UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Removal of natural organic matter for drinking water treatment using electrocoagulation and ultrafiltration Froese, Emily 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2019_may_froese_emily.pdf [ 3.07MB ]
Metadata
JSON: 24-1.0377512.json
JSON-LD: 24-1.0377512-ld.json
RDF/XML (Pretty): 24-1.0377512-rdf.xml
RDF/JSON: 24-1.0377512-rdf.json
Turtle: 24-1.0377512-turtle.txt
N-Triples: 24-1.0377512-rdf-ntriples.txt
Original Record: 24-1.0377512-source.json
Full Text
24-1.0377512-fulltext.txt
Citation
24-1.0377512.ris

Full Text

REMOVAL OF NATURAL ORGANIC MATTER FOR DRINKINGWATERTREATMENT USING ELECTROCOAGULATION ANDULTRAFILTRATIONbyEmily FroeseB.Sc., The University of British Columbia, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Chemical and Biological Engineering)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)March 2019© Emily Froese, 2019The following individuals certify that they have read, and recommend to theFaculty of Graduate and Postdoctoral Studies for acceptance, the thesis entitled:Removal of Natural Organic Matter for drinking water treatment using electro-coagulation and ultraltration.Submitted by Emily Froese in partial fulllment of the requirements forthe degree of Master of Applied Sciencein Chemical and Biological Engineering .Examining Committee:Madjid Mohseni, Chemical and Biological EngineeringCo-supervisorPierre Bérubé, Civil EngineeringCo-supervisorAnthony Lau, Chemical and Biological EngineeringSupervisory Committee MemberiiAbstractThis work combined iron electrocoagulation (EC) and ultraltration (UF) to treatsynthetic and natural surface waters to remove Natural Organic Matter (NOM).Fixed EC conditions were applied to the feed water in a continuous ow EC re-actor, at a ow rate of 1 LPM and an applied current of 2 A. These test conditionsresulted in an average DOC and UVA-254 reduction of 33 ± 4% and 57 ± 8%respectively for the synthetic feed water.The EC euent acted as the feed water for the EC/UF systems. Two UF mem-brane types were tested: (1) a 300kDa ceramic disk membrane; and, (2) a 0.04µmPVDF hollow-bre membrane. Both systems were operated at a constant ux of50 LMH. Periodic backwash cycles were applied to evaluate the eect of back-washing on the overall membrane fouling rates. Backwashing with a 30 minuteltration cycle and 5 minute backwash reduced the fouling rate by (50±3)% and(2±6)% in the ceramic and PVDF membranes, respectively. Applying backwashto the EC/UF test with natural feed water resulted in a (95 ± 0.5)% lower foul-ing rate compared to the EC/UF test with synthetic feed water test under thesame conditions. Therefore, water composition has a signicant eect on themembrane fouling rates of the EC/UF system.Air sparging was also applied to the PVDF membrane system and resulted inreducing the overall fouling rate by (64± 2)%. Comparatively, the combinationof air sparging and backwashing reduced the overall fouling rate by (98± 1)%.The EC/UF test with air sparging and backwash also resulted in additional NOMremoval of the UF permeate. Air sparging has never been combined with anEC/UF system and these are promising results in the quest to develop a EC/UFdrinking water treatment system for small and remote communities.iiiLay SummaryNatural Organic Matter (NOM) is often found in drinking water sources. Re-moval of NOM from drinking water is important because NOM interferes withdisinfection processes and can lead to the formation disinfection byproducts,which are heavily regulated. Disinfection is an important part of water treat-ment, and removal of NOM before disinfection is needed to prevent these byprod-ucts from forming. Electrocoagulation (EC) uses electricity to add particles to thewater, which can capture NOM.The key goal of this work was to combine EC with a membrane ltrationstep to remove these particles from the water. Two dierent lters were usedand dierent ltration settings were tested, such as applying air sparging andbackwashing to improve ltration performance. These ltration settings havenever before been tested when combining ltration with EC pre-treatment fordrinking water. Results from theses tests indicate that air sparging improvesmembrane performance and water quality.ivPrefaceThe work presented in this thesis was completed by the author, Emily Froese,under the supervision of Professors Madjid Mohseni (Chemical and BiologicalEngineering) and Pierre Bérubé (Civil Engineering) at the University of BritishColumbia.The author was responsible for literature review, experiment design, data col-lection and data analysis. Dr Mohseni and Dr Bérubé contributed to developmentof the experimental plan and provided guidance for development of research ob-jectives and data analysis.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . xivList of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xixDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xx1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Research motivation . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Background and literature review . . . . . . . . . . . . . . . . . 21.2.1 Drinking water treatment and Natural Organic Matter . 21.2.2 Removal of NOM from drinking water . . . . . . . . . . 41.2.3 Electrocoagulation . . . . . . . . . . . . . . . . . . . . . 5vi1.2.4 Membrane ltration and the EC/UF hybrid system . . . 91.2.5 Membrane fouling mitigation . . . . . . . . . . . . . . . 171.3 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . 201.4 Scope of research . . . . . . . . . . . . . . . . . . . . . . . . . . 211.5 Research signicance . . . . . . . . . . . . . . . . . . . . . . . . 212 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . 232.1 Preparation of synthetic feed water . . . . . . . . . . . . . . . . 232.1.1 Preparation of NOM stock solution . . . . . . . . . . . . 242.2 Collection and properties of natural surface water . . . . . . . . 252.3 Experimental apparatus . . . . . . . . . . . . . . . . . . . . . . . 262.3.1 EC experimental apparatus and procedure . . . . . . . . 262.3.2 UF experimental apparatus and procedure . . . . . . . . 282.4 Experimental matrix . . . . . . . . . . . . . . . . . . . . . . . . . 342.5 Analytical methods . . . . . . . . . . . . . . . . . . . . . . . . . 352.5.1 Total and dissolved organic carbon . . . . . . . . . . . . 362.5.2 UVA-254 . . . . . . . . . . . . . . . . . . . . . . . . . . . 372.5.3 Specic UV-absorbance . . . . . . . . . . . . . . . . . . . 372.5.4 Apparent Molecular Weight by HPSEC . . . . . . . . . . 382.5.5 Total and dissolved iron . . . . . . . . . . . . . . . . . . 402.5.6 Analysis and logging of trans-membrane pressure . . . . 402.5.7 Analysis of resistance recovery from backwashing . . . 412.6 Data analysis and condence intervals . . . . . . . . . . . . . . . 423 Results and discussion of synthetic water tests:UF membrane Type 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 453.1 Filtration of raw water . . . . . . . . . . . . . . . . . . . . . . . 463.2 Filtration of EC treated water . . . . . . . . . . . . . . . . . . . . 493.2.1 Preparation of EC pre-treated feed water . . . . . . . . . 493.2.2 Membrane fouling during ltration of EC pre-treated feed 533.2.3 Permeate water quality . . . . . . . . . . . . . . . . . . . 59vii4 Results and discussion of synthetic water tests:UF membrane Type 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 624.1 Filtration of raw water . . . . . . . . . . . . . . . . . . . . . . . 634.2 Filtration of EC treated water . . . . . . . . . . . . . . . . . . . . 674.2.1 Membrane fouling during ltration of EC pre-treated feed 674.2.2 Permeate water quality . . . . . . . . . . . . . . . . . . . 724.3 Comparison of UF membrane systems . . . . . . . . . . . . . . . 774.3.1 Membrane fouling comparison . . . . . . . . . . . . . . 774.3.2 Permeate water quality comparison . . . . . . . . . . . . 795 Results and Discussion of natural surface water test . . . . . . . 835.1 Membrane fouling . . . . . . . . . . . . . . . . . . . . . . . . . . 855.2 Permeate water quality . . . . . . . . . . . . . . . . . . . . . . . 896 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916.1 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . 916.2 Recommendations for future work . . . . . . . . . . . . . . . . . 93References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94viiiList of Tables2.1 Target concentrations for SR-NOM synthetic feed water . . . . . 242.2 Natural surface water properties . . . . . . . . . . . . . . . . . . 252.3 EC test conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 272.4 UF membrane properties . . . . . . . . . . . . . . . . . . . . . . 292.5 Experimental matrix: Synthetic feed water using SR-NOM . . . 352.6 TOC instrument accuracy . . . . . . . . . . . . . . . . . . . . . . 372.7 Identied NOM fractions, based on HPSEC results . . . . . . . . 393.1 Feed and permeate water quality results for raw water ltrationtest using UF membrane Type 1, without backwash . . . . . . . 463.2 Linear regression of raw water ltration using UF membrane Type 1,with 95% condence intervals calculated based on the slope . . . 483.3 EC test results, with an operating current density of 6.43mA/cm2and target metal loading of 35mg/L Fe . . . . . . . . . . . . . . . 503.4 Average percent (%) removal of dierent AMW fractions duringltration of SR-NOM feed . . . . . . . . . . . . . . . . . . . . . . 513.5 Water quality results for EC euent settling test of SR-NOM feedwater . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523.6 Linear regression of EC pre-treated SR-NOM water ltration us-ing UF membrane Type 1 . . . . . . . . . . . . . . . . . . . . . . 554.1 Feed and permeate water quality for raw water ltration usingUF membrane Type 2 . . . . . . . . . . . . . . . . . . . . . . . . 644.2 EC treated UF feed water quality . . . . . . . . . . . . . . . . . . 67ix4.3 Linear regression for UF membrane Type 2 ltration tests . . . . 714.4 Linear regression and R2 values for resistance accumulation forltration with UF membranes Type 1 and Type 2 . . . . . . . . . 785.1 EC test result summary for treatment of natural surface water . 845.2 Linear regression of EC pre-treated synthetic and natural feedwater ltration using UF membrane Type 2 . . . . . . . . . . . . 85xList of Figures1.1 Membrane classication parameters . . . . . . . . . . . . . . . . 111.2 Diagram of cross-ow conguration, for inside-out ltration mode 121.3 Diagram of dead-end conguration and accumulation of foulants 122.1 Process ow chart for EC/UF hybrid system . . . . . . . . . . . . 262.2 Process ow chart for UF ltration of raw water . . . . . . . . . 262.3 Schematic for EC system . . . . . . . . . . . . . . . . . . . . . . 282.4 Process schematic for UF membrane Type 1 (ceramic disk) . . . 302.5 Process schematic for UF membrane Type 2 (hollow-bre) . . . 323.1 UF membrane Type 1 fouling resistance accumulation for rawwater ltration tests with and without backwash . . . . . . . . . 473.2 AMW of untreated water and EC euent for tests using syntheticSR-NOM feed water . . . . . . . . . . . . . . . . . . . . . . . . . 503.3 Particle Size Distribution results of EC euent and settled super-natant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533.4 Snapshot of resistance accumulation for ltration of EC pre-treatedfeed using UF membrane Type 1 . . . . . . . . . . . . . . . . . . 543.5 Average resistance recovery between 200-400L/m2 for UF mem-brane Type 1 tests with EC pre-treated feed and backwash . . . 583.6 Normalized DOC and UVA-254 for UF Membrane Type 1 perme-ate samples during ltration of EC pre-treated SR-NOM feed . . 603.7 SUVA of permeate samples in UF membrane Type 1 ltration tests 61xi4.1 MWD of Suwannee River NOM feed water and UF permeate us-ing UF membrane Type 2, by HPSEC . . . . . . . . . . . . . . . . 644.2 Fouling resistance during ltration of Raw water using UF mem-brane Type 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664.3 Fouling resistance during continual ltration of EC pre-treatedwater using UF membrane Type 2 . . . . . . . . . . . . . . . . . 684.4 Fouling resistance accumulation for ltration tests using UF mem-brane Type 2 and EC pre-treated feed . . . . . . . . . . . . . . . 694.5 Normalized DOC and UVA-254 for UF Membrane Type 2 perme-ate samples during ltration of EC pre-treated SR-NOM feed . . 724.6 AMW of feed and permeate samples for UF membrane Type 2 tests 744.7 Water quality of feed and permeate samples for ltration testswith air sparge. . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.8 Resistance accumulation comparing UF membrane Types 1 and2 for Raw and EC pre-treated feed water for cases without back-wash (a) and with backwash (b) . . . . . . . . . . . . . . . . . . 774.9 Permeate water quality during raw water ltration: comparisonof UF membranes Type 1 and Type 2 . . . . . . . . . . . . . . . . 794.10 Permeate water quality during EC pre-treated water ltration:comparison of UF membranes Type 1 and Type 2, during ltrationwith and without backwash (30 min cycle + 5 min backwash) . . 804.11 SUVA of UF permeate: comparison of UF membranes Type 1 andType 2, during ltration with and without backwash (30 min cy-cle + 5 min backwash) . . . . . . . . . . . . . . . . . . . . . . . . 815.1 Resistance accumulation for UF test with EC pre-treated naturalsurface water with 30 minute ltration cycles and 5 min backwash 865.2 Reversible fouling for UF membrane Type 2 experiment with ECpre-treated natural surface water . . . . . . . . . . . . . . . . . . 875.3 Average resistance recovery for synthetic and natural feed watersusing UF membrane Type 2 . . . . . . . . . . . . . . . . . . . . . 88xii5.4 Normalised DOC and UVA-254 for UF Membrane Type 2 perme-ate samples during ltration of EC pre-treated natural surfacewater feed and synthetic feed water . . . . . . . . . . . . . . . . 90xiiiList of AbbreviationsAMW Apparent Molecular Weight; the estimated molecular weight of acompoundBWAs Boil Water AdvisoriesCC Chemical CoagulationCOD Chemical Oxygen DemandCP Concentration PolarizationDBP Disinfection By-ProductDI Deionized WaterDOC Dissolved Organic CarbonEC ElectrocoagulationHAAs Halo-Acetic Acids, a DBPHPSEC High Performance Size Exclusion Chromatography; an analyticalmethod to determine the MWD of an organic sampleICP-OES Inductively Coupled Plasma-Optical Emission Spectroscopy; ananalytical method to determine the concentration of metals in anaqueous sampleIHSS International Humic Substances SocietykDa kilo-Dalton, unit to express molecular weight of organic com-poundsLMH Litres per Meter squared per Hour (Lm−2h−1); Unit for ltrationuxLOQ Limit of QuanticationxivLPM Litres Per Minute (L min−1); Unit for volumetric ow rateMDL Method Detection LimitMF MicroltrationMWCO Molecular Weight Cut-O; The nominal molecular weight abovewhich particles are rejected by a membraneMWD Molecular Weight DistributionNDIR Non-Dispersive I0frared Detection; A method to detect gases suchas CO2NF NanoltrationNOM Natural Organic MatterOP Operating PotentialPSD Particle Size DistributionRO Reverse OsmosisRPD Relative Percent Dierence between the measured value and thetrue value of the standard.SHA Sulfonated Humic AcidSR-NOM Suwannee River NOM, a standard NOM source provided by theIHSSSRFA Suwannee River Fulvic AcidSRHA Suwannee River Humic AcidSTP Standard Temperature and Pressure (25◦C and 1 atm)SUVA Specic UV-absorbanceTHMs Trihalomethanes, a DBPTMP Trans-Membrane PressureTOC Total Organic CarbonUF UltraltrationUVA UV-AbsorbanceUVA-254 UV-Absorbance at 254nmxvList of SymbolsA The membrane surface areaAl Atomic symbol, aluminumAl(OH)3(s) Aluminum hydroxide, an aluminum precipitateC Concentration of speciesδ Inter-electrode gap∆E◦ Overall reaction potential∆P Trans-membrane pressure∆Pt Term for the measurement of TMP as a function of time∆z Membrane pore lengthE◦ Standard reduction potentialex Error associated with measured value x, calculated from the 95%condence intervals of replicate measurements%ex Relative error of measured value xF Faraday’s constantFe Atomic symbol, ironFe2+ Ferrous ironFe3+ Ferric ironFe(OH)3(s) Ferric hydroxide, an iron precipitateI Applied currenti Applied current densityJ Filtration uxM Molar mass of compoundxvim Mass or metal released from electrolytic reactionMLmeasured The measured metal loading from ECMLtheory Theoretical metal loading or concentration of iron in the cell ef-uentµ Dynamic viscosity of the liquidN Number of samples used in condence interval calculations anddata analysisPpore Membrane pore density (or porosity)φ Current eciencyPXRA The signal output of pressure transducer APXRB as A The output of pressure transducer B, expressed in equivalentunits to pressure transducer Aq The total charge passed through the electrodeQ Volumetric ow rateR Membrane resistancer Membrane pore radiusRend, i−1 The fouling resistance at the end of the previous cycleRf The accumulated fouling resistance of a membraneRm The intrinsic hydraulic membrane resistance of a clean mem-braneRr, i The resistance recovery from backwash at the start of the i-thcycleRstart, i The resistance measured at the begining of the current cycleRT Total resistance of the membranes Standard deviation of replicate measurementssm Standard deviation of the slope, from linear regressiont Timet95% 2-sided t-value for calculation of 95% condence intervalsτ Membrane pore tortuosityxviiz Valence of ion released from anode, or the number of electronsexchangedxviiiAcknowledgementsI would like to thank my supervisors Madjid Mohseni and Pierre Bérubé for tak-ing me on as a student and providing support over the last 2+ years. I trulyappreciate the opportunity to have been part of RES’EAU WaterNET, and thisexperience has provided me with an education beyond the technical aspects ofwater treatment and chemical engineering.I would also like to thank my lab mates for sharing their knowledge and labequipment. Having someone to chat with in the lab while I watched water dropsslowly ll a sample vial always made the process more enjoyable. Thanks toDoug in the CHBE workshop and Scott in the CIVL department workshop forhelping me build my setup (and make it work). And thank you to Joerg andShona, for answering my questions about membranes and pressure transducersas I tried to assemble a setup that was completely new to me.The team at BQE Water was so supportive of my decision to return to UBC,for which I am very grateful. Thank you to my parents for raising me to lovescience and encouraging me throughout my education to achieve my goals. Andnally – Mark. Your support throughout this degree has been invaluable. Thankyou.xixThis work is dedicated to kids who love sciencexxChapter 1Introduction1.1 Research motivationThe main objective of drinking water treatment is to produce clean water thatis safe to drink. Canada is perceived as a country rich in freshwater sourcesdue to the abundant lakes and rivers that cover the country’s surface. Thesewater sources supply potable water for over 90% of Canadian municipalities(Environment Canada, 2011). However, this freshwater is not pristine, and thecurrent methods of treatment and distribution are not perfect. At the time ofwriting, there are over 700 drinking water advisories in place across the country(see http://www.watertoday.ca/ to access the most recent numbers). Althoughthe majority of these advisories are issued as preventative measures to addressequipment and process issues, unacceptable water quality accounted for 16.5%of all Boil Water Advisories (BWAs) issued in 2017 and over 70% of these advi-sories were issued to small systems that serve communities of 500 people or less(Environment and Climate Change Canada, 2018).Unacceptable water quality can be caused by insucient treatment, whichmay be due to poor or deteriorating source water quality or inadequate oper-ation of the water treatment plant. A review of waterborne disease outbreaksin British Columbia, Canada, between 1980 and 2002 concluded that the most1common outbreak was caused by Giardia, a pathogen commonly found in sur-face water (Sierra Legal Defense Fund, October 2003). In most cases, the solutionimplemented to treat for Giardia contamination was an increased chlorine dose,which is not desirable for many reasons. However, in small systems, chlorine isone of the most common methods of disinfection.Chlorine is inexpensive. It is capable of inactivating many pathogens of con-cern, and provides a secondary disinfection by maintaining a residual concentra-tion in the treated water through the distribution system until the water reachesthe consumer’s tap. However, chlorine is also associated with unwanted tasteand odour issues, corrosion of pipes and ttings, and increased risk of formingdisinfection byproducts (DBPs). Furthermore, in addition to the higher dosagerequired to kill pathogenic Giardia cells, chlorine is also ineective in killingother protozoa, such as Cryptosporidium cells when they are in the oocyst form(Sierra Legal Defense Fund, October 2003). Therefore, chlorine is not an idealdisinfectant.A reliable water supply and treatment system is important for any commu-nity. However, small and remote communities have the added challenge of meet-ing the needs of the community with limited resources. The motivation behindthis research is to investigate a water treatment technology that can be appliedto small and remote water systems which rely on surface water as the drinkingwater source. A successful treatment system would eectively meet treatmenttargets and would be user friendly, low cost and robust.1.2 Background and literature review1.2.1 Drinkingwater treatment andNatural OrganicMatterThere are many dierent approaches to drinking water treatment, but the over-all objective is always the same: produce a safe and reliable supply of drinkingwater for the consumer. Drinking water regulations are often primarily focusedon removal or inactivation of pathogens in water, but there are other chemical2contaminants that can also have adverse health eects on consumers. DBPs area type of chemical contaminant that are heavily regulated in Canada and aroundthe world but are not the primary focus of drinking water treatment.DBPs are formed when water containing natural organic matter (NOM) un-dergoes a disinfection process such as chlorination or chloramination (Rook,1976). Common DBPs include Trihalomethanes (THMs) and Halo-acetic Acids(HAAs), however, these compounds only represent a small fraction of possiblecompounds formed during disinfection. DBPs are cause for concern because theyare suspected to have negative health eects (Richardson, 2003). Therefore, pre-vention of DBP formation is a very important part of drinking water treatment.Preventing DBP formation involves removing DBP precursors, like NOM, priorto disinfection.NOM is a term used to describe the complex organic molecules that are pro-duced by the breakdown of plant and animal matter in water. The chemicaland physical properties of NOM will vary, depending on source organic mate-rial, and time of year (Thurman (1985) cited in Brinkman and Hozalski (2015)).The molecular weight and character of the NOM can be separated into dier-ent classes depending on its variable molecular structure. NOM can be classiedbased on hydrophilic, transphilic or hydrophobic character as well as its acidity(Leenheer & Croué, 2003). Two common methods for determining the concen-tration of NOM in water is by measuring the dissolved oxygen concentration(DOC) and the UV-absorbance of a sample at 254nm (UVA-254).The variable chemical and physical properties of NOM impacts the eciencyof the treatment process applied (Collins et al., 1986). Dierent fractions of NOMare preferentially removed by the dierent treatment technologies (Sharp et al.,2006; Aschermann et al., 2016) and dierent fractions of NOM are correlatedto dierent types of DBP formation (Hua et al., 2015). The hydrophobic, largermolecular weight and aromatic fractions have been cited as more likely to leadto DBP formation (Lin & Wang, 2011; Rook, 1976). Therefore, in the applicationof drinking water treatment of surface water, removal of these fractions of NOM3prior to disinfection is a priority.1.2.2 Removal of NOM from drinking waterA typical North American surface water treatment system involves coagulation(and rapid mixing phase), occulation (and slow mixing phase), some sort ofsolid/liquid separation, and a nal disinfection step. The solid/liquid separationcould include sedimentation and granular media ltration of the settled super-natant or direct ltration of the coagulation/occulation euent.Each unit in this process can be studied and described in detail and properoperation of each unit is key to successful process performance. The coagula-tion/occulation stage is responsible for removing suspended material, includingpathogens such as bacteria and protozoa, from the water. Enhanced Coagulationis the term used to describe coagulation with the objective of removing the NOMdissolved in the water. This is done by adding excess coagulant and following aprotocol for ash mixing and occulation (Crozes et al., 1995). To better under-stand enhanced coagulation for NOM removal and how it can be improved, it isuseful to understand the basic principles of coagulation.Chemical coagulation and occulationCoagulation is an important process in water treatment. In this process, chargedspecies are added to water to remove suspended and dissolved charged parti-cles. Most of these particles have a negative surface charge and this chargedlayer can be neutralized by adding positively charged ions to the water, whichreduces the repulsive forces between the particles and allows then to come to-gether (Crittenden et al., 2012a).Typical coagulants are aluminum or iron salts, which are added as a liquidreagent to the source water and are hydrolysed to form ferric or aluminum hy-droxide species. This is known as chemical coagulation (CC). The coagulant isfully dispersed during a rapid mix phase and then the water is transferred to a4slower, gentle mixing phase that allows particles to grow in size. This is calledocculation. If excess coagulant is added and the solution is at a suitable pH, thehydrolysed iron or aluminum ions will exceed their solution solubility and pre-cipitate in the form of aluminum or ferric hydroxides (Fe(OH)3(s) or Al(OH)3(s)),which can further capture soluble NOM (Crittenden et al., 2012a). This is knownas the ‘sweep oc’ mechanism of coagulation/occulation and is an importantpart of enhanced coagulation for NOM removal from surface water.Enhanced coagulation is a complex process and determination of optimal co-agulant dose is best done using lab-scale jar tests with the real water to be treated(Crozes et al., 1995). Technology drawbacks and issues include high sludge pro-duction, process requirements for pH adjustment and control, and high reagentconsumption compared to regular coagulation processes. Furthermore, reagentssuch as ferric chloride are classied as dangerous goods, and therefore requireregulated handling and storage of the reagent when shipping it to site. The draw-backs of this process has led to the research and development of other technolo-gies to remove NOM.Electrocoagulation (EC) is a promising alternative to chemical coagulation.Other non-coagulation based technologies capable of NOM removal also exist,including ion-exchange, adsorption, and membrane ltration (or a combinationof these processes) (Humbert et al., 2007; Velten et al., 2011; Stoquart et al., 2012;Winter et al., 2017). Each process has its own set of strengths and weaknesses andwill not be discussed in detail. It is only important to know that many dierentapproaches exist to solve this complex problem.1.2.3 ElectrocoagulationEC follows the same principles as chemical coagulation however the means ofintroducing the coagulant species to the water is dierent. EC has been widelyapplied to dierent types of water to remove hardness, chemical oxygen demand(COD), turbidity, viruses, heavy metals, and NOM (Zhao et al., 2014; Tanneru &Chellam, 2012; Nariyan et al., 2017; Vik & Carlson, 1984).5Aluminum and iron are the most common materials used in EC (Dubrawski& Mohseni, 2013; Särkkä et al., 2015); however, iron is preferred due to lower costand higher durability (Moreno C. et al., 2009). In EC, an electric current is passedthrough a pair of electrodes immersed in the water to be treated. This results ina set of electrochemical reactions happening on the surfaces of the electrodes. Inan iron EC unit, these reactions are:Anode:Fe 0s Fe2+ + 2 e– E◦ = 0.447 V (1.1)Cathode:2 H2O + 2 e– H2(g) + 2 OH– E◦ = −0.8277 V (1.2)The standard reduction potential cited (E◦) is measured at STP (25◦C and 1 atm)(Vanysek, 2005). This coupled reaction corresponds to an overall reaction po-tential of ∆E◦ = −0.3807 V. In theory, Fe3+ could also be produced from theanode, but the prediction of ferrous iron production corresponds to observationsin literature (Bagga et al., 2008; Tanneru & Chellam, 2012).Direct formation of Fe3+ is likely not often observed because the standardreduction potential for ferric iron is much higher than that for ferrous iron, re-sulting in a greater overall reaction potential. The electrochemical reaction forthe direct production of ferric iron on the anode is:Fe 0s Fe3+ + 3 e– E◦ = 0.037 V (1.3)which would result in an overall reaction potential of ∆E◦ = −0.7907 V). Thispotential may not be in the range of current densities commonly investigated inthis application.Lakshmanan et al. (2009) completed an in-depth study on the ironspecies generated during iron EC. Their results suggested that Fe2+ was formeddirectly through electrolysis and was then oxidized to Fe3+ in solution. How-6ever, these reactions are complex, and are aected by system conditions such asdissovled oxygen (DO) concentration and water matrix (Dubrawski et al., 2015;Dubrawski & Mohseni, 2013)The amount of Fe2+ added to the system can be predicted by Faraday’s Lawof Electrolysis:m =qMFz(1.4)where m is the mass of material released from the electrode (g), q is the totalcharge passed through the electrode (Coulombs),M is the molar mass of the ma-terial (g/mol), F is Faraday’s Constant (96485 Coulombs/eq), and z is the numberof electrons transferred (eq/mol). In the case of Fe EC, M = 55.85 g/mol and z= 2. When operating under constant current operation, the total charge (q) isproportional to the applied current multiplied by time:q = It (1.5)where I is the applied current (Amp) and t is the time (sec).In a continuous ow reactor that is operated with a known volumetric owrate, Q, the theoretical metal loading (MLtheory) or concentration of iron in thecell euent can be predicted and controlled using the following equation:MLtheory =I ×MQ× F × z =I × 55.85 g/molQ× 96 485 C/mol× 2 (1.6)Using this equation, ML is expressed as the total concentration of iron asmass/volume (mg/L). The actual metal loading (MLmeasured) can be measuredby analyzing a sample for concentration of total iron, which can be used to de-termine the current eciency (φ) of the system.φ =MLmeasuredMLtheory(1.7)A new anode is expected to operate at 100% current eciency, however as an an-7ode ages, the current eciency will decrease if the anode is not cleaned regularly(Lakshmanan et al., 2009).After Fe2+ is released into solution, a series of chemical reactions can occurin the bulk, including oxidation and hydrolysis of the iron ions. The oxidationrate of Fe2+ to Fe3+ and precipitated iron species solubility in EC experiments de-pends on pH (Ben-Sasson et al., 2009). Lakshmanan et al. (2009) observed a fasteroxidation rate of Fe2+ in a EC batch reactor compared to a Fe2+ chemical batchreactor, due to the localized high pH on the cathode. However, the oxidationrate could be aected by mass transfer limitations, caused by the overall cur-rent density and charge loading rate in the EC reactor (Dubrawski & Mohseni,2013). Therefore, depending on the reactor conditions, the nal EC-producedspecies may include other ferric and ferrous (oxy)hydroxides, beyond Fe(OH)3(s).The species formed, and their subsequent transformation as they age, dependson the ions available for complexation in the water, as well as the presence ofdissolved oxygen (Dubrawski et al., 2015). Dubrawski and Mohseni (2013) ob-served many iron species including magnetite, goethite and green rust underdierent EC conditions, which appeared to inuence NOM capture by EC. Thisis a fundamental dierence between CC and EC: iron CC involves hydrolysis ofthe coagulant species (such as Fe3+) but does not typically include oxidation ofFe2+ in the solution bulk, since iron is not commonly introduced as Fe2+ for thispurpose.It was thought that the formation of Fe(OH)3(s) ocs drives NOM capture dur-ing EC, as it does in CC. Bagga et al. (2008) hypothesized that the presence ofNOM inhibits Fe2+ oxidation, by complexing with the ferrous species and pre-venting formation of Fe(OH)3(s) and reducing the overall eciency for NOM re-moval through EC. However, Timmes et al. (2010) identied an important fea-ture of the system used by Bagga et al.: the batch reactor used did not accountfor the depletion of oxygen in the system. In the work completed by Timmeset al. (2010), when adequate oxygen was available, there was comparable per-formance between CC and EC for NOM removal in ltered, EC treated samples.8Further research into the mechanism of NOM removal by EC indicated that thesolids produced during EC are able to capture and remove contaminants, eitherby sweep occulation or charge neutralization mechanisms, depending on thedose of iron added to solution (Timmes et al., 2010).The majority of EC NOM removal studies are focused on lab scale or batchreactors. McBeath (2017) investigated a continuous ow pilot scale EC unit forremoval of NOM from surface water, up to a ow rate of 10 LPM. He concludedthat even with high metal loading (up to 66 mg/L Fe) the energy requirements donot outweigh the benets of the process. McBeath (2017) was able to achieve anaverage DOC and UVA-254 reduction of 37.2±4.2% and 54.7±0.9%, respectivelyusing humic acid as a model NOM reagent.These are promising results for applying electrocoagulation to drinking wa-ter treatment. However, the nal product must be a water that meets drinkingwater quality guidelines, which means that the coagulant added and ocs formedmust be removed. Therefore, a method of solid/liquid separation is required ifEC is to become a process that can produce clean, safe drinking water.1.2.4 Membrane ltration and the EC/UF hybrid systemA possible method for removing the solids produced in the EC unit is throughultraltration (UF), which is one classication in the larger eld of membraneltration. Membrane ltration is a way of physically separating materials fromwater based on the size or molecular weight of the material. There are manydierent types, classications, and congurations for membrane ltration tech-nologies. The following section will provide an overview of membrane ltrationand previous research combining EC pre-treatment with membrane ltration.This will provide background information on the membrane ltration setupsused in this research. Additional details on the experimental setup are availablein Section 2.3.2.9Membrane ltration - basic principlesA key principle of membrane ltration is the relationship between the volumet-ric ow rate, membrane surface area and the trans-membrane pressure (TMP).The following equation can be used to represent this relationship, which is amodied form of Darcy’s law (Cheryan (1998) cited in American Water WorksAssociation (2008)):J =QA=∆PµRT(1.8)Where J is the ltration ux (m s−1), Q is the volumetric ow rate (m3s−1), A isthe membrane surface area (m2), ∆P is the trans-membrane pressure (Pa), µ isthe dynamic viscosity of the liquid (kg m−1 s−1) and RT (m−1) is the total resis-tance term of the membrane. RT accounts for the internal membrane resistanceas well as the accumulation of resistance as water is ltered. The accumulatedresistance is known as membrane fouling and will be discussed later in this chap-ter.Membrane systems can be classied by membrane material, system hydro-dynamic properties (such as ow conguration and geometry) and membranepore size. These classications are outlined in Figure 1.1.Polymeric and ceramic materials are two classications of membrane mate-rial. The membrane material may inuence physical and chemical interactionsbetween the membrane and the feed water. This may be due to to propertiessuch as hydrophilic or hydrophobic character, surface roughness, surface chargeand pore shape (Crittenden et al., 2012b). Polymeric membranes are often usedin drinking water treatment because they are considered easy to install and rel-atively inexpensive. Ceramic membranes were previously considered too ex-pensive due to high capital costs but are gaining attention as the eld develops.Research suggests that ceramic membranes are less susceptible to irreversiblefouling and would therefore support a system with higher throughput (S. Lee etal., 2013). However, for experimental design of this project it was assumed that10Figure 1.1: Membrane classication parametersthese interactions are minimal and the membrane ltration is purely a physicalprocess of size exclusion. This assumption is discussed further in Section 4.1.Both ceramic and polymeric membranes were used in this research.The hydrodynamics of a membrane system is aected by the ow congura-tion and geometry, which in turn, aects the membrane performance. In termsof operation, membrane systems can either be operated under dead-end or cross-ow conguration. Figures 1.2 and 1.3 present a simplied schematic of the twodierent ow congurations.Cross-ow operation is benecial for feed waters that have high concentra-tions of suspended solids, due to the sweeping movement of ow across the sur-face of the membrane, preventing a buildup of large cake layer (Crittenden et al.,2012b). Dead-end ltration systems have the benet of simple design and oper-ation. However, without the tangential force of a cross-ow, severe fouling andconcentration polarization are likely to occur, leading to an increase in the forcerequired to lter the water (Belfort et al., 1994). In drinking water applications,the source water is typically low in suspended solids the dead-end congurationis more common. Therefore, a dead-end ltration conguration was used for this11Figure 1.2: Diagram of cross-ow conguration, for inside-out ltration modeFigure 1.3: Diagram of dead-end conguration and accumulation of foulantsproject.Membranes can also be operated under constant pressure or constant uxconguration. Constant pressure systems are common in the lab scale but con-stant ux systems are more common for full scale plants. Miller et al. (2014)compared constant pressure and constant ux operation for UF membranes. Thethreshold ux of the system was identied (the ux above which membrane re-sistance rapidly increases, as dened by Field et al. (1995)). Tests were completedat uxes and equivalent pressures above and below this critical ux value. Re-sults indicated that at low ux, constant ux and constant pressure tests demon-strated similar fouling behaviour. However, at high uxes, results deviated andmembrane resistance rapidly increased for constant high ux experiments, butnot for constant high pressure experiments. This implies that constant pressureand constant ux systems may not yield the same results when scaling up a ltra-tion system during technology development. A constant, low ux conguration12was used in this project, as it is more likely to resemble a large system.Membrane geometry denes how the driving force for ltration is applied.Submerged systems are typically operated at constant ux while external geom-etry typically refers to a pressurized system. Geometry refers to the physicalshape of the membrane module. Examples of dierent module shapes include:hollow-bre, atsheet, tubular, or spiral wound systems. The membrane geom-etry aects the uid dynamics of the system, which is extremely important interms of fouling and process design (Belfort et al., 1994). Lab-scale atsheet (disk)and hollow-bre membrane modules were used in this research, as these geome-tries resemble those used in drinking water applications. Schematics of the UFmembrane systems used in this project are available in Section 2.3.2.Membrane pore-size or molecular weight cut o (MWCO) is the most im-portant parameter in membrane classication, as it denes what material is re-jected by the membrane. Both terms are used to classify the size of particles ormolecules that will be removed from the water (Crittenden et al., 2012b).Microltration (MF) and ultraltration (UF) systems are known as low pres-sure membranes and are designed for suspended particle and colloid removal.The approximate pore-sizes of these membranes are 0.1−0.2µm (nominally 0.2)and 0.01−0.05µm for MF and UF membranes, respectively (Allgeier, November2005). Nanoltration (NF) and Reverse Osmosis (RO) systems are known as highpressure membrane systems and are used for small molecule and ion removal toproduce ultra-pure water. With respect to the contaminants commonly found insurface water sources, low pressure membrane systems are sucient for meet-ing treatment objectives when combined with a subsequent disinfection step toinactivate any bacteria or viruses that may be able to permeate the membrane(Jacangelo et al., 1995). Low pressure membranes require less energy than highpressure membranes to operate, which is favourable when designing a simplesystem for a small community. In this research, two UF membranes were testedwith the EC pre-treatment step.Selection of membrane pore size has a signicant impact on membrane per-13formance in terms of permeate water quality and changes in membrane resis-tance (known as membrane fouling). NOM has long been identied as a mem-brane foulant for both high pressure and low pressure membranes. N. Lee et al.(2004) compared UF and MF membranes for ltration of NOM-containing waterand concluded that MF membranes are more prone to fouling than UF mem-branes due to the dierence in pore sizes relative to the size of particles removedby the membrane. Therefore, the combination of EC (which increases the aver-age particle size) with a UF membrane system is expected to reduce membranefouling.Previous research that has incorporated EC/membrane hybrid systems hasprimarily focused on MF systems. Bagga et al. (2008), Ben-Sasson et al. (2013),and Gamage and Chellam (2011) all investigated hybrid EC/MF systems for NOMremoval. However, very little work has been completed using hybrid EC/UF sys-tems. Han et al. (2015) applied an EC/UF process to investigate sulfonated humicacid (SHA) removal from water and achieved up to 95% SHA removal with a5kDa UF membrane. This process, however, incorporated sedimentation of theocs and well as a pre-ltration step with a bag lter prior to ultraltration. Hanet al. did not consider direct ltration of the EC euent. Direct ltration is de-sirable as it may simplify the total footprint of a full-scale treatment plant.Membrane foulingAs mentioned throughout this chapter, membrane fouling is an important con-sideration in membrane operation and system design. A simplied descriptionof membrane fouling is to characterize it as two types: internal fouling (wheremolecules will become attached to the inner pores of the membrane structure)and external fouling (where a cake layer of particles build up on the outer surfaceof the membrane). In both cases, the required pressure to push water throughthe membrane at a constant ux will be increased.Recall the resistance term, RT in equation 1.8. Membrane fouling can beevaluated by treating fouling as a ‘resistance in series’ model, where the total14resistance,RT , is the sum of each factor that contributes to the overall resistance(American Water Works Association, 2008):RT = Rm +Rf (1.9)Rm is the intrinsic hydraulic membrane resistance of a clean membrane and Rfis the accumulated fouling resistance. Rm is related to physical properties of themembrane, as outlined in equation 1.10Rm =8× τ ×∆zpi × r4 × Ppore (1.10)where τ , ∆z, Ppore, and r refer to the pore tortuosity, pore length, pore density(or porosity), and pore radius of the membrane, respectively (American WaterWorks Association, 2008). Rm can be determined by measuring the ux or TMPof the system when ltering deionized water through a clean membrane. This iscalculated through the following equation:Rm =∆P◦µJ(1.11)∆P◦, µ and J are the clean water TMP (Pa), dynamic viscosity (Pa.s), and ltra-tion ux (m/s), respectively. Rm may change as the membrane ages and under-goes periodic chemical cleans. Gao et al. (2016) suggested that chemical clean-ing of PVDF membranes using NaOCl resulted in irreversible compaction of themembrane material properties, causing sharp changes inRm. However, typicallyit is Rf that changes as material is ltered out of solution.Rm and RT must be used to determine Rf . In a constant ux system RT canbe determined by measuring the TMP as a function of time (∆Pt):RT =∆PtµJ. (1.12)15Using equations 1.9, 1.11, and 1.12 we can dene an equation and solve for Rf :Rf =∆Pt −∆P◦µJ(1.13)which can be used to evaluate the system in terms of membrane fouling rate andfouling mechanism.The membrane fouling rate (or, the rate at which Rf increases) dependson many factors such as mass of foulants approaching the membrane, massof foulants being pushed away from the membrane through back-transport orcross-ow mechanisms, and adsorption of material onto the surface or innerpores.Four main mechanisms are accepted as the dominant mechanisms for fouling,and a fouling model has been proposed to correspond to each of these mecha-nisms to express the relationship between resistance accumulation and volumeltered (Hermia, (1983), cited in Bolton et al., (2006)). The four mechanisms are:• Complete blocking (external): inverse relationship, particles seal pores,area blocked is proportional to area ltered• Incomplete blocking (external): exponential relationship, particles accu-mulate on area already covered by other particles, or on remaining pores• Cake fouling (external): linear relationship, a porous cake accumulates,proportional to the volume ltered• Standard blocking (internal): power function, material adsorbed onto theinner pore decreases the actual pore radiusIt is accepted that membrane fouling may not follow a single mechanismthroughout a ltration cycle. Fouling is a complex process, and the dominantmechanism may change over time or more than one mechanism may be occur-ring during ltration of material (Jermann et al., 2007; Bolton et al., 2006).16As NOM varies in size and structure, dierent constituents of NOM will foulmembranes in dierent ways. Huber (1998) was able to dierentiate betweenthe dierent fractions of NOM that cause irreversible fouling in RO membranes.Knowledge of which fractions cause fouling helps determine which processesare necessary for treatment prior to membrane ltration. This is important, asdierent fractions of NOM are preferentially removed by dierent treatment pro-cesses.In the application of coagulation pre-treatment of membrane ltration, thedominant fouling mechanism observed is often described as cake fouling. Baggaet al. (2008) observed incomplete blocking during raw water ltration and cakefouling when the feed water was pre-treated by chemical coagulation. This im-plies that coagulation pre-treatment changes the particle properties and aectsthe fouling mechanism.Gamage and Chellam (2011) also observed cake fouling as the dominant foul-ing mechanism during microltration of aluminum EC pre-treated feed water,which was an improvement over direct ltration of raw feed water. EC pre-treatment improved the overall fouling rate. However, these works did not in-vestigate other means of fouling mitigation.1.2.5 Membrane fouling mitigationThere are three approaches to fouling mitigation during water treatment:1. Pre-treatment to capture or remove foulants before they reach the mem-brane;2. Hydraulic controls, such as backwashing, turbulence or cross-ow, to dis-lodge foulants from the surface; and,3. Chemical controls, such as chemical cleans with citric acid or NaOCl toremove internal foulants.17Application of these controls is designed to either prevent fouling from occur-ring, or temporarily remove the fouling from the membrane before it begins tore-accumulate in later ltration cycles. Fouling that can be removed in this wayis known as ‘reversible’. Chemical cleaning is designed to remove ‘irreversible’fouling, such as adsorbed material or internal foulants that would otherwise notbe removed through backwashing.The EC/UF hybrid system is an example of pre-treatment to remove NOMbefore it reaches the membrane. In municipal wastewater systems, direct mi-croltration is problematic due to the colloidal fraction of particles suspended inthe water causing fouling and lowering the overall ux of the system. Pouetand Grasmick (1994) applied aluminum electrocoagulation-otation as a pre-treatment step in a municipal wastewater MF system, which improved the mem-brane ux of the system from 20LMH to 350LMH.Ben-Sasson et al. (2013) investigated the eect of pH and anode type onEC pre-treatment for NOM removal and membrane fouling mitigation. The re-searchers observed that improved NOM removal did not correlate with improvedmembrane eciency, yet both parameters were dependent on initial pH, totalmetal loading, and type of electrode used (Ben-Sasson et al., 2013).Gamage and Chellam (2011) applied Al EC to MF for surface water treatmentand observed cake fouling in all cases, which exacerbated fouling at high TMPsdue to cake compression. The authors suggested that EC pre-treatment requiresa balance between total metal loading and large oc generation (Gamage & Chel-lam, 2011). EC causes higher total ltration resistance, as the total mass of ma-terial in the feed water is increased, but the increased particle size decreases thespecic resistance of the cake. This agrees with results reported by Ben-Sassonand Adin (Ben-Sasson & Adin, 2010): the researchers claim that the sweep coagu-lation mechanism is responsible for fouling mitigation, as it changes the particlesizes in such a way that the cake fouling mechanism dominates, instead of theother blocking models.18These examples suggest that EC pre-treatment has the potential to improvemembrane eciency, however very little work has been completed on how hy-draulic and chemical controls can be applied to improve fouling mitigation in ahybrid EC/UF system. Timmes et al. (2010) observed high hydraulic recovery ofmembrane resistance when using EC treated water. However, cyclic backwash-ing was not investigated. Ben-Sasson and Adin (2010) applied dead end, constantpressure MF to Fe EC treated water. Flux restoration was observed by rinsing themembrane with tap water, due to easy removal of the cake layer. However, the re-versible fouling was not investigated further than that observation (Ben-Sasson& Adin, 2010). Reversible fouling can be removed by backwashing or introduc-ing turbulence into the system, yet very little research has been completed onthe eect of hydraulic controls for UF membrane mitigation of EC pre-treatedwater.A review by Böhm et al. (2012) highlights the importance of uid dynamicsin membrane fouling mitigation. Understanding of the system uid dynamicshelp to optimize the hydraulic controls, such as induced turbulence and back-wash eciency. Furthermore, backwash is important but so are other methodsof induced turbulence, such as air sparge. The use of air bubbles is commonin membrane ltration fouling control as means to introduce turbulence in thesystem without the expenses of a pressurized cross-ow system (Cui et al., 2003).In the application of combining EC with membrane ltration for drinking wa-ter, the majority of research has focused on pre-treatment and evaluation of per-meate water quality, not hydraulic fouling control. There has been no researchinvestigating the use of air sparge on UF feed water or subsequent membraneperformance. This is a knowledge gap that is addressed in this research: Theeect of cyclic backwash and induced turbulence by air sparging on an EC/UFsystem.191.3 Research objectivesThe objectives of this research are as follows:1. Apply iron EC to synthetic surface water containing Suwannee River NOM.• Evaluate EC performance by measuring total iron, dissolved organiccarbon and UVA-254 in EC-treated water, under xed EC settings.• Analyze the molecular weight distribution of NOM in raw water (un-treated) and EC treated water (ltered to 0.45µm).2. Filter EC pre-treated water using a ceramic disk UF membrane.• Analyze membrane fouling rates under dierent ltration cycle andbackwash cycle duration settings. Compare these results to a ltra-tion test without backwash.• Analyze DOC, UVA-254 and total iron in the UF permeate, as mea-surement of permeate water quality. Compare these values to rawwater values to determine the eectiveness of the iron EC/UF hybridsystem to capture and remove NOM from the source water. Evaluatepermeate water quality at the start and end of the ltration test.3. Filter EC pre-treated water using a PVDF hollow-bre UF membrane. Eval-uate membrane fouling and permeate water quality.• Analyze membrane fouling rates when air sparging and/or backwash-ing is applied. Compare these results to ltration test where no foul-ing mitigation is applied.• Analyze DOC, UVA-254 and total iron in the UF permeate, as mea-surement of permeate water quality. Compare these values to rawwater values to determine the eectiveness of the iron EC/UF hybridsystem to capture and remove NOM from the source water. Evaluatepermeate water quality at the start and end of the ltration test.204. Apply the EC/UF process to a natural surface water and evaluate the treatedwater quality (DOC, UVA-254, and Fe) and fouling rate for a single UF op-eration mode. Compare results to tests using synthetic surface water.1.4 Scope of researchA set of experiments were designed to address the research objectives outlinedabove, using two dierent UF membranes. The membrane fouling and perme-ate water quality were monitored and analysed. The eect of backwash wasinvestigated with both membranes by evaluating the overall rate of resistanceincrease across multiple cycles. Air sparging, backwash duration and lter cycleduration were all investigated, however, a complete factorial experimental de-sign was not implemented due to the limited time available to complete a MAScresearch project.1.5 Research signicanceThe benet of a hybrid EC/UF system is that it can treat water without havingto deal with signicant reagent addition or pH adjustment. The EC/UF systemcan capture and remove NOM from surface water prior to chlorination, whichreduces the risk of DBP formation. Membrane ltration can remove the ocs pro-duced in the EC unit and has the added benet of providing disinfection creditsby also ltering out some pathogens. These combined processes then decreasethe overall demand of chlorine required for disinfection, and decrease the tasteand odour issues associated with chlorine disinfection.The combination of electrocoagulation and ultraltration has the potential tohave signicant impact on the water treatment options available to small com-munities where reliable, safe drinking water is still a requirement but a highcapacity/high volume water treatment system may not be feasible.This research is signicant because little work has been done on this type of21integrated process. Limited research into UF fouling with EC pre-treatment hasbeen completed, and in the research that has been published, the focus has beenon water quality and overall membrane performance in dead-end, constant pres-sure, lab-scale systems. There is very little research into the eect of hydraulicfouling mitigation techniques like cyclic backwashing and induced turbulencethrough air sparging on EC pre-treated water for drinking water treatment andNOM removal. I will address these gaps and propose additional work to helpfurther the development of an EC/UF process for drinking water treatment.22Chapter 2Materials and MethodsThis chapter describes details of the experimental methods and apparatus usedduring the project and the methods used to analyse samples and water quality.All tests were completed at room temperature (22◦C).2.1 Preparation of synthetic feed waterTo account for the seasonal variability of NOM, a synthetic feed water was usedthroughout this research, which modeled natural surface water but was preparedusing a reference NOM material (Suwannee River NOM). Suwannee River NOM(SR-NOM) is a RO isolate that is used as a reference material for NOM research.SR-NOM can be further separated into Suwannee River Humic Acid (SRHA) andSuwannee River Fulvic Acid (SRFA) fractions.Her et al. (2002) analyzed the molecular weight distribution of SRHA andSRFA solutions using techniques combining UV-absorbance and DOC with chro-matography. In all cases, the signal response curve began at an apparent molec-ular weight (AMW) less than 30kDa (Her et al., 2002). The MWCO for UF mem-branes range from 1 - 500kDa (Crittenden et al., 2012b). Therefore, depending onmembrane pore size or MWCO, a UF membrane may be able to reject large MWfractions of NOM.23Synthetic surface water was prepared to model a surface water site in BritishColumbia, Canada. Table 2.1 outlines the parameters of interest and target con-centrations in the synthetic feed water used in these tests.Table 2.1: Target concentrations for SR-NOM synthetic feed waterParameter Target valuepH 7.07± 0.42UVA-254 0.153± 0.010Alkalinity, mg/L as CaCO3 76.8± 2.7DOC, mg/L C 5.80± 0.31Chloride, mg/L Cl– 2.94± 0.11Sulphate, mg/L SO –24 32.5± 1.3The synthetic feed water was prepared using deionized (DI) water (18mS/cm2)and stock solutions of SR-NOM, sodium sulphate, sodium chloride and sodiumbicarbonate. Synthetic feed was prepared in small batches as required for the dayof use (6L to 10L batches). A graduated vessel was partially lled with DI wa-ter, then the sulphate, chloride, and NOM stock solutions were added and mixed.The solution was topped up to the target volume and the pH in the vessel wasmeasured and recorded using a Hanna pH electrode. The sodium bicarbonatestock solution was then added and the solution was once again mixed. The nalpH was measured and recorded. All reagents used were ACS or reagent grade. Ifneeded, pH of the feed batch was adjusted to 7.07±0.42 using dilute sodium hy-droxide or sulphuric acid solutions. Feed water pH, DOC and UVA was measuredfor every feed batch prepared.2.1.1 Preparation of NOM stock solutionThis work evaluated the removal of soluble organic matter from surface water.A 400mg/L (as C) NOM stock solution was prepared to spike the synthetic waterwith a known concentration of dissolved organic carbon. IHSS Suwannee Rive24NOM RO isolate (lot 2R101N) was used as the dissolved organic carbon source.The stock was prepared in 500mL batches. 400mg of SR-NOM was accuratelyweighed and transferred to a 500mL volumetric ask. The ask was lled withdeionized water and shaken to fully suspend the material. The solution was thentransferred to a 1L Erlenmeyer ask and mixed using a magnetic stir plate atroom temperature for 24 hours. The stock was then ltered through a 0.45µmmembrane lter to ensure that only dissolved organic carbon was collected in thenal stock. Concentration of the NOM stock solution was conrmed using a TOCanalyzer (Shimadzu carbon analyzer, non-purgeable organic carbon method).2.2 Collection and properties of natural surfacewaterTable 2.2: Natural surface water propertiesParameter Reported valuepH 7.40± 0.13Turbidity, NTU 1.59± 0.64UVA-254 0.15± 0.01Alkalinity, mg/L as CaCO3 36.4± 6.0TOC, mg/L C 5.31± 0.64Chloride, mg/L Cl– 0.71± 0.02sulphate, mg/L SO –24 4.12± 0.22Nitrate, mg/L NO –3 0.56± 0.18To test the EC/UF process on natural surface water, feed water was collectedfrom a drinking water source in Middle River, British Columbia, Canada. Thesample was collected in a 20L jug and delivered to the laboratory within 48 hoursof collection. The sample was stored in the dark at 4◦C until required for test-ing. Testing was completed within one week of sample collection. Feed water25properties are listed in Table 2.2. Prior to the start of the test using this water,the water was allowed to warm to room temperature.2.3 Experimental apparatusThis research evaluated a hybrid electrocoagulation-ultraltration (EC/UF) tech-nology for the treatment of drinking water. The EC unit was operated as a con-tinuous ow system, to produce a composite volume of EC euent, which actedas the feed for the UF unit. Raw feed water was also ltered directly through theUF unit. A general schematic of the overall process is presented in Figures 2.1and 2.2. Stars in the gure indicate sample points throughout the process. Exper-iments evaluated permeate water quality and membrane fouling in these systemsduring ltration and evaluated the eect of cyclic backwash and air sparge.Figure 2.1: Process ow chart for EC/UF hybrid systemFigure 2.2: Process ow chart for UF ltration of raw water2.3.1 EC experimental apparatus and procedureThe removal of NOM using EC for drinking water has been extensively studied.In this project, EC conditions were xed to provide a UF feed water of similarcomposition. These conditions were selected based on those that were most suc-cessful in NOM removal according to previous work (McBeath, 2017). The ECconditions applied are outlined in Table 2.3. See Figure 2.3 for a schematic of the26Table 2.3: EC test conditionsParameter Settinginter-electrode gap, δ 2 mmCell ow rate, Q 1 LPMAnode material Cold-rolled steelCathode material Stainless Steel (304)Operating current, I 2 AOperating current density, i 6.42 A/m2Target metal loading, ML 35 mg/LEC unit. Additional details of EC reactor geometry and design are provided byMcBeath (2017).Prior to the start of each experiment the iron electrode was manually cleanedwith dilute sulphuric acid (0.657N) and scrubbed with a steel wool scrub pad. Thiswas to remove any ferric hydroxide from the surface of the anode and maximizecurrent eciency. The electrode was then rinsed with DI water and placed intothe EC apparatus with an inter-electrode gap of 2mm. A DC power supply wasused to drive the electrochemical reactions. It was operated in constant currentmode, with variable voltage. The continuous ow reactor was operated at 2Aand a ow rate of 1LPM to provide a targetML of 35mg/L Fe. Current eciencyand metal loading was conrmed through measurement of Total iron in the ECeuent sample (see Section 2.5.5). A peristaltic pump was used to push waterthrough the EC unit. The volumetric ow rate was measured and adjusted to1.0±0.1 LPM before the start of each EC experiment using a stopwatch and a 1Lgraduated cylinder.At the start of the EC test the system was operated until it reached steadystate and the voltage stabilized. Then, the EC euent composite volume wascollected in a single bucket. Once the desired volume of euent was produced,the process stopped and the euent composite was sampled. The solution wasmixed briey and allowed to occulate for approximately 10 minutes prior to27Figure 2.3: Schematic for EC systemsampling to ensure all of the iron particles were suspended. A 50mL sample wasthen collected and ltered through a 0.45µm PVDF syringe lter to measure theDOC concentration in the euent. This mimics direct ltration of EC euentin a full-scale process. McBeath (2017) was unable to correlate strong evidenceof occulation time with NOM removal while using the same system and it wastherefore assumed that the EC euent could be treated as having constant com-position during the UF stage of the test (McBeath, 2017).2.3.2 UF experimental apparatus and procedureTwo dierent membrane types were used to meet the objectives of this project.The rst was a ceramic disk dead-end membrane and the second was a hollow-28bre polymeric membrane. Important properties of each membrane type areoutlined in Table 2.4.Table 2.4: UF membrane propertiesParameter Type 1 Type 2Material type Ceramic PolymericMembrane material ZrO2 PVDFMembrane shape Disk, at sheet Hollow-breConguration dead-end, submerged dead-end, submergedNominal pore size (or MWCO) (300kDa) 0.04 µmNominal ltration ux, J 50 LMH 50 LMHNominal surface area 13.2 cm2 34 cm2Volume of membrane module 3.3 cm3 622 cm3The nominal pore size or MWCO is listed as provided by the membrane man-ufacturer. Both membranes were tested using continuous ltration and cyclicbackwash, using an automated timer system. The eect of air sparge was alsoevaluated using UF membrane Type 2. Two membrane units (noted as A or B dur-ing testing) were operated in parallel to allow for duplicate tests for each mem-brane type. Due to the dierence in geometry for the two lters, the systems arenot identical, as evident by the nominal ltration surface area and volume of themembrane module. See Figures 2.4 and 2.5 for schematic diagrams of the two UFmembrane units.UF Membrane Type 1During ltration tests using UF membrane Type 1, feed water was held in a sep-arate feed tank, which was gently mixed (60rpm) throughout the duration of theexperiment to ensure consistent composition of water being fed to the mem-brane. This was meant to model direct ltration, without a sedimentation stepbefore the membrane.29Figure 2.4: Process schematic for UF membrane Type 1 (ceramic disk)Filtration using UF membrane Type 1 required re-purposing of a CeramicDisk Membrane lter holder designed for pressure ltration using lab-scale ce-ramic disk lters (47mm, Tami Industries). This involved removal of the pressuresystem and applying suction to the permeate side of the membrane through aperistaltic pump. Through these changes, the system was modied to model adead-end, at sheet, submerged membrane system.UF membrane Type 1 was housed in a module with a 47mm diameter andsealed with a rubber o-ring. The available surface area for the membrane was13.2cm2. The average available volume in the housing above the membrane was3.3cm3. The ow rate of solution into the lter was set at 50LMH, which cor-responds to a volumetric ow rate of 1.1ml/min. The volumetric ow rate wasmeasured at the beginning, middle and end of each UF test.Filtration was driven by a peristaltic pump applying suction to the mem-brane module. Backwash stages used a second peristaltic pump, owing in theopposite direction. The pumps are labeled as P1 and P2 in Figure 2.4. An au-30tomated timer controlled the timing of ltration and backwash; solenoid valvescontrolled the direction of ow. During ltration, V1 was open and P1 was ON.During backwash, V2 was open and P2 was ON. Backwash reject was collectedin a separate container. Trans-membrane pressure (TMP) was measured using apressure transducer connected to datalogger (labeled as PT-A/B on Figure 2.4). Apressure gauge was also included in the line to allow for visual conrmation ofsuction and ow (labeled as PG-A/B on Figure 2.4). UF permeate samples werecollected from the euent line from P1. Ten minute and ve minute backwashtimes were tested. These times are longer than what would typically be used ina full scale process but the extended backwash was selected because it was ex-pected that the conguration of UF membrane Type 1 would require extra timeto completely ush the system with backwash water, due to the shape and ori-entation of the membrane module.Brand new 300kDa ceramic membranes were cleaned using an adaptationof the manufacturers recommended cleaning procedure. The adapted procedureis as follows: soak the membrane in 0.5N NaOH solution at 70-80◦C for 1 hourand rinse the membrane and system with DI until pH runs neutral. Then soakthe membrane with 75% w/w phosphoric acid solution at 40-50◦C and rinse themembrane and system with DI until pH runs neutral. This was done to removeany residual manufacturing residues.At the end of each UF test, the membrane was removed from the module,rinsed with DI water and gently wiped with a soft tissue to remove the accumu-lated cake on the surface of the membrane. The membrane was then re-installedinto the module and underwent a chemical clean that is similar to proceduresfollowed in drinking water treatment plants.The rst step in the cleaning cycle was a 1 hour backwash and soak using a400mg/L NaOCl and 0.1N NaOH solution. The system was then rinsed with DIwater until the wash was neutral. Then the system underwent a 1 hour back-wash and soak using a 1% (w/v) citric acid solution and rinsed with DI until therinse ran neutral. Between experiments the system was left at room temperature,31primed with DI water.Prior to the start of each UF test, the system was rinsed with DI water for aminimum of 2 hours. The TMP was logged during this time and dened at theclean water TMP (∆P◦, used to calculate the initial membrane resistance, denedas Rm).UF Membrane Type 2Figure 2.5: Process schematic for UF membrane Type 2 (hollow-bre)UF membrane Type 2 used a hollow-bre membrane module, constructedfrom commercially available membranes (0.04µm, Suez ZeeWeed membranes,PVDF, non-ionic and hydrophilic). Each module was composed of three 19cmlengths of the hollow bre membranes, with the top end of the module open forsuction and the bottom end sealed. During ltration the module was submergedin a graduated cylinder. Feed solution was pumped into the cylinder at a rateslightly faster than the ltration suction rate to ensure that the solution level(and therefore the hydraulic head) remained constant throughout the test. Thisis identied as P3 in Figure 2.5.32UF feed water was held in a 20L feed tank, which held graduated cylindersthat contained the submerged membrane modules. The feed in the large tank wasmixed throughout the test (at 120 rpm) to ensure all iron ocs were suspendedand to provide consistent composition of water being fed to the membrane. Dueto the geometry of the tank, a higher mixing speed was required to suspend allparticles compared to the mixing speed used to test UF membrane Type 1.The available surface area of each module was 34cm2. The average availablevolume in the cylinder housing the module was 622cm3. The ow rate of solutioninto the lter was set at 50LMH, which corresponds to a volumetric ow rate of2.8ml/min. The volumetric ow rate was measured at the beginning, middle andend of each UF test.Cyclic ltration and backwash cycles were controlled using an automatedtimer and peristaltic pumps (P1 and P2 in Figure 2.5). Due to the limited volumeof feed water available and simplicity of the UF system design, backwash rejectwent directly into the feed cylinder and was recombined with the fresh feed.Solenoid valves were not used.Trans-membrane pressure was measured using a pressure transducer con-nected to datalogger (labeled as PT-A/B on Figure 2.5). A pressure gauge wasalso included in the line to allow for visual conrmation of suction and ow (la-beled as PG-A/B on Figure 2.5). UF permeate samples were collected from theeuent line from P1.During air sparging tests, an air line was submerged to the bottom of thegraduated cylinder. Air was sparged at a pressure greater than 30cm H2O, whichis the height of the column of water. This value was not measured and the airow rate was set at a pressure high enough to overcome the hydraulic headin the cylinder to provide consistent, vigorous air ow to both cylinders. Thecompressed air used was provided from the building air system.At the end of each UF test, the hollow-bre modules were removed from thesystem, and rinsed with DI water and gently wiped with a soft tissue to removeany accumulated cake on the surface of the membrane. The modules were then33re-installed and underwent a chemical clean that is similar to procedures fol-lowed in drinking water treatment plants.The rst step in the cleaning cycle was a 1 hour backwash and soak usinga 50mg/L NaOCl solution. The system was then rinsed with DI water until therinse ran neutral. Then the system underwent a 1 hour backwash and soak usinga 1% (w/v) citric acid solution and rinsed with DI until the rinse ran neutral.Between experiments the system was left at room temperature, primed with DIwater.Prior to the start of each UF test, the system was rinsed with 50mg/L NaOClsolution for 20 minutes and then washed with DI water for a minimum of 20minutes. The TMP was logged during the DI wash and was dened as the cleanwater TMP (∆P◦, used to calculate the initial membrane resistance, Rm).2.4 Experimental matrixA series of experiments were designed to evaluate membrane fouling in EC pre-treated water and compare it to water that had not been pre-treated (called rawwater in this text). Two dierent membrane types were studied (See Section 2.3.2for details on the UF setup). See Table 2.5 for the conditions tested.The tests were designed to evaluate the performance in terms of water qual-ity and resistance accumulation with or without the use of EC pre-treatment,backwash or air-sparge. In the table, Xdenotes with treatment applied, × de-notes without treatment applied. Additional tests on the eect of lter cycleduration and backwash duration were also completed for UF membrane Type 1,indicated by a X* symbol. Air sparge was only available using UF membraneType 2. Note: this is not an exhaustive matrix and it could easily be expandedfor future research.The results from these experiments were reviewed and one setting from thismatrix was selected and applied to natural surface water collected from MiddleRiver, British Columbia. The setting selected was:34Table 2.5: Experimental matrix: Synthetic feed water using SR-NOMEC pre-treatment Backwash Air spargeType 1 X × n/aType 1 X X* n/aType 1 × × n/aType 1 × X* n/aType 2 X × ×Type 2 X X ×Type 2 × × ×Type 2 X × XType 2 X X X• UF membrane Type 2;X With EC pre-treatment;X With backwash;× Without air sparge.Chapter 5 presents the results from this experiment.2.5 Analytical methodsWater quality and membrane fouling were measured throughout this work. Wa-ter quality parameters include: Dissolved Organic Carbon (DOC), UV-absorbance(UVA-254) and total and dissolved iron. Membrane fouling was evaluated bymeasuring and logging the TMP and using it to calculate the accumulated mem-brane resistance, Rf . These methods are discussed in the following sections.352.5.1 Total and dissolved organic carbonIn this work, DOC is synonymous with TOC (Total Organic Carbon) as all sam-ples were pre-ltered to 0.45µm or less, prior to analysis.DOC was analysed using a Shimadzu ASI-V Total Organic Carbon analyser.Reagent blanks, calibration check standards and duplicate samples were run witheach batch of samples to ensure reliable results. The non-purgable organic car-bon method was used for these samples. This method allows for determinationof organic carbon without interference from carbonate species that were part ofthe feed solution.In the method, the sample is acidied and sparged with compressed ultra-pure air to convert inorganic carbon to CO2 which is then purged from the sam-ple. The sample then undergoes combustion over a catalyst heated to 680◦C,oxidizing all non-purgeable organic carbon to CO2, which is then detected usingan NDIR sensor.TOC Method Detection Limit and AccuracyThe method detection limit (MDL) and limit of quantication (LOQ) were de-termined following the procedure described in (US EPA Oce of Water, 2016).Following these guidelines, the LOQ of TOC analysis by the Shimadzu TOC in-strument was determined to be 0.20mg/L C. The MDL of the instrument wasdetermined to be 0.21mg/L C.Standard solutions at dierent known concentrations were also analyzed atdierent occasions to determine the precision and accuracy of the TOC method.Table 2.6 outlines the results of these samples. These results indicate that theTOC analysis is accurate at concentrations of 2.5mg/L DOC or higher, whichis the working range for this research. At the higher concentrations (2.5mg/Land 5.00mg/L) there is no signicant dierence between the measured value andthe true value of the TOC standard, as calculated based on 95% condence in-tervals. Details on how the condence intervals were calculated is presented inSection 2.6.36Table 2.6: TOC instrument accuracy. Measured value is reported with the calcu-lated 95% condence intervals. N is the number of samples,RPD is the RelativePercent Dierence between the measured value and the true value of the stan-dard.True value Measured value N RPDmg/L DOC mg/L DOC0.20 0.23± 0.01 9 6.8%0.50 0.52± 0.01 9 2.2%2.50 2.43± 0.15 11 1.3%5.00 5.04± 0.12 61 0.4%2.5.2 UVA-254UV-absorbance was analyzed using a Cary UV-Vis spectrophotometer, at a wave-length of 254nm. A 1cm quartz cell was used to hold each sample during mea-surement. Method blanks and duplicate samples were run with each batch of 10samples to ensure reliable results.The estimated MDL of this method was determined to be 0.0040, which isthe value of the 99th percentile of a set of 100 method blank samples analysed bythe UV spectrophotometer at 254nm. A detailed MDL and LOQ study was notperformed, as UVA-254 was reported as an absolute value and was not used tocalculate the concentration or quantify the amount of a substance in the sample.2.5.3 Specic UV-absorbanceSpecic UV-absorbance (SUVA) is a method of relating the ratio of UV-absorbingcompounds to the total organic carbon concentration in a sample. The followingequation is used to calculate SUVA:SUV A =UV A254DOC× 100 (2.1)SUVA is useful because provides an indication of the aromatic character of the37NOM and is correlated to the DBP formation potential of water (Hua et al., 2015).The uncertainty of SUVA measurements were calculated based on propagationof error of UVA-254 and DOC measurements, as discussed in Section 2.6.2.5.4 Apparent Molecular Weight by HPSECThe apparent molecular weight (AMW) of NOM was determined using a Wa-ters 2695 XE Separations Module, coupled with a Waters 2487 dual λ absorbancedetector at 260nm, as described by Dubrawski and Mohseni (2013). The carriersolution was a phosphate buer of 0.01M KH2PO4, 0.01M K2HPO4 and 0.06M NaCl(Certied A.C.S. reagent, Fisher Scientic). The column ow rate was 0.7mL/min.The chromatogram produced by the instrument gives the UVA in relation toelution time. The corresponding molecular weight of the compounds were deter-mined based on a calibration performed using a set of polysulphonate standards,following a procedure described in literature (Sarathy & Mohseni, M., 2007). Thefollowing equation was obtained from the calibration and used to relate elutiontime to the apparent molecular weight of NOM in the sample:log(MW ) = −0.2857t+ 6.9205 (2.2)where t is the elution time (in minutes). The R2 value for this equation is 0.9911.The calibration data used to develop this equation is historical data and acalibration was not performed with each sample batch. Therefore, the calculatedAMW is only an estimate of the MWD of NOM samples, and is not expectedto represent exact values. Nevertheless, HPSEC chromatograms are useful inhelping quantify the fractions of NOM removed from or present in a sample. Togain further insight to the fractionation of NOM during treatment, an analysisof UV chromatogram peaks was performed using Systat Peakt®v4.12.Peakt®was applied to identify dierent AMW fractions using a deconvolu-tion method assuming symmetrical Gaussian peaks. Five AMW fractions wereidentied with the purpose of quantifying the change in peak area during treat-38ment. In a HPSEC chromatogram, the low MW compounds (such as proteinbuilding blocks) elute rst, followed by low MW humics, non-polar humics, andthen much larger MW inorganic colloids and biological residues (Chow et al.,2008). The assumed NOM type for each fraction was labeled based on the peakassignment made by Chow et al., in a dierent body of work (2008). The charac-teristics and properties of the fractions were not investigated further. Table 2.7presents the identied AMW range for the dierent NOM fractions assumed inthis research.Table 2.7: Identied NOM fractions, based on HPSEC resultsFraction AMW range, Da Assumed NOM typeF1 <500 Building blocksF2 500-700 Low MW humicsF3 700-900 Low MW humicsF4 900-1100 Low MW humicsF5 >1100 Non-polar humic substances,inorganic colloids,biological residuesThese dened fractions are qualitative. The peaks identied may not repre-sent the true nature of elution of NOM fractions, as they may not be symmetrical,Gaussian peaks. NOM varies in chemical and physical properties, and the inter-actions of each type of NOM with the HPSEC column may vary. Zhou et al.(2000) suggest that UVA determination of MWD can vary from 10-20%, there-fore the identied peaks using Peakt®are most useful in identifying a change, ifany, in the MWD of NOM in samples, and not measure accurate concentrationsof NOM.392.5.5 Total and dissolved ironAnalysis of iron concentration was performed using ICP-OES. Calibration stan-dards were prepared in the same matrix as the sample to be analyzed within theexpected concentration range of the samples (0.100 - 100.0mg/L Fe). Sampleswere acidied using nitric acid to a nal concentration of 2% HNO3 and wereheld at room temperature for a minimum of 24 hours prior to analysis to ensurethat all iron was solubilized. Samples with anticipated high iron concentrationswere diluted with acidied matrix blank diluent and held at room temperatureuntil all iron was solubilized.2.5.6 Analysis and logging of trans-membrane pressureTrans-membrane pressure (TMP) data was logged using a pressure transducerthat translated the pressure as a signal (in V). Two dierent pressure transducerswere used when running duplicate tests in parallel. These transducers operatedin two dierent output ranges, therefore, the signal from one had to be convertedto match the signal from the other.A calibration was performed, where the two transducers were connected inseries and pressure was applied. The signal from PXR-A was matched to PXR-Band the following equation was obtained:PXRB as A = PXRA × 0.4005 + 0.4025 (2.3)TheR2 value for this data was 0.9993. These signals were then converted to unitsof pressure (psi) used to calculated the system TMP (∆Pt). Pressure TransducerA had a maximum voltage of 2.5V, corresponding to +15 psi. The minimum volt-age was 0.0V, which corresponded to -15psi. Using these values, the followingequation was obtained to convert signal output values to psi.∆Pt = PXRA × 29.9550− 27.0580 (2.4)40The TMP data was logged using a HOBO datalogger at 10 second intervalsthroughout the UF tests. This data was then processed using a Python script toconvert the logger output signal (in V) to pressure values (in kPa or psi).To evaluate the accumulated fouling throughout the test, the ∆Pt term wasconverted to the accumulated resistance term, Rf , as described in Section 1.2.4,using equation 1.13, repeated here for clarity:Rf =∆Pt −∆P◦µJ(2.5)The dynamic viscosity, µ, was assumed to stay constant at 0.9544mPa.s, cor-responding to a temperature of 22 ◦C. For the purpose of calculations, the mem-brane ux, J , was assumed to stay constant at the rate measured at the beginningof the test. Using these values, Rf was calculated as a function of time. The timevariable was then translated to the more useful term, throughput. Throughputallows for easy comparison between dierent membrane types and the variationin ow rate to account for changes in surface area at constant ux.Throughput is calculated by calculating the volume ltered as a function oftime in relation to the surface area (SA) of the membrane.Throughput = Q× tSA= J × t (2.6)where Q is the volumetric ow rate and t is time.Linear regression was performed on the dierent datasets to evaluate the rateof resistance accumulation for each test. The calculated rate of accumulation wasexpressed as the slope, ± the corresponding 95% condence interval, based onthe calculated standard deviation of the slope.2.5.7 Analysis of resistance recovery from backwashingBackwashing eciency can be quantied by measuring the resistance recoverybetween ltration cycles. The following formula was applied to complete this41calculation:Rr, i = 1− Rend, i−1 −Rstart, iRend, i−1(2.7)where Rr, i is the resistance recovery from backwash at the start of the i-th cycle,Rend, i−1 is the fouling resistance at the end of the previous cycle andRstart, i is theresistance measured at the begining of the current cycle.2.6 Data analysis and condence intervalsDetermination of condence intervals is an important part of data analysis, as ithelps identify which test conditions cause signicant dierences in test results.95% condence intervals were calculated for water quality results based on theassumption of a Gaussian distribution of random error for duplicate tests. 2-sidedt-values were used in the calculations, based on the number of samples (N ) anddegrees of freedom (N − 1). The general formula used for calculating the 95%condence interval, e, is:e =t95% × s√N(2.8)where s is the standard deviation of the replicate measurements and N is thenumber of measurements. t95% is the 2-sided t-value for 95% condence. Thepercent relative error for each measured value is:%e =eMeasured Value (2.9)In this work, every experiment was completed in duplicate and the resultsfor water quality measurements were determined by calculating the condenceinterval for two duplicate tests. Error bars in gures were calculated using 2-tailed t-values for 95% condence limits, unless otherwise noted.Some reported values, such as Rf or the normalized sample concentrations,C/C◦, are calculated values and were not based on direct analysis. In these cases,42propagation of error had to be considered to determine the error associated withthe calculated value. The following formulas were applied to determine the prop-agation of error when simple mathematical operations were applied. These stepsare outlined in reference material (Harris, 2007a).Propagation of error when the operation is addition or subtraction (ie, y = x1 +x2):ey =√(ex1)2 + (ex2)2 (2.10)Propagation of error when the operation is multiplication or division, such as cal-culation of normalized feed concentration or SUVA (ie, w =x1x2):%ew =√(%ex1)2 + (%ex2)2 (2.11)Calculation of error from relative percent error:ew = %ew × w (2.12)Propagation of error works under the assumption that systematic error hasbeen detected and corrected. The 95% condence intervals were calculated foreach replicate test and sample, and the error of the ratio was calculated to provideerror bars for graphs representing this data.To determine the error of the calculated accumulated resistance, Rf (fromequation 1.13), equations 2.11 and 2.10 were applied. Error bars on the gurespresenting Rf data represent the calculation of Rf error of the measured maxi-mum and minimum TMP of parallel systems.Linear regression was performed to determine the slope and rate of changefor resistance increase with respect to throughput. The standard deviation of theslope (sm) was calculated following steps provided in a reference text (Harris,2007b). The 95% condence interval was calculated using equation 2.8, withthe 2-sided t-value at 95% condence interval and N − 2 degrees of freedom,where N is the number of data points used in the regression. Calculated slopes43were deemed signicantly dierent if the determined condence intervals didnot overlap. It is important to note that the calculated slopes did not account forany uncertainty in the x or y range.44Chapter 3Results and discussion ofsynthetic water tests:UF membrane Type 1Two types of UF membranes were evaluated in this work, however because theyvary in material, geometry and pore size (MWCO), they will be discussed sepa-rately. This chapter presents water quality and membrane fouling results for UFmembrane Type 1.As described in Section 2.3.2, UF membrane Type 1 is a 300kDa, ceramic diskmembrane with dead-end ow conguration. The disk is housed in a modulewith a working volume of 3.3cm3 and suction is applied to drive ltration andmodel a submerged, at-sheet membrane system. The schematic and descriptionof this setup is available in Section 2.3.2. The eect of backwash duration andltration cycle duration was investigated, using EC pre-treated synthetic feedwater.During the test, the ltration ux was kept constant at 50 ± 2.5 LMH. Theinitial grab sample began after 90 minutes and nished once the required vol-ume was collected, which is equivalent to a throughput of 60-100L/m2. A sec-ond sample was collected after approximately 18-20 hours of ltration, which45is equivalent to a throughput of approximately 650-1100L/m2 (depending on themeasured ow rate of the test and the time of the sample).3.1 Filtration of raw waterRaw water ltration tests were completed to set a baseline for ltration of thefeed water without EC pre-treatment. Tests with continual ltration and a 60minute ltration cycle with a 10 minute backwash were performed. The MWCOof the membrane used in this test is 300kDa. Therefore, given the SR-NOM MWDupper limit of 30kDa (as discussed in Section 2.1), removal of DOC by the mem-brane was not expected.During the test, DOC, UVA-254 and Fe were measured in the feed and perme-ate samples. Table 3.1 presents the feed and permeate water quality for the UFmembrane Type 1 continual ltration test using raw water. Unfortunately, thefeed sample for the raw water ltration test with backwash was contaminatedwith EC euent from other tests and water quality results for this test are notreported.Table 3.1: Feed and permeate water quality results for raw water ltration usingUF membrane Type 1, without backwash. Reported as average± 95% condenceinterval.DOC UVA-254 Fe - totalmg/L C - mg/L FeRaw feed 5.50± 0.20 0.228± 0.005 < 0.10Initial sample @75-100 L/m2 5.54± 0.09 0.231± 0.010 < 0.10Final sample @900-1100 L/m2 5.52± 0.18 0.230± 0.003 < 0.10These results indicate that there is no signicant dierence between the feedand permeate samples in terms of the measured water quality parameters. Thissupports the assumption that NOM will not be removed by the membrane and46any observed NOM removal in UF tests using EC pre-treated feed water is dueto the EC capture of NOM and not this membrane.The TMP was also monitored throughout the tests to determine whether theSR-NOM feed water would cause any fouling. It was hypothesized that foulingwould not be observed, as the MWCO of the membrane is much larger than theexpected MWD of SR-NOM and it was assumed that membrane ltration waspurely a physical process.The accumulated membrane resistance was calculated as described in Sec-tion 2.5.6. The fouling resistance during the raw SR-NOM water ltration testsis presented in Figure 3.1. The rate of fouling resistance accumulation with re-spect to throughput is referred to as the ‘fouling rate’ in this text.Figure 3.1: UF membrane Type 1 fouling resistance accumulation for raw wa-ter ltration tests with and without backwash. Error bars represent calculatedmaximum and minimum Rf for duplicate tests.Despite the overlap of the error bars of the calculated Rf values, these re-sults indicate that resistance is increasing during ltration of raw water for bothtests, at dierent rates. Although fouling was not expected during ltration of47the synthetic feed water, the apparent resistance may be due to the fraction ofdissolved or ‘soluble’ NOM that passed through the 0.45µm lter during solutionpreparation, but was rejected by the 300kDa membrane.Examining the results of these two tests, backwashing appears to reduce theoverall fouling rate. Linear regression analysis was performed on the resistanceresults to determine if this dierence was signicant. Determination of the slopein the linear region of ltration provides information on the relative fouling rateto compare the two experiments. For the test with backwash, regression wasperformed with data points collected at the end of each ltration cycle, as this isthe point at which maximum fouling would occur. Table 3.2 presents the resultsfrom this analysis.Table 3.2: Linear regression of raw water ltration using UF membrane Type 1,with 95% condence intervals calculated based on the slopeTest setting slope, m−1L.m−2R2without backwash (6.18± 0.30)× 108 0.9644with backwash (2.37± 0.63)× 108 0.7769These results suggest that the backwash is reducing the overall fouling ratefor the raw water ltration tests with UF membrane Type 1, as there is no over-lap of condence intervals between the calculated slopes. However, as noted inTable 3.1, the dierence in DOC and UVA-254 between feed and permeate sam-ples is not signicant, which implies that despite this increase in resistance, the300kDa membrane does not signicantly remove NOM from feed water.Given the R2 value for the two test conditions and the spread of data pointsfrom the parallel tests, linear regression may not be an appropriate method toevaluate the fouling rate for these tests conditions. A true evaluation of the foul-ing rate for the tests with backwash would be to identify the fouling rate for eachindividual ltration cycle, instead of collecting data points at the end of each cy-48cle. However, the purpose of this exercise was to evaluate the overall increasein membrane resistance as water is ltered. A detailed evaluation of the foulingmechanism was not within the scope of this research.3.2 Filtration of EC treated waterUF tests were completed with EC pre-treated feed water to evaluate the eectof ltration cycle duration and backwash duration on membrane performancein terms of permeate water quality and accumulation of membrane resistance.Results of EC pre-treated feed water preparation are presented in Section 3.2.1and UF test results are presented in Sections 3.2.2 and 3.2.3.3.2.1 Preparation of EC pre-treated feed waterThe EC performance for NOM removal depends on a number of factors such asmetal loading, current density and concentration of dissolved oxygen. To tryand mitigate variability of results, EC experimental conditions were xed, andselected based on results observed by McBeath (2017).This report will not provide a detailed discussion of NOM removal results ormechanism as it was not part of the scope of the project. Nevertheless, under-standing the EC euent water quality is valuable as this water is the incomingfeed for the UF system. Table 3.3 outlines the EC test performance and results fortreatment of SR-NOM synthetic surface water. EC euent samples were lteredthrough a 0.45µm lter after approximately 10 minutes of occulation.These test conditions resulted in an average DOC and UVA-254 reduction of33 ± 4% and 57 ± 8% respectively. The decrease in SUVA indicates that thearomatic fraction is preferentially removed by EC, which is similar to results re-ported in literature for EC and CC treatment using iron for NOM removal (Baggaet al., 2008). The average metal loading for all EC tests was 34.2 ± 1.3mg/L Fe,corresponding to a average current eciency of 95%. In all cases an increase49Table 3.3: EC test results, with an operating current density of 6.43mA/cm2 andtarget metal loading of 35mg/L Fe.Feed EuentDissolved Organic Carbon, mg/L C 5.50± 0.20 3.66± 0.37UV absorbance at 254 nm 0.225± 0.004 0.097± 0.014pH 7.32± 0.17 8.54± 0.24Total iron, mg/L Fe <0.10 34.2± 1.3Dissolved iron (0.45µm) <0.10 0.25± 0.10SUVA 4.08± 0.17 2.65± 0.46in pH was observed, which was expected given the anticipated reaction on thecathode.HPSEC analysis was also performed to determine the AMW and change inMWD of NOM in the EC pre-treated water. Results of HPSEC analysis are pre-sented in Figure 3.2.Figure 3.2: AMW of untreated water and EC euent for tests using syntheticSR-NOM feed water50These results appear to agree with results reported in literature that the largerMW faction of NOM is removed using EC (Dubrawski et al., 2013). Table 3.4presents the percent removal of the dierent fractions identied using Peakt®.Results indicate that increasing percent removal occurs with increasing AMW,Table 3.4: Average percent (%) removal of dierent AMW fractions during l-tration of SR-NOM feed, as calculated through deconvolution of peaks usingPeakFit®, reported with 95% condence intervals (N = 7)Fraction AMW range, Da Average % removalF1 <500 (13± 18)%F2 500-700 (37± 4)%F3 700-900 (53± 5)%F4 900-1100 (71± 5)%F5 >1100 (91± 3)%with over 90% removal of the higher AMW fraction (F5) and only (13± 18)% ofthe low MW fraction (F1).An analysis of the particle size distribution (PSD) of the EC euent was alsocompleted, to conrm whether UF is an appropriate method to remove the ocsproduced. A sample of EC euent was measured for PSD directly, as well as asample that was allowed to settle in a beaker for 22 hours. Table 3.5 presents thewater quality for the two samples. The reported values for Dissolved iron andDissolved Organic Carbon are based on ltration of the samples with a 0.45µmsyringe lter. As this was only an exploratory test, it was not performed in du-plicate and strong conclusions cannot be made from the resulting water qualitydata. Nevertheless, the high concentration of total iron and particle concentra-tion by % volume in the Settled supernatant sample suggests poor settleability ofthe EC produced ocs. Although both the total iron and particle concentrationdecreased after 22 hours of settling, over 20mg/L total Fe and a detectable levelof particles remained in the supernatant sample. This is unacceptable for treateddrinking water.51Table 3.5: Water quality results for EC euent settling test of SR-NOM feed water.EC euent Settled supernatant (22h)Total iron, mg/L Fe 32.4 21.1Dissolved iron, mg/L Fe) <0.10 <0.10Dissolved Organic Carbon, mg/L C 2.65 3.30UV absorbance at 254 nm 0.082 0.117pH 9.06 8.88SUVA 3.09 3.53Particle concentration by % volume 0.086 0.019Furthermore, the observed increase in DOC and UVA-254 in the ltered sam-ples may indicate possible changes to oc stability and ltered water quality withrespect to time. However, further analysis would be required to understand ocformation and stability for a specic water sample, as these behaviours vary de-pending on the water matrix (Crittenden et al., 2012a). EC was also applied tonatural surface water, which had a dierent matrix than the synthetic SR-NOMfeed water. The results of this test are discussed in Chapter 5.Particle size distribution analysis provides insight into what happened to theocs produced from EC. The PSD results for the EC euent and the Settled su-pernatant are presented in Figure 3.3. The PSD results are reported as the %volume of each fraction, normalized to the particle concentration by % volume.Therefore, although the relative peak height of the supernatant sample is higherthan that of the non-settled EC euent sample, the total concentration is stilllower. The distribution reported through the PSD analysis suggests that onlyparticles greater than 724µm settled out of solution. These results indicate thatgiven these test conditions, settling is insucient as means of solid-liquid sepa-ration of EC treated water.A bi-modal distribution of particle size diameters is also observed. This maybe due to the breakup and regrowth of occulated aggregates during the PSDanalysis (Maguire, 2009). The aggregate particle size during ltration may be52Figure 3.3: Particle Size Distribution results of EC euent and settled super-natant reported as % volume in the sample analyzed. The total particle concen-tration by % volume in EC euent and settled supernatant samples was 0.0864and 0.0193 respectively.larger than the diameter detected, which is useful when combining EC pre-treatmentwith membrane ltration.Furthermore, all particles detected were reported as greater than 2µm andless than 2200µm in diameter. Therefore, the membranes selected for this re-search (with MWCO or pore size of 300kDa and 0.04µm) are appropriate, as thesmallest detected particle size is around two orders of magnitude greater thanthe pore size of the membranes used.3.2.2 Membrane fouling during ltration of EC pre-treatedfeedUF tests were completed with and without backwash and at dierent ltrationcycle and backwash cycle durations. The test without backwash was used to de-termine the upper limit of fouling during ltration of EC pre-treated feed. Threedierent backwash and ltration cycle duration settings were tested, as denedin Table 2.5. The remainder of this section will discuss membrane fouling from53these tests. Permeate water quality results will be discussed in Section 3.2.3.Figure 3.4 presents a snapshot of the accumulated resistance data, Rf , forthese tests between 200-400L/m2. Each ltration cycle is connected by a line,with points plotted at the beginning, middle and end of each cycle. The continualltration tests (without backwash) is one continuous line.(a) Without backwash (b) 30 min ltration & 5 min backwash(c) 60 min ltration & 5 min backwash (d) 60 min ltration & 10 min backwashFigure 3.4: Snapshot of resistance accumulation for ltration of EC pre-treatedfeed using UF membrane Type 1, from 200-400L/m2. Each ltration cycle is rep-resented by its own line. Error bars represent calculated maximum and minimumRf for duplicate tests.In the test without backwash (Figure 3.4a), accumulated resistance appears toincrease linearly with increasing throughput, indicating that the dominant foul-54ing mechanism is through cake formation. Resistance accumulation rates for allfour tests appear to be of the same magnitude. However, the absolute resistanceaccumulation does not appear to dier signicantly between the dierent testswith backwash. Linear regression and analysis of the overall fouling rates pro-vides more information as to whether there is a signicant dierence in overallfouling rate between the dierent treatments.For the tests that included backwash, linear regression was performed on datapoints collected at the beginning, middle and end of each ltration cycle. Anal-ysis of the variance was performed on these results to compare the calculatedslopes. Results from this analysis indicate that there is no signicant dierencebetween the calculated overall rates of fouling, regardless of the point at whichthe slope was measured (ie. at the beginning, middle, or end of the ltration cy-cle. ANOVA results for this comparison are: Fcalc = 0.21, p = 0.81; Fcrit = 5.15,α = 0.05. In terms of practical applications of this data, the data point collectedat the end of the ltration cycle provides useful information for determining themaximum rate of resistance accumulation. The linear regression results, calcu-lated based on the accumulated resistance at the end of the ltration cycle arepresented in Table 3.6, along with the results from the test without backwash,for comparison. These results indicate that membrane ltration with backwashTable 3.6: Linear regression of EC pre-treated SR-NOM water ltration usingUF membrane Type 1, presented with 95% condence intervals. For tests withbackwash, linear regression was performed on data points collected at the endof each ltration cycle.Test setting slope, m−1L.m−2R2without backwash (5.11± 0.05)× 109 0.998730 min ltration & 5 min backwash (2.57± 0.15)× 109 0.966660 min ltration & 5 min backwash (1.67± 0.27)× 109 0.896860 min ltration & 10 min backwash (3.96± 0.22)× 109 0.973955using UF membrane Type 1 is able to reduce the total rate of resistance accumula-tion, as every test including backwash had a signicantly lower slope comparedto the test without backwash. Therefore, we can conclude that cyclic backwashsignicantly reduces the overall fouling rate during ltration of EC pre-treatedwater using a dead-end, ceramic membrane. The overall fouling rate was reducedby (23± 4)%, (67± 5)%, and (50± 5)% for the 30-5, 60-5 and 60-10 ltration andbackwash cycle settings, respectively.The test with 60 min ltration cycle and 10 min backwash had a higher foul-ing rate than the 30-5 and 60-5 tests. This may be due to the initial conditions ofthe test; the membrane module contained an air bubble for the rst 4 hours ofltration, which resulted in a higher initial TMP at the start of the test. Perhapsthis high TMP caused a change within the cake, making it more compact andcause increased overall resistance. This work combined iron electrocoagulation(EC) and ultraltration (UF) to treat synthetic and natural surface waters to re-move Natural Organic Matter (NOM). Fixed EC conditions were applied to thefeed water in a continuous ow EC reactor, at a ow rate of 1 LPM and an appliedcurrent of 2 A. These test conditions resulted in an average DOC and UVA-254reduction of 33± 4% and 57± 8% respectively for the synthetic feed water.The EC euent acted as the feed water for the EC/UF systems. Two UF mem-brane types were tested: (1) a 300kDa ceramic disk membrane; and, (2) a 0.04µmPVDF hollow-bre membrane. Both systems were operated at a constant ux of50 LMH. Periodic backwash cycles were applied to evaluate the eect of back-washing on the overall membrane fouling rates. Backwashing with a 30 minuteltration cycle and 5 minute backwash reduced the fouling rate by (50±3)% and(2±6)% in the ceramic and PVDF membranes, respectively. Applying backwashto the EC/UF test with natural feed water resulted in a (95 ± 0.5)% lower foul-ing rate compared to the EC/UF test with synthetic feed water test under thesame conditions. Therefore, water composition has a signicant eect on themembrane fouling rates of the EC/UF system.Air sparging was also applied to the PVDF membrane system and resulted in56reducing the overall fouling rate by (64± 2)%. Comparatively, the combinationof air sparging and backwashing reduced the overall fouling rate by (98± 1)%.The EC/UF test with air sparging and backwash also resulted in additional NOMremoval of the UF permeate. Air sparging has never been combined with anEC/UF system and these are promising results in the quest to develop a EC/UFdrinking water treatment system for small and remote communities. This mayalso be a trait of EC produced ocs. Harif et al. (2012) observed that aluminumEC generated ocs were more prone to compaction and restructuring, comparedto those produced from CC. However, additional testing would be required toconrm these results.ANOVA was also applied to compare each test setting to determine if therewas a signicant dierence between the dierent backwash settings. To do this,the slopes calculated from early, middle and end points of the ltration cycleswere averaged and compared. ANOVA results for the comparison of the dierentbackwash test results are as follows: Fcalc = 16.0, p = 0.004; Fcrit = 5.15,α = 0.05. This analysis supports the conclusion that all three backwash andltration cycle timing conditions tested resulted in signicantly dierent overallfouling rates, as Fcalc is less than Fcrit and p is less than α.However, this conclusion may not represent the true eect of backwash andltration cycle duration, as other confounding variables may be inuencing theoverall fouling rate. Ye et al. (2011) studied fouling and mitigation in hollow-breltration with periodic backwash. They found that short periodic backwash onlyexpanded the foulant cake, which re-compressed after ltration began. Further-more, changes in cake height, composition and structure were aected by thehydrodynamics of the backwash (Ye et al., 2011). Therefore, given the limitedconditions tested, a clear conclusion regarding the optimal backwash and ltra-tion cycle durations for this system cannot be made. Further research would berequired to condently optimize the fouling rate for UF of EC pre-treated water.Another parameter of ltration performance is determination of the resis-tance recovery between backwash cycles. This approach allows us to quantify57whether the lter cake is eectively removed or merely expanded. The aver-age resistance recovery from backwashing was calculated for the period of 200-400L/m2 for the UF membrane Type 1 ltration tests with EC pre-treated feed.These results are presented in Figure 3.5.Figure 3.5: Average resistance recovery measured during cycles between 200-400L/m2 throughput for UF membrane Type 1 tests with EC pre-treated feed andbackwashThese results indicate that there is no signicant dierence in the relativeresistance recovery between the three dierent settings tested. This is furthersupported by the results of ANOVA analysis. Comparing the mean resistancerecoveries for the three backwash settings for this system resulted in a calculatedF-value of 1.39 and a p-value of 0.28. The F-critical value for this analysis was3.63 (assuming 95% condence, α = 0.05). Therefore, these calculated meansare not signicantly dierent and the backwash and ltration cycle times testeddid not aect the resistance recovery between cycles.An important part of fouling mitigation is the hydrodynamics of the systemand the degree of back-transport at the membrane surface. To improve the re-covery during backwash, a couple of things could be done. First, increase theintensity of the backwash. For these tests, the backwash intensity was xed at58approximately 120% of the ltration ux, however higher backwash intensitiesmay provide improved fouling control (Shi et al., 2014). Varying the intensity ofthe backwash will change the hydrodynamics of the backwash cycle and may im-prove fouling mitigation. Second, introduce some sort of additional shear forceor cross-ow into the system, such as an air sparge. As mentioned earlier, Ye etal. (2011) studied periodical backwash in a hollow-bre membrane system. Theyfound that air scouring during backwash assisted in membrane cleaning and in-troduced forces necessary to carry foulants away (Ye et al., 2011). Limitationsof the UF membrane Type 1 setup did not allow for investigation into this typeof turbulence for this membrane, but it was studied using UF membrane Type 2(presented in Chapter 4).3.2.3 Permeate water qualityThe permeate water quality was monitored throughout each test. The initialgrab sample began after 90 minutes of ltration and nished once the requiredvolume was collected, which is equivalent to a throughput of 60-100L/m2. Asecond sample was collected after approximately 18-20 hours of ltration, whichis equivalent to approximately 650-1100L/m2 (depending on the measured owrate of the test and the time of the sample).DOC, UVA-254 and Fe were measured in the feed and permeate samples. Re-sults for DOC and UVA-254 are presented in Figure 3.6. DOC and UVA-254 val-ues are expressed with respect to the initial value of the raw feed water, priorto EC pre-treatment. The measured concentrations of Fe in the permeate wereall below 0.100mg/L Fe, indicating complete removal of iron from the feed water.Results do not conclusively demonstrate a signicant dierence in water qual-ity between the initial and nal permeate samples for the UF membrane Type 1tests.It was hypothesized that the formation of a lter cake may enhance NOMremoval through formation of a dynamic membrane, as observed in literature(Chellam & Sari, 2016), however this does not appear to happen in this case. A59(a) DOC (b) UVA-254Figure 3.6: Normalized DOC and UVA-254 for UF Membrane Type 1 permeatesamples during ltration of EC pre-treated SR-NOM feed.dynamic membrane is formed when particles are deposited on the membranesurface and assist in additional particle removal through interception and Brow-nian motion (Wiesner et al., 1989).There is no signicant dierence in DOC results between initial and nalpermeate samples. There appears to be a dierence between initial and nalUVA-254 values for the test with continual ltration and the 60 minute ltrationcycle with 5 minute backwash. However, recall that the (60-5) test had the lowestrate of resistance accumulation, which would imply that lter cake was the mostporous.If the lter cake had a signicant impact on oc and NOM removal, then itwould be expected that the nal UF permeate sample in the continuous ltrationtest would have a signicantly lower DOC and UVA-254 than all other tests, giventhat a continual ltration test would result in the most signicant cake layer. Asthere is no strong pattern regarding additional NOM removal, this was not thecase and these results are inconclusive.There was also no detectable dierence in water quality in the permeate whencomparing continuous ltration tests and tests with backwash. This indicatesthat backwash does not cause signicant changes to oc properties or perme-60ate water quality. The results of these tests indicate that within the parameterstested, all signicant NOM removal occurs in the EC pre-treatment step and isnot aected by backwash and ltration cycles settings.Finally, SUVA was calculated for the permeate samples. SUVA is an expres-sion of the ratio of DOC and UVA-254 in a water sample and is a measure of theDBP formation potential of the water (Hua et al., 2015). SUVA results for theinitial and nal permeate samples are presented in Figure 3.7.Figure 3.7: SUVA of permeate samples in UF membrane Type 1 ltration testsThese results suggest that there is no signicant dierence in SUVA betweenthe initial and nal UF permeate samples for these tests. The SUVA of the un-treated, raw feed water was 4.08 ± 0.17, which is signicantly higher than thereported permeate values. Therefore, the EC/UF system is able to produce waterwith a signicantly lower SUVA than the initial value, and thus, with an expectedlower DBP formation potential.61Chapter 4Results and discussion ofsynthetic water tests:UF membrane Type 2Two types of UF membranes were evaluated in this work, however because theyvary in material, geometry and pore size (MWCO), they will be discussed sep-arately. This chapter presents water quality and membrane fouling results forUF membrane Type 2. A brief discussion comparing the test results for the twomembrane types is presented in Section 4.3. A detailed comparison between thetwo membranes is not feasible as the dierences in hydrodynamics and pore sizeand material prevent accurate or representative conclusions to be made.UF membrane Type 2 is a 0.04µm, PVDF membrane with hollow-bre ge-ometry and outside-in ow conguration. Each membrane module consisted of3 bres, 19cm in length, that were submerged in a 500mL graduated cylinderkept at constant level throughout the duration of the test. The average workingvolume of the cylinder was 622cm3. Suction was applied to the module with aperistaltic pump to drive ltration through the membrane. The schematic anddescription of this setup is available in Section 2.3.2.The eect of backwash and cross-ow through air sparging was investigated,62using EC pre-treated SR-NOM feed water. Natural surface water was also testedusing UF membrane Type 2, with these results being presented separately inChapter 5.During the tests, the ltration ux was kept constant at 50 ± 2.5 LMH. Theinitial grab sample began after 90 minutes and nished once the required vol-ume was collected, which is equivalent to a throughput of 60-100L/m2. A sec-ond sample was collected after approximately 18-20 hours of ltration, whichis equivalent to a throughput of approximately 650-1100L/m2 (depending on themeasured ow rate of the test and the time of the sample).4.1 Filtration of raw waterAs with UF membrane Type 1, an initial test was completed using UF membraneType 2 to evaluate membrane performance and fouling rate during ltration ofRaw water. The Raw water ltration test was completed to identify whether anyNOM was removed by the lters themselves, without EC pre-treatment. Themembrane pore size used in these tests was 0.04µm. As mentioned in Section 2.1,the synthetic feed was prepared using a stock solution of Suwannee River NOM(SR-NOM), that had been pre-ltered with a 0.45µm lter. Therefore, becauseUF membrane Type 2 had a pore size that was an order of magnitude smallerthan that of the lter used to prepare the solution, it was expected that some ofthe higher MW fractions of NOM may be rejected by the membrane. Table 4.1presents the water quality results for the direct ltration of Suwannee River NOMfeed solution. These results indicate that there is no detectable signicant dier-ence in DOC for feed and permeate samples. UVA-254 values, however, suggestthat the membrane is removing a small portion of the UV-absorbing compounds,as there appears to be a dierence between the feed and permeate samples forUVA-254. HPSEC analysis of the feed and permeate samples is able to providefurther insight to this observation, as presented in Figure 4.1.There appears to be a dierence in the permeate peaks in the higher MW63Table 4.1: Feed and permeate water quality for raw water ltration using UFmembrane Type 2. Reported as average ± 95% condence interval.DOC UVA-254 Fe - totalmg/L C - mg/L FeRaw feed (batch) 4.94± 0.20 0.232± 0.004 < 0.10Initial UF permeate @75-100L/m2 4.66± 0.25 0.193± 0.015 < 0.10Final UF permeate @900-1100L/m2 4.84± 0.09 0.198± 0.005 < 0.10Figure 4.1: MWD of Suwannee River NOM feed water and UF permeate usingUF membrane Type 2, by HPSEC.fractions F4 and F5. Calculation of the % reduction in AMW fraction area (calcu-lated as the sum of the areas of peaks assigned to each fraction) suggests a 11%removal of F4 and 17% removal of F5. These results suggest that the membranemay be able to remove a small fraction of the larger MW NOM, according to theMWD from HPSEC analysis. This was not observed during the raw water l-tration using UF membrane Type 1. The dierence in UVA-254 during ltration64with UF membrane Type 2 may be due to the dierence in membrane pore sizeor MWCO. Both membrane types are classied as UF membranes. However, theyvary in terms of nominal MWCO and pore size. UF membrane Type 1 is reportedas having a MWCO of 300kDa; UF membrane Type 2 has a reported pore size of0.04µm. The relationship between MWCO and pore size is non-linear and the ac-tual MWCO may dier from the manufacturer stated value depending on NOMcharacteristics and properties (Cho et al., 2000). Therefore, pore size and MWCOwould need to be measured directly to allow for direct comparison between thetwo membrane types. However, given the dierence in permeate UVA-254, thepossible dierence in pore size between these two membranes may explian thedierence in perfomance during ltration of raw water.Another possibility is the membrane material. Gray et al. (2007) suggest thathydrophobic compounds are more likely to adsorb onto the inner pores of themembrane while hydrophilic compounds will form a gel layer and cause surfacefouling. S. Lee et al. (2013) completed research comparing ceramic and polymericmembranes during ltration of NOM from source water. Part of the analysis wasto compare the adsorption of Suwannee River Humic acid (SRHA) onto the mem-brane material in a batch test. Results indicated that the ceramic membrane hada much lower tendency to adsorb SRHA compared to the polymeric membranesused (S. Lee et al., 2013). Therefore, although it was assumed that NOM removalby membrane ltration was only based on size exclusion, the chemical and phys-ical properties beyond MWCO may also play a role in membrane performance.Further investigation into these properties was beyond the scope of this project.The TMP was also monitored throughout the raw water UF test to determinethe extent of membrane fouling without EC pre-treatment during ltration of SR-NOM feed. The measured TMP and the initial membrane resistance were usedto calculate the accumulated fouling resistance, as described in Section 1.2.4.Experiments were completed at room temperature (22±2◦C) and the dynamicviscosity was assumed constant at 0.9544mPa.s. The measured volumetric owrate from each individual test was used to determine the membrane ux, J . The65accumulated fouling resistance during the raw water ltration tests is presentedin Figure 4.2.Figure 4.2: Fouling resistance during ltration of Raw water using UF membraneType 2. Error bars represent calculated maximum and minimumRf for duplicatetests. Variation in rst 150L/m2 is due to changes in water level of membranemodule cylinder, before automated water level was applied.These results suggest that some degree of fouling is occurring during ltra-tion of Raw SR-NOM feed water, as the general trend of fouling resistance isincreasing. Linear regression of the UF membrane Type 2 SR-NOM raw waterltration test resulted in a calculated fouling rate of (3.34±0.17)×108 m−1/Lm−2with an R2 value of 0.9678. This slope is (46± 4)% less than the calculated foul-ing rate of UF membrane Type 1 (a ceramic membrane). This was unexpected asceramic membranes are reportedly less prone to fouling compared to polymericmembranes (S. Lee & Kim, 2014; S. Lee et al., 2013). Additional replicate rawwater ltration tests are recommended to conrm these observations. However,a detailed analysis of the raw water ltration mechanism was beyond the scopeof the the experimental design.664.2 Filtration of EC treated waterUF tests were completed with EC pre-treated feed water to evaluate the eectof ltration cycle backwash and air sparging on permeate water quality andUF membrane fouling rates. The procedure for preparing EC treated feed wasdiscussed in Section 3.2.1. The only dierence between preparation of EC pre-treated feed for UF membrane Type 2 tests was that a larger volume was required.Table 4.2 outlines the EC performance and results for EC pre-treated feed tests,presented and discussed previously in Section 3.2.1.Table 4.2: EC treated UF feed water quality, presented previously in Section 3.2.1.EC treated UF FeedDissolved Organic Carbon, mg/L C 3.66± 0.37UV absorbance at 254nm 0.097± 0.014pH 8.54± 0.24Total iron, mg/L Fe 34.2± 1.28Dissolved iron (0.45µm) 0.25± 0.10SUVA 2.65± 0.46Given the results from previous tests, it was expected that the DOC and UVAconcentrations would not change signicantly through UF membrane Type 2ltration tests. The particulate iron was expected to be completely removed.4.2.1 Membrane fouling during ltration of EC pre-treatedfeedUF tests were completed using EC pre-treated feed to determine the eect of airsparging as a fouling mitigation technique. The combination of backwash andair sparge was also investigated.Figure 4.3 presents the resistance accumulation data from the continual l-tration tests using UF membrane Type 2. The resistance of raw water ltrationusing the same membrane is included in the graph for reference.67Figure 4.3: Fouling resistance during continual ltration of EC pre-treated wa-ter using UF membrane Type 2. Error bars represent calculated maximum andminimum Rf for duplicate tests.These results indicate that air sparging has a signicant eect during contin-ual ltration in terms of fouling mitigation. EC pre-treatment causes a signicantincrease in resistance accumulation compared to ltration of raw water, whichagrees with results observed in the tests with UF membrane Type 1. Again, this isto be expected, given the concentration of iron particles in the EC treated water.After approximately 700 L/m2 of throughput, there appears to be a zonewhere the resistance accumulation rate seems to decrease for the test with ECpre-treated water (without air sparge). This is the point of system failure. Themembrane had exceeded the TMP limit for suction of water and was unable tocontinually lter water at the dened ux of 50LMH.The eect of backwash was also investigated. UF membrane Type 2 ltrationtests were completed using backwash with and without air sparge. The TMPwas monitored throughout the test and the accumulated resistance, Rf was cal-culated according to the method described in Section 2.5.6. The resistance accu-68mulation results for UF membrane Type 2 tests using EC pre-treated SR-NOMfeed is presented in Figure 4.4.(a) Without air sparge(b) With air spargeFigure 4.4: Fouling resistance during ltration of EC pre-treated water using UFmembrane Type 2, with and without backwash, in presence and absence of airsparge. Filtration cycle was 30 minutes with period of 5 minute backwash. Totalcycle duration was 35 minutes. Error bars have been omitted for clarity.Results indicate that backwash improves fouling mitigation in cases with and69without air sparge. In the case without air sparge (Figure 4.4a), the use of back-wash appears to successfully mitigate fouling for the rst 100L/m2, but then theaccumulated resistance begins to increase at a rate similar to that of the contin-uous test without backwash. After this point, the degree of reversible hydraulicfouling appears to increase: the nal Rf point is increasingly greater than thepoint at the start of the cycle, for each subsequent cycle. However, the over-all rate of resistance accumulation follows the same trend as the test withoutbackwash. This suggests that the lter cake that accumulated on the membraneduring ltration was released from the membrane during the backwash stage,but was re-established once the next ltration cycle began. This suggests thatthe cake fouling with EC pre-treated feed is reversible, but another process maybe required to sweep the cake away from the membrane before the next ltrationcycle begins.The eect of this sweeping force is evident in the tests with air sparging(Figure 4.4b). The overall fouling rate for both tests with air sparge is much lowerthan the tests without air sparge. However, the combination of backwash andair sparge eectively eliminates fouling compared to the other test conditions.This improved mitigation with air and backwash could be because air bubblesfrom the sparge were ineective at removing small particles that may have beenattracted to the membrane. The backwash step may be responsible for dislodgingparticles that were stuck to the membrane surface, improving the overall foulingmitigation for the process.Linear regression was performed on the linear portion of the data for all tests.Results from linear regression are presented in Table 4.3. Examination of theseresults suggest that the optimal method of fouling mitigation for this system is acombination of backwash and air sparge. Backwash resulted in minimal foulingrate decrease compared to the test without backwash (2 ± 6)%. Air spargingalone resulted in a (64 ± 2)% decrease in overall fouling rate (slope) while airsparging combined with backwash resulted in a decrease of (98± 1)%.Air sparging is eective in fouling mitigation but it may also have an eect70Table 4.3: Linear regression for UF membrane Type 2 ltration tests reportedwith 95% condence interval. For tests with backwash, regression is performedon data point logged at the end of the ltration cycle.Test Setting Slope, m−1L.m−2R2EC, No backwash, no air (5.47± 0.30)× 109 0.9599EC, With backwash, no air (5.30± 0.16)× 109 0.9923EC, No backwash, with air (1.97± 0.03)× 109 0.9961EC, With backwash, with air (0.09± 0.032)× 109 0.4598Raw water, no backwash, no air (0.33± 0.017)× 109 0.9678on the size and properties of the ocs. Vigorous air sparging may introduceenough turbulence in the UF feed to break apart the EC ocs. Jarvis et al. (2005)investigated oc strength and regrowth of NOM ocs using dierent kinds ofchemical coagulants. The authors concluded that the type of coagulant used mayaect oc properties such as size, resistance to breakage, and ability to regrowafter breakage occurs (Jarvis et al., 2005). Floc size, breakage, and regrowth wasnot investigated here, but given the chemical complexity of EC produced ocs,it would be valuable to investigate this topic further.Nevertheless, these results are promising as they suggest that a water treat-ment system using conventional fouling mitigation techniques would be ableto operate for a long time without stopping to perform more intensive chemi-cal cleaning. There does not appear to be any fatal aws in the process whenpre-treating with iron EC. However, in this work, only one sparge conditionwas tested. More research could be done to optimize the conditions for foulingmitigation for EC pre-treated feed using air sparge. For example, pulse bubblesparging may be even more eective than continuous bubble sparging (Jankhah& Bérubé, 2014). Furthermore, introducing air to a system with EC pre-treatedsolids may cause chemical and physical changes in the properties of the ocs.This is discussed further in Section 4.2.2.714.2.2 Permeate water qualityThe permeate water quality was monitored throughout each test. The initialgrab sample began after 90 minutes of ltration and nished once the requiredvolume was collected, which is equivalent to a throughput of 60-100L/m2. Asecond sample was collected after approximately 18-20 hours of ltration, whichis equivalent to approximately 650-1100 L/m2 (depending on the measured owrate of the test and the time of the sample). DOC, UVA-254 and Fe were measuredin the feed and permeate samples. Results for DOC and UVA-254 are presentedin Figures 4.5a and 4.5b respectively. DOC and UVA-254 values are expressedwith respect to the initial value of the raw feed water, prior to EC pre-treatment.The measured concentrations of Fe in the permeate were all below 0.100mg/L Fe,indicating complete removal of iron from the feed water.(a) DOC (b) UVA-254Figure 4.5: Normalized DOC and UVA-254 for UF Membrane Type 2 permeatesamples during ltration of EC pre-treated SR-NOM feed.For the tests without air sparge, there is no observable trend in changes toDOC between the initial and nal UF permeate samples. There does appear tobe a dierence between initial and nal UVA-254 values for the test with contin-ual ltration. This was also observed during the continual ltration test usingUF membrane Type 1. This suggests that the lter cake may be contributing toimproved removal of UV-absorbing compounds. However, additional replicate72tests would be required to make a stronger conclusion.As for the tests with air sparge, results suggest a signicant increase in DOCand UVA-254 in the UF permeate between the initial and nal samples. Theresults from these two tests are very interesting. The average UF feed DOCC/C◦was 0.72 ± 0.11. The average UVA-254 C/C◦ was 0.45 ± 0.18. These valuesare much higher than the respective measurements for the initial UF permeatesamples for the tests with air sparge. This suggests that the air sparge has anadditional eect on NOM removal at the start of the test.The additional initial NOM removal may be due to two possible explanations.First, the turbulence from the air sparge may cause improved mixing of the ocsin the UF feed cylinder, allowing for better mixing and adsorption of NOM bythe produced EC solids. However, oc strength and size was not measured. Thesecond possibility is that the air sparge is introducing additional oxygen to thesystem, oxidizing any ferrous iron in the feed water to ferric iron. This wouldinduce formation of ferric hydroxide species. As mentioned in Chapter 1, previ-ous researchers have suggested that a steady ow of oxygen is required for aneective electrocoagulation process (Timmes et al., 2010). This could be furtherinvestigated by studying the speciation of iron in the UF feed over time and re-peating the test with a nitrogen sparge, isolating the UF feed from oxygen andobserving the results of performing the test under an inert atmosphere.The error bars for the tests with air sparge are large, indicating high variabil-ity in the measured results. An improvement of this test would be to repeat thetest conditions but with better control on the degree of air sparging and improvedpurity of the air line, to better control systematic error.The low permeate concentrations do not stay constant throughout the tests,as the nal UF permeate samples for the tests with air sparge are much higherthan the initial values. This trend is exhibited in both DOC and UVA-254 data,suggesting that although additional NOM removal is initially occurring, withtime, the captured NOM is either re-released into solution or the rejected NOMis building up in the feed cylinder and ends up permeating the membrane.73To understand this change in water quality throughout the ltration tests,the UF feed samples were analysed. HPSEC analysis of the feed and permeatesamples was used to understand this change in NOM in the feed and permeateconcentrations. The apparent molecular weight of the feed and permeate sam-ples was examined for UF membrane Type 2 tests with and without air sparge.These results are presented in Figure 4.6.(a) Initial sample; without air sparge (b) Final sample; without air sparge(c) Initial sample; with air sparge (d) Final sample; with air spargeFigure 4.6: AMW of feed and permeate samples for UF membrane Type 2 testswith backwash, with and without air sparge. Raw water AMW distribution isincluded for reference.Figures 4.6a and 4.6b conrm the observations made for the tests without airsparge: there does not appear to be a signicant dierence between the UF feedand UF permeate samples. Figures 4.6c and 4.6d, however, suggest a dierent74story for the tests with air sparge. The initial sample (presented in Figure 4.6c)appears to have a signicant dierence between the early UF feed and the UFpermeate samples, across all AMW fractions . However, at the end of the test,the results indicate that the NOM fractions that are causing the increased feedand permeate concentrations are the lower MW fractions. The fraction capturedby the EC treatment is not released back into solution. Therefore, the lower MWfractions are building up in the feed concentration and permeating the mem-brane. Because this was not observed in the tests without air sparge, we mustconclude that the air sparge is the source of this change in behaviour.DOC and UVA-254 as a function of throughput is presented in Figure 4.7 forUF membrane Type 2 tests with air sparge. Additional samples, beyond the initialand nal samples, were collected for the test with backwash to track the feedand permeate water quality in greater detail throughout the test. These resultshighlight a number of interesting features. First, the feed sample for the testwithout backwash had higher reported DOC and UVA-254 than the test withbackwash. This is likely due to the inherent dilution of the backwash setup.Backwash was performed with permeate water, that was pumped directly intothe feed cylinder without a bleed and rell step between each ltration cycle.Therefore, the feed for the test with backwash would have a lower concentrationthan the test with backwash.Second, the DOC and UVA-254 increased in the feed and permeate through-out both tests and there is a signicant dierence between the feed and permeateconcentrations for both tests. Therefore, in this case, UF membrane Type 2 is re-moving additional NOM compared to the standard 0.45µm syringe lter used toprepare the UF feed/EC euent samples. As discussed previously, the air spargemay be causing a chemical change in the ocs, allowing them to capture moreNOM. However, because the enhanced NOM removal is only detected in the per-meate samples, the particles responsible for capturing the additional NOM mustbe small enough to permeate the standard syringe lter, but too large to escapethe 0.04µm membrane.75(a) DOC, without backwash (b) DOC, with backwash(c) UVA-254, without backwash (d) UVA-254, with backwashFigure 4.7: Water quality of feed and permeate samples for ltration tests withair sparge.Therefore, the EC/UF system with air sparge appears to successfully improveNOM removal compared to other ltration systems without air. Operation of thesystem at a pilot scale with continual bleed from the UF cylinder would providemore information about the changes in the UF feed and permeate composition.Further analysis of the system and investigation into the eect of air sparge isrecommended.764.3 Comparison of UF membrane systemsA comparison of the two UF membrane systems was also performed, to evaluatethe dierence in performance between the two membrane types.4.3.1 Membrane fouling comparisonFigure 4.8 presents the accumulated resistance data for the ltration tests withmatching conditions using UF membranes Type 1 and Type 2. These matchingconditions include: continual ltration of raw water, continual ltration of ECpre-treated water, and ltration with 30 minute ltration cycle and 5 minutebackwash for EC pre-treated water.(a) Without backwash (b) With backwashFigure 4.8: Resistance accumulation comparing UF membranes Type 1 and 2 forRaw and EC pre-treated feed water for cases without backwash (a) and with back-wash (b). Error bars represent the calculated Rf error, based on the maximumand minimum value of the parallel tests.The result of linear regression analysis of theses six tests is presented in Ta-ble 4.4. According to the 95% condence intervals for the calculated values, thefouling rates for each of these test settings is uniquely dierent and there is asignicant dierence in the fouling accumulation rates of UF membranes Type 177and Type 2. The fouling rate for UF membrane Type 1 is (5.1± 0.3)% less thanthat of UF membrane Type 2 for the continuous ltration tests (without back-wash). The fouling rate for UF membrane Type 1 is (51 ± 3)% less than that ofUF membrane Type 2 for the tests with backwash.Table 4.4: Linear regression and R2 values for resistance accumulation for ltra-tion with UF membranes Type 1 and Type 2.Slope, m−1L.m−2R2UF1-EC-continuous (5.11± 0.05)× 109 0.9987UF2-EC-continuous (5.47± 0.30)× 109 0.9599UF1-raw-continuous (0.62± 0.03)× 109 0.9487UF2-raw-continuous (0.33± 0.02)× 109 0.9678UF1-EC-BW(30-5) (2.57± 0.15)× 109 0.9666UF2-EC-BW(30-5) (5.30± 0.16)× 109 0.9923The dierence between the two membrane types could be explained by anumber of factors. First, the membrane material is dierent, and dierent mem-brane materials have dierent fouling tendencies (S. Lee & Kim, 2014). Second,the pore sizes are dierent; UF membrane Type 2 has a tighter pore size than UFmembrane Type 1, which would result in a higher overall membrane resistance.Third, the hydrodynamics of the two modules are completely dierent, whichwill aect membrane fouling and backwash eciency (Shi et al., 2014).UF membrane Type 1 is a small chamber that holds approximately 3.3mLof feed water. Water is fed through a 1/4” diameter opening from the centretop of the module and the ow then rapidly expands to the full diameter of themodule, approximately 41mm. Rapid expansion, even at low ow rate, results ina turbulent ow pattern. The ow would not be considered fully developed untila distance of 60X the initial pipe diameter. This turbulence induces shear andback-transport on the surface of the membrane, which would lower the overallfouling rate.UF membrane Type 2 is housed in a large cylinder that holds approximately78622mL of water. Feed water was fed through a 1/4” hose immersed approxi-mately 10cm below the water level in the cylinder. Without the use of inducedshear through air sparge or mechanical agitation, hollow bre dead-end mem-branes have very poor back-transport (Belfort et al., 1994; Ye et al., 2011). Theinertial lift caused by particles and turbulent ow in the UF membrane Type 1module may be the the cause of the lower fouling rates during ltration of ECpre-treated feed, but a more in-depth study on the properties of the membranesis required to make strong conclusions.These factors may help explain the overall dierence in performance betweenthe two membranes. However, complete calculation and characterization of theuid dynamics of these membrane systems is beyond the scope of this project.4.3.2 Permeate water quality comparisonThe dierence in permeate water quality was presented previously when eachmembrane type was discussed individually. Comparison here is made to high-light the dierences, if any, in membrane performance in terms of water quality.Water quality for raw water ltration tests is presented in Figure 4.9. Results(a) DOC (b) UVA-254Figure 4.9: Permeate water quality during raw water ltration: comparison ofUF membranes Type 1 and Type 279indicate that there is no signicant dierence between the two membranes interms of DOC removal but there appears to be a slight dierence in terms ofUVA removal between UF membranes Type 1 and Type 2. This dierence wasdiscussed previously, in Section 4.1, this could be due to the dierences in thechemical properties of the membranes. However, further analysis of these prop-erties would be required to make a stronger conclusion.Water quality for EC pre-treated UF tests is presented in Figure 4.10. Here,there does not appear to be a signicant dierent between the two types, how-ever there does appear to be a dierence between the initial and nal UVA-254values for the tests with continual ltration. This was already discussed in Sec-tion 4.2.2. Comparison of the permeate results for EC pre-treated feed to the rawwater permeate (Figure 4.9) suggests that the fraction of UV-absorbing materialthat was removed by UF membrane Type 2 is captured in the EC pre-treatmentstep. There is no signicant dierence in the permeate water quality of thesemembranes under the conditions tested.(a) DOC (b) UVA-254Figure 4.10: Permeate water quality during EC pre-treated water ltration: com-parison of UF membranes Type 1 and Type 2, during ltration with and withoutbackwash (30 min cycle + 5 min backwash)Finally, SUVA was calculated for the permeate samples. SUVA results for theinitial and nal permeate samples are presented in Figure 4.11. These results80suggest that there is no signicant dierence in SUVA between the initial andnal UF permeate samples for these tests. The SUVA of the untreated, raw feedwater was 4.08± 0.17, which is signicantly higher than the reported permeatevalues. Therefore, regardless of membrane type, the EC/UF system is able toproduce water with a signicantly lower SUVA than the initial value, and thus,with a potentially lower DBP formation potential.(a) UF membrane Type 1 (b) UF membrane Type 2Figure 4.11: SUVA of UF permeate: comparison of UF membranes Type 1 andType 2, during ltration with and without backwash (30 min cycle + 5 min back-wash)The results from these tests are important because they suggest that the ECpre-treatment step is the more important step in the hybrid EC/UF process. Thereis no signicant dierence in membrane permeate water quality, and thereforewhen designing a drinking water treatment system, the decision for which mem-brane to use does not have to be based on permeate water quality for NOM re-moval. It should instead be based on membrane fouling rates and simplicity offouling control. In order to optimize the membrane system, other EC settingswould need to be investigated as well as further study to optimize backwash.81Furthermore, these results are based on treatment of a synthetic surface wa-ter, that was prepared using only the soluble fraction of Suwannee River NOM.As the chemical and physical properties of NOM vary signicantly dependingon water source, it is also important to observe how the system responds to realwater samples. One EC/UF test was completed using natural surface water. Theresults from this test are presented in Chapter 5.82Chapter 5Results and Discussion of naturalsurface water testNatural surface water was collected from Middle River, British Columbia andwas treated with the EC/UF hybrid system. A single test condition was appliedto evaluate the general performance of the EC/UF process using natural sur-face water as the feed water. See Section 2.2 for a description of the feed watercollection and source water quality results. A summary of the results for ECpre-treatment of natural surface water are presented in Table 5.1, alongside thesynthetic surface water results for comparison.These results indicate that the EC settings used were not optimal for thistype of water, as the DOC removal was low for the natural surface water feed(approximately 12%). McBeath (2017) observed (33.7± 4.8)% reduction in DOCand a (48.1± 2.7)% reduction in UVA-254 when treating surface water with thesame continuous ow EC reactor as the one used in this work, at a ow rate of10 LPM. The OP for McBeath’s system was 7V to 14V and the current densitywas 7.35 mA/cm2. The DOC removal of the natural surface water is less thanthe removal achieved by McBeath, however UVA-254 removal was signicantlyhigher. High UVA-254 removal indicates that there may be variation in the NOMstructure and properties, which aect NOM removal by EC.83Table 5.1: EC test result summary for treatment of natural surface water, with anoperating current density of 6.43mA/cm2 and target metal loading of 35mg/L Fe.Synthetic surface water Natural surface waterFinal DOC, mg/L C 3.66± 0.37 4.89± 0.20% removal DOC 33% 12%Final UVA-254 0.097± 0.014 0.063± 0.044% removal UVA-254 57% 67%Final pH 8.54± 0.24 7.07Initial SUVA 4.08 3.38Final SUVA 2.65 1.28EC current eciency, φ 95% 89%The natural surface water EC test had a lower current eciency than thesynthetic surface water. The measured operation potential (OP) during the ECtest with the natural surface water was in the range of 11.9V to 12.5V. The OPfor the synthetic surface water EC tests ranged from 5.7V to 9.4V. Therefore,the resistance in the natural surface water EC test was much higher than thesynthetic surface water tests.An important dierence between the two water sources is the water ma-trix. The synthetic surface water was prepared using only sodium salts of sul-phate, chloride and bicarbonate. The natural surface water likely contained othercations such as magnesium and calcium (not measured). The natural surfacewater also contained signicantly lower sulphate and chloride concentrations,which would explain the higher OP during the EC test; the operation poten-tial (or voltage) is inversely related to resistance (as dened by Ohm’s Law).High ion concentrations imply higher ionic strength, which would result in re-duced repulsive forces between molecules (ζ-potential) and improved occula-tion (Crittenden et al., 2012a). The composition of the two dierent feed watersare available in Tables 2.1 and 2.2.Other compounds found in surface water may also aect NOM removal byEC. For example, Dubrawski et al. (2013) observed reduced NOM removal with84increasing carbonate concentration in EC tests with an iron anode and SuwanneeRiver NOM. Nevertheless, the EC euent from the natural surface water feedunderwent ltration by UF membrane Type 2, which is discussed in the nextsection.5.1 Membrane foulingOne UF test setting was completed to investigate the general eect of feed wa-ter on the EC/UF system performance. UF membrane Type 2 was used, with a30 minute ltration cycle and 5 minute backwash. Figure 5.1 presents the mem-brane resistance accumulation for the natural surface water test, compared to UFmembrane Type 2 tests completed with EC pre-treated synthetic feed water andbackwash.These results highlight the signicant dierence between the ocs producedwith the dierent feed types. When only backwash was applied, the naturalfeed water had signicantly lower fouling compared to the synthetic feed water.Membrane fouling for the natural surface water test with only backwash wascomparable to the synthetic surface water results that incorporated both back-wash and air sparge. Table 5.2 reports the linear regression results for determin-ing the rate of resistance increase during ltration of EC pre-treated syntheticand natural feed water.Table 5.2: Linear regression of EC pre-treated synthetic and natural feed wa-ter ltration using UF membrane Type 2, with 95% condence intervals. Linearregression is based on data points collected at the end of each ltration cycle.Water source Slope, m−1L.m−2R2Synthetic feed water SR-NOM, EC + BW (5.30± 0.16)× 109 0.9923Synthetic feed water SR-NOM, EC + BW + air (0.089± 0.032)× 109 0.4598Natural Surface Water, EC + BW (0.239± 0.024)× 109 0.919785Figure 5.1: Resistance accumulation for UF test with EC pre-treated natural sur-face water with 30 minute ltration cycles and 5 min backwash. Presented withsynthetic feed water test results (with and without air sparge). Error bars repre-sent calculated maximum and minimum Rf for duplicate testsThese results suggest that feed water quality has a signicant impact on UFfouling rates. The fouling rate of the natural surface water test was (95± 0.5)%lower than the synthetic feed water test, under identical conditions. The testwith synthetic feed water and air sparge (Synthetic feed water, SR-NOM, EC +BW + air) still had the lowest fouling rate of the three tests, suggesting that airsparge is still required to optimise fouling minimisation.During the experiment, the foulant layer was observed as a large uy layerof ocs that accumulated on the surface of the membrane. At the end of theltration cycle the ocs were adhered to the bre, however the cake must havebeen extremely permeable because the resistance accumulation rate of the syn-thetic surface water was signicantly higher than the natural surface water. Thiscould be due to the structure of the ocs, causing a more permeable cake. Ad-86ditional tests with the natural surface water and air sparge would provide moreinformation on the eect of air sparge to help optimise fouling mitigation.Resistance accumulation for the natural surface water is presented in Fig-ure 5.2. These results indicate that both reversible and irreversible fouling is oc-Figure 5.2: Reversible fouling for UF membrane Type 2 experiment with EC pre-treated natural surface water. Error bars represent calculated maximum and min-imum Rf for duplicate testscurring. The calculated average resistance recovery for this test was (22± 4)%,which is signicantly higher than the calculated recovery of the synthetic feedwater tests. The resistance recovery results of the UF membrane Type 2 back-wash tests are presented in Figure 5.3. These results support the observationthat there is a signicant dierence in the relative resistance recovery betweenthe three dierent settings tested. Comparing the mean resistance recoveries forthe three backwash settings for this system resulted in a calculated F-value of13.0 and a p-value of < 0.001. The F-critical value for this analysis was 3.42(assuming 95% condence, α = 0.05). Therefore, these calculated means are sig-87Figure 5.3: Average resistance recovery for synthetic and natural feed watersusing UF membrane Type 2nicantly dierent and feed water matrix and oc properties have a signicantinpact on membrane recovery during ltration of EC pre-treated feed.The higher relative resistance recovery for the natural surface water test in-dicated that backwash eciency is also aected by oc characteristics and feedwater matrix. These results imply that complicated air sparge systems may notbe necessary depending on the type of oc produced. To further expand thisresearch and understand the eect of source water quality on EC/UF processperformance, it is recommended to investigate further the oc properties suchas ζ-potential and particle size distribution in the UF feed water and how theseproperties aect membrane fouling.Ulu et al. (2014) evaluated oc formation through measurement of ζ-potentialand particle size distribution measurements in a 50ppm humic acid solution treatedwith iron and aluminum EC. They concluded that the oc size evolution does notonly depend on solution chemistry, but also on occulation conditions such asshear and occulation time. For iron EC produced ocs, a maximum oc sizewas observed, at which point the ζ-potential underwent reversal and the shear88force exceeded oc strength and led to breakage (Ulu et al., 2014). A valuablecontribution to the eld would be to further examine the eect of EC-producedoc size and characteristics on UF process eciency.5.2 Permeate water qualityThe permeate water quality was monitored throughout the test. The initial grabsample began after 90 minutes of ltration and nished once the required vol-ume was collected, which is equivalent to a throughput of 60-100L/m2. A sec-ond sample was collected after approximately 18-20 hours of ltration, whichis equivalent to approximately 650-1100L/m2 (depending on the measured owrate of the test and the time of sampling).DOC, UVA-254 and Fe were measured in the feed and permeate samples. Re-sults for DOC and UVA-254 are presented in Figure 5.4. DOC and UVA-254 val-ues are expressed with respect to the initial value of the raw feed water, priorto EC pre-treatment. The measured concentrations of Fe in the permeate wereall below 0.100mg/L Fe, indicating complete removal of iron from the feed wa-ter. Results are presented alongside the EC pre-treated synthetic feed permeatewater quality results for comparison.These results indicate that there is no signicant dierence in permeate wa-ter quality when comparing the two source feed waters. The lter cake doesnot appear to contribute to NOM removal, as there is no signicant dierencebetween initial and nal DOC or UVA-254 for either test.However, recall that the DOC removal by EC pre-treatment was 12% (as mea-sured in a 0.45µm ltered sample immediately after EC treatment). This wouldindicate that NOM removal is enhanced through membrane ltration. As notedin Figure 5.4a, the DOC removal in the permeate samples was approximately 44%(compared to DOC in raw feed water). However, a feed sample was taken duringthe nal permeate sample and the measured concentration was 3.89± 0.20. TheDOC of the corresponding permeate sample was 3.12±0.20 mg/L. This suggests89(a) DOC (b) UVAFigure 5.4: Normalised DOC and UVA-254 for UF Membrane Type 2 permeatesamples during ltration of EC pre-treated natural surface water feed and syn-thetic feed waterthat the UF feed DOC may be decreasing with time, but since only one UF testcondition was evaluated with this feed, additional tests should be completed toconrm these results.Nevertheless, the source feed water type does not seem to aect the UF per-meate water quality when comparing synthetic and natural surface water sources.This is indicative of a robust system, however it is recommended that additionalEC and UF conditions be tested with natural surface water sources to better un-derstand the system.90Chapter 6ConclusionThis research combined iron electrocoagulation (EC) and ultraltration (UF) totreat synthetic and natural surface waters in order to remove Natural OrganicMatter (NOM). This work is especially valuable for the development of smallwater systems for remote communities, which have historically been subject todrinking water advisories due to poor water quality or insucient treatment.The EC/UF system has the potential to eectively treat surface water, however,very little research has been completed on the eect of hydraulic controls for UFmembrane fouling mitigation of EC pre-treated water. This knowledge gap hasbeen addressed in this work.6.1 Summary of resultsFixed EC conditions were applied to the feed water in a continuous-ow EC reac-tor, at a ow rate of 1 LPM and an applied current of 2 A. The EC euent actedas the feed water for the UF systems. The membranes types used were: (1) A300kDa ceramic disk membrane, operated in a dead-end ow conguration; and,(2) a 0.04µm PVDF hollow-bre membrane operated as a submerged membranesystem with outside-in ow. Both systems were operated at a constant ux of50 LMH. Periodic backwash cycles were applied to evaluate the eect of back-91washing on the overall membrane fouling rates. The backwash intensity waskept constant at approximately 120% of the ltration ux. Air sparging was alsoapplied to the hollow-bre PVDF membrane as a method of fouling mitigation.EC treatment of synthetic surface water resulted in average DOC removal of33±4% and UVA-254 removal of 57±8%. Average SUVA went from 4.08±0.17in the feed to 2.65 ± 0.46 in the permeate. The decrease in SUVA is due to re-moval of the UV-absorbing fraction of NOM, which could reduce the disinfectionbyproduct formation potential of the treated water. The average concentration oftotal iron in the EC euent was 34.2±1.3mg/L, corresponding to a 95% currenteciency.For the ceramic disk membrane, periodic backwash was demonstrated to de-crease the overall fouling rate during ltration of EC pre-treated feed, comparedto tests without backwash. Filtration of raw water indicated that NOM was notremoved by the membrane, and instead was captured by EC ocs, which werethen ltered out. There was no signicant dierence between the initial and nalUF permeate samples, indicating that the development of a lter cake layer didnot contribute to additional oc or NOM removal.For the PVDF hollow-bre membrane, both backwashing and air spargingwere applied as fouling mitigation techniques. Backwashing on its own resultedin minimal fouling rate decrease. However, air sparging resulted in reducing theoverall fouling rate by (64 ± 2)%. The combination of air sparging and back-washing reduced the overall fouling rate by (98 ± 1)%. The EC/UF test withair sparging and backwash also resulted in additional NOM removal from theUF feed by the membrane (not observed in the tests without air sparging). DOCin the treated water was reduced from 4.69 ± 0.20mg/L in the UF feed (EC ef-uent) sample to 1.24 ± 0.11mg/L in the UF permeate. UVA-254 was reducedfrom 0.170 ± 0.04 to 0.020 ± 0.008. Therefore, air sparging appears to have asignicant eect on membrane fouling and permeate water quality.A single test condition using natural surface water was also completed. Ap-plying backwash to the EC/UF test with natural surface water resulted in a (95±920.5)% lower fouling rate compared to the EC/UF test with synthetic feed watertest under the same conditions. Therefore, water composition has a signicanteect on the membrane fouling rates of the EC/UF system.6.2 Recommendations for future workThis research provides promising results for the development of a EC/UF watertreatment system. However, this research can be expanded to better understandthe system. The eects of air sparging on the feed water chemistry should beinvestigated further. This could be done by examining oc size and strengthas well as changes in ζ-potential during air sparge of Fe EC treated water. Itis recommended that compressed air, high purity air and nitrogen sparging beinvestigated to evaluate the eect of oxygen on the sparging process.Development of a pilot scale UF system is also recommended. The EC unitused in this work has already been investigated at the pilot scale (McBeath, 2017).Addition of a UF module to the pilot system would provide the opportunity toevaluate the EC/UF process in a mode more closely resembling real-life appli-cations. A pilot-scale EC/UF unit would also provide more information on theeect of feed water quality on process performance, as the pilot would be appliedto natural feed water sources. The membrane type did not appear to have an ef-fect on the permeate water quality, therefore selection of the membrane to beused in the pilot system should be motivated by the ability to easily implementhydraulic controls such as air sparging.Finally, remote communities often have unreliable power supplies, and mayrely on diesel generators for power production. Investigation into using alter-native energy sources to power the EC unit and methods of low-energy, passiveltration may make this process more attractive to remote communities.93ReferencesAllgeier, S. (November 2005). Membrane Filtration Guidance Manual (Tech. Rep.).United States Environmental Protection Agency.American Water Works Association. (2008). Microltration and UltraltrationMembranes for Drinking Water (M53). Journal / American Water WorksAssociation, 100(12), 84–97.Aschermann, G., Jeihanipour, A., Shen, J., Mkongo, G., Dramas, L., Croué, J. P., &Schäfer, A. (2016). Seasonal variation of organic matter concentration andcharacteristics in the Maji ya Chai River (Tanzania): Impact on treatabilityby ultraltration. Water Research, 101, 370–381.Bagga, A., Chellam, S., & Cliord, D. A. (2008). Evaluation of iron chemicalcoagulation and electrocoagulation pretreatment for surface water micro-ltration. Journal of Membrane Science, 309(1-2), 82–93.Belfort, G., Davis, R. H., & Zydney, A. L. (1994). The behavior of suspensions andmacromolecular solutions in crossow microltration. Journal of Mem-brane Science, 96, 1–58.Ben-Sasson, M., & Adin, A. (2010). Fouling mitigation by iron-based electrooc-culation in microltration: Mechanisms and energy minimization. WaterResearch, 44(13), 3973–3981.Ben-Sasson, M., Calmano, W., & Adin, A. (2009). Iron-oxidation processes in anelectroocculation (electrocoagulation) cell. Journal of Hazardous Materi-als, 171(1-3), 704–709.Ben-Sasson, M., Zidon, Y., Calvo, R., & Adin, A. (2013). Enhanced removal of94natural organic matter by hybrid process of electrocoagulation and dead-end microltration. Chemical Engineering Journal, 232, 3380–345.Böhm, L., Drews, A., Prieske, H., Bérubé, P. R., & Kraume, M. (2012). The impor-tance of uid dynamics for MBR fouling mitigation. Bioresource Technol-ogy, 122, 50–61.Bolton, G., LaCasse, D., & Kuriyel, R. (2006). Combined models of membranefouling: Development and application to microltration and ultraltrationof biological uids. Journal of Membrane Science, 277 (1-2), 75–84.Brinkman, B. M., & Hozalski, R. M. (2015). Temporal eects variation on mem-brane of NOM and its treatment. Journal - American Water Works Associ-ation, 103(2), 98–106.Chellam, S., & Sari, M. A. (2016). Aluminum electrocoagulation as pretreatmentduring microltration of surface water containing NOM: A review of foul-ing, NOM, DBP, and virus control. Journal of Hazardous Materials, 304,490–501.Cho, J., Amy, G., & Pellegrino, J. (2000). Membrane ltration of natural organicmatter: Factors and mechanisms aecting rejection and ux decline withcharged ultraltration (UF) membrane. Journal ofMembrane Science, 164(1-2), 89–110. doi: 10.1016/S0376-7388(99)00176-3Chow, C. W., Fabris, R., Leeuwen, J. v., Wang, D., & Drikas, M. (2008). As-sessing Natural Organic Matter Treatability Using High Performance SizeExclusion Chromatography. Environmental Science & Technology, 42(17),6683–6689.Collins, M. R., Amy, G. L., & Steelink, C. (1986). Molecular weight distribution,carboxylic acidity, and humic substances content of aquatic organic matter.Implications for removal during water treatment. Environmental Scienceand Technology, 20(10), 1028.Crittenden, J. C., Trussell, R. R., Hand, D. W., Howe, K. J., & Tchobanoglous, G.(2012a). Coagulation and Flocculation. InMwh’s water treatment: Principlesand design, third edition (pp. 541–639).95Crittenden, J. C., Trussell, R. R., Hand, D. W., Howe, K. J., & Tchobanoglous, G.(2012b). Membrane Filtration. In Mwh’s water treatment: Principles anddesign, third edition (Third ed., pp. 819–900). John Wiley & Sons, Inc.Crozes, G., White, P., & Marshall, M. (1995). Enhanced coagulation: Its eecton NOM removal and chemical costs. Journal / American Water WorksAssociation, 87 (1), 78–89.Cui, Z. F., Chang, S., & Fane, A. G. (2003). The use of gas bubbling to enhancemembrane processes. Journal of Membrane Science, 221(1-2), 1–35.Dubrawski, K. L., Fauvel, M., & Mohseni, M. (2013). Metal type and naturalorganic matter source for direct ltration electrocoagulation of drinkingwater. Journal of Hazardous Materials, 244-245, 135–141.Dubrawski, K. L., & Mohseni, M. (2013). In-situ identication of iron electro-coagulation speciation and application for natural organic matter (NOM)removal. Water Research, 47 (14), 5371–5380.Dubrawski, K. L., Van Genuchten, C. M., Delaire, C., Amrose, S. E., Gadgil, A. J.,& Mohseni, M. (2015). Production and transformation of mixed-valentnanoparticles generated by Fe(0) electrocoagulation. Environmental Sci-ence and Technology, 49(4), 2171–2179.Environment and Climate Change Canada. (2018). Canadian Envi-ronmental Sustainability Indicators: Drinking water advisories (Tech.Rep.). Retrieved from https://www.ec.gc.ca/indicateurs-indicators/default.asp?lang=en&n=2C75C17A-1Environment Canada. (2011). Municipal Water Use, 2009 Statis-tics (Tech. Rep.). Ottawa. Retrieved from http://www.ec.gc.ca/Publications/B77CE4D0-80D4-4FEB-AFFA-0201BE6FB37B/2011-Municipal-Water-Use-Report-2009-Stats_Eng.pdfField, R. W., Wu, D., Howell, J. A., & Gupta, B. B. (1995). Critical ux concept formicroltration fouling. Journal of Membrane Science, 100(3), 259–272.Gamage, N. P., & Chellam, S. (2011). Aluminum electrocoagulation pretreatment96reduces fouling during surface water microltration. Journal of MembraneScience, 379(1-2), 97–105.Gao, F., Wang, J., Zhang, H., Zhang, Y., & Hang, M. A. (2016). Eects of sodiumhypochlorite on structural/surface characteristics, ltration performanceand fouling behaviors of PVDF membranes. Journal of Membrane Science,519, 22–31.Gray, S. R., Ritchie, C. B., Tran, T., & Bolto, B. A. (2007). Eect of NOM character-istics and membrane type on microltration performance. Water Research,41(17), 3833–3841.Han, N., Huang, G., An, C., Zhao, S., Yao, Y., Fu, H., & Li, W. (2015). Removal ofSulfonated Humic Acid through a Hybrid ElectrocoagulationâĂŞUltral-tration Process. Industrial & Engineering Chemistry Research, 54(21), 5793–5801.Harif, T., Khai, M., & Adin, A. (2012). Electrocoagulation versus chemical coagu-lation: Coagulation/occulation mechanisms and resulting oc character-istics. Water Research, 46(10), 3177–3188.Harris, D. (2007a). Chapter 3: Experimental Error. In Quantitative chemicalanalysis, seventh edition (pp. 39–52).Harris, D. (2007b). Chapter 4: Statistics. InQuantitative chemical analysis, seventhedition (pp. 53–77).Her, N., Amy, G., Foss, D., & Cho, J. (2002). Variations of molecular weightestimation by HP-size exclusion chromatography with UVA versus onlineDOC detection. Environmental Science and Technology, 36(15), 3393–3399.Hua, G., Reckhow, D. A., & Abusallout, I. (2015). Correlation between SUVA andDBP formation during chlorination and chloramination of NOM fractionsfrom dierent sources. Chemosphere, 130, 82–89.Huber, S. A. (1998). Evidence for membrane fouling by specic TOC constituents.Desalination, 119(1-3), 229–234.Humbert, H., Gallard, H., Jacquemet, V., & Croué, J. P. (2007). Combination ofcoagulation and ion exchange for the reduction of UF fouling properties of97a high DOC content surface water. Water Research, 41(17), 3803–3811.Jacangelo, J. G., Adham, S. S., & Laine, J.-M. (1995). Mechanism of Cryptosporid-ium, Giardia, and MS2 virus removal by MF and UF. Journal - AmericanWater Works Association, 87 (9), 107–121.Jankhah, S., & Bérubé, P. R. (2014). Pulse bubble sparging for fouling control.Separation and Purication Technology, 134, 58–65.Jarvis, P., Jeerson, B., Gregory, J., & Parsons, S. A. (2005). A review of ocstrength and breakage. Water Research, 39(14), 3121–3137.Jermann, D., Pronk, W., Meylan, S., & Boller, M. (2007). Interplay of dierentNOM fouling mechanisms during ultraltration for drinking water pro-duction. Water Research, 41(8), 1713–1722.Lakshmanan, D., Cliord, D. A., & Samanta, G. (2009). Ferrous and ferric ion gen-eration during iron electrocoagulation. Environmental Science and Technol-ogy, 43(10), 3853–3859. doi: 10.1021/es8036669Lee, N., Amy, G., Croué, J. P., & Buisson, H. (2004). Identication and under-standing of fouling in low-pressure membrane (MF/UF) ltration by natu-ral organic matter (NOM). Water Research, 38(20), 4511–4523.Lee, S., Dilaver, M., Park, P. K., & Kim, J. H. (2013). Comparative analysis of foul-ing characteristics of ceramic and polymeric microltration membranesusing ltration models. Journal of Membrane Science, 432, 97–105.Lee, S., & Kim, J. (2014). Dierential natural organic matter fouling of ceramicversus polymeric ultraltration membranes. Water Research, 48, 43–51.Leenheer, J. A., & Croué, J.-P. (2003). Characterizing aquatic dissolved organicmatter. Environmental Science & Technology, 37 (1), 18A-26A.Lin, H. C., & Wang, G. S. (2011). Eects of UV/H2O2 on NOM fractionation andcorresponding DBPs formation. Desalination, 270(1-3), 221–226.Maguire, I. (2009). The Eect of Coagulation Conditions on Floc Size Distributionand Morphology: Implications for Downstream Treatment Processes (Unpub-lished doctoral dissertation). Oregon State University.McBeath, S. T. (2017). Pilot capacity iron electrocoagulation scale-up for natural98organic matter removal for drinking water treatment (Unpublished doctoraldissertation). The University of British Columbia.Miller, D. J., Kasemset, S., Paul, D. R., & Freeman, B. D. (2014). Comparison ofmembrane fouling at constant ux and constant transmembrane pressureconditions. Journal of Membrane Science, 454, 505–515.Moreno C., H. A., Cocke, D. L., Gomes, J. A. G., Morkovsky, P., Parga, J. R., Peter-son, E., & Garcia, C. (2009). Electrochemical Reactions for Electrocoagula-tion Usin Iron Electrodes. Industrial & Engineering Chemistry Research(48),2275.Nariyan, E., Sillanpää, M., & Wolkersdorfer, C. (2017). Electrocoagulation treat-ment of mine water from the deepest working European metal mine âĂŞPerformance, isotherm and kinetic studies. Separation and PuricationTechnology, 177 , 363–373.Pouet, M. F., & Grasmick, A. (1994). Electrocoagulation and otation. Applica-tions in crossow microltration. Filtration and Separation, 31(3), 269–272.Richardson, S. D. (2003). Disinfection by-products and other emerging contam-inants in drinking water. TrAC - Trends in Analytical Chemistry, 22(10),666–684.Rook, J. (1976). Haloforms in Drinking Water. Journal - American Water WorksAssociation(March), 168–172.Sarathy, S., & Mohseni, M. (2007). The impact of UV/ H2O2 advanced oxidationon molecular size distribution of chromophoric natural organic matter. En-vironmental Science & Technology,, 36(xx), 160–169.Särkkä, H., Vepsäläinen, M., & Sillanpää, M. (2015). Natural organic matter(NOM) removal by electrochemical methods âĂŤ A review. Jeac, 755, 100–108.Sharp, E. L., Parsons, S. A., & Jeerson, B. (2006). Seasonal variations in naturalorganic matter and its impact on coagulation in water treatment. Scienceof the Total Environment, 363(1-3), 183–194.Shi, X., Tal, G., Hankins, N. P., & Gitis, V. (2014). Fouling and cleaning of ultra-99ltration membranes: A review. Journal of Water Process Engineering, 1,121–138.Sierra Legal Defense Fund. (October 2003). Watered down: A report on waterbornedisease in British Columbia 1980-2002 (Tech. Rep.).Stoquart, C., Servais, P., Bérubé, P., & Barbeau, B. (2012). Hybrid membraneprocesses using activated carbon treatment for drinking water: a review.Journal of membrane, 412, 1–12.Tanneru, C. T., & Chellam, S. (2012). Mechanisms of virus control during ironelectrocoagulation - Microltration of surface water. Water Research, 46(7),2111–2120.Timmes, T. C., Kim, H. C., & Dempsey, B. A. (2010). Electrocoagulation pre-treatment of seawater prior to ultraltration: Pilot-scale applications formilitary water purication systems. Desalination, 250(1), 6–13.Ulu, F., Barişçi, S., Kobya, M., Särkkä, H., & Sillanpää, M. (2014). Removal ofhumic substances by electrocoagulation (EC) process and characterizationof oc size growth mechanism under optimum conditions. Separation andPurication Technology, 133, 246–253.US EPA Oce of Water. (2016). Denition and Procedure for the Determination ofthe Method Detection Limit, Revision 2: EPA 821-R - 16-0 06 (No. December).Retrieved from https://www.epa.gov/sites/production/files/2016-12/documents/mdl-procedure_rev2_12-13-2016.pdf%0Awww.epa.govVanysek, P. (2005). Electrochemical Series. In D. R. Lide (Ed.), Crc handbook ofchemistry and physics, internet version, 2005 (p. 8-23). CRC Press.Velten, S., Knappe, D. R. U., Traber, J., Kaiser, H.-P., von Gunten, U., Boller, M., &Meylan, S. (2011). Characterization of natural organic matter adsorptionin granular activated carbon adsorbers. Water research, 45, 3951–3959.Vik, E. A., & Carlson, D. A. (1984). Electrocoagulation of Potable Water. WaterResearch, 18(11), 1355–1360.Wiesner, M. R., Clark, M. M., & Mallevialle, J. (1989). Membrane Filtration of100Coagulated Suspensions. Journal of Environmental Engineering, 115(1), 20–40.Winter, J., Barbeau, B., & Bérubé, P. (2017). Nanoltration and tight ultraltrationmembranes for natural organic matter removalâĂŞContribution of foulingand concentration polarization to ltration resistance. Membranes, 7 (3).doi: 10.3390/membranes7030034Ye, Y., Chen, V., & Le-Clech, P. (2011). Evolution of fouling deposition andremoval on hollow bre membrane during ltration with periodical back-wash. Desalination, 283, 198–205. doi: 10.1016/j.desal.2011.03.087Zhao, S., Huang, G., Cheng, G., Wang, Y., & Fu, H. (2014). Hardness, COD and tur-bidity removals from produced water by electrocoagulation pretreatmentprior to reverse osmosis membranes. Desalination, 344, 454–462.Zhou, Q., Cabaniss, S. E., & Maurice, P. A. (2000). Considerations in the useof high-pressure size exclusion chromatography (HPSEC) for determiningmolecular weights of aquatic humic substances. Water Research, 34(14),3505–3514.101

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0377512/manifest

Comment

Related Items