THE SHERMAN WATER TREATMENT PLANT MF/UF AND RO PILOT STUDY: PILOTING DURING THE 100 YEAR FLOOD

THE SHERMAN WATER TREATMENT PLANT MF/UF AND RO PILOT STUDY: PILOTING DURING THE 100 YEAR FLOOD Evan C. Ged, CH2M HILL, 12750 Merit Dr., Suite 1100, Dallas, TX 75251 Evan.ged@ch2m.com, Ph: (972) 663-2388 Brian J. Fuerst, CH2M HILL, Dallas, TX Jim C. Lozier, CH2M HILL, Phoenix, AZ Joseph R. Elarde, CH2M HILL, Naples, FL Abstract This paper summarizes the MF/UF and RO pilot study conducted on behalf of the City of Sherman for the upcoming expansion to their water treatment plant. The pilot took place March June of 2015 and involved piloting four MF/UF membranes and one RO system to treat the combined MF/UF filtrate. During May and June of 2015 the north Texas region experienced record rainfall that resulted in flooding. The source water for the plant, Lake Texoma, ultimately overtopped the spillway, designed to withstand the 100 year flood, for only the fourth time since construction in 1944. This phenomenon resulted in a dramatic change in water quality as runoff entered the lake and the overflow created a vortex at the spillway. The feed water to the pilot units experienced a dramatic increase in solids as evidence from turbidity increasing from <1 NTU to >200 NTU. The feed water also exhibited increased organics, metals, and algae content, all impacting the downstream membrane processes. The paper describes membrane performance throughout the periods of changing water quality and demonstrates how this technology has several advantages over conventional water treatment for making rapid process adjustments to combat the effects of the flooding. Introduction The City of Sherman, Texas, currently operates a 10 million gallon per day (MGD) water treatment plant (WTP) (Public Water System I.D. No. 0910006) to treat raw water from Lake Texoma. The plant uses conventional treatment (flocculation/sedimentation/filtration) for compliance with federal and state drinking water regulations followed by split-stream electrodialysis reversal (EDR) to reduce the level of hardness and dissolved solids. As a result of current regulations and growing water demands, the City of Sherman is expanding its Lake Texoma Surface Water Treatment Plant by 10 MGD of finished water capacity. Microfiltration/ultrafiltration (MF/UF) followed by reverse osmosis (RO) has been selected as the best apparent technology to meet the City s operational and water quality goals. This treatment process is the optimum solution for the City s water treatment plant expansion for the following reasons: The dual-membrane process produces the highest quality finished water possible Best cost/benefit ratio of alternative processes examined MF/UF provides highest quality RO feed water Modular and is easily expandable to match anticipated growth rate, capital improvement funding, and future expansion needs 1

Within the last few years, MF/UF has matured as a process ideally suited to meeting emerging federal and state surface water drinking regulations. Additionally, MF/UF is the pretreatment of choice for reverse osmosis systems because it provides a RO feed water with the lowest fouling potential. Despite its more than 20-year history for potable water production in the United States, the Texas Commission on Environmental Quality (TCEQ) still classifies MF/UF as innovative technology. As a result, pilot testing for a minimum of 90 days is required before the process can be permitted for use as part of the expanded City of Sherman WTP. In addition, pilot testing provides valuable information for both MF/UF system design and vendor selection. MF/UF modules from four different manufacturers were tested during the City of Sherman pilot study. In addition to the 90-day MF/UF pilot test, a 60-day RO pilot test was also performed to evaluate the effectiveness of the MF/UF pretreatment for providing stable RO performance. A singlestage, three-element RO unit was operated to simulate the lead portion of the first stage of a full scale system. The pilot unit used three low energy brackish water RO elements. During the course of piloting, the north Texas region experienced periods of intense, prolonged rainfall that led to multiple instances of flooding. As a result, the levels in Lake Texoma reached the 100 year flood elevation and overtopped the spillway. The dramatic increase in lake level, and additional particulate load from runoff, altered the composition of the lake and increased the concentration of organics and particulate matter while simultaneously reducing the concentration of dissolved solids. These water quality changes, turbidity in particular, can pose operational issues with the membrane processes by causing increased rates of fouling and loss of production. The following sections describe the results of the MF/UF and RO pilot testing for treatment of Lake Texoma surface water throughout the course of the flooding events. Pilot Study Objectives The City of Sherman pilot study was conducted to assess the ability of the MF/UF and RO system to meet the City s water quality goals, to provide a comparison of alternative MF/UF products, to collect sufficient performance information for bidding and design of the full-scale MF/UF system and to confirm design criteria for the full-scale RO system. Specific objectives of the pilot study were as follows: Organic/Inorganic Removal Improve MF/UF performance (operate at increased flux) through coagulant dosing to the raw water. MF/UF vendors were provided with the option to use ferric sulfate to coagulate a portion of the dissolved organics present in Lake Texoma water. Such coagulation has been shown to reduce membrane fouling. Where a coagulant was used, the capital and operating cost associated with its use would be accounted for in the life cycle cost of the MF/UF supplier s bid. If necessary, oxidize soluble iron (Fe) and manganese (Mn) to particulate form using KMnO4 or NaOCl. The need for reduction in these contaminants will depend on the level of total Fe and Mn in the raw water, how much Fe and Mn can be removed by the MF/UF system removal without pre-oxidation, whether pre-oxidation enhances or reduces membrane 2

performance and, where used, whether pre-oxidation results in the formation of oxidized Fe and Mn in the MF/UF filtrate (which would cause RO performance degradation). Particle (Pathogen) Removal Provide a positive barrier to the passage of Cryptosporidium and Giardia Provide filtrate turbidity of less than 0.1 NTU Demonstrate membrane integrity through daily integrity tests. Membrane Hydraulic Performance Assess impact of any chemical pre-treatment (potassium permanganate, ferric sulfate) on membrane performance, and effectiveness of maintenance clean (MC) and clean in place (CIP) Evaluate rate of membrane fouling (as measured by rate of transmembrane pressure (TMP) increase, required MC and CIP frequency and effectiveness of each in maintaining or restoring performance). Demonstrate overall MF/UF recovery of 95%. Membrane Operational Considerations Verify design and operational criteria, allowing refinements in capital and operating cost estimates. In many cases, pilot testing can be used to justify a less conservative design, resulting in significant capital and operating cost savings. Allow owners and regulators the opportunity to become familiar with the process and equipment firsthand. Compare performance of four MF/UF membrane module types and develop the operational data necessary to evaluate each on a 20-year life cycle cost basis. Specific RO Pilot Objectives Confirm suitability of MF/UF filtrate (combined from all four systems) to minimize RO fouling. Verify design flux. Compare required feed pressure and permeate quality with manufacturer s software projections. MF/UF and RO Pilot Units Characteristics of the membrane modules used in this study are presented in Table 1. To provide for a functional pilot unit for testing, a single module was housed on a skid together with other 3

equipment including necessary tankage, piping, valving and instrumentation and controls to conduct all necessary functions. Table 1 - MF/UF Module Characteristics Parameter Membrane A Membrane B Membrane C Membrane D Membrane Material PVDF PVDF PVDF PVDF 7.1 x 6.3 8.9 x 7.75 6 x 6.6 8.5 x 7.1 Module Area (ft 2 ) 600 829 538 775 Nominal Pore Size (µm) 0.02 µm (UF) 0.03 µm (UF) 0.1 µm (MF) 0.01 µm (UF) TMP Range (psi) 0-40 0-30 7-45 0-29 Backwash Methods Module Flow Range (gpm) Flow Path Coagulant Dose BW/AS with periodic MC BW/AS with periodic CEB RF/AS with periodic EFM BP/AS with periodic MC 8-33 13.6 40.9 0-40 10-48 Module Dimensions (Dia-in x Lengthft) Outside-to- Inside 0.5 ppm ferric sulfate Outside-to-Inside Outside-to- Inside Outside-to- Inside None used None used None used The RO pilot unit treated filtrate from the MF/UF units, as is anticipated for the full-scale facility. All four MF/UF pilot filtrate streams were combined into a common filtrate pipe that flows into a 500 gallon break tank. The break tank feeds the RO booster pump and flows to the RO pilot unit. The RO pilot unit is configured into a 3 element single-stage arrangement (1:1:1). The pilot was loaded with the low energy elements and allowed to run for 60 days. These elements are manufactured with a spiral wound polyamide material designed for high productivity at low pressures. A summary of the membrane characteristics and pilot operating conditions is presented in Table 2. 4

Table 2 - Characteristics of Reverse Osmosis Pilot Membrane Parameter Value Application Low pressure brackish water Configuration Spiral wound Composition Composite Polyamide Length 40 inches Diameter 3.95 inches Active Surface Area 85 square feet Maximum Pressure 600 psig Average Salt Rejection 99.2% Targeted Permeate Flow 3.7 gpm Targeted Concentrate Flow 4.5 gpm Targeted Feed Pressure Range 70 100 psig Targeted Flux 21 gfd Targeted Recovery 46% Maximum Temperature 113 F ph, Operating 6.5-7.0 (Optimum Rejection) ph, Cleaning 2.0-10.0 Feed Turbidity <1 NTU Feed SDI <5 Chlorine Tolerance <0.1 ppm Water Quality Impacts of Record Rainfall The pilot effort took place from March 6 through June 29, 2015 and was split into three 30+ day stages of operation. The first stage consisted of an optimization period where the vendors could vary the flux and cleaning protocols to optimize membrane performance. The second stage of operation involved running at a constant flux and cleaning interval and was used to determine the maximum instantaneous flux that would be allowed for bidding. The third, and final, stage of testing allowed for further data collection where the vendors could adjust operations and collect data that may be helpful in bidding the membrane system. Raw (lake) water quality remained fairly constant during the first two months of testing with raw water turbidity generally < 1 NTU and TOC typically around 4.3 mg/l. As part of the pilot 5

testing protocol, a turbidity spike was required to challenge the MF/UF modules with a higher load of suspended solids. During Stage 3 testing, several heavy rainfall events caused elevated levels of turbidity in Lake Texoma, which satisfied this requirement. The heavy rainfall in the north Texas region caused Lake Texoma to reach the flood gauge elevation, equivalent to the 100 year flood. As the flows entered the Denison Dam spillway a giant vortex was created that reached 8 feet in diameter. The combination of runoff into the lake and the turbulence caused by the vortex greatly increased the suspension of lake sediment and turbidity levels. The raw water turbidity began to rise on May 19, 2015 reaching levels as high as 200 NTU on June 1, 2015. The conventional plant was shut down for half a day to allow the turbidity to stabilize as the operations staff had concerns that the conventional dual media filters would not be able to process the 200+ NTU water, however, the pilot systems remained in operation. Within a twelve hour period the turbidity stabilized around 30 NTU and remained at this level for the remaining 28 days of pilot testing. All rainfall events for Lake Texoma were recorded throughout the pilot study and presented in Figure 1, cumulative rainfall for this period is presented in Figure 2. All data was collected by the U.S. Army Corp of Engineers and presents precipitation measured at the Lake Texoma Dam station as well as the average from all monitoring stations in the Lake Texoma watershed. In addition to the change in raw water turbidity, several other water quality parameters were affected by the rainfall events. Table 3 summarizes minimum and maximum values observed before and after the rainfall events. The increases in metals, such as aluminum, iron, and manganese, can cause fouling of the RO elements when present in concentrations observed during the post-rainfall period. Having MF/UF ahead of the RO provides an effective barrier against these metals as the MF/UF filtrate metals concentrations were not impacted by the increases in raw water metals concentrations. Table 3 - Pre- and Post-Rainfall Raw Water Quality Water Quality Parameter Pre-rainfall Post-rainfall Aluminum (mg/l) 0.02 2.02 Total Iron (mg/l) < 0.01 1.03 Total Manganese (µg/l) < 2.0 35 TDS (mg/l) 1,200 500 UVA254 (cm -1 ) 0.058 0.413 TOC (mg/l) 4.25 5.68 SUVA254 (L/mg/m) 1.50 7.72 Turbidity (NTU) 0.370 >200 Total hardness (mg/l as CaCO3) 376 196 6

4.5 4 3.5 Dam (in) Basin (in) 3 Rainfall (inches) 2.5 2 1.5 1 0.5 0 3/1/2015 4/1/2015 5/1/2015 6/1/2015 Figure 1 Daily precipitation data for Lake Texoma 40 35 30 Dam (in) Basin (in) Rainfall (inches) 25 20 15 10 5 0 2/25/2015 3/17/2015 4/6/2015 4/26/2015 5/16/2015 6/5/2015 6/25/2015 Date Figure 2 Cumulative precipitation data for Lake Texoma 7

Another factor impacting the performance of the membranes was the large amount of algal material and zebra mussels conveyed to the plant. The existing WTP routinely switched between operating two separate sets of intake pumps. When the larger pumps, also deeper in the lake, were turned on after prolonged periods of not operating they would send slugs of algae and zebra mussels to the plant. It is common practice for the operators to also let the first flush of material bypass the plant, but still deal with a substantially increased solids load throughout the operation of these intake pumps. This phenomenon was made worse during the period of heavy rainfall as the vortexing lake enhanced suspension of the filamentous algae. During this period of operation the strainers on the pilot units were challenged to remove this material. Staff had to clean the wye strainer ahead of the raw water feed tank up to 3 times per day. Even with screening this large material, a substantial load was reaching the pilot units. With effective pre-filters in place there were no detrimental impacts on the membranes. However, the wedge wire type strainers encountered more operational issues compared to a polypropylene stacked disc type. Following the pilot study it was decided to implement the stacked disc prefilters as well as a balancing basin into the full scale design which would serve multiple purposes: Allow large debris such as filamentous algae and zebra mussels to settle out ahead of the membrane process; Create a consistent flow and upstream head condition for feeding the membrane pumps; And serve as an application point for pre-oxidants if they are needed in the future. A picture of the incoming material is shown in Figure 3. Zebra Mussels Figure 3 Filamentous algae and zebra mussels in the raw water feed tank strainer 8

MF/UF Membrane Performance Performance of the MF/UF membranes was continuously monitored throughout the entire 90 day pilot study. The systems were operated using a PLC/PC based automated control system with data logging and remote system monitoring. These were programmed to automatically control all filtration operations, including fluid cleansing of the membrane. In normal operation, the PLC/PC monitored and recorded trans-membrane pressure, flow rates, turbidity for performance monitoring. In addition, the PLC/PC also monitored operating temperature, air scrub parameters (delivery pressure, flow rate and duration) and other parameters useful for optimizing operation. The PLCs were equipped with remote-monitoring capability, allowing adjustments to the unit s operating parameters via modem connection to the pilot unit PC. The pilot procedure for all membranes involved optimizing flux rate and chemical cleaning concentrations during Stage 1 testing. This period was followed by Stage 2 operation at the maximum allowable instantaneous flux that would be used for bidding the membrane systems. All membrane suppliers were required to maintain this flux throughout the 30 day Stage 2 period and for the first 15 days of Stage 3, prior to making any operational changes. The maintenance clean (also referred to as chemically enhanced backwash (CEB) or enhanced flux maintenance (EFM)) procedure required hypochlorite soak and recirculation at 3 day intervals using concentrations determined by each of the vendors. A more extended recovery clean (also referred to as clean-in-place (CIP)) was performed at the end of each stage using hypochlorite and citric acid for the removal of organic and inorganic foulants, respectively. To assess the effectiveness of the recovery cleans and determine the amount of irreversible fouling that occurred during each stage of testing, a clean water permeability test (CWPT) was performed following each CIP, prior to the start of the subsequent stage of testing. CWPT s were done using the cleanest water available at the plant site. This water had undergone dual media filtration as well as pumped through a 10 µm cartridge filter ahead of the existing WTP electrodialysis reversal system (EDR). Each membrane vendor tabulated the results of the CWPT at 70%, 100%, and 130% of the maximum instantaneous flux and compared the post-stage 1, post-stage 2, and post-stage 3 results to assess loss of permeability. The performance of the individual membranes is discussed in the sections below. Membrane A During the optimization period this manufacture tested fluxes ranging from 47 72 gfd and chose 72 gfd as the optimal flux to use during the Stage 2 test period. To enhance performance they applied a 0.5 ppm ferric sulfate coagulant in a mixing tank ahead of the pilot skid. The precoagulation was used to reduce fouling of the UF membrane by coagulating the larger molecular weight (colloidal) fraction of organic matter. This is the portion of organics that is most capable of adsorbing to the membrane surface or in the membrane pores and causing a reduction in permeability. This in turn allowed the membrane to operate at a much higher flux than without pre-coagulation. During the first two stages of testing the system was able to operate at 96.3% recovery (with no backwash recycle). During the same time, the average feed water turbidity was less than 2 NTU and the membrane produced an average filtrate turbidity of 0.016 NTU. Feed and filtrate turbidity data for Membrane A is presented in Figure 4. 9

Figure 4 - Membrane A turbidity and raw water temperature data As indicated in Figure 4, the turbidity levels were stable around 1 2 NTU until early May. The turbidity began rising throughout May and on June 1, 2015 the incoming turbidity to the plant reached values over 200 NTU and was unable to be detected by the turbidimeters on the pilot units, but was confirmed in the lab at the plant. The online monitoring system for Membrane A recorded a maximum turbidity measurement of 100 NTU, but stabilized around 40 NTU after about 12 hours. The membrane was able to process the high turbidity water, as indicated by maintaining an average filtrate turbidity of 0.015 NTU throughout Stage 3. However, the greatly increased solids loading on the fibers caused a process upset that resulted in the system reaching the terminal TMP of 40 psi. Figure 5 shows the TMP, permeability, and flux recorded throughout the entire pilot study. Figure 6 shows the TMP, permeability, and flux temperature corrected 20 C. The TMP rise rate was very rapid, increasing from 10 psi to 40 psi over the course of about 12 hours. Upon reaching the terminal TMP the system proceeded with an automatic shutdown. The next morning a recovery clean was performed to remove the foulants that may have caused the TMP increase. Upon restarting the system the TMP rise rate remained around 5 psi/day. Even though the TMP increase was not sustainable, the filtrate water quality was not impacted. It was decided that the aggressive flux of 72 gfd was not appropriate for these water quality conditions. The manufacturer decided to decrease the flux to 61 gfd and increase maintenance clean frequency to daily for the remainder of the pilot study. The TMP was able to be managed at these conditions and stabilized between 15 25 psi for the remaining 3 weeks of piloting. This demonstrated the ability of Membrane A to still run at a high flux rate and produce high quality 10

filtrate. The ability to quickly and easily make process adjustments to the rapid change in water quality is one of the main advantages of the membrane systems. Figure 5 Membrane A TMP, Permeability, and Flux Figure 6 Membrane A TMP, Permeability, and Flux temperature corrected to 20 C 11

Overall, the membrane performance was excellent through the first two stages of testing and Stage 3 performance was adequate for incorporation into a full scale plant in the very rare event of flooding of this magnitude. This membrane demonstrated the ability to maintain permeability following the high turbidity event as evidence by the post Stage-3 clean water permeability test results. The membrane achieved 96.3% of the clean water permeability when compared to post- Stage 1 results. Additionally, the flux used in this system was 26% greater than the next highest flux rate used in the other membrane systems (57 gfd) and was sustainable under typical water quality conditions. Process adjustments such as increased cleaning frequency or concentrations and adjustments to flux are easily manipulated through the control system and provide a major advantage over conventional water treatment. Membrane B During the optimization period, this membrane supplier performed a critical flux test where they varied flux between 35 50 gfd to determine where the TMP rise rate became non-linear. It was determined that the sustainable (sub-critical) operating flux was 45 gfd and would be used in Stage 2 testing. Through the first few weeks of testing the monitoring data exhibited a high degree of scatter and inconsistency. This was believed to be caused by a malfunctioning 3-way valve causing the system to not return to operation following a backwash. The system was shut down from March 24 26 to replace the valve, after which the flux, turbidity, and temperature readings appeared more stable. During the remaining 2 weeks of Stage 1 the filtrate turbidity was consistently below 0.01 NTU and the TMP ranged from 10 15 psi. Figure 7 shows feed turbidity, filtrate turbidity and temperature data for Membrane B. It should also be noted that the turbidimeter for this pilot unit was scaled to read a maximum of 20 NTU in the PLC. Stage 2 and Stage 3 values for feed turbidity were measured at 30-40 NTU in the lab. Figure 7 Feed turbidity, filtrate turbidity, and water temperature for Membrane B 12

The system operated at 45 gfd and 96.5% recovery through Stage 2. Feed and filtrate turbidity measurements were consistent with Stage 1 testing until April 30 when an increase in filtrate turbidity was noticed. From April 30 till May 6 the filtrate turbidity increased from less than 0.01 NTU to over 0.1 NTU. It was discovered that a faulty drain valve was causing the issue and was also resulting in higher TMPs and decreased membrane permeability. The valve was replaced and a CIP was performed on May 7 to remove the solids built up on the membrane. To meet the requirement for 30 days of continuous testing during Stage 2, the vendor elected to restart Stage 2 testing following the May 7 CIP to allow for 30 days of uninterrupted data collection. Upon resuming operation the TMP stabilized around 8 16 psi and filtrate turbidity dropped to about 0.01 NTU. Figure 8 illustrates the TMP, permeability, and flux for Membrane B and Figure 9 shows these parameters temperature corrected to 20 C. Figure 8 TMP, permeability, and flux for Membrane B This system performed best during Stage 3 while water quality was at its worst. TMP ranged from 6 10 psi and permeability was highest at 4 7 gfd/psi indicating that the heavy rains in the Lake Texoma catchment had no discernable impact on these parameters. The filtrate turbidity increased during this phase of testing, correlating closely with feed turbidity (see Figure 7), but these levels were still acceptable for a full-scale system operation at <0.05 NTU 95-percent of the time. Clean water permeability test results also demonstrated this membrane s ability to maintain permeability throughout the 90 day test period where the post-stage 3 clean water permeability was 99.4% of the post-stage 1 values. 13

Figure 9 TMP, permeability, and flux for Membrane B temperature corrected to 20 C Membrane C During Stage 1 this manufacturer experimented with ferric sulfate doses, ranging from 1 18 ppm. However, after 10 days of operation ferric sulfate pre-dosing was terminated as the manufacturer did not see that such dosing enhanced operating flux nor reduced TOC to the degree that would justify the capital and operating cost of coagulant use. Flux was also varied between 40 60 gfd to determine the most appropriate flux to provide a fouling rate that allowed for CIPs no more than once every 30 days and maintenance cleans once every 3 days. It was determined that 50 gfd would be the flux used for Stage 2 and bidding. The pilot system was able to run successfully throughout Stage 2 testing at a flux of 50 gfd and a recovery of 96.1%. The membranes maintained a permeability of at least 2 gfd/psi during this test period even though it was discovered that the hypochlorite feed pump was not functioning properly. During the 34 day Stage 2 testing, eight of the maintenance cleans circulated warm water (90 100 F) only. When manually adding hypochlorite to the filtrate tank, the maintenance cleans effectively reduced the TMP to 5 7 psi. Even with non-chlorinated maintenance cleans, the Stage 2 test conditions demonstrated this membrane would be acceptable for use in a full scale application with average temperature corrected permeability of 4.1 gfd/psi, average feed turbidity of 1.1 NTU and average filtrate turbidity of 0.017 NTU. During Stage 3 testing the system performed better than Stage 2, despite the increase in average feed turbidity from 1.1 to 11.2 NTU. Under these conditions the average temperature corrected permeability increased to 4.8 gfd/psi and the filtrate turbidity remained relatively constant at 0.018 NTU. Additionally, the clean water permeability following Stage 3 was 96.6% of the values measured after Stage 1, demonstrating the membranes ability to maintain permeability 14

through water quality excursions. Figure 10 shoes turbidity and water temperature data for Membrane C, Figure 11 shows TMP, permeability, and flux, and Figure 12 shows temperature corrected data. Figure 10 Feed turbidity, filtrate turbidity, and water temperature for Membrane C Figure 11 TMP, permeability, and flux for Membrane C 15

Figure 12 TMP, permeability, and flux for Membrane C temperature corrected to 20 C Membrane D Membrane D performed well throughout the entire 90 day pilot study, running at a flux of 57 gfd without the addition of coagulant. The flux was originally set at 47 gfd during Stage 1, but nearzero TMP rise was observed over the first 25 days of testing. The raw water temperature was also at its coldest during the first 25 days with temperatures ranging from 5 11 C, corresponding to a temperature corrected flux (20 C) of 72 gfd. Prior to starting the Stage 2 test period the flux was adjusted to 57 gfd, corresponding to a 72 gfd temperature corrected flux when the raw water temperature was at 14 C at the start of Stage 2. Throughout Stage 2 the system operated at 57 gfd with a recovery of 96.2% and maintained TMP between 5 10 psi. A slight increase in TMP was observed just prior to the completion of this Stage of testing due to a failure to initiate two maintenance cleans. Regardless of missing these cleaning events, the membrane demonstrated high permeability in the range of 5 12 gfd/psi and maintained an average filtrate turbidity of 0.014 NTU. During Stage 3, and the period of flooding, there was no noticeable change in membrane performance. Average filtrate turbidity remained at 0.015 NTU, TMP remained between 5-10 psi, and permeability remained between 5 12 gfd/psi. The stable performance clearly demonstrated the robustness of Membrane D when operated at the selected flux, recovery, and cleaning conditions established in Stage 3. The results of the clean water permeability test showed a loss of permeability of approximately 15% when compared to Stage 1. This could be a result of a less aggressive CIP protocol (shorter recirculation/soak time and lower chemical concentrations) compared to Membranes A and C. Figure 13 shows turbidity and water 16

temperature data, Figure 14 shows TMP, permeability, and flux, and Figure 15 shows the temperature corrected data. Figure 13 Feed turbidity, filtrate turbidity, and water temperature for Membrane D Figure 14 TMP, permeability, and flux for Membrane D 17

Figure 15 TMP, permeability, and flux for Membrane D temperature corrected to 20 C RO Pilot Results The goal of the RO study was to quantify performance of the lead portion of a full-scale RO system in order to confirm the quality of filtrate produced by the membrane filtration systems was suitable to limit fouling and provide for stable RO performance. The single-stage pilot was ran in a single-pass configuration (no concentrate recycle) to simulate the lead elements in the full scale system. The system was monitored continuously over the 60 day period and data was recorded once daily on manual log sheets. Membrane performance was tracked using the manufacturer s proprietary normalization software and evaluated based on salt passage and permeate flow. Feed, concentrate and permeate pressures, feed and concentrate flows, feed temperature and feed and permeate conductivity were recorded manually and on a daily basis by reading values off the associated instrumentation on the RO unit. These data were used by the manufacturer-provided data normalization software to calculate and trend the following performance parameters: Recovery Normalized permeate flow Normalized salt rejection and passage Differential pressure (feed-concentrate) Heavy rainfall that occurred beginning at 30 days into the two-month testing period caused the conductivity to decrease from a maximum of 1868 µs/cm to a minimum of 858 µs/cm, nearly a 55% reduction. This caused permeate conductivity to decrease from a maximum of 150 µs/cm to 18

a minimum of 69 µs/cm. The three elements exhibited an initial normalized salt passage (NSP) of 4%, significantly greater than the manufacturer s rated passage of 0.8-1.0%. As importantly, NSP increased to ~8% over the course of the study, which corresponds to a 100% increase in permeate TDS. The NSP increase was considered to be the result of unplanned exposure of the elements to free chlorine (or other unidentified oxidant). Free chlorine may have been present in the RO feed water as a result of residual concentrations following a MC or CIP from the MF/UF pilots. To prevent such exposure, sodium bisulfite dosing to the RO feed water was initiated on June 1, 2015 to quench any oxidant that might be present. NSP remained stable for the first two weeks following the implementation of SBS dosing, however NSP began to rise again on June 14, 2015. Normalized permeate flow (NPF) also began to decline about 6 weeks into the RO pilot testing (around June 10, 2015). This decline coincided with an increase in the organic loading in the RO feed water as measured by steady increases in TOC and UV254. Although MF/UF pretreatment will remove particulates from the RO feed water, these membranes only effectively reduced TOC by approximately 10% and UV254 by about 50%. After completion of the pilot testing, the three RO elements were removed, packaged and shipped to a third party to conduct wet testing (measurement of product flow, differential pressure, and salt rejection at the manufacturer s standard test conditions). The data confirmed the increase in NSP and autopsy results indicated: The membrane was free from mechanical damage based on physical appearance and vacuum testing. The membrane surface and feed spacer contained an orange-brown gelatinous foulant, composed primarily of clays and organic material. Loss on Ignition (LOI) showed 43% of the foulant was organic in nature with the remainder being inorganic. Fujiwara test results were positive, indicating halogenation of the membrane structure. Halogenation was also indicated through dye testing, which showed uniform dye uptake by the membrane and penetration through the entire membrane structure. The wet test results, when considered together with the Fujiwara and dye test results, clearly indicate the elements were exposed to free chlorine, causing degradation of the polyamide rejection layer and the loss of salt rejection. Chlorine exposure most likely resulted when a hypochlorite maintenance clean was conducted on one or more of the MF/UF pilot units and residual chlorine solution was not sufficiently flushed from the unit prior to its return to service, resulting in chlorine containing filtrate passing into the RO unit. With a full-scale MF-UF/RO system, on-line instrumentation (e.g., ORP, free chlorine analyzer) would be installed in the RO feed piping to prevent such exposure. Cleaning frequency could also be increased to remove organic foulants that occur during water quality excursions. Permeate Blending A key aspect of the RO pilot study was to generate RO permeate that could be blended with MF/UF filtrate in order to provide a treated water that would simulate the finished water quality produced by the future MF/UF and RO facility. This was accomplished by preparing a 50/50 (v/v) blend of MF/UF filtrate and RO permeate. It should be noted, however, that the finished water TDS will be slightly higher in the full scale plant since the pilot unit did not account for multiple stages and arrays of RO elements. The blend water samples were analyzed for ph, conductivity, hardness, TOC, UVA254, and simulated distribution system (SDS) total 19

trihalomethanes (TTHMs) and regulated haloacetic acids (HAA5). The SDS procedure involved the following steps: 1. adjusting the blend water sample ph to 8.0; 2. dosing the sample with free chlorine to achieve a residual of 5.0 mg/l for a contact time of 5 minutes; 3. dosing the sample with ammonia to convert the free chlorine to chloramines and achieve a target combined chlorine residual of 3.5 3.7 mg/l; and 4. quenching the chlorine residual with sodium thiosulfate after 1 and 5 days contact time. For the seven samples sent for analysis, the TOC ranged from 2.13 2.45 mg/l, a total reduction of organics of over 50% compared to the raw water. TTHMs and HAA5 measured after 1- and 5- day contact times represent the range of DBP levels anticipated based on the minimum and maximum residence times in the distribution system. For the four sample dates (April 15 to June 10), DBP formation potential steadily increased due to the progressively higher TOC concentrations in raw water (and hence UF filtrate). TTHMs ranged from 22.1 37.5 µg/l and HAA5 from 13.6 32 µg/l. The maximum concentrations were approximately half of the maximum contaminant levels. These results, along with those presented earlier for MF/UF, indicate that the combination of MF/UF and partial RO treatment provides an effective and robust treatment system for the removal of both pathogens and DBP precursors to provide a finished water fully compliant with EPA and State of Texas drinking water regulations. To gather more information for detailed design of the RO system, additional pilot testing was conducted in the fall of 2015 following the award of the MF/UF contract. The selected MF/UF supplier was required to provide a pilot unit for up to 3 months that was used to generate RO feed water. The MF/UF unit was operated without the use of hypochlorite maintenance cleans to avoid exposure of the RO membranes to free chlorine. The same RO element type was used so that data from both the first and follow-on testing can be directly compared to show that the loss of salt rejection experienced during the first test was caused by chlorine exposure and not by the quality of the MF/UF filtrate to be produced at full-scale. Summary and Conclusion A summary and comparison of the MF/UF pilot results is presented in Table 4. As evidence by the consistent filtrate water quality, these membranes can effectively produce high quality water even during periods of poor source water quality. The conclusions drawn from the pilot study include: Membrane B, C, and D were successfully operated at a stable flux and with a recovery >95%. Membrane A operated at > 95% recovery but required a 15% reduction in flux during the flood events. The systems were able to achieve a 30 day CIP and 3 day maintenance clean interval when operated at the flux and recovery established in Stage 2. Filtrate turbidity, TMP, and permeability were unaffected by the changing water quality for membranes B, C, and D. For Membrane A, TMP rise rates were able to be managed by adjusting flux and cleaning intervals. 20

All membranes provided an effective barrier against the passage of pathogens such as Cryptosporidium and Giardia as demonstrated by passing daily integrity tests. All four products were found to be acceptable for implementation in a full scale design, and all vendors were invited to bid on the project. 21

Table 4 Summary of MF/UF Piloting Parameters (Stage 2 only) Parameter Membrane A Membrane B Membrane C Membrane D Instantaneous flux 72 gfd 45 gfd 50 gfd 57 gfd System recovery 96.2% 96.5% 96.1% 96.2% Feed water temperature 14 21 C 13 23 C 13 20 C 13 20 C Permeability 6.8 11.8 gfd/psi 3 6 gfd/psi 2 11 gfd/psi 4.2 12.2 gfd/psi Permeability (@ 20 C) 6.6 10.1 gfd/psi 2 6 gfd/psi 0.6 11.4 gfd/psi 4.7 13.6 gfd/psi TMP 7-12 psi 8 20 psi 5 25 psi 7.4 13.2 psi TMP (@ 20 C) 6 13 psi 5 22 psi 4 24.5 psi 3.6 11.7 psi Average feed turbidity Average filtrate turbidity 2.2 NTU 7.2 NTU 1.5 NTU 0.29 NTU 0.016 NTU 0.015 NTU 0.017 NTU 0.014 NTU Coagulant dose 0.5 ppm ferric sulfate None None None BW interval 14 minutes 40 minutes 20 minutes 30 minutes BW air flow 6 7 scfm 30 seconds at 4 scfm followed by 30 second module drain BW liquid flow 60 second backpulse at 8 gpm, 30 second feed water flush at 15 gpm 30 second reverse flow at 33 gpm, followed by 30 seconds of forward flush at 12 gpm 3 scfm 30 60 seconds at 4 scfm 8 gpm (simultaneous with air scour) 30 60 seconds at 34 gpm BW duration 90 seconds 180 seconds 60 seconds 120 seconds 22

Table 4 Summary of MF/UF Piloting Parameters (Stage 2 only) Parameter Membrane A Membrane B Membrane C Membrane D MC intervals Every 72 hours Every 72 hours Every 72 hours Every 72 hours MC duration 15 minutes with chemical solution recirculated at 5 gpm 20 gpm flow for 30 seconds followed by a 15 minute soak 30 minutes 20 minutes with chemical soak MC chemical NaOCl at 500 mg/l NaOCl at 350 mg/l NaOCl at 500 mg/l NaOCl at 300 mg/l CIP interval Every 30 days Every 30 days Every 30 days Every 30 days CIP solution CIP duration NaOCl at 500 mg/l Citric Acid at 2,000 mg/l HCl to bring ph to 2.1 Temperature set at 95 F NaOCl soak for 6 hours Citric Acid soak for 6 hours NaOCl at 2,000 mg/l with caustic at ph 12 HCl at ph 2.0 Temperature of 95 F NaOCl circulation for 45 minutes followed by soak for 1 hour then recirculation again for 45 more minutes. Repeat for HCl. 0.2% NaOCl, 1% NaOH, 2% citric acid NaOCl recirculation for 2 hours at 90 100 F followed by Citric Acid recirculation for 1 hour at 90 100 F NaOCl at 3,000 mg/l citric acid at 5,000 mg/l Temperature set at 31-35 C NaOCl circulation for 3 hours followed by citric acid circulation for 2 hours 23

The ability of the MF/UF systems to produce filtrate with stable and very low turbidity while treating drastically changing lake quality, without the need for chemical clarification and requiring little or no operational adjustments, clearly demonstrates the major benefit of membrane filtration over conventional treatment. This was illustrated by need for the City to briefly stop distribution of finished water from their existing WTP because of excessive solids loading to the granular media filters. Moreover, the stable MF/UF filtrate ensured the downstream RO unit was able to operate without a concern about fouling from increased particulate concentrations. As the United States and other parts of the world experience more intense weather events, such as that experienced during the City of Sherman pilot testing, the importance of treatment technologies like MF/UF to operate effectively and with little operator attention will become more important. 24