory set-up. The feed iron and A new membrane coupon was used for each
manganese suspension is stored in a tank and experiment.
maintained at 8°C with a cooling bath to simulate The tangential flow was always laminar since
the actual temperature of the natural ground- the tangential Reynolds number (Re) ranged from
water. The suspension is circulated through a MF 100 to 400. The pressure difference across the
cell by a positive displacement pump with membrane ranged from extremely low pressure
controlled flow rate. The suspension is circulated (6kPa) to low pressure (100kPa), and the
tangentially to the membrane surface to minimize temperature was maintained close to 8°C. As
previously mentioned, two iron feed concen-
trations (5 and 10 mg/L) were tested in the case
Paw waterumpling
of the artificial groundwater, in combination with
two manganese concentrations (1 and 2 mg/L).
Each MF experiment started with a membrane
rinsing and pure water permeability measure-
ments performed with deionized water. During a
pure water permeability test, the pressure
difference across the membrane was increased
from 6 to 100 kPa and then back-decreased from
Pump PermlllllI Op 100 to 6 kPa. After this initial permeability test,
Pemlete urn~l~ ...J the iron and manganese suspension, which was
Fig. ]. Microfiltration set-up. obtained after oxidation of the natural or artificial
 D. Ellis et al. / Desalination 130 (2000) 255-264 259

groundwater (the procedure is described above), not occur in the absence of iron, as shown by
was microfiltered for 1 to 2.5 h depending on the Stumm and Morgan [12]. The adsorption of
membrane fouling rate. At the end of each run, manganous ions on iron oxides or the copreci-
the membrane was rinsed and cleaned with a pitation of iron and manganese may explain this
10% nitric acid solution for 9 min. Finally, the phenomenon. As indicated by the absence of any
membrane was rinsed and its pure water significant amount of dissolved manganese, the
permeability was measured again. The permeate KMnO4 oxidation of manganese was very
iron and manganese concentrations and the quickly completed after K M n O 4 was added to the
permeation flow rate were measured during the water, in agreement with Knocke et al. [4] who
MF of the suspension. studied manganese oxidation kinetics with
different oxidants. The pH after iron and manga-
nese oxidation completions was approximately
3. Results and discussion 8.5.
3.1. Oxidation The evolution of the particle size distribution
after oxidation is shown in Fig. 2 for both natural
As expected from the work of Stumm and Lee and artificial water samples. There are many
[11], and as indicated by the absence of any fewer manganese particles than iron particles in
significant amount of dissolved iron (-<0.45 ~tm), the oxidized natural water sample since its iron
the oxidation of iron with oxygen, at pH 8 and at concentration is much greater than its manganese
8°C, was achieved within a few minutes both concentration. The sizes of manganese oxides are
with natural and artificial groundwater. then much easier to appreciate in the curve
Surprisingly, 20% of the dissolved manganese corresponding to the artificial manganese
(_<0.45 ~tm) disappeared in the same conditions, solution oxidized with K M n O 4. For the purpose
both with natural and artificial groundwater. of comparison, the results of an artificial iron
Experiments with artificial solutions containing solution oxidation are also shown in Fig. 2. From
only manganese reveal that such phenomena do this figure, and from all other measurements, it

100 ' 1 +;
(

ii .
z
I

10 .2 10 "I 1 10 10 2 10 3
Particle dlameter (pro)

Fig. 2. Particle size distributions of oxidized natural and artificial groundwater samples.
260 D. Ellis et al. / DesaBnation 130 (2000) 255-264

appears that, from oxide particle sizes, there is no 2O

significant difference between the natural

groundwater and the artificial groundwater.
Actually, all the particle size analyses reveal that
both iron and manganese oxide particle sizes
range from 1.5 (see the zoom in Fig. 2) to a |i,2
maximum of 50 ~tm both for natural and artificial
8
groundwater. This means that MF with 0 30 60 90 120 150 180
Time (mln)
membranes with an absolute porosity lower than
1 ~tm should be an appropriate separation process Fig. 3. Microfiltrationblanktest (deionizedwater; APM=
for removal of iron and manganese oxide 95 kPa; Re = 260; pH = 6; T=21°C).
particles.

No significant flux decline was observed during

3.2. MF: permeate quality
this blank test (see Fig. 3) which means that the
In all the experiments conducted with both lost of permeability during the MF of iron and
artificial and natural groundwater, the iron manganese oxide particles is truly coming from
concentration of the MF permeate stayed below membrane fouling. Since the permeate quality
0.1 mg/L, which is the detection limit. The vast analysis reveals that neither iron nor manganese
majority of the manganese permeate concen- passes through the membrane, it is also
trations stayed below 0.03 mg/L, which is the concluded that membrane fouling is due to oxide
detection limit. A number of permeates were particle deposit at the membrane surface.
slightly above 0.03 mg/L, but these concentration The effect of tangential flow velocity on
values can probably be explained as sampling permeation flux decline is shown in Fig. 4 for a
errors since they were randomly dispersed natural groundwater sample. From these results
throughout the analyses. These results are it appears that an increase in tangential flow does
coherent with those of C6t6 et al. [9] who not lead to a significant increase in the
reported total retention of iron and manganese permeation flux even at higher pressure. It should
oxides in a full scale plant with a 0.1 ~tm MF be noted that the Reynolds number range is quite
membrane. This confirms that MF is a very low. A much higher tangential flow rate will
efficient separation process for iron and increase energy consumption but may also lead
manganese oxide particles, even for groundwater to a decrease in the membrane fouling rate since
with high level of these metals. a high tangential flow rate promotes the back-
transport of particles from membrane surface to
bulk flow. For example, Bourdon et al. [13]
3.3. MF: effects o f operating variables on
showed that, for the MF of karstic groundwater,
membrane fouling rate with natural groundwater
a two-fold change in tangential flow Reynolds
During MF of the iron and manganese oxide number (from 8,000 to 16,000) approximately
particles, a permeation flux decline was doubled the permeation flux.
observed, the extent of which was strongly The effect of pressure can also be seen in
dependent on the operating conditions. A MF Fig. 4. It is clearly shown that at higher
blank test with deionized water was run to transmembrane pressure, the permeation flux
establish which part of this permeate flux decline decreases much more rapidly than at lower
should be attributed to membrane compaction. pressure. Another experiment was performed at
 D. Ellis et al. / Desalination 130 (2000) 255-264 261

5 (a)
i i i I
4
z NetundJ I I I
• P s 10,7 kPIK Re.,d141

i~3 i
•
B
•
o
o
P-. 12.1 kPs; R e - 95
PmSlS.I kPs; R e . 6 4 6
p s 102,?ldPll:/:b~ 111
I j°
| • Pa 10.TkPa; RRm441
[ ] P.95.1kPll; RO=646
O p . 0.4 kPII; Re m2~S
[]
| m

1 |* m
mI

Ill
I i
1 ""
e I
•e
0 15 30 4S 60 7S g0 106 120 135 150
r o 8 A o o o

~ 30 e0 90 120 1S0 ~80 210 240

Time (rain) T~* (m~)

Fig. 4. Effect of tangential flow velocity and pressure on (b)

membrane fouling rate (natural water samples: 9 < Fe <

11 mg/L, 1.6 < Mn < 1.8 mg/L, pH 8.5-8.7, T = 8°C).

'"
very low pressure (APM = 8.4 kPa) and at medium
tangential flow rate (Re = 270). In this case the
membrane was cleaned after 2 h o f filtration, and
i
then the filtration continued for another 2 h. The
results of this experiment are shown in Fig. 5a
,,
o o
.,.,,
~ ~ go 120 ~r~ ~8o zlo
I
2,o
rime (m~)
and 5b. Under these conditions, the permeation
flux, or filtration rate, stays close to 0.5 m/h over Fig. 5. Effect of pressure on membrane fouling rate
(natural water samples: 9 < Fe < 11 mg/L, 1.6 < Mn <
two consecutive periods of 2 h. It shows that, at
1.8 mg/L, pH 8.5-8.7, T = 8°C.)
least for 2 h, MF can be performed at very low
pressure without any significant permeation flux
decline. Also, Fig. 5a shows that a slight increase induced transport of particles from the membrane
in the pressure difference across the membrane surface back to the bulk flow is stronger than the
(8.4-10.7 kPa) makes a significant difference advective flux of particles toward the membrane
since at 10.7 kPa the permeation flux decreases [16]. According to Altmann and Ripperger [18],
with time, which is not the case at 8.4 kPa. This this may indicate that, for a permeation velocity
result is in agreement with those of Field et al. of 0.5 m/h, the iron and manganese particle sizes
[14] and Howell [15] who worked on the MF are higher than the critical size below which
separation of yeast and who proposed the concept particles should deposit onto the membrane
o f a critical permeation flux under which the surface.
permeation flux should stay almost constant. It is also very interesting to note that the MF
Field et al. [14] used a membrane with pure filtration rate, which is obtained at very low
water permeability of 0.78 m 3m -2 h-] bar-~ which pressure, ranges between those of conventional
yielded a constant permeation flux of 0.025 m/h rapid and slow sand filtration (see Fig. 5a) where
at a pressure of 9 kPa, a Reynolds number of 50 similar or higher pressures are used and for
and for a 5% dry weight yeast suspension. The which the filter medium thickness is much higher
absence o f permeation flux decline at very low than for MF. This indicates that a MF system
pressure indicates a lesser deposit of particles on operated at very low pressure could be much
the membrane surface. This may indicate that, more compact than a conventional slow sand
below a specific permeation flux, the shear- filter.
262 D. Ellis et al. / Desalination 130 (2000) 255-264

1.0 iO | eNatural: P=8.4kPa; Re = 266; Fe= 10.0 mg/L; Mn = 1.68 mg/I.

0'51I I I T
O Artlficiel; P = 8.7 kPa; Re = 284;
T Y
Fe = 8.3 rag/1.; Mn = 1.73 mg/L
g° 1.5,
I

!
I

! a
I ,, I

i
I

i
I

~. 1.0 ~ ' " ' • N,quml; P = 10.710¢ Re=441; F e - 11,0mprl.; Mn= 1.80 mg/l.
i :-. II ii Ii i •
& ArUflckll; P-11.1 kPa; Re=443; Fe= 8.4 mg/L; Mn= 1.76 mg/t.
0
4.0 D Natural; P= 95.1 kPa; Re= 646; Fe = 8.7 mg/t.; Mn= 1.71 mg/L
~lm_ /1 Artificial; P = 95.1 kPe; Re = 644; Ice= 8.3 mg/L; Mn,, 1.71 mg/L

0
0 30 60 90 120 150 180 210 240
Time (rain)

Fig. 6. Effect o f tangential flow velocity and pressure on membrane fouling rate: comparison o f artificial and natural
groundwater behaviors (pH $.5-8.7, T = 8°C).

[ e P = 8 . 7 lope; Re=28~ Fe= 8.3 mg/l.; Mn.

l o,0Q 2 - o 1.73
2 mg/L
o ¢~

0.5 I T T - T "- 1" "- I

0 P = 8.5 kPa; Re ,, 288; Fe ,, 4.8 rag/L; Mn = 0.90 mg/L /
g° 1.5
I I I I I

7. " P=11~1kPr' R~'443; ~'8"4m~L; Mn= I'78mg/L

1.0 i m ~ " I J| • • • J

i 0.5
0
o. 4.0
• P = 11.2 kPa; Re = 443; Fe = 4.6 rnl~t.; Mn ,, 0.91 mg/L

2.0

0
0 30 60 90 120 150 180 210 240
Time (rain)

Fig. 7. Effect o f iron and manganese concentrations on membrane fouling rate for artificial groundwater (pH 8.5-8.7,
T = 8°C).

The effect of pressure on fouling rate is shown pressure, indicating that it may be more
in Fig. 5b in terms of membrane permeability profitable to operate at low pressure.
variation during MF. The permeability corre-
sponds to the permeation flux divided by the
3.3. MF: comparison of natural and artificial
pressure difference across the membrane. This
behaviors
shows how much permeate can be produced per
unit of membrane surface and pressure, and so it Similar MF experiments were repeated with
is an indication of energy efficiency of the both natural and artificial groundwaters. A
process. Results presented in Fig. 5b show that comparison of such experiments is shown in
the energy efficiency is higher at very low Fig. 6 where three different operating condition
 D. Ellis et al. / Desalination 130 (2000) 255-264 263

sets were used. This figure shows clearly that pressures MF permeation flux is comparable to
natural and artificial groundwaters have similar slow sand filtration rate at a similar pressure
behaviors. Therefore, it was concluded that difference but with a much thinner filtering
artificial groundwater simulates natural ground- medium. The effect of tangential flow and iron
water behavior satisfactorily. Further experi- and manganese concentrations is much weaker
ments were conducted with artificial groundwater than the pressure effect within the experimental
samples which contained less iron and limits of the present study. Behaviors of the
manganese than natural water in order to study artificial and natural groundwaters are always
the effect of iron and manganese concentrations very similar, indicating that artificial waters can
on permeation flux. Results shown in Fig. 7 be used to conduct more basic studies.
indicate that a two-fold reduction in concentra-
tion leads to a slightly slower decrease in
permeation flux. When the particle concentration Acknowledgment
is reduced, fewer particles are brought close to This study was funded by the National
the membrane surface and so fewer particles may Sciences and Engineering Research Council of
be deposited on the membrane surface, resulting Canada (NSERC) and by Hydro-Quebec's LTEE
in less fouling. In our case this phenomenon does (Laboratoire des Technologies l~lectrochimiques
not greatly affect the membrane fouling rate. et des ]~lectrotechnologies).
Further experiments should be conducted on a
broader concentration range with artificial
groundwaters in order to better understand the
effect of concentration on fouling kinetics. References
[1] Government of Canada, Guidelines for Canadian
Drinking Water Quality, 6th ed., Health and Welfare,
4. Conclusions Canada, 1996.
For both natural and artificial groundwaters, [2] American Water Works Association, Water Quality
ferrous ion oxidation with oxygen and and Treatment, 5th ed., McGraw-Hill, New York,
1999.
manganous ion oxidation with potassium
[3] C. Lessard, D. Ellis, J. S&odes and C. Bouchard,
permanganate result in the formation of particles Vecteur Environnement, 32(3) (1999) 30.
ranging in size from 1.5 to 50 p.m. These oxide [4] W.R.Knocke, J.E. Van Benschoten, M. Kearney, A.
particles were filtered continuously at the Soborskiand D.A. Reckhow, Alternativeoxidants for
laboratory scale on a MF polymeric membrane the removal of soluble iron and manganese, Report
with an absolute porosity of 0.2 lam. The MF for the AWWA Research Foundation, USA, 1990.
permeate analysis do not reveal any significant [5] P. Mouchet, J. AWWA, 84(4) (1992) 158.
presence of metal, indicating that the particles are [6] D. Lapointe and J. Barrette, Sciences et Techniques
very efficiently filtered even at relatively high de I'eau, 17(2) (1984) 189.
concentrations of iron and manganese (10 and [7] M.M. Middleton and W.K. Hendershaw, Proc.,
2mg/L, respectively). The permeation flux Membrane Technology Conference, AWWA, Reno,
USA, 1995, pp. 699-710.
decline is strongly affected by transmembrane
[8] D. Mourato and C. Smith, Proc., 6th Workshop on
pressure. A significant permeation flux decline Drinking Water, AQTE, Montr6al, 1994, pp.
occurs at higher operating pressures whereas a 705-716.
constant permeation flux of 0.5 m/h is obtained at [9] P. C6t~, D. Mourato, C. Gtingerich, J. Russell and
lower pressures, i.e., below 10 kPa. Under low E. Houghton, Immersed membrane filtration for the
264 D. Ellis et al. / Desalination 130 (2000) 255-264

production of drinking water: case studies, ISWA J. Membr. Sci., 100 (1995) 259.
Conference, Membranes in Drinking and Industrial [15] J.A. Howell, J. Membr. Sci., 107 (1995) 165.
Water Production, Amsterdam, 1998. [16] M. Wiesner and P. Aptel, in: Water Treatment/
[10] D. Ellis, Master's Thesis, Civil Engineering, Laval Membrane Processes, J. Malleviale, P.E. Odenaal and
University, Quebec, Canada, 1998. M. Wiesner, eds., McGraw-Hill, New York, 1996.
[11] W. Stumm and G.F. Lee, Industrial and Engineering [17] American Public Health Association, American
Chemistry, 53 (1961) 143. Water Works Association, Water Pollution Control
[12] W. Stumm and J.J. Morgan,Aquatic Chemistry, 3rd Federation, Standard Methods for the Examination of
ed., Wiley-Interscience, New York, 1996. Water and Wastewater, 17th ed., American Public
[13] F. Bourdon, M.-M. Bourbigot and M. Faivre, L'eau, Health Association, Washington, DC, 1989.
I'industrie, les nuisances, 121 (1988) 35. [18] J. Altmann and S. Ripperger, J. Membr. Sci., 124
[14] R.W. Field, D. Wu, J.A. Howell and B.B. Gupta, (1997) 119.