Sign up for our newsletter

Direct Membrane Treatment of Anaerobic High Iron and Manganese Groundwaters

We have had the pleasure of working with the city of Grimes for almost 20 years. Our work there has spanned all facets of civil, transportation, water, and wastewater. Recently, thoughtful preparation went into planning for a new water treatment facility. This involved pilot testing to ensure the right and most economical methods were chosen before proceeding with design.

The following article evolved when FOX Engineering worked with Harn R/O Systems, Inc. and local contract operations firm, USW Utility Group, to evaluate reverse osmosis (RO) treatment for the city’s water supply to improve water quality and meet the long-term treatment needs.

For more information, contact:

Julia Nemeth-Harn, PE; Harn R/O Systems, Inc.; JuliaNemeth@harnrosystems.com;

Steve Troyer, P.E., BCEE; FOX Engineering Associates, Inc.; stroyer@foxeng.com

Introduction

Although there has been an industry-wide perception that feedwater sources containing dissolved metals must be pre-treated prior to reverse osmosis (RO) or nanofiltration (NF) spiral-wound membrane treatment because of their potential to foul the spiral-wound membranes, in actuality there have been many membrane plants – some that are decades old – that have successfully treated groundwaters with high levels of dissolved metals with direct membrane treatment. To evaluate whether or not oxidation and filtration pretreatment might or might not be recommended, there are several site-specific factors that should be assessed including: age and condition of raw water feed facilities, distance of wells from treatment facility, level of metal ions and state of metal ions (dissolved or particulate) and quantity of potential complexing ions such as phosphate, amount of hydrogen sulfide in the raw water, and amount of bacteria or organics in the raw water. After evaluating these considerations, a pilot study should be performed to confirm the proposed approach.

A high level of metals in a raw water source should not automatically mean that a feedwater is not suited for membrane treatment or that extensive pretreatment will be required. Often times the pretreatment process not only adds capital and operating and maintenance cost, but actually is detrimental to the membrane treatment operation.

Anaerobic Groundwater and Aerobic Pretreatment Considerations

Groundwater sources usually contain a low level of dissolved oxygen versus surface waters.  A dissolved oxygen (DO) level less than 1.0 mg/l is considered anaerobic. In an anaerobic groundwater common metals such as iron and manganese primarily exist in their soluble forms, Fe2+ and Mn2+. In an oxidizing environment iron and manganese will precipitate. They will release another electron and become triple-charged ions, Fe3+ and Mn3+. In this precipitated, particulate form they are a foulant to a spiral-wound RO or NF membrane system. Fouling may be avoided by preventing oxidation of the metals or completely removing the precipitated iron or manganese after oxidation. Parameters that affect metal oxidation include DO level, water temperature, pH and the presence of complexing ions such as phosphate, organic matter, silica, bacteria or suspended solids.

In the early days of membrane treatment RO feedwater containing iron – even at low levels of less than 1 mg/l – was commonly pretreated with oxidation and filtration pretreatment. The oxidation step could be accomplished either through chemical addition with chlorine, potassium permanganate, other oxidants, or aeration. Then after adequate mixing and detention time the water would go through media filtration to remove the oxidized, particulate iron. When properly designed and operated, the pretreatment should reduce the oxidized metals down to a level of less than 0.1 mg/l. After filtration an additional step would be required prior to feeding the pre-treated water to the reverse osmosis system which would be removing the residual oxidant from the feedwater. RO and NF membranes are made from a thin film composite polyamide chemistry that cannot tolerate exposure to oxidants. Oxidants will destroy the ion rejection characteristics of the membrane surface and will permanently damage the membranes. Therefore, an oxidant reducing agent such as sodium bisulfite must be injected into the RO feed stream to neutralize any oxidant carryover. If there is an interruption in the injection of the neutralizing chemical then rapid degradation of the membrane salt rejection can occur.

These pretreatment process steps add complexity, capital and operating cost, and risk to the membrane treatment process. Another detrimental effect of oxidation pretreatment is the greatly increased potential for membrane biofouling. Membrane biofouling will proliferate in the aerobic atmosphere of a pre-oxidized feedwater. The biological activity will have more time and room to grow through all of the pretreatment equipment and piping. Even if chlorine is used as the oxidizing agent, biofouling will not be reduced. Studies have shown that the chlorination/dechlorination process required for membrane pretreatment does not inhibit biofilm formation in the membrane system. In fact, biofilm formation may even be accelerated because the biological mass that is killed by the chlorine may become a food source to permit the surviving biology to proliferate (Paul 1990).

For these reasons, the preferred approach for cost savings, ease of operations, reduced membrane cleaning frequency and increased membrane longevity is to directly treat anaerobic groundwater containing dissolved metals without aerobic pretreatment.  The potential benefits warrant site-specific investigation and pilot testing where feasible. A case study on a project in Grimes, Iowa is presented herein, as well as another brief case study on an RO system in Illinois. Design and operational recommendations to help keep the water source anaerobic are included in the conclusion section of this paper.

 

Grimes Pilot Study – Phase 1, Jordan Well Membrane Treatment

Background

Grimes, Iowa is a growing community located 15 miles northwest of Des Moines. Due to increases in potable water demand in the Grimes service area and limited water resources, the City of Grimes decided to expand their existing conventional water treatment plant that has been operating since 2000. The Grimes WTP was originally built to treat water from the Shallow wellfield with lime softening and filtration for hardness and iron removal.

The City decided to drill a deep well, approximately 2000 feet deep, into the Jordan Aquifer in 2012. The Jordan water source has a higher level of total dissolved solids than the Shallow wellfield, but a lower level of iron. The City was interested in having their Engineer, FOX Engineering Associates investigate membrane treatment for this water source. Their objectives were to supplement the capacity being provided from the conventional WTP and to improve the water quality being provided to their customers. The average iron level of the Jordan well was 2.3 mg/l and for the Shallow wellfield it was over 12 mg/l. Iowa Department of Natural Resources requires a ninety-day pilot study for approving the implementation of membrane potable water treatment. The pilot study for Grimes consisted of two phases: the first phase involved direct membrane treatment of Jordan well feedwater. The second phase involved direct membrane treatment of a blend of the Jordan well and Shallow well feedwater.

The Phase 1 RO pilot study on Jordan aquifer feedwater commenced in August 2015. The main goal for the RO treatment was softening, iron removal and other mineral reduction.  The feedwater hardness averaged 432 mg/l as CaCO3.  Generally, utilities prefer to target a finished water hardness of 100-200 mg/L as CaCO3.  Also, radionuclides in the feedwater were relatively high and sulfates were also undesirably high. Pilot testing RO treatment of this water source would confirm the reduction of the above-described minerals and would confirm important design parameters.

Description of RO Pilot System

Performing a full-scale pilot test (rather than a single element or bench scale) using the equivalent of a 2:1 array of 4” diameter, seven element-long pressure vessels enables the testing and verification of the proposed design on a small, inexpensive scale before the production system is built. The pilot study also allows for testing and optimization of operating parameters such as chemical dosage, blend rates and cartridge filter replacement frequency. The pilot system operation also allows for raw water blending and testing so the Engineer can evaluate the actual blended finished water, not just the RO permeate. Also, actual concentrate is produced – not simulated or recycled concentrate.  This can be useful for concentrate disposal facility design and permitting.

The pilot system (Figure 1) was initially set up at the Grimes WTP to produce 17.0 gpm of permeate at a system recovery of 85% (ratio of permeate produced to total water treated).

Figure 1. Full-Scale RO Pilot System

Phase 1 RO Pilot System Operation

The pilot system was started on August 19, 2015, shut down on November 24, 2015 and operated for approximately 1,991 hours over a 3-month period.The permeate from the pilot unit was blended with lime softened water at a ratio of 58% permeate to 42% lime softened water. Blending with lime softened water will help stabilize the permeate –raising the pH and alkalinity and improving the Langelier Saturation Index (LSI). This also helps reduce the capital cost of the treatment process since the membrane system can be down-sized, and the operating cost will also be reduced since less post-treatment chemicals are required. Some utilities maximize their blend capability by incorporating conventional oxidation/filtration iron removal on the blend stream to keep the iron level as low as possible in the blended finished water.

Phase 1 Pilot Results

The pilot unit was set up at a conservative average flux rate of 13.7 gallons per day per square foot (GFD). A good hydraulic balance was obtained between the first and second stages by incorporating interstage boost pumping. The system operated at 85% recovery throughout the test period. The pilot operation was very stable throughout the test until the last week of testing when a sudden second stage scaling event occurred due to loss of scale inhibitor injection. The permeate conductivity and iron rejection were stable throughout the test also.

The membrane autopsy performed by Avista Technologies showed the scale formation was stunted and unnatural and consisted primarily of calcium, and trace amounts of iron, silicon and magnesium. These shapes likely indicate there was some inhibition of the scale formation by the scale inhibitor, however, as the system continued to operate without scale inhibitor injection, or possibly after the skid shut down and the supersaturated concentrate did not get adequately flushed out, some degree of scale did form. The product used during the pilot study was a scale inhibitor/dispersant: Avista Vitec 3000, injected at 5 mg/l dosage. No acid pretreatment feed was used.

The autopsy noted a minute iron presence relative to the calcium.  The iron was not a significant foulant. The laboratory flat sheet clean-in-place study performed by Avista indicated that cleaning with Avista 130 cleaner restored the flow back to the manufacturer’s specification.  The Avista 130 cleaner is a low pH powdered cleaner designed to remove metal foulants such as iron, manganese, and aluminum and to remove calcium carbonate scale deposits from spiral-wound membranes.

Phase 1 Pilot Conclusions

In conclusion, the Phase 1 pilot study was very stable except for one recoverable scaling event and the permeate quality indicated excellent total dissolved solids removal. Important design parameters were confirmed for implementation in the full-scale plant design such as flux rate, flux balance, recovery, and scale inhibitor and cleaning recommendations.

The Owner, Engineer and Iowa regulators were confident that direct membrane treatment of the Jordan feedwater would be successful, sustainable and cost effective. The Jordan feedwater had an average total iron level of 2.31 mg/l. The permeate iron level averaged <0.100 mg/l, below the detection limit.  Table 1 below summarizes the water quality results from Phase 1.

Ammonia, mg/l as N Chloride, mg/l Sulfate, mg/l Alkalinity, mg/l as CaCO3 Tot. Hard., mg/s as CaCO3 TDS, mg/l Iron, mg/l (total)
Feed 1.46 98.9 619 268 432 1303 2.31
Permeate 0.08 3.55 <1.00 42.901 <7.00 35.43 <0.100
Soft. + Perm. 0.41 24.6 193.3 33.0 145.8 364.8 0.12

Table 1 –  Phase 1 Testing Water Quality Data

 

Grimes Pilot Study – Phase 2, Blended Jordan and Shallow Well Membrane Treatment

Background

The City of Grimes has a Shallow wellfield that is currently blended with Jordan well water and treated through conventional lime softening. The Owner and Engineer wanted to determine if the Shallow wells could be used as a back-up/supplement for the Jordan well as a feedwater source for the RO system. The characteristics of the Shallow wells are significantly different than the Jordan well. Most notably, the total iron level in the Shallow wells was reported to be up to 12 mg/l and the manganese was reported to be 0.65 mg/l. Although it is typically undesirable to blend two different water sources upstream of a reverse osmosis plant, this operating scheme would only be instituted as a back-up condition. Normally, with iron levels as high as projected from the Shallow wells, iron removal pretreatment would be considered mandatory for successful membrane treatment, however, it was determined that, for back-up treatment purposes, it would be worthwhile to test direct membrane treatment of the blended feedwater.

The feedwater blend ratio was 55% Shallow wellfield water and 45% Jordan well water. The blended incoming total iron level averaged 7 mg/l. At this phase of the testing it became relevant to start differentiating between dissolved, ferrous iron (Fe2+) and particulate, ferric iron (Fe3+). Additional testing for the iron states demonstrated that about two-thirds of the total iron in the blended feed was already in the precipitated, ferric form. The blended feed hardness was also about 20% higher than the Jordan well water.

The same membranes, Hydranautics ESPA2-4040, were equally well suited to treat the blended feedwater, therefore, the same membranes were used in the Phase 2 test as had been used in the Phase 1 test. Knowing that the blended feedwater would have a much higher fouling potential from the particulate iron and a higher scaling potential from the higher hardness level, the pilot target recovery was reduced from 85% to 80%. The flux balance between stages 1 and 2 was again maintained evenly through the use of an interstage boost pump. A bag filter was installed upstream of the cartridge filter to help remove some of the particulate iron. A 10 micron bag filter and 1 micron cartridge filters were used. A pretreatment acid feed was not instituted initially. The same finished water blend ratio of 58% permeate to 42% lime softened water was maintained. The same scale inhibitor and injection rate was used.

Phase 2 RO Pilot System Operation

The Phase 2 study operated for approximately 1,760 hours. The membranes in the second stage of the pilot exhibited signs of scaling – increased 2nd stage feed pressure and differential pressure – within a few days of the Phase 2 testing starting up. The recovery was reduced to 75% and that recovery was maintained successfully throughout the remainder of the testing.

The pressures and flows were variable during the first few weeks of operation. The logged data indicated operating pressure spikes every time the pilot would start-up after a down period. Due to hydraulic constraints, if the water plant was not running, the pilot could not run. It was evident that shut-downs were detrimental to the operation, the pressure spikes after a shut-down period were likely due to inadequate post-flushing due to hydraulic disposal limitations at the pilot site.

Acid pretreatment feed utilizing 93% sulfuric acid was implemented to reduce the RO feed pH to about 6.5. This would help re-solubilize the particulate iron and help minimize the calcium carbonate scaling. Prior to instituting the acid feed the pilot unit was cleaned with Avista P903 (formerly known as 127).  This is a low pH powdered cleaner designed to remove iron, manganese, and aluminum deposits from spiral-wound membranes. The cleaning was very effective, dissolving and removing a significant amount of black material. The second stage feed pressure was reduced from 275 psi to 121 psi.  The membrane salt rejection was still excellent.

The bag and cartridge filters were effective at reducing the particulate iron loading. The bag filters typically required changing about every two weeks and the cartridge filters about once a month. Figure 2 is a photo of the spent bag and cartridge filter elements depicting heavy iron loading, but the solid white core of the 1 micron cartridge filter elements clearly indicates the filter removal efficiency was successful at protecting the membrane system from pass-through of particulate iron.

Figure 2. Grimes Phase 2 Pilot Spent Bag and Cartridge Filter Elements

An autopsy of the tail end element from the second stage of the pilot unit was performed at the end of the study. The element examination indicated that there was very little foulant on the membrane, and what was measurable was primarily iron and silica and some biological constituents.

Phase 2 Pilot Conclusions

Biological activity was seen in all the membrane autopsies, in both Phase 1 and Phase 2 testing, therefore, it can be speculated that this system would likely have the same potential for biofouling as other similar applications. The anaerobic operation during the pilot and in the full scale operation appears to mitigate the potential for biofouling.  The blended feedwater contained an average of 7 mg/l total iron and 2.6 mg/l of dissolved iron. The permeate iron level averaged <0.1 mg/l, below the detection limit. The concentrate iron level averaged 26 mg/l. Table 2 below summarizes the water quality results from Phase 2.

Ammonia, mg/l as N Chloride, mg/l Sulfate, mg/l Alkalinity, mg/l as CaCO3 Tot. Hard., mg/s as CaCO3 TDS, mg/l Iron, mg/l (total) Iron, mg/l (diss)
Feed 1.0 45.0 477.0 287.9 555.6 1046.3 7.0 2.6
Permeate <0.050 1.66 3.46 11.86 <6.62 50.25 <0.10 <0.10
Soft. + Perm. 0.33 24.3 182.5 36.7 124.5 353.3 <0.10 <0.10

Table 2 – Phase 2 Testing Water Quality Data

 

Upon the completion of the pilot phase, the City and their Engineer moved forward in implementing a full-scale membrane plant at their existing water plant. The basis of design was a 1.6 million gallon per day (MGD) permeate production system operating on Jordan well water. The design approach was to install one RO train initially with space to add a second RO train. The first train was installed and started up in June 2016. The second train was installed and started up in May 2017.  Both are currently operating well.

Toluca, Illinois – Aerobic Converted to Anaerobic Membrane Treatment

The Toluca, Illinois RO plant started operating in June 1992.  The well water was high in hardness, radionuclides and iron (up to 2.2 mg/l, depending on the well) and contained some manganese (about 0.025 mg/l). The original plant design for RO pretreatment incorporated chlorine feed on the raw water to oxidize the metals, then aeration and detention, then booster pumps, media filtration and RO pretreatment chemical injection. After chlorine injection and aeration, the pH of the feedwater increased from the natural pH of 7.6 to 8.2. Therefore acid pretreatment injection was required to lower the pH back down to prevent calcium carbonate scaling, sodium bisulfite injection was required to neutralize the chlorine residual, and scale inhibitor was also injected. Caustic was injected as a post-treatment chemical to bring the pH back up. The chemical consumption for the treatment process comprised 68% of the total cost to treat the water. The system had to be cleaned every three to six months. Membranes lasted an average of two years. Figure 3 shows the original treatment train.

Figure 3. Case Study 2. Toluca Illinois. Treatment Train before Retrofit

In 2009 the City decided to modify and upgrade their RO plant. They switched to a direct membrane treatment, anaerobic feed – bringing the raw water directly in to the cartridge filters and membrane system (Figure 4). They converted the existing aeration/detention equipment to degasify the permeate after the membrane process. The retrofit was completed in 2010. The original membranes are still installed and working well. They have never required a cleaning. The cost for water treatment chemicals was reduced by 75% and other process improvements were implemented in the retrofit reducing the operating cost in total by over $137,000 per year, which would allow the City to recoup the total retrofit cost in less than three years.

Figure 4. Case Study 2. Toluca Illinois. Treatment Train after Retrofit

Conclusion

As demonstrated from these case studies as well as numerous other successful direct treatment high iron membrane projects, spiral wound membrane treatment can be utilized directly on anaerobic feedwater containing high levels of metals. If aerobic pretreatment is required because other ions such as boron or arsenic dictate a need for aerobic pretreatment, or because the concentrate disposal method cannot accommodate high metals levels, then care must be taken in the design of the pretreatment to minimize the carryover of oxidized particles, oxidizing chemicals, or polymers to the membrane process. Pilot testing is highly recommended to confirm that direct treatment will be successful. Aerobic pretreatment processes add capital and O & M cost, and are detrimental to the membrane treatment operation, will cause increased biofouling, increased cleaning frequency and reduced membrane life. Below are a few recommendations to help minimize air introduction and biofouling potential:

  • Use non-metallic or 316 stainless materials of construction in the well and raw water transmission piping
  • Use submersible well pumps with foot valves to keep pump column full of water and minimize aeration from stuffing boxes on vertical turbine pumps
  • Minimize elevation variations in raw water transmission piping and use manual air relief valves on well discharge and piping high points, automatic air release valves scale and can pull air in
  • Provide well flush capabilities at the well heads and the water treatment plant
  • Do not oversize raw water transmission piping which could permit settling and stagnation
  • Do not permit well driller to use biologically-active driller’s mud, ensure that thorough well development is performed – preferably with air sparging, and that silt density index tests to predict the colloidal fouling potential of the well are performed

References

Castle, Ron J. and Julia Nemeth-Harn, PE (2006), Case Studies: Aerobic vs. Anaerobic Pretreatment of Groundwater, AMTA 2006 Biennial Conference and Exposition

Galjaard, Gilbert (2008), 8 Years RO Experience at WTP Heemskerk – Biofouling Aspects, AMTA/SEDA Joint Conference, Naples, FL

Hiemstra, Peter, J. Van Passen, B. Rietman, and J. Verdouw (2001), Aerobic Versus Anaerobic Nanofiltration: Fouling of Membranes, AWWA Trends in Water Series: Membrane Practices for Water Treatment. pp. 55-82

Missimer, Thomas M., Water Supply Development for Membrane Water Treatment Facilities, CRC Press, 1994

Paul, David (1990), Reverse Osmosis Membrane Fouling – The Final Frontier, Ultra Pure Water, Vol. 7, No. 3 pp. 25-36

Robinson, Ken, Jason Bailey and Julia Nemeth-Harn (2015), The RO Retrofit: Making Things Right, AWWA/AMTA 2015 MTC Conference

FOX Engineering is an environmental engineering firm based in Ames, Iowa. We specialize in water and wastewater solutions for our diverse municipal and industrial clients. Our work varies in size and scope and can be found throughout the Midwest and beyond.