Microfiltration for Removal of Manganese from Surface Water

Microfiltration for Removal of Manganese from Surface Water Carl Schneider, Ph.D., P.E., Senior Process Engineer Wiedeman & Singleton, Inc., Atlanta, GA Peter Johns, P.E., Vice-President Wiedeman & Singleton, Inc., Atlanta, GA Robert P. Huehmer, Process Manager US Filter, Timonium, MD Abstract Experiments were conducted to investigate the alternatives for manganese removal when microfiltration was used instead of media filtration. Comparative bench studies were conducted with a variety of oxidants, including: chlorine, chlorine dioxide, permanganate and hydrogen peroxide. Pilot-scale microfiltration experiments were conducted at three sites with hypochlorite, chlorine dioxide, ozone and permanganate. Pilot-scale microfiltration experiments indicated that the effectiveness of each oxidant on manganese removal was site specific. For the most difficult water tested, 99% of the manganese was removed when 0.5 mg/l of chlorine dioxide was dosed using a twenty-minute reaction time. 1.0 Introduction 1.1 Chemistry of Manganese Removal In some regions of the United States, manganese is a common element in the soil where it exists primarily as manganese dioxide, which is very insoluble in water. Under anaerobic conditions manganese is reduced from an oxidation state of IV to II and becomes soluble 1. Similar reactions can occur with iron at the same time. If manganese can be re-oxidized to an oxidation state of IV, the manganese will become insoluble again. While manganese can be oxidized by chlorine, this reaction may proceed slowly and may not occur until after filtration or in the distribution system. Manganese can cause staining of clothes and plumbing fixtures, and incrustation of water mains, which can result in black water and debris at the customer s tap 2. Manganese is more difficult to oxidize than iron 3, hence if the treatment process oxidizes and removes manganese; it also oxidizes and removes iron. The raw water manganese is often in excess of the Environmental Protection Agency s recommended level of 0.05 mg/l for finished water. Water from impounded reservoirs often has increased concentrations of manganese and iron in the summer time due to thermal stratification of the reservoir and organic loads in the ambient water. 1 of 11

Manganese removal in potable water treatment plants is normally achieved by the following mechanisms:? Oxidation of manganese and removal of the insoluble oxide by sedimentation or filtration; and,? Adsorption to a manganese selective media such as manganese dioxide coated media or ion exchange resin. 1.2 Manganese Removal Utilizing Adsorption and Oxidation Removal of manganese by adsorption to a commercially available zeolite mineral known as greensand was developed in the last decades. It is now apparent that removal of manganese in conventional water treatment plants by adsorption followed by oxidation can occur without planning as the granular media filters function as naturally occurring green sand filters. Knocke, Occiano and Hungate 4 quantified the presence of the manganese in the coating by extraction of the coating from samples of filter media from thirteen water treatment plants in Virginia and North Carolina. They used a 0.5 % nitric acid solution, a strong reducing agent (hydroxylamine sulfate) and a reaction time of two hours. The amount of manganese extracted ranged from nil to 60 milligrams per gram of media (mg/gm) (6%). By comparison, an extraction of virgin greensand by the same method yielded 4.3 mg/gm (0.4 %). A similar extraction was conducted on a sample of media from a water treatment plant in South Carolina 5. Nitric acid was used, but sodium thiosulfate was substituted as the reducing agent. Before the extraction, the sand was brown and after the extraction, the sand was restored to the original off-white color. Knocke et al. 4 determined that the capacity of the media to remove manganese (without the presence of chlorine) was a function of the amount of manganese in the coating and the ph. If chlorine residual is present, the rate constants for adsorption of manganese increased. In addition, the presence of chlorine extended the capacity of the media for adsorption of manganese without showing any indications of exhaustion of the media capacity. 2.0 Case Studies 2.1 Study A: Manganese Removal by Ozonation/Microfiltration: Avon, CO The cost of real estate in the resort communities of Colorado makes the development of conventional water treatment plants less attractive. The source water is low in dissolved organic matter, although high turbidity and color often occur during high flow periods. Membrane treatment has been increasingly used to produce potable water in situations where a smaller footprint is desirable and the source water is relatively free of dissolved organics that may result in disinfection by products. Recent residential and golf course development in the Vail, Beaver Creek and Avon communities has reached the limit of the capacity of the existing Avon Water Treatment Plant. The Eagle River, which runs through these communities, is typical of Colorado mountain rivers, in that it is composed primarily of snowmelt and rainfall from the mountainous terrain. Possessing seasonal turbidity spikes, the Eagle River also 2 of 11

possesses concentrations of manganese exceeding EPA s Secondary Maximum Contaminant Level of 0.05 mg/l. The existing water treatment plant utilizes ozonation, rapid mix, flocculation, sedimentation, and filtration to provide finished water. A membrane pilot study was initiated with ozone to oxidize manganese prior to removal by microfiltration. The ozone concentration was adjusted by varying the voltage applied to the cell. Ozone was educted into a plug-flow contactor with a nominal 3 minute detention time. The oxidized water was then stored in a tank containing a jet aerator to strip residual ozone. The detention time of the aeration step was approximately 10 minutes. The nominal ozone dose was maintained between 0.5 and 1 mg/l of ozone applied. Removal of the ozone prior to microfiltration was required to prevent oxidative damage to the membrane. The ozonated and stripped water was then filtered using a dead-end flow microfiltration system. The membrane pilot unit consisted of a feed/clean-in-place tank, membrane feed pump, three membrane modules, and associated piping and automatic valves. Each membrane module contained approximately 20,000 hollow fibers fabricated of polypropylene (PP), approximately 1 meter in length. The nominal pore size of the PP membranes was 0.2?m. During the demonstration, a series of paired water samples (raw water and MF filtrate) were taken for analyses of manganese (Figure 1). The mean manganese concentration in the raw water was 0.13 mg/l. The mean of the treated water was 0.024 mg/l with all samples below the EPA guideline of 0.05 mg/l. Due to the variability in the data two statistical analyses were performed to draw inference from the data. 0.3 Figure 1 Removal of Manganese by Ozone in Pilot Study A Manganese Concentration (mg/l) 0.25 0.2 0.15 0.1 0.05 Raw Water Filtrate EPA Guideline 0 0 1 2 3 4 5 6 7 8 Sample No. A paired student t-test was performed to compare the means of the raw and treated water manganese concentrations. The test indicated that the removal of manganese 3 of 11

observed using ozonation and microfiltration is significant to a 95% confidence level. Additionally, a test was performed to determine whether the mean filtrate manganese concentration was significantly less than the EPA recommendation of 0.05 mg/l. The analysis indicated that the filtrate manganese concentration was less than 0.05 mg/l with a confidence exceeding 95%. The demonstration successfully indicated that ozone might be used to oxidize manganese found in surface water prior to removal using microfiltration. The demonstration period was not of long enough duration to evaluate the extent of membrane fouling incurred as a result of the manganese. However, the use of polypropylene membranes necessitated the inclusion of an ozone stripping process step before the membranes that may not be necessary with an oxidant resistant membrane material. Backwashes were performed every 22 minutes with compressed air and feed water to remove particulate matter from the membrane surface. During normal operation of the unit, some membrane fouling was observed that was not mitigated by the backwash sequence. Clean in place (CIP) procedures were implemented to remove foulants from the membrane surface and restore the system trans-membrane pressure to the original state prior to fouling. Citric acid and sodium hydroxide with surfactant were used as the clean-in-place chemicals. 2.2 Study B: Manganese Removal by Free Chlorine/Microfiltration: Fairmont, WV The Fairmont Water Treatment Plant treats surface water withdrawn from the Tygart River that has been impounded in a raw water reservoir adjacent to the water treatment plant. Several miles upstream of the river pump station, the Tygart River is dammed with a resultant lake. The raw water quality of the impounded water is variable with an average turbidity of 6.5 NTU and range of between 1 and 100 NTU. The existing water treatment plant practices pre-chlorination, and operators were unaware of the presence of manganese in the raw water. Once the presence of manganese was confirmed, bench-sale studies were performed to evaluate the free chlorine concentration and reaction time required to oxidize the soluble manganese. Based upon these results, the pilot study was modified to incorporate pre-oxidation of manganese using sodium hypochlorite. Raw water from the adjacent reservoir was to a 600-gallon tank. Sodium hypochlorite was dosed prior to the tank to maintain a free chlorine residual of 1 mg/l into the microfiltration units. The tank was configured with an overflow to permit constant flow from the chemical treatment building. The approximate detention time of the water was 30 minutes prior to filtration. The membrane pilot unit consisted of a feed/clean-in-place tank, membrane feed pump, three membrane modules, and associated piping and automatic valves. Each membrane module contained approximately 20,000 hollow fibers fabricated of polyvinylidene fluoride (PVDF), approximately 1 meter in length. The nominal pore size of the PVDF membranes was 0.1?m. Backwashes were performed every 22 minutes by compressed air with feed water to remove particulate matter from the membrane surface. Occasional clean in place procedures were implemented to remove foulants from the membrane surface and restore the system trans-membrane pressure to the original state prior to fouling. Citric acid and sodium hypochlorite were used as the clean-in-place chemicals. 4 of 11

Manganese concentrations in the raw water were routinely measured (Figure 2). Seasonal variations of dissolved manganese levels of up to 0.38 mg/l were observed and were correlated to the release of water from the lake. It is hypothesized that reducing conditions at the lake bottom account for the increases in manganese. Figure 2 Manganese Removal by Hypochlorite in Pilot Plant B Manganese Concentration 0.200 Raw Water 0.180 Filtrate 0.160 0.0 mg/l 0.140 Free Chlorine 0.120 0.100 0.080 0.060 0.040 0.020 0.000 4/15/99 4/25/99 5/5/99 5/15/99 5/25/99 2.5 2 1.5 1 0.5 0 Free Chlorine Concentration The mean raw water manganese concentration was 0.0947 mg/l. The mean filtrate manganese concentration was 0.0223 mg/l. A paired student-t statistical analysis confirms the results indicated in the Figure 2. The manganese concentrations in the filtrate are lower than the feed water concentrations with a 95% confidence level. Furthermore, the filtrate manganese concentration is less than EPA s recommendation manganese concentration of 0.05 mg/l. In order to evaluate whether permanent fouling of the membrane was occurring, data was collected for the determination of the hydraulic resistance (Rm). The calculation of hydraulic resistance corrects for differences in the flow in the membranes and for changes in the viscosity due to temperature. The membrane resistance was calculated using the following formula: R m =??????J) where, R m = membrane resistance m -1?P = transmembrane pressure??? = absolute viscosity J = Flux, m 3 /m 2 /day This data (Table 1) indicates that permanent fouling of the membrane was not occurring. The apparent decrease in the clean membrane resistance after additional cleanings (Figure 3) was noted, but not understood. The vendor indicated that improved performance with successive CIPs is not unusual for the PVDF membranes. It is hypothesized that successive CIPs remove manufacturing residuals from the membranes. 5 of 11

Clean No. Table 1 Effect of Membrane Cleaning in Pilot Study B CIP Interval (Operating Hours) Chemicals Used Resistance before CIP (x10 12 m -1 ) Resistance after CIP (x10 12 m -1 ) Start-up TMP and Resistance 3.31 1 469 2% citric acid followed by 12.3 4.75 2 534 3% citric acid followed by 12.6 3.58 3 293 2.5% citric acid followed by 12.6 3.16 4 249 2.5% citric acid followed by 10.6 3 5 322 2% citric acid followed by 8.9 3.34 6 190 2% citric acid followed by 9.9 3.17 7 226 2% citric acid followed by 8.8 3.34 8 151 2.5% citric acid followed by 10.7 2.98 9 359 2% citric acid followed by 8.0 2.78 10 142 2% citric acid followed by 8.9 2.61 11 364 2% citric acid followed by 11.8 2.51 Figure 3- Decline in Clean Membrane Resistance after Successive Cleanings Clean Membrane Resistance (m^-1) 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 2 4 6 8 10 12 14 CIP # 6 of 11

3.3 Study C: Manganese Removal by Chlorine Dioxide/Microfiltration: Monroe, GA 3.3.1 Pilot Study Background The existing water treatment plant in Monroe, Georgia is a 6.4 conventional surface water treatment plant treating water from the Alcovy River. The majority of the existing filters have been in service for over 30 years without major modification or replacement of media. In order to meet the provisions of the Interim Enhanced Surface Water Treatment Rule (IESWTR), the existing filters may require rehabilitation and improvements. In addition, the owners of the plant wanted to expand the plant to 10 MGD. Due to the increased demands in the community for safe drinking water, one possible alternative to upgrading the filters in the existing surface water plant was to install a membrane process to treat the raw surface water. To investigate this alternative, a pilot study was conducted at the Monroe Water Treatment Plant between November 23, 1999 and March 12, 2000. The Alcovy River experiences seasonal manganese excursions in excess of the levels recommended by EPA. The existing water treatment system successfully removes the manganese to below this concentration due to the manganese dioxide coating on the filtration media. The mechanism of manganese removal by microfiltration requires the oxidation of soluble manganese to particulate form prior to filtration and can not rely on adsorption to the coating on the media since to filters would no longer be used. 3.3.2 Bench-scale Testing Bench-scale testing was conducted to determine which of the possible oxidants might improve manganese removal by microfiltration. Membrane filtration disks (47 mm diameter) were obtained with approximately the same pore size as the membrane in the pilot plant, 0.1 micrometers (um). A one-liter sample of raw water was treated with each oxidant and then sampled over an interval of four hours. After sampling, each sample was filtered as quickly as possible by a membrane filtration apparatus as typically used for Total Coliforms analysis by the membrane filtration technique. After each sample had been filtered, manganese was measured in the filtrate. In the bench-scale tests, the time required for removal of manganese by pre-oxidation and microfiltration was compared for a variety of oxidants, including chlorine, chlorine dioxide, permanganate and hydrogen peroxide (Figure 4). The authors would have included ozone in this list if a source of this oxidant had been readily available. These experiments indicated that chlorine dioxide was the preferred oxidant for the oxidation of manganese. 7 of 11

Figure 4 Results of Bench Tests for removal of Manganese 0.2 0.18 0.16 0.14 0.12 0.1 0.08 With Chlorine With Chlorine Dioxide With Hydrogen Peroxide Manganese Goal 0.06 0.04 0.02 0 0 50 100 150 200 250 3.3.3 Pilot Scale Testing Reaction time, minutes A pilot study was initiated to determine design parameters for the plant expansion. Raw water from the water treatment plant intake was pumped from the existing chemical treatment building to the membrane pilot unit. A chemical dosing point was located immediately downstream of the pump. The estimated length of pipe from the chemical treatment building to the membrane unit was four hundred feet. A 500-gallon polyethylene tank was installed prior to the membrane system inlet to provide additional contact time for the oxidation reaction to be completed. The tank was configured with an overflow to permit constant flow from the chemical treatment building. The approximate detention time of the water was 20 minutes prior to filtration. The membrane pilot unit selected consisted of a feed/clean-in-place tank, membrane feed pump, three membrane modules, and associated piping and automatic valves. Each membrane module contained approximately 20,000 hollow fibers fabricated of polyvinylidene fluoride (PVDF), approximately 1 meter in length. The nominal pore size of the membranes studied was 0.1?m. The flow averaged 20.6 gpm, which resulted in an average membrane flux of 70 gallons per square foot per day. Backwashes were performed every 22 minutes by compressed air with feed water to remove particulate matter from the membrane surface. During normal operation of the unit, some membrane fouling was observed that was not mitigated by the backwash sequence. Occasional clean in place procedures were implemented to remove foulants from the membrane surface and restore the system transmembrane pressure to the original state prior to fouling. Citric acid and sodium hypochlorite were used as the clean-in-place chemicals. 8 of 11

Table 2 illustrates the operational period when each oxidant was applied to the membrane pilot system. Figure 5 shows the changes in transmembrane pressure (TMP) versus time. Each ramp up of the TMP over time was ended by a cleaning event. After cleaning the TMP slowly increases as deposits built on the membrane that were not be removed by the backwash sequence every 22 minutes. Due to the changes in water quality observed during the study, no inferences should be drawn upon the fouling effects of each oxidant upon the membrane system. Table 2: Dates when Pre-oxidant was added in Pilot Study C Phase Pre-oxidant Start Stop CIP 1 none 11/23/99 2/3/00 12/4/99 12/30/99 1/27/00 2 Hypochlorite 2/3/00 2/7/00 3 Chlorine Dioxide 2/7/00 3/3/00 2/15/00 4 Permanganate 3/3/00 Figure 5 - Membrane Operating Parameters 25.00 20.00 15.00 10.00 Flow 5.00 TMP Temperature 0.00 11/6 11/20 12/4 12/18 1/1 1/15 1/29 2/12 2/26 3/11 3/25 Date Table 3 illustrates the effectiveness of the clean-in-place regime utilized during the study. The initial membrane resistance was measured as 3.23x10 12 m -1. New membranes typically take several weeks of use to adequately wet-out all of the membrane pores. The clean-in place (CIP) performed on December 3 rd did not result in adequate recovery of the membrane resistance. Citric acid was added to the CIP regime to remove scales and precipitates of multivalent cations. The subsequent three CIPs resulted in a continued decrease in the membrane resistance to values expected by the manufacturer. This indicates that there is no indication of permanent fouling of the microfiltration membranes. 9 of 11

Table 3: Membrane Resistance after Membrane Cleaning in Pilot Study C Run Date CIP Performed Chemicals Used Resistance after CIP (x10 12 m -1 ) 1 03-Dec 1000 mg/l NaOCl 4.28 2 30-Dec 2% citric acid; 1000 mg/l NaOCl 3.70 3 27-Jan 2% citric acid; 1000 mg/l NaOCl 3.04 4 16-Feb 2% citric acid; 1000 mg/l NaOCl 2.82 The pre-oxidants (Table 4) were injected into the process pipeline with twenty minutes of retention time ahead of the microfilters. Chlorine Dioxide effectively reduced the manganese concentration to less than 0.05 mg/l (i.e., desired treatment goal) at a dose rate of 0.5 mg/l. According to the manufacturer, PVDF was resistant to all the oxidants that were tested. No adverse effects on the membranes were noted during the month that chlorine dioxide was applied. Table 4: Average Manganese and Iron Removal in Pilot Study C Pre-oxidant Sample Manganese, mg/l none Feed 0.166 0.872 Filtrate 0.129 0.004 Chlorine Feed 0.140 0.650 Filtrate 0.100 0.01 Chlorine Dioxide Feed 0.094 0.84 Filtrate 0.001 0.00 Permanganate Feed 0.099 0.90 Filtrate 0.030 0.02 WTP Average Finished water 0.027 0.00 Iron (II and III), mg/l Bench studies performed at the existing water treatment plant indicated that chlorine dioxide was the best oxidant for manganese removal for Alcovy River water. Chlorine dioxide provided removal of manganese to less than 0.05 mg/l with less than twenty minutes of reaction time. During membrane pilot study C, the application of 0.5 mg/l of chorine dioxide prior to the microfiltration system resulted in the removal of 99% of the manganese. As a result of pilot study C, the municipality is proceeding with the design and construction of a 10-MGD microfiltration facility. Provision shall be made in the design for the application of chlorine dioxide for the oxidation of soluble manganese. 10 of 11

4.0 Conclusions 1. If microfiltration is used to replace existing sand filters, the existing media filtration may be removing manganese with assistance of an adsorption mechanism. Since the adsorption mechanism will be eliminated when the media filters are replaced, design of membrane facilities for the removal of soluble manganese must include an effective oxidation step to ensure adequate manganese removal. 2. The use of oxidation/microfiltration is an effective process for the removal of manganese from surface water. Care must be taken to determine the most effective oxidant. As the case studies demonstrate, although sodium hypochlorite was an effective oxidant for the Fairmont Study, it was not effective at the Monroe Water Treatment Plant. Pilot studies for oxidation/microfiltration plants should investigate the efficacy of multiple oxidants. 5.0 Acknowledgements: The authors would like to acknowledge the contributions of Aaron Balczewski, USFilter and David Michelsen, SEA Engineering (formerly of USFilter). The authors would also like to acknowledge the staffs of the Avon, Fairmont, and Monroe Water Treatment Plants for operating the pilot equipment during the studies. 6.0 References 1. Sawyer, C. N., P. L. McCarty and G. F. Parkin, 1994. Chemistry for Environmental Engineering, Fourth Edition, McGraw-Hill, New York, p578. 2. White, G. C. 1999. Handbook of Chlorination and Alternate Disinfectants, Fourth Edition, John Wiley and Sons, Inc., New York, p500-501. 3. AWWA, 1999. Water Quality and Treatment: A Handbook of Community Water Supply, 5 th Edition, McGraw-Hill, New York. 4. Knocke, W. R., S. Occiano and R. Hungate, 1990. Removal of Soluble Manganese from Water by Oxide Coated Filter Media, AWWARF, Denver. 5. Schneider, C. G., P. J. Johns and R. P. Huehmer, 2001. Removal of Manganese by Microfiltration in a Water Treatment Plant, Proceedings of the 2001 Membrane Technology Conference, AWWA & IWA, March 4-7, 2001, San Antonio, Texas. 6. Jimbo, Y. and Goto K., 2000. Iron and manganese removal by membrane filtration system, Proceedings of the Conference on Membranes in Drinking and Industrial Water Production, Volume 2, October 2000, Desalination Publications, L Aquila, Italy, p371 381. 7. Takizawa, S., Fu, L., Pradhan, N., Ike, T., Ohtaki, M., Ohgaki, S., 2000. Pretreatment processes for membrane filtration of raw water containing manganese, Proceedings of the Conference on Membranes in Drinking and Industrial Water Production, Volume 2, October 2000, Desalination Publications, L Aquila, Italy, p 355 362. 11 of 11