BIOFOULING IN A SEAWATER REVERSE OSMOSIS PLANT ON THE RED SEA COAST, SAUDI ARABIA 1 Mohamed O. Saeed, A.T. Jamaluddin and I. A. Tisan Saline Water Conversion Corporation P.O.Box 8328, Al-Jubail -31951, Saudi Arabia Tel: + 966-3-343 0012, Fax: + 966-3-343 1615 Email: [email protected]D. A. Lawrence and M. M. Al-Amri Satellite Plants Department, SWCC, PO Box 6538, Jeddah 21452 Kamran Chida DuPont Permasep® Co., PO Box 1049, Jeddah 21431 ABSTRACT This paper reports on a study that investigated the environmental and pretreatment impact on biofouling in a seawater reverse osmosis (SWRO) plant. The effect of a pretreatment chemical (chlorine), and certain alterations of chemical dosing on membrane biofouling was investigated. The paper also reports on the biofouling potential of the source water, and the effect of chlorination on this biofouling potential. Experiments were carried out to study biofouling in a SWRO Plant, on the Red Sea coast, under a set of four pretreatment modes. These included the normal operation mode, where coagulant is dosed immediately before the media filter and where sodium metabisulfite (SBS) is dosed after the media filter. Secondly, the operation with the coagulant dosing point shifted back to the pressure side of the seawater intake pump. Thirdly, with the SBS dosing point shifted to after the micron cartridge filter, and fourthly while the plant was operating without chlorination/dechlorination. Bacterial generation time and biofilm attachment slides were used to evaluate biofouling. Generation (doubling) times were lowest (higher multiplication capacity) nearest the intake, and they increased gradually along the pretreatment line, becoming the highest, closest to the membranes and in the brine reject. When the SBS was shifted, chlorine removal became closer to the membranes. Following this, generation time in the water samples, taken after the dual media filter (ADMF), after the micron cartridge filter (AMCF) and immediately before the membranes, decreased significantly, reflecting more biofouling potential in the membranes. This correlates well with operational data 1 Published in Desalination, 128 (2000) 177-190.
21
Embed
Biofouling in a Seawater Reverse Osmosis Plant on the Red Se
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
BIOFOULING IN A SEAWATER REVERSE OSMOSIS PLANT ON THE RED SEA COAST, SAUDI ARABIA1
Mohamed O. Saeed, A.T. Jamaluddin and I. A. Tisan
Saline Water Conversion Corporation P.O.Box 8328, Al-Jubail -31951, Saudi Arabia Tel: + 966-3-343 0012, Fax: + 966-3-343 1615
where the SBS shift has resulted in doubling the frequency of membrane cleaning. The
generation times were higher when no chlorine was used, indicating less membrane
biofouling potential. Water samples from the plant's intake in the sea had 24 h-
generation time values less, but close to those of chlorinated seawater. This indicated
high nutrient load and questionable water quality of the intake. The bacterial
attachment to the biofilm slides showed the general trend exhibited by the generation
time of planktonic bacteria. The biofilm formation in the brine was generally to a
lesser extent than the preceding sampling stations ADMF and AMCF, indicating
removal of nutrients along the pretreatment line. However, when the SBS dosing point
was shifted closer to the membranes, the number of attached bacteria in the brine
increased significantly. Bacteria attached to the biofilm slides randomly and in
microcolonies with vesicles in between. Although chlorination enhances biofouling, the
current experiments indicate that the questionable quality of source water is one cause
of the operational problems in this plant.
Key words : Biofouling; Chlorination; Bacteria; Generation time; Water quality. 1. INTRODUCTION A seawater reverse osmosis (SWRO) plant at Al-Birk, in the southern region of the Red
Sea coast of Saudi Arabia, is facing major operational problems which were thought to
stem from biofouling. These problems include: increased ∆P build up across
membranes at alarming rates that necessitate frequent membrane cleaning (at times
cleaning is needed at 10 days intervals or less), increased membrane flux decline rates,
elevated permeate conductivity, and clogging of media and micron cartridge filters. The
main cause for the decline in SWRO plant performance in the Middle East has been
identified as topical biofilm accumulation on membranes, aided particularly by the large
surface areas of the membranes [1].
The bacteriological activity, and hence biofouling problems in SWRO plants, vary with
each site. There are site dependent factors which govern the degree of biological
activity, such as: temperature, nutrient load, pollution, and the depth and location of the
intake. The temperature of the southern region of the Red Sea around the intake of the
SWRO plant in Al-Birk is high. For example, it varied between 27.1 to 34.5oC during
the period of March -November, 1997, the period in which this study was undertaken.
As indicated by the high bacterial multiplication rate, the nutrient load of the intake
seemed to be high, while water currents are minimal and the water column over the
intake seems stagnant [current study]. All these factors are conducive to higher bacterial
activity. Biofouling of the hollow fine fibers (HFF) membranes which are in use at Al-
Birk plant has been reported [2,3&4]. Enhanced aftergrowth rates of bacteria were
observed by Winters [5].
In conclusion, the biofouling of SWRO plants with surface intake is too well known to
need further documentation. To control the biofouling in these plants, chemical
disinfection is used. Conventionally, chlorine has been used as a chemical disinfectant.
It was reported [6] that chlorine degrades humic acids and similar high molecular
weight compounds present in coastal seawater to smaller molecules that can be
assimilated by bacteria. The chlorine suppressed bacterial activity, but when the
sodium metabisulfite (SBS) is added to remove chlorine, the surviving bacteria quickly
take advantage of the nutrients furnished by the degradation of larger molecules and
enter into a cycle of tremendous growth (termed by some as "aftergrowth"). The
significant increase in the biomass of bacteria after dechlorination continued with slime
development on the surfaces of the piping and RO membranes [7].
To overcome the aftergrowth problem encountered upon dechlorination, alternate
disinfectants have been proposed such as chloramine and copper sulfate [8]. The Umm
Lujj SWRO plant, northern Red Sea, Saudi Arabia, has been operating since 1986 with
copper sulfate as disinfectant without unexpected operational troubles or membrane
fouling.
Another approach is to stop chlorination/dechlorination altogether. The no chlorine RO
operation has been tried in a commercial facility, operating in warm water with high
total dissolved solids. The performance of the plant has been improved and membrane
cleaning, which had reached an alarming rate, was cut down significantly [9]. An
SWRO plant in the Caribbean was operated successfully without ever using
chlorination/dechlorination [10]. Several other plants receiving surface seawater feed of
differing characteristics with regard to salinity and temperature have operated
successfully without chlorine. Intermittent chlorination is used only to protect the
intake structures from marine fouling [9].
The reported success of the no chlorine RO operation has led to speculation that a
similar operation may be an avenue to overcome the biofouling problem at Al-Birk
SWRO Plant. This approach looks attractive, particularly when considering the
involvement of M/s DuPont in the reported successful operation without chlorine [9].
DuPont is the manufacturer and supplier of the Permasep® B-10 single bundle
permeators used at Al-Birk.
A joint project between the Saline Water Conversion Corporation (SWCC), represented
by its Research and Development Center, and DuPont, was drafted to study the
operational problems of the SWCC Al-Birk SWRO Plant. This part of the experiments
deals with biofouling which is a major task in this research.
2. OBJECTIVES (a). Investigate status and the potential of biofouling in seawater, and pretreated feed
water at different operational modes with chlorination/dechlorination, and without
chlorine.
(b). Suggest possible measures to alleviate the biofouling problem.
3. EXPERIMENTAL
Experiments were designed to study the biofouling in the SWCC Al-Birk SWRO Plant.
3.1 Project Phases
The study was divided into 4 operational phases with respect to chemical pretreatment
as follows: 3.1.1 Plant Operating as is In this mode of regular operation, the coagulant "magnifloc® C-573" is injected just
ahead of the coarse sand filters, chlorine (Calcium hypochlorite) is dosed in the intake
pit, while sodium metabisulfite (SBS) is dosed to remove chlorine after dual media
filters ahead of micron cartridge filters.
3.1.2 Coagulant Dosing Point Shift The dosing of coagulant was shifted from before coarse sand filters up stream to the
pressure side of the seawater pump (a distance of about 85 m from the regular dosing
point mentioned above).
3.1.3 SBS Dosing Point Shifts The SBS dosing point was shifted from before to after the micron cartridge filter (a
distance of about 13 m ahead of the high pressure pump). A second point was selected
by moving the SBS dosing to a location just ahead (1 m) of the high pressure pump.
This shift was meant to ascertain the effect of maintaining chlorinated flow over the
low pressure section between the micron cartridge filter and the high pressure pump.
This piping section was suspected to be an additional biofouling source.
3.1.4 Plant Operating Without Chlorination/Dechlorination The coagulant dosing point remained as in 3.1.2 above.
For all phases; analyses were carried out using water samples for enumeration of
bacteria, and biofilm samplers with studs (coupons) were fixed along the pretreatment
line to study bacterial attachment.
Water samples were taken from 6 sampling stations:
(1) The pressure side of the seawater intake pump, (2) Before the media filter, (3) After
the media filter, (4) After the micron cartridge filter, (5) Ahead of the membranes (not
sampled during coagulant shift), (6) The brine reject of B-10 permeators.
The biofilm samplers were fixed in 4 locations as follows:
(1) The pressure side of the seawater intake pump, (2) After the media filter, (3) After
the micron cartridge filter, (4) Brine reject.
3.2 Bacteria in Water Samples In the chlorinated sampling stations, water samples were taken in sterile plastic
sampling bags containing sodium thiosulfate as dechlorinator, while plain sterile bags
without the dechlorinator were used to collect water samples from dechlorinated
sampling stations. Bacteria were counted immediately after sampling and this count
was designated 0-h count. Further counting was carried out after 24h (24-h count) and
72 h (72-h count) following the incubation of samples at a temperature of 30 oC in a
thermostatically controlled incubator. The samples were first mixed well on a vortex
mixer, and a pour plate count in marine agar was employed to reveal the colony forming
units (CFU) [11,12]. Briefly; the samples were first serially diluted in ten-fold steps in
filtered (5 µm) and heat sterilized seawater. Three dilutions from each sample were
plated to enable a suitable number of bacteria to be counted. For each dilution, three
replicates of 1 ml were seeded in separate sterile petri dishes. Thirteen milliliters of
Molten marine agar at 46 oC were added to each dish. In each dish, the agar was mixed
well with the 1 ml sample. The dishes were then allowed to cool for 30 min at room
temperature and then incubated (inverted) in the constant temperature incubator. The
colony forming units were counted after 96-h of incubation.
Water samples from the vicinity of the plant intake in the sea at 5 m depth, which is the
depth of the intake chamber's inlet, were also analyzed for comparison with chlorinated
intake seawater.
Zero-hour counts were used as a base to calculate the generation time for 24h and 72 h
of incubation as per the following formula [13]:
Generation time (h) = ∆ t K / (ln Nt - ln Nt0), where:
∆t = 24 for 24 h - generation time and 72 for 72 h - generation time, K = 0.693
Nt = count at 24 or 72 h and Nt0 count at 0h.
The generation time reflects the speed of bacterial multiplication and was used as an
index of nutrient load of water samples.
3.2 Biofilm Bacteria Biofilm samplers were installed in four locations as mentioned above. Each sampling
unit contained six holders, each with a stud (coupon) of glass slide measuring 2.5 x 2.2
cm. In each sampling station, water was diverted to flow through the sampler at a rate
of 10 l/min. After 20 days the slides were retrieved and the biofilm attached was
scraped off for enumeration of attached bacteria. The slide was aseptically removed
from the holder, placed into a sterile petri dish, and then flooded with a measured
quantity of sterile seawater. The biofilm was scraped first by a sterile dissecting knife
and then by a sterile cotton swab. The slide was removed and the biofilm suspension
was transferred to a test tube and mixed vigorously on a vortex mixer. The suspension
was then serially diluted and grown in marine agar, employing the pour plate technique
as described. Following 96 h of incubation at 30 oC the number of bacteria (CFU) was
obtained and expressed as CFU/cm2. A scanning electron microscope was used to
reveal bacteria attached to biofilm slides. Slides with intact biofilms were air dried and
placed on a brass plug using double sided carbon (conductive) tape. They were then
coated with 30 nm of gold, using JEOL-1100E ion sputtering device. The specimens
(slides) were placed in the JEOL model JSM-5300LV scanning electron microscope
and were observed at 2000, 3500, and 5000 magnification using 20 KV accelerating
voltage at 10 mm working distance.
3.3 Analysis of Data Data was subjected to rigorous statistical comparisons using an analysis of variance to
reveal differences between means and LSD test to distinguish differing means.
4. RESULTS
4.1 Bacteriological Analysis of Water Samples Bacterial generation times (doubling times) were computed for all phases of the project
and compared statistically (Table 1). The general trend is that the time needed for
bacterial cells to divide (generation time), increases along the pretreatment line, is
lowest in the chlorinated seawater feed and is highest in the brine reject. The table also
shows that generation times from the no chlorine phase are consistently higher than
from the other three phases when the plant was running with chlorination/
dechlorination. Table 2 shows the bacterial density and the generation times of water
samples from Al-Birk SWRO Plant intake site at a depth of 5 m. The water currents
around the intake were minimal and the water column seemed stagnant. A comparison
of the generation time of the intake samples and those of chlorinated seawater shows
longer generation times for the intake samples than the chlorinated water samples
(Table 2). Table 3 gives a comparison of bacterial counts and generation times during
the two SBS dosing shifts.
In chlorinated seawater the 0-h count was low when compared to sampling stations with
no chlorine. However, upon dechlorination there was a rapid bacterial growth which
resulted in lower generation times as compared to samples with no chlorine (whether
chlorine was removed prior to these sampling stations or no chlorine was used).
4.2 Biofilm Studies Table 4 presents bacterial attachment to biofilm sampling slides at four stations along
the pretreatment line during the four phases of the project. The table shows there was a
2-log reduction of bacteria attached to biofilm slides after the micron cartridge filter
(MCF) when this location became chlorinated following the SBS dosing shift. The
number of attached bacteria was consistently and significantly lower in the brine than
either after the dual media filter (DMF) or the MCF, when the SBS dosing point was
before the MCF. However, when the SBS dosing point was shifted to after MCF,
significantly more bacteria were attached in the brine than after DMF or after MCF. In
this phase (SBS shift), the number of bacteria attached to the slides in the brine reject
was higher than those attached in the same sampling station during the first two phases
of the project, i.e., when the SBS dosing point was before the MCF. During the no
chlorine phase, the bacterial attachment was steady at 5 logs in all sampling stations
except in the after DMF where it registered an increase of 1 log over the other sampling
stations.
Figure 1 shows scanning electron micrographs of the bacteria attached to microscope
slides at different stations during various phases of the project. The biofilm is not a
homogeneous continuous lawn, rather, it is an aggregation of cells in twisting patches
creating a network of vesicles or channels between them. Cells are rod shaped, more or
less uniform in size, and many take a curved (vibrioid) shape. Many cells show
biopolar bodies and possess capsules.
5. DISCUSSION Biofouling is not a problem unique to a particular SWRO plant but is a major concern
in many such plants operating on surface seawater feed. The Al-Birk SWRO Plant is
more vulnerable to biofouling because the geographic location is in a hot climate which
is more conducive to biofouling. One way to study this biological activity is by
quantitative estimation of bacterial biomass. Whilst the estimation of numbers of free
bacteria in water (planktonic) may initially appear a relatively simple task, it is in
practice fraught with problems. There is a wide natural variation in the number of
bacteria in seawater because many factors make their numbers change constantly, such
as light, temperature, tide, currents, turbidity and nutrients. An added variation source
along the pretreatment line is the untimed detachment of bacteria from biofilm in the
system into water lines and possibly in the water samples. In this study the variation in
the same location was very noticeable, giving a standard deviation of close to the
average count. Other studies have shown a difference in the count of more than one log
in seawater. This was encountered even in consecutive samples [12].
The constant variation in number makes comparison by bacterial density in water
column unreliable. Since generation time deals with an initial and a later count of the
same sample it becomes a very suitable tool for comparison. However, a general trend
in bacterial count is significant. The 72-h counts were in the vast majority of samples
more than the 24-h counts in all phases except the no chlorine phase where it was less.
This is clear evidence that with no chlorine in water, the nutrients were exhausted
following the first 24-h of growth. This has lead to a significant decrease in the
biofouling potential during the no chlorine phase. Plant performance data also support
the above-mentioned evidence of decreased biofouling potential when no chlorine was
used. The average rise in ∆P across the membranes was 0.1 bars/day during the no
chlorine phase compared to an average rise of 0.3 bars/day during the chlorinated
phases. The membrane flux decline rate was on an average <0.40%/day compared to
0.86% in earlier phases when the plant was operated with chlorine.
When generation times of the four phases of the project are compared (Table 1), a
general trend could be noticed. There is an increase in generation times from the intake
water towards the membranes. This means bacteria attached to pipings and filters or
floating in the water column are removing nutrients from the flowing water prolonging
the multiplication time of bacteria. Therefore, control of nutrients is one way of
* Generation time, (g) = ∆t k / (ln Nt- ln Nt0 ) ; where ∆ t =24 for 24h-generation time and 72 for 72h-generation time, K = 0.693, Nt count at 24 or 72 h and Nt0 count at
0h. 1 Baseline ( as is ) plant operating with the coagulant dosing before the coarse sand filter and the SBS dosing point after the dual media filter. 2Coagulant dosing point shifted back to the pressure side of the intake pump . 3Sodium metabisulfite ( SBS ) dosing point shifted to after the micron cartridge filter 4± Standard deviation. a,b,c,d,e,fFor the same sampling station within different phases (horizontal rows) and for the same generation time (24 or 72 h ) , means sharing any common single letter superscript are not different; others are different . For the same project phase and within different sampling stations (vertical columns), means (of either 24 or 72h - generation times) sharing any 2 common letter superscripts are not different ; others are different (P≤ 0.05) ; Analysis of Variance and LSD test.
Table 2. Comparison of bacterial count (colony-forming units/ml) and generation timesb(hours) of the intake area and chlorinated intake feed seawater (n=5)
Sampling station 0-h count 24-h
count 24 h - generation
time 72-h
count 72 h-generation
time Intake raw water 6.46x102 1.93x105 2.88 ± 0.31b 1.47x105 9.58 ± 0.25 Chlorinated intake water 2.69x102 2.36x105 2.47 ± 0.19c 4.41x105 6.78 ± 0.95c
a Sample taken from 7 m depth close to the intake chamber in the open sea. b Each generation time is statistically different from its corresponding one , (P ≤ 0.05); LSD test. c± Standard deviation. Table 3. Comparison of bacterial count (colony-forming units/ml) and generation times (hours) when the
sodium metabisulfite (SBS) dosing point was shifted to two locations after the micron cartridge filter (n=15)
Sampling Station 0-h Count 24-h Count 24-h Generation
1For each sampling station, means with same letter superscript are not different while those with different ones are different. (P≤ 0.05) ; Analysis of variance and LSD test. 2 Shifted to a point 13 m ahead of the high pressure pump. 3 Shifted to a point just ahead (1 m) of the high pressure pump. ± Standard deviation Table 4. Bacterial attachment to biofilm sampling slidesa at four stations along the pretreatment line
Sampling Stationb Bacteria ( CFU) / cm2
Project phase1 Intake Seawater After dual media filter
No Chlorine (1.23± 0.34)x105 (2.20± 0.30)x106 (1.70± 0.33)x105 (3.45± 0.99)x105 a Glass slides (coupons) measuring 2.5 x 2.2 cm fixed in biofilm samplers at each of four sampling station as
indicated in table. b Within each project phase (horizontal rows), number of bacteria attached at each sampling station is different
from others. In the same sampling station during different phases of the project (vertical columns), the number of attached bacteria is different from others (P≤ 0.05); Analysis of variance and LSD test.
± Standard deviation
Fig. 1a. An Electron micrograph of a biofilm slide place in SBS-dechlorinated feed
after the micron cartridge filter showing cells attached in irregular patches. Some cells with bipolar bodies appearing as white spots at the poles of the cells. Sheathes around the cells could also be noticed. Cells are rod-shaped and many are curved (bar=5 µµµµm).
Fig.1b. A micrograph from a slide placed in SBS dechlorinated feed after the micron cartridge
filter showing more cells attached than in above case, Fig.1a. Also, cells take more elongated shape than in the above figure (bar = 5 µm).
Fig.1c. A micrograph from a slide in the chlorinated intake water during the first phase of the
project. Cells are few and smaller in size when compared to cells from dechlorinated stations in Fig. 1a & b above. Note the formation of a base layer on which cells are attached (bar = 5 µµµµm).
8. REFERENCES 1. Ahmed, S, Al-Ansari, M. and Kannari, T., (1989), Desalination, 74 69. 2. Osta, T. and Bakheet, L., (1987), Desalination, 63, 71-80. 3. Hassan, A., Al-Jarrah, S., Al-Luhaibi, T., Al-Hamdan, A., Bakheet, L. and Al-Amri, M.,
(1989), Desalination,74, 37-51. 4. Hassan, A., Abanmy, A., Al-Thobaiety, M., Maani, T., Al-Luhaibi, T., Al-Masudi, I., Al-
Gherier, A., Bakheet, L., Al-Amri, M. and Atiya, K., (1991), Performance Evaluation of SWCC SWRO plants-Part III, Research Activities and Studies (SWCC), 2, 94-108, [5]
5. Winters, H., (1994), Biofouling Status of the Saline Water Conversion Corporation (SWCC) Reverse Osmosis (RO) Plants in the Kingdom of Saudi Arabia, A Consultancy Report, 23 pp.
6. Applegate, L., Erkembrecher Jr. C., and Winters, H., (1989), Desalination, 74, 59-72. 7. Winters, H. and Isquith, I., (1995), A Critical Evaluation of Pretreatment to Control
Fouling in Open Seawater Reverse Osmosis - Has it been a success? Proceedings of the IDA World Congress on Desalination and Water Sciences, Abu Dhabi, UAE, Vol. I, 321, Nov. 18-24.
8. Winters, H., (1995), Biofouling: Its history and how it affects today's Desalination
industry, Proceedings of the IDA World Congress on Desalination and Water Sciences, Abu Dhabi, UAE, Vol. I, 255-264, Nov. 18-24.
9. Moch Jr., I., Ben Hamida, A., Pohland, H., (1995), New Technology to Control
Biofouling, Proceedings of the IDA World Congress on Desalination, Abu Dhabi, UAE, Vol. IV, 59-72, Nov. 18-24.
10. Barendsen, W. and Moch, I., (1995), Design and Operation of 1.2 mgd (4542 cmd) Open
Sea Intake Plant in the Caribbean, Proceedings of the IDA World Congress on Desalination, Abu Dhabi, UAE, Vol. III, 25-40, Nov. 18-24.
11. American Public Health Association, (1995), American Water Works Association and
Water Environment Federation (Eds), Standard Methods for the Examination of Water and Wastewater, 19 ed. APHA, Washington, DC.
12. Al-Tisan, I., Chandy, J., Abanmy, A. and Hassan, A., (1995), Optimization of Seawater
Reverse Osmosis Pretreatment: Part III - A Microbiological Approach, Proceedings of the IDA Water Congress on Desalination, Abu Dhabi, UAE, Vol. IV, 3-17, Nov. 18-24.
13. Brooks, G. F., Butel, J. S. and Ornston, L. N., (1991), (Eds) Jawetz, Melnick & Adelberg's
Medical Microbiology, 19th edition, Appleton & Lange, East Norwalk, Connecticut. 14. Bott, T.R., Bohnet, In: M., (1990), (Ed.) Fouling of Heat Exchanger Surfaces, VDI
publication, Düsseldorf, pp. 5.1-5.20. 15. Winters, H , (1987), Desalination, 110, 12-20.
16. Khedr, M., (1995), Analysis of Complex Fouling Problems Encountered in Gulf Water RO, Proceedings of the IDA World Congress on Desalination, Abu Dhabi, UAE, Vol. IV, 19-31, Nov. 18-24.
17. Saeed, M., Al-Amoudi, M. and Al-Harbi, A., (1987), Diseases of Aquatic Organisms 3,
177-180. 18. Hamida, A. and Moch, Jr., I., (1996), The International Desalination and Water Reuse,
Quarterly, 6/3, 40-44. 19. Le Chevallier, M., Cawthon, C. and Lee, R., (1988), Applied and Environmental
Microbiology, 54, 2492-2499. 20. New Scientist (editorial), (1996), 31 August, 31-36. 21. Osterton, J. W., (1994), In: Biofouling and Bio-corrosion in Industrial Water System, G.
Geesey, Z. Lewandowski, and H.-C. Flemming, (Ed.) Lewis Publisher, London, pp. 1-14. 22. Winters, H., (1997), Desalination, 110, 93-96. 23. Dudley, L. and Darton, E., (1997), Desalination, 110, 12-20.