Hypolimnetic oxygenation of water supply reservoirs using bubble plume diffusers

Abstract

Mobley M, Gantzer P, Benskin P, Hannoun I, McMahon S, Austin D, Scharf R. 2019. Hypolimnetic oxygenation of water supply reservoirs using bubble plume diffusers. Lake Reserv Manage. 35:247–265.

Hypolimnetic oxygenation of water supply reservoirs improves water quality by preventing anoxia. This article summarizes the operational results using linear bubble plume diffuser hypolimnetic oxygenation systems installed in water supply reservoirs. The results obtained for 8 sites demonstrate that diffuser technology was effective at increasing hypolimnetic dissolved oxygen (DO) and spreading it to blanket sediments. The diffuser systems maintained increased DO in the hypolimnion during successive years of operation at every site. Improved oxygen levels reduced anoxic products and nutrients to mitigate the causes of taste and odor in the water supplied to water treatment facilities. This yielded additional treatment capacity, reduced treatment costs, and provided alternatives to water treatment plant modifications. The diffuser systems provide a simple and effective hypolimnetic oxygenation system, with 19 in operation. The hypolimnetic oxygenation systems (HOS) functions without a water pump and can obtain favorable oxygen transfer efficiencies. The diffusers are installed and maintained from the surface without divers. Diffusers provide an economical means to distribute oxygen input over large areas of the reservoir hypolimnion. Diffuser system installation costs run between $0.5M and $2.5M ($40 to $800 per hectare meter), with annual operating costs between $30K and $140K ($5 to $36 per hectare meter). Capital and operational costs vary depending on site specific conditions. With diffuser oxygenation, customer complaints were reduced and, in some cases, substantial monetary savings were realized by reduced treatment costs. Our results demonstrate that diffusers are a cost-effective treatment option for water supply reservoirs in which anoxia induces water quality problems.

Keywords:

• Bubble plume
• diffuser
• hypolimnion
• oxygenation
• water supply

Anoxic conditions in the hypolimnion of water supply reservoirs can cause many issues and concerns related to water quality and taste and odor (Cooke and Kennedy 2001). Dissolved metals and algae blooms deplete water treatment plant capacity and increase treatment costs (Jung et al. 1999). In extreme cases, metals and algae can make the sources unusable, even with increased treatment (SFPUC 2019). Hypolimnetic oxygenation systems (HOS) are defined as systems that add dissolved oxygen (DO) to the hypolimnion using pure oxygen gas while preserving thermal stratification. A successful HOS can eliminate anoxic products (Gantzer et al. 2009), reduce nutrient cycling and algae blooms (Beutel and Horne 1999), and control issues related to taste and odor (Jung et al. 1999). The addition of HOS can increase overall water treatment capacity and reduce water treatment operating costs (Benskin 2018). There are several HOS designs that have been applied successfully, including side stream supersaturation, submerged contact chambers, and diffused oxygen bubble plumes (Speece and Malina 1973, Ashley 1985, Little 1995, Burris and Little 1998, McGinnis and Little 2002), with no single system type ideal for all applications (Wagner 2015). Of the HOS in operation for water supply reservoirs reported by Wagner (2015), 19 currently utilize bubble plume line diffusers.

When designed properly, diffusers successfully add oxygen to the deep waters of the hypolimnion, while preserving thermal stratification and avoiding sediment disturbance. Bubble plumes, near-field mixing patterns related to plumes, and flow regimes within the plume have been the subject of several studies (Miller et al. 2001, McGinnis et al. 2004, Socolofsky and Adams 2003, and Singleton et al. 2007).

Bubble plume diffuser HOS are little reported in the peer reviewed literature, mainly consisting of review of diffusers as a form of HOS (Beutel and Horne 1999, Singleton and Little 2006, Wagner 2015), but limited with regard to water quality results. This work is the first survey of water quality effects of bubble plume diffuser HOS improvements across a range of reservoirs with corresponding customer satisfaction and cost benefits. Moreover, very little work has been performed in regard to oxygen distribution or to water quality improvements in water supply reservoirs following diffuser operation. A concurrent focus of this work, therefore, is to demonstrate oxygen spreading beyond direct contact with the diffuser.

Study sites

Of the 19 water supply reservoirs currently utilizing bubble plume diffuser HOS, 8 sites were selected that provided the most complete data available for this study (Table 1).

Table 1. Summary of study sites showing characteristics of each reservoir and oxygenation system.

Reservoir	Storage volume	Maximum depth	HOS installation date	HOS capacity	Total length of diffuser
	(hectare meter)	(m)		(kg/d)	(m)
Spring Hollow	1210	64	1997/2004	1100	610
Western Virginia Water Authority
Salem, VA
Carvins Cove	2430	21	2005	3600	1220
Western Virginia Water Authority
Roanoke, VA
Calaveras	11,800	37	2005	3400	610
San Francisco Public Utility Commission
Sunol, CA
San Antonio	6780	40	2009	8200	1150
San Francisco Public Utility Commission
Sunol, CA
Vadnais Lake	1110	18	2011	6500	910
Saint Paul Regional Water Services
Saint Paul, MN
Pleasant Lake	1220	15.2	2013	7500	560
Saint Paul Regional Water Services
Saint Paul, MN
Lake Casitas	29,330	70	2015	27,300	1650
Casitas Municipal Water District
Oak View, CA
Aurora Reservoir	3820	27.4	2015	2300	700
City of Aurora
Aurora, CO

Virginia

Spring Hollow Reservoir and Carvins Cove Reservoir are human-made water-supply reservoirs operated by the Western Virginia Water Authority (WVWA) that serve the city of Roanoke and surrounding counties (Gantzer et al. 2009). Spring Hollow Reservoir is a pumped storage reservoir that was supplied by withdrawing water from the Roanoke River during high flow periods with a storage capacity of 12 million cubic meters (3.2 billion gallons) and a maximum depth of 64 m (210 feet). Spring Hollow Reservoir was the first water supply application of this bubble plume line diffuser design when it was installed in 1997. It had one 610 m (2000 feet) long diffuser installed with 1100 kg/d (1.2 tons/d) oxygen delivery capacity. Carvins Cove Reservoir was supplied by 2 natural tributaries that flow through agriculturally dominated lands and by 2 creeks from an adjoining watershed that are routed through diversion tunnels with a storage capacity of 24 million cubic meters (6.1 billion gallons) and maximum depth of 21 m (70 feet). Carvins Cove Reservoir was equipped with two 610 m (2000 feet) long line diffusers that were installed in 2005, capable of distributing 3600 kg (4 tons) of oxygen per day into the hypolimnion.

California

Calaveras and San Antonio reservoirs are operated by the San Francisco Public Utilities Commission (SFPUC). Calaveras Reservoir is fed by Arroyo Hondo and Calaveras Creek. The reservoir was equipped with two 305 m (1000 feet) long lengths of line diffuser installed in 2005, capable of distributing 3350 kg (3.7 tons) of oxygen per day into the hypolimnion. When the diffuser was installed, Calaveras Reservoir was operated with a seismic restricted storage capacity of 42 million cubic meters (11 billion gallons) and a maximum depth of 18 m (60 feet). Since then, a new replacement dam has been constructed that is currently refilling to return the reservoir to 118 million cubic meters (31 billion gallons) capacity and a maximum depth of 37 m (120 feet) (SFPUC 2018).

San Antonio Reservoir was fed by San Antonio and Indian creeks. It has a capacity of 68 million cubic meters (18 billion gallons) and a maximum depth of 40 m (130 feet). San Antonio Reservoir was equipped with 2 line diffusers installed in 2005 with a total length of 1150 m (3780 feet), capable of distributing 3350 kg (3.7 tons) of oxygen per day into the hypolimnion.

Lake Casitas, operated by the Casitas Municipal Water District (CMWD) located in Ventura County, has a capacity of approximately 300 million cubic meters (78 billion gallons). The maximum water depth is 70 m (230 feet) when full; however, recent drought conditions reduced the current maximum depth to less than 50 m and adversely impacted water quality conditions. Lake Casitas receives its inflows from the surrounding watershed including flows from the Ventura River Diversion and Santa Ana and Coyote Creeks. Withdrawals from the lake are made to supply a water treatment plant, which provides water for domestic and agricultural uses. Lake Casitas was equipped with 3 diffusers positioned at 4 different elevations, with a total length of 1650 m (5400 feet). The system was installed in 2015 and could distribute 27,000 kg (30 tons) of oxygen per day into the hypolimnion.

Colorado

Aurora Reservoir is a Colorado Front Range terminal storage reservoir for the city of Aurora, Colorado. The reservoir is a part of the Aurora Water Prairie Waters Project and is utilized by 2 City of Aurora water purification facilities. The reservoir provides 38 million cubic meters (10 billion gallons) of storage and has a maximum depth of 27 m (90 feet). Aurora Reservoir was equipped with one 700 m (2300 feet) long diffuser that was installed in 2015 to distribute up to 2300 kg (2.5 tons) of oxygen per day into the hypolimnion.

Minnesota

Saint Paul Regional Water Services (SPRWS) pumps Mississippi River water from north of Saint Paul, Minnesota, through a chain of lakes: Charley, Pleasant, Sucker, and Vadnais (Walker et al. 1989). This water supply system has been in operation for more than a century. Two lakes are deep enough to have stable thermal stratification in the summer: Vadnais Lake has a storage capacity of 11.1 million cubic meters (3 billion gallons) and a maximum depth 16.5 m (54 feet). Pleasant Lake has a storage capacity of 12.2 million cubic meters (3.2 billion gallons) and a maximum depth of 15 m (49 feet). Vadnais Lake was equipped with 2 diffusers with a total length of 915 m (3000 feet) that were installed in 2011 and could distribute 6500 kg (7 tons) of oxygen per day into the hypolimnion. Pleasant Lake was equipped with 2 diffusers with a total length of 553 m (1845 feet) that were installed in 2013 and could distribute 7500 kg (8.3 tons) of oxygen per day into the hypolimnion.

Materials and methods

HOS consist of 3 main components: an oxygen supply facility, flow control, and the diffuser.

Oxygen supply facilities and equipment

Hypolimnetic oxygenation systems require a land-based facility to supply and control oxygen flow to the diffusers in the reservoir. Oxygen supply is usually provided by truck delivery of bulk liquid oxygen (LOx) or onsite generation commonly using pressure swing adsorption technology. Flow control can be simple and operated manually or equipped for remote control.

A liquid oxygen supply facility requires tanker truck access and includes an equipment pad, tank foundations, an insulated storage tank, and ambient air vaporizers. The liquid oxygen is converted to gas as it passes through the vaporizers. The vaporization process and corresponding expansion create the necessary pressure to supply the oxygen gas flow to the diffusers in the reservoir. Because of the driving pressure created from gaseous expansion, no pumps or compressors are needed, even for deep reservoirs. The LOx facility equipment can be leased or purchased, with the contracted bulk gas supplier providing maintenance and monitoring. As a part of the contract, the supplier will usually monitor the tank level and dispatch trucks to refill the facility as needed. Liquid oxygen systems are available with very large delivery capacities and are being used to supply reservoir diffusers with up to 180,000 kg/d (200 tons/d). Even larger capacity LOx systems are available.

Onsite oxygen generation utilizes a pressure swing adsorption (PSA) or vacuum swing absorption (VSA) process to isolate oxygen from a compressed air stream using a zeolite sieve. A PSA/VSA oxygen supply system requires an equipment building, air compressor, air receiver tank, electric supply, molecular sieve tanks, control valves, and an oxygen receiver tank. During operation, the compressed air stream is fed through one of the molecular sieves. During the pressurization phase nitrogen is trapped and oxygen passes through. At the end of the pressurization cycle, the sieve is depressurized and vented, releasing the trapped nitrogen to the atmosphere. PSA/VSA systems commonly consist of 2 molecular sieves, of which 1 sieve is in the pressurization phase and the other is being depressurized and vented. PSA produces oxygen gas with a nominal purity of 93% up to 4.5 bar (65 psig), whereas a VSA delivers the same purity at pressures of 1 bar (15 psig). An oxygen booster to increase oxygen supply pressures may be required. PSA systems are readily available at capacities up to 4500 kg/d (5 tons/d) for off-the-shelf units. Larger industrial PSA and VSA systems are also available.

A LOx supply facility will have lower capital costs and less maintenance than a similarly sized onsite generation system, but onsite systems can usually provide oxygen at less cost than that delivered to a LOx facility. LOx systems can supply oxygen for a short term at higher applied gas flow rates than originally designed and can economically be oversized or modified to supply higher gas flow rates, which can be valuable if circumstances dictate unforeseen high oxygen demands. All of the study sites used LOx oxygen supply facilities for the time periods in this report; however, the 2 sites in Virginia are currently being converted to PSA.

Flow control

A flow control manifold is used to regulate the applied gas flow rate to each diffuser. A flow control manifold consists of a flow meter, flow control valve, isolation and vent valves, and pressure gauges for each diffuser. The flow control can be as simple as a rotameter and manual valves or can utilize electronic flow control with remote operation depending on client requirements.

Bubble plume diffusers

The bubble plume diffusers in the water supply reservoirs of this study are based on a linear design originally developed by the Tennessee Valley Authority for hydropower reservoir release improvements (Mobley and Brock 1995, Mobley 1997). The linear diffusers are constructed of high-density polyethylene (HDPE) piping, porous hose, concrete anchors, and stainless steel connecting components. All HDPE connections are joined by a heat fusion procedure, including all anchor and gas piping connections. Flow control orifices along the length of the diffuser are used to provide a uniform bubble pattern along the full length of the porous hose sections. Diffusers are often more than 1000 m long. Pressure requirements for operating the diffuser include hydrostatic pressure of the water depth of the reservoir and the head loss across the flow control orifices and supply pipes. The HDPE working pressure rating is reduced for contact with oxygen gas and for expected ambient temperatures. Anchor tethers are constructed of nylon-coated stainless steel cable. The tether cable lengths can be designed to hold the diffuser at a specific elevation or distance above the bottom. The diffuser lines are deployed and retrieved without need for divers, utilizing a buoyancy pipe to raise and lower the diffuser in the reservoir. The porous hose is manufactured from linear low-density polyethylene and rubber from recycled car tires. The hose has been shown to provide high oxygen transfer efficiency (DeMoyer et al. 2001) and is capable of distributing oxygen in reservoirs for up to 15 yr without excessive degradation or clogging. The diffuser lines require no maintenance during that time unless the porous hose or piping is damaged by boat anchors or other means.

Design and layout

In a thermally stratified reservoir when (oxygen) gas is applied to the diffuser, the gas bubbles rise from the diffuser and entrain cold water from the elevation at which the diffuser is installed. The bubble–water mixture forms a plume and moves upward. As the plume rises, oxygen is transferred from the bubbles to the water. The momentum imparted to the plume is eventually used up as the entrained (cold) water moves up into warmer and less dense water near the top of the hypolimnion below the thermocline. This is the elevation of maximum plume rise (EMPR), where the water in the plume detrains and falls away from the plume (Figure 1). As the oxygenated water falls away from the plume, a portion in the near field plunges to a depth at or below the diffuser elevation (McGinnis et al. 2004), while the remaining portion falls to an elevation of equal density (EED), because the water from the plume is colder than the surrounding water but warmer than it was where it was entrained, and at that elevation it spreads laterally (Figure 1). Any remaining bubbles continue to rise toward the surface, entraining ambient water and creating a new plume.

Figure 1. Images of DO measurements collected in Spring Hollow Reservoir after 24 h of diffuser operation, comparing an idealized plume (top) and a modeled plume (bottom).

Bubble plumes have been studied in thermally stratified reservoirs (Wuest et al. 1992, McGinnis et al. 2004), leading to the development of bubble plume models (Hauser 2004, Singleton et al. 2007). Bubble plume models are an integral part of the design process and are used to predict oxygen placement, plume dynamics related to the EMPR and EED, and oxygen addition to the water column. The EMPR is critical in the design to identify whether a single plume can adequately mix the full depth of the hypolimnion as well as ensure the thermocline is preserved. The EED is important to identify lateral and longitudinal regions of mixing. The initial vertical oxygen placement is well predicted by bubble plume models, as can be seen when the plume model predictions are overlaid on the field measurements for comparison (Figure 1).

Once the oxygen input and plume characteristics have been identified, the diffuser layout is designed using available bathymetry and compressible flow modeling software. The flow model is used to identify required pressures, head losses, and desired pipe sizing to optimize the gas flow of the diffuser. For most installations, line diffusers are positioned along the thalweg at a consistent distance above the bottom, such as Spring Hollow Reservoir and Aurora Reservoir. Depending on the oxygen input needed and the reservoir bathymetry, multiple diffusers may be required to place oxygen in the desired volumes of the reservoir, as at Vadnais and Pleasant lakes. For wide, relatively flat bottom conditions, such as in Carvins Cove Reservoir, diffusers were positioned in parallel and evenly spaced to promote uniform lateral mixing.

In deeper reservoirs, an engineered vertical plume placement can increase the effectiveness of the system. Placing diffusers in the deepest portion of the reservoir can result in oxygen plumes that are completely adsorbed before the plume reaches the thermocline. Additional diffusers at higher elevations may be required to spread oxygen higher in the water column. With an engineered plume placement, the diffusers are installed at set elevations with varying length anchor cables over the bottom topography. With multiple diffuser elevations, the individual diffuser capacity can be sized for the specific volume and oxygen demands for each level. For example, the lowest reservoir elevation zone usually will have the smallest water volume while higher elevations may have significantly more water volume and require more diffuser capacity. A multiple diffuser elevation design was chosen for Lake Casitas. The Casitas design includes 4 diffuser elevations selected with plume model predictions for oxygen placement (Figure 2). Verification of the plume model predictions was obtained with field measurements during initial operation of the system at Lake Casitas (Figure 3). The diffuser elevations are shown and the vertical oxygen placement and lateral density spread are delineated by measured DO increase. Engineered plume placement can be especially beneficial to reservoirs with multilevel outlet structures, such as Casitas, so that operators can choose the best water quality layer available and operate the oxygen system to target specific layers.

Figure 2. Example of Casitas Reservoir diffuser layout, showing multiple diffusers installed at different elevations and their model predicted zone of oxygenation. Oxygenation is expected in the zone above each dot.

Figure 3. Initial oxygen input observed in Casitas Reservoir following 24 h of operation, showing lateral distribution from diffuser operation. Numbered circles represent plant withdrawal elevations, solid lines show modeled plumes, and dashed lines show corresonding elevation of maximum plume rise (EMPR) and elevation of equal density (EED) for each plume.

Data collection

All data were provided by each municipality/utility and collected using required protocol and standards per regulatory and reporting guidelines. Oxygen contour plots collected for Casitas, Spring Hollow Reservoir, and Carvins Cove Reservoir were obtained using a SeaBird Electronics 19PlusV2 high-resolution profiler. The 19PlusV2 has a 4 Hz sample rate, provides a fast response, and can collect data down to the sediments because the internal pump design continuously circulates water over the sensors.

Aurora Water uses the following methods to obtain data presented in this article:

• Dissolved oxygen —YSI EXO2 Multiparameter Sonde equipped with EXO optical Dissolved Oxygen Smart Sensor.

• Metals (Mn)—USEPA Method 200.8 metals by inductively coupled plasma, mass spectrometry.

• Total phosphorus (as P)—Standard Methods 4500-P G 21st edition and Hach QuikChem Method 10-115-01-1-F. Determination of total phosphorus by flow injection analysis colorimetry (acid persulfate digestion method).

• Soluble reactive phosphorus (dissolved PO₄-P)—Standard Methods 4500-P G 21st edition and Hach QuikChem Method 10-115-01-1-M, determination of orthophosphate in waters by flow injection analysis colorimetry.

Operation

Hypolimnetic oxygenation systems are typically operated during thermal stratification. The diffuser systems are designed to meet highest oxygen demands in the reservoir, with additional capacity to recover from a shutdown or if start-up occurs following the onset of stratification. With the capacity to dramatically increase DO content, the diffusers are rarely operated at maximum capacity during a normal operating season. For example, Carvins Cove Reservoir was originally designed to deliver 2000 kg/d; however, following the first year of operation in 2005, applied gas flow rates were closer to 1000 kg/d for subsequent years. Most installations adjust oxygen flows every few days or weekly based on feedback from water quality data obtained from the reservoir or treatment plant operation. During recent years, several bubble plume diffuser systems (Spring Hollow, Carvins Cove, Pleasant, Vadnais Casitas, and Aurora) are being operated year-round, making seasonal adjustments to the applied gas flow rate.

Results and discussion

The first observation following a successful installation and operation of a bubble plume diffuser is increased oxygen content. This was ubiquitous with all 8 systems. During continued operation, dissolved oxygen spreads throughout the hypolimnion for a significant distance from the diffuser and dissolved oxygen concentrations are maintained at high levels year-round. With maintenance of high DO levels in the hypolimnion, anoxic by-products were reduced or eliminated, such as soluble manganese (Mn), iron (Fe), and phosphorus (P). Consecutive years of operation at the study sites resulted in consistently improved water quality to the water treatment plant, and in some cases the annual oxygen usage to maintain hypolimnetic DO has decreased. Reduction in oxygen capacity requirements over time has allowed the Spring Hollow and Carvins Cove installations the option of changing to on-site oxygen generation.

Oxygen content

When properly designed, bubble plume diffusers have been observed to be very effective at increasing and maintaining oxygen content throughout the hypolimnion, both in the bulk water and down to the sediment. Lateral oxygen distribution was predicted using the plume model, as well as being observed during initial operation of diffuser systems such as in Spring Hollow Reservoir (Figure 1) and Casitas Reservoir (Figure 3). For these systems, start-up of the diffuser system occurred during low DO conditions, which allowed the DO increase to be observed. DO was observed to increase almost immediately with initial operation and increased steadily following continued operation of the diffuser system (Figure 4).

Figure 4. Initial operation DO Increases in Calaveras Reservoir. Dashed lines show historical averages without oxygenation.

During subsequent years of operation, DO was observed to remain at elevated levels in the hypolimnion. Casitas Municipal Water District operated the diffuser system to achieve DO levels greater than 8 mg/L over the sediments in the portion of the reservoir hypolimnion near the outlet structure in front of the dam (Figure 5). This goal has continued to be achieved with strategic operation of the diffusers.

Figure 5. Dissolved oxygen concentrations near the dam at Lake Casitas, showing seasonal anoxia before HOS installation and elevated DO following HOS installation in September 2015.

Aurora Water has been operating the diffuser system to maintain DO levels greater than 7 mg/L throughout the hypolimnion (Benskin 2018). Dissolved oxygen improvements in the hypolimnion, measured 1 m off the bottom in Aurora Reservoir, were observed to achieve the target concentration with hypolimnion DO measurements observed to match surface measurement year-round (Figure 6).

Figure 6. Surface and bottom DO in Aurora Reservoir, showing sustained DO in the bottom waters following diffuser start-up in 2016.

Figure 7. DO by depth for August 2017. In the 2017 operating season, there is a sag in DO in the thermocline layer.

Figure 8. DO by depth for August 2018. In 2018, oxygen input was increased to maintain DO kevels >2.0 mg/L in the thermocline layer.

There was an oxygen sag in the metalimnion in 2017, most severe in August, with DO dropping below 2 mg/L (Figure 7). Cause of the DO sag is likely decay of settled algae and seston perched on a density gradient at the top of the hypolimnion, which was aerobic. The oxygen delivery rate was 1905 kg/d 01–10 August and 2014 kg/d 10–31 August. To maintain metalimnetic DO concentrations greater than 2.0 mg/L, the oxygen delivery rates were increased in 2018: 2014 kg/d 01–09 August and 2177 kg/d 10–31 August. Because of increased oxygen delivery, median DO concentrations were significantly higher (p < 0.0001) in August 2018 than in August 2017 (Figure 8) for the epilimnion (7.5 and 7.6 mg/L), metalimnion (3.2 and 4.9 mg/L), and hypolimnion (7.2 and 9.2 mg/L).

Pleasant and Vadnais lakes experienced sharp improvements to median hypolimnetic DO with hypolimnetic oxygenation. With hypolimnetic oxygenation, the median DO was 4.2 and 8.5 mg/L in Pleasant and Vadnais lakes, respectively. These values were significantly higher (p < 0.0001) than hypolimnetic DO in previous periods without hypolimnetic aeration or oxygenation (0.11 and 0.2 mg/L) and with hypolimnetic aeration (1.7 and 2/7 mg/L).

Oxygen distribution in the hypolimnion

A key benefit of bubble plume diffuser technology is the ability to add large quantities of oxygen over large areas. As previously discussed, the plume generated by the diffuser promotes circulation within the hypolimnion as water is continuously entrained by the plume. Oxygen distribution has been observed to spread along density layers both laterally and horizontally away from the diffuser as identified in field observations.

During the first year of oxygenation in Carvins Cove Reservoir, water column data were collected at 2-m increments across the reservoir before and after diffuser operation. Data were comprised of 90 profiles, providing the initial signature of DO placement and subsequent spreading throughout the hypolimnion. Prior to diffuser operation, DO was observed to be ≤2.5 mg/L throughout the hypolimnion and <2.0 near the sediments (Figure 9). The thermocline was identified to be in the vicinity of elevation 346 m msl. After 45 d of operation, applying 32 Nm³/h (20 SCFM) to each diffuser, results showed a dramatic increase in DO to greater than 5 mg/L throughout the entire hypolimnion and down to and along the sediments, as shown in data collected at Carvins Cove (Figure 10). The oxygen placement was distributed bank to bank throughout the vertical elevation of the hypolimnion. This represented spreading over 100 m perpendicular to the placement of the diffuser. Additionally, the water with 4 mg/L at the thermocline was still present, demonstrating that the bubble plumes had not destratified the reservoir. Oxygen was observed to blanket the sediments including in the original channel, which is below the elevation of the diffuser.

Figure 9. DO data collected 180 m laterally across Carvins Cove Reservoir at the onset of diffuser operation, applying 32 Nm³/h (20 SCFM) to each diffuser. Arrows at ∼45 and 110 m represent diffuser locations.

Figure 10. DO data collected 180 m laterally across Carvins Cove Reservoir after 45 d of diffuser operation, applying 32 Nm³/h (20 SCFM) to each diffuser, showing spreading of DO throughout the hypolimnion and over sediment. Arrows at ∼45 and 110 m represent diffuser locations.

Carvins Cove Reservoir had 2 diffusers that were relatively short compared to the overall length of the reservoir. The diffusers in Carvins Cove Reservoir were 600 m long and positioned in the deepest part of the reservoir between 100 and 700 m upstream of the withdrawal structure. This left a large portion of the hypolimnion not in direct contact with the diffusers. Water column profiles collected along the length of the reservoir documented elevated DO levels throughout the hypolimnion and oxygen in excess of 8.0 mg/L at the sediments spreading over 2000 m upstream from the diffusers (Figure 11).

Figure 11. DO data collected horizontally along Carvins Cove Reservoir during 2006, showing horizontal spreading over 2000 m upstream of the diffuser. The diffuser is represented by the white line that extends along the bottom from about 100 to 700 m.

Anoxic by-products

As previously shown, diffuser operation was observed to successfully increase and maintain oxygen conditions in the hypolimnion. As a result of increased DO, anoxia and elevated levels of anoxic by-products such as soluble iron (Fe), manganese (Mn), and phosphorus (P) were mitigated. The study site diffuser systems have worked so well in preventing Fe and Mn from going into solution that some sites rarely even sample for them anymore (SFPUC 2019).

Iron

Iron levels for Carvins Cove Reservoir were observed to decrease from the onset of diffuser operation and remain low throughout consecutive years (Figure 12). Review of the pre-oxygenation data shows total Fe concentrations in the hypolimnion to exceed 1.5 mg/L with levels observed greater than 0.3 mg/L throughout the water column each year following fall turnover. Ambient DO of less than 5 mg/L was observed to correspond to elevated Fe levels (Figure 12). Since the diffuser system was installed, Fe levels were consistently observed to be less than 0.3 mg/L in the hypolimnion and 0.1 mg/L or less throughout the entire water column following the fall turnover during isothermal conditions.

Figure 12. Total Fe concentrations reported throughout the water column in Carvins Cove Reservoir between 2000 and 2009, showing elevated Fe levels each year corresponding to DO dropping below 5 mg/L outlined as DO 5 followed by significant decrease in Fe concentrations after diffuser start-up in 2005.

Manganese

Total Mn concentrations were also observed to decrease with diffuser operation in Carvins Cove Reservoir. Prior to the diffuser installation, total Mn concentrations were observed to exceed 3.0 mg/L in the bottom 6 m of the hypolimnion, with levels near 0.5 mg/L throughout the water column later in the year following fall turnover. After the installation of the diffuser in 2005, (1) elevated Mn levels were isolated to the bottom meter and were observed to range between 1.0 and 2.0 mg/L, (2) levels in the bulk hypolimnion were observed between 0.1 and 0.5 mg/L, and (3) Mn levels throughout the entire water column during isothermal conditions following fall turnover were less than 0.1 mg/L (Figure 13).

Figure 13. Total Mn concentrations reported throughout the water column in Carvins Cove Reservoir between 2000 and 2009, showing elevated Mn levels each year corresponding to DO below 5 mg/L outlined as DO 5 followed by decreased Mn concentrations throughout the water column after diffuser start-up in 2005.

Decreased levels of total Mn were also observed in the Aurora Reservoir throughout the water column, including a dramatic shift in the amount of dissolved Mn contributing to total Mn. Results observed in Aurora showed that prior to diffuser operation nearly all Mn from samples collected near the bottom consisted of the dissolved form of Mn, with levels being observed in excess of 0.50 mg/L. With diffuser operation, total Mn was observed to decrease by more than half, but more importantly, dissolved Mn contribution to total was significantly less than before oxygen addition. During the first year of diffuser operation in 2016, maximum dissolved Mn concentration was observed just over 0.10 mg/L, with subsequent years observed less than 0.10 mg/L (Figure 14). Increased oxygen input to maintain DO values >2.0 mg/L at the thermocline was utilized in 2018 and may have been responible for decreasing the manganese release in 2018 compared to 2017. With the large reduction in dissolved Mn, water from the hypolimnion can be treated at both Aurora Water water purification facilities year-round as needed, and pre-oxidant chemical demand has been greatly reduced.

Figure 14. Total and dissolved Mn concentrations collected near the bottom in Aurora Reservoir, showing Mn spikes in excess of 0.500 mg/L prior to diffuser operation, consisting mainly of the dissolved form of Mn, followed by both decreased total Mn concentrations and decreased contribution from the dissolved form of Mn after diffuser installation.

In Lake Casitas, prior to diffuser installation, maximum Mn levels were observed as high as 0.86 mg/L. Following installation and operation of the diffuser, maximum Mn levels were observed to be reduced by 90%, with a maximum concentration of 0.09 mg/L in 2017 (Table 2).

Table 2. Summary of water quality parameters assessed at Lake Casitas from 2012 to 2017.

Year	Bubbler^1 HOS operation	End-of-year WSEL (m)	Measured DO < 1.5 mg/L^2 (at 36.6 m)	Max. total dissolved phosphorus (μg/L)	Max. ammonia-N (mg/L)	Avg. growing season^3 Secchi (m)	Max. manganese (mg/L)	Yearly taste/ odor customer complaints^4
2012	Bubbler	164.0	16/34 (47%)	290	0.423	3.3	0.070	48
2013	Bubbler	159.7	18/38 (47%)	280	0.3	2.9	0.350	22
2014	Bubbler	156.1	28/41 (68%)	390	1.35	2.2	0.350	100
2015	Bubbler/HOS	152.1	51/63 (81%)	350	1.29	2.5	0.860	140
2016	HOS	148.1	0/41 (0%)	140	0.088	3.5	0.050	0
2017	HOS	149.4	0/27 (0%)	50	0.087	2.7^5	0.090	0

¹ Bubbler refers to a series of aerators installed in 2005 to provide vertical mixing above elev121 m.

² A percentage of DO measurements <1.5 mg/L is presented in addition to the number of recorded measurements satisfying this criterion.

³ Secchi depth is averaged over the algal growing season (Feb–Oct).

⁴ Customer complaints related to lake water quality issues.

⁵ No data from July through September for 2017.

Phosphorus

Internal phosphorus loading was identified as one of the primary sources of nutrient loading in Aurora Reservoir. Huisman et al. (2004) identified that cyanobacteria can access nutrients via diel vertical migration to a critical depth (∼20 m). For water supply reservoirs with a thermocline less then 20 m deep, such as Aurora Reservoir, oxidation of the hypolimnion could reduce a nutrient availability to cyanobacteria. Therefore, controlling internal phosphorus loading, especially soluble reactive phosphorus, was a primary objective of the oxygenation system for Aurora Water.

Prior to diffuser installation in Aurora Reservoir, total and soluble reactive phosphorus samples collected near the bottom were observed to be greater than 150 and 130 μg/L, respectively (Figure 15). After diffuser start-up in May 2016, maximum total phosphorus levels were observed between 50 and 60 μg/L for 2016 and 2018 and 100 μg/L for 2017. Soluble reactive phosphorus levels were observed between 40 and 50 μg/L for 2016 and 2018. In 2017 soluble reactive phosphorus concentration peaked at 86 μg/L, higher than in 2016 and 2018 but still representing a 35% reduction from pre-oxygenation values. Higher oxygen input in 2018 may have contributed to lower phosphorus levels. As a result of the decreased phosphorus in Aurora Reservoir, the Trophic State Index for total phosphorus was reduced from 48.9 (2015) to 41.1 (2018) (Benskin 2018).

Figure 15. Total (top) and soluble reactive (bottom) phosphorus collected at the surface and near the bottom in Aurora Reservoir, showing decreased levels of both following diffuser start-up in 2016.

At Calaveras and San Antonio reservoirs, SFPUC found that even when keeping the bulk hypolimnion volume oxygenated, some phosphorus was going into solution. This was attributed to letting the DO at the bottom get too low before increasing the oxygen input rate. SFPUC was not increasing the input rate until the bottom measurements reached 2 mg/L or lower, which may have left some anoxic or very low DO volumes where phosphorus was released from the sediments. In 2018, SFPUC changed its operating criteria and is now using a trigger of 5 mg/L to increase the oxygen input rate. No results from this change are yet available (SFPUC 2019).

The common effect of diffuser operation as reported by the municipalities contributing to this work has been a significant suppression of internal loading of Fe, Mn, and P.

Taste and odor

Several factors contribute to taste and odor in water supply reservoirs, such as geosmin and 2-MIB from cyanobacteria in surface waters (Srinivasan and Sorial 2011) and Mn and hydrogen sulfide (H₂S) from hypolimnetic anoxia in bottom waters (Gerling et al. 2014). Taste and odor issues in water-supply reservoirs often lead directly to customer complaints.

Prior to diffuser installation in Lake Casitas, drought conditions were observed to have had a negative impact on water quality. This was reflected in the raw water where Mn and H₂S levels increased to 0.4 mg/L and 2 mg/L, respectively (Water Quality Solutions 2018). Customer complaints during this time period increased dramatically (WQS 2018). During 2014 there were 104 complaints. In 2015, there were 140 complaints. The diffuser system was initially operated in late September 2015. Manganese concentrations in the hypolimnion, lake raw effluent, and treatment plant effluent dropped significantly within a week of the start of oxygenation (WQS 2018). Hydrogen sulfide concentrations in the hypolimnion were also observed to drop significantly within a few weeks of the start of oxygenation (WQS 2018). Customer complaints related to lake water quality issues dropped to zero (WQS 2018). These are complaints due to lake water quality issues, not miscellaneous customer complaints. In subsequent years with increased hypolimnetic DO levels, concentrations of phosphorus, ammonia-N, and manganese have been decreased in Lake Casitas (Table 2). CMWD does not monitor for MIB/geosmin, but there have been no seasonal algal-related taste/odor complaints since installation of the diffusers, and no algaecide treatments for cyanobacteria since installation of the diffusers. As a result of continued diffuser operation, and despite continued drought conditions, customer complaints related to taste and odor have nearly disappeared.

With initial operation of the hypolimnetic oxygenation, Aurora Water also experienced a dramatic decrease in taste and odor complaints related to the use of Aurora Reservoir source water. With diffuser system operation, water could be withdrawn from the hypolimnion year round, allowing the utility to avoid geosmin-producing cyanobacteria that are present in the epilimnion during the warmer months of the year. Complaints were reduced from approximately 100 per summer to fewer than 10 total complaints related to the use of Aurora Reservoir water in the last 3 yr with oxygenation (Benskin 2018).

Turbidity

One of the existing turbidity standards for the CMWD filtration plant at Lake Casitas is that effluent turbidity shall be less than or equal to 0.2 NTU in at least 95% of the measurements taken each month based on 4-h readings with a turbidity performance goal of 0.10 NTU. Additionally, when any individual filter is placed back into service following a backwash or other interruption event, the filtered water turbidity of the effluent from that filter shall not exceed 0.2 NTU after the filter has been in operation for 4 h. Following the installation of the diffuser in 2015, CMWD conducted a diffuser operational test to identify the impact on effluent filter water turbidity. Approximately 1 week after beginning the continuous operational test, the average turbidity of the combined effluent dropped from 0.10 to 0.07 NTU (Table 3). Also, with the diffuser operating continuously, the treatment staff members reported that they were able to more easily achieve the 0.2 NTU standard. For each year that the diffuser system has been fully operational (2016–2018) the percentage of turbidity measurements equal to or less than 0.20 NTU increased to 100% (Table 4).

Table 3. Summary table of averaged treatment water turbidity at Lake Casitas during 2015 diffuser operational testing.

Date	Turbidity (NTU)*	Diffuser operation
08/01/15–08/25/15	0.10	Pre start-up
08/26/15–09/08/15	0.15	Start-up; sporadic operation
09/09/15–09/23/15	0.08	Start-up; continuous operation
09/24/15–10/28/15	0.06	35 d trial

* Averaged treatment water turbidity (4 h increments).

Table 4. Summary of turbidity measurements collected in 4 h increments reported by CMWD for 2014–2017. Data provided show minimum, maximum, average, and percentage <0.20 measured.

Year	Turbidity (NTU)
	Minimum	Maximum	Average	% <0.20
2014	0.01	0.40	0.05	99.6
2015	0.02	0.26	0.05	99.8
2016	0.02	0.11	0.03	100.0
2017	0.00	0.10	0.03	100.0
2018	0.01	0.07	0.03	100.0

Installation and operation costs

Installation costs vary widely for both the oxygen supply and reservoir diffuser system, depending on site-specific conditions and requirements. A requirement for a new truck access road, architectural requirements around the tank, and the distance from the facility to the reservoir were observed to dramatically affect the installation costs of the oxygen supply facility. Similarly, the distance from the reservoir piping access point to the diffuser location(s) in the reservoir was observed to affect the installation costs of the diffuser. Actual costs from installations at the study sites show a wide variation of costs, mostly due to site-specific requirements (Table 5). These projects have an average installation cost of about $200/kg/d of oxygen addition capacity with a range of $40 to $400/kg/d.

Table 5. Installation and operating costs.

Reservoir	Installation costs					Operating costs
	HOS capacity	Oxygen supply	Diffuser	HOS total	Cost per capacity	Annual operating costs	Cost per volume
	(kg/d)	(1000 $)	(1000 $)	(1000 $)	($/kg/d)	(1000 $)	($ per hectare m)
Spring Hollow	1100	NA	NA	NA	NA	NA	NA
Western Virginia Water Authority
Salem, VA
Carvins Cove	3600	NA	195	NA	NA	NA	NA
Western Virginia Water Authority
Roanoke, VA
Calaveras	3400	736	241	977	287	108	9.15
San Francisco Public Utility Commission
Sunol, CA
San Antonio	8200	580	416	996	121	35	5.16
San Francisco Public Utility Commission
Sunol, CA
Vadnais Lake	6500	336	380	716	110	33	29.73
Saint Paul Regional Water Services
Saint Paul, MN
Pleasant Lake	7500	558	380	938	125	44	36.07
Saint Paul Regional Water Services
Saint Paul, MN
Lake Casitas	27,300	670	530	1200	44	140	4.77
Casitas Municipal Water District
Oak View, CA
Aurora Reservoir	2300	803	197	1000	435	37	9.69
City of Aurora
Aurora, CO

In some cases, costs for water treatment plant (WTP) modifications or upgrades to treat problematic anoxic products were avoided by the installation of a diffuser system in the reservoir. At Crystal Lake Reservoir the City of Cheyenne, Wyoming, avoided an expensive WTP upgrade with an oxygenation system installed in 2009 (Brandhuber et al. 2010). In the southeastern United States, an undisclosed client avoided a tens of million dollars expenditure on a WTP modification with an oxygenation system installed in 2016.

Operation costs include the costs for the oxygen used and maintenance of the oxygen supply and diffusers. Oxygen costs were observed to vary depending on the availability in the region, the delivery distance, and the contract amount. Current bulk liquid oxygen costs are around $0.09/kg ($84/ton) (current price at Aurora in 2019), $0.11/kg ($96/ton) (current price at CMWD in 2018), and $0.15/kg ($136/ton) (current price at SPRWS in 2019). Maintenance of a liquid oxygen facility is minimal as there are very few moving parts. Maintenance and monitoring of the facility were often contracted to the bulk gas supplier providing the oxygen deliveries. Maintenance is part of the unit costs given for CMWD and SPRWS. The cost of on-site oxygen generation depends on the available electricity cost but was usually less expensive than liquid oxygen delivery at about $0.06/kg ($52/ton) for lo-pressure delivery (1.0 barg, 15 psig) to $0.10/kg ($90/ton) for high-pressure delivery (3.0 barg, 65 psig) using an average electrical cost of $0.12/kWh (AirSep 2016). Maintenance of a PSA system will include maintenance of the air compressor, air filtration systems, and solenoid valves. Typical diffuser maintenance is porous hose replacement every 12 to 15 yr at a cost of about $3.30 per meter of diffuser length.

Oxygen usage was observed to vary at each reservoir due to ambient conditions such as seasonal weather conditions and inflow water quality. How accurately the operators maintain a desired DO level through monitoring and flow adjustments and any other variation in operation pattern was also observed to have a big impact on oxygen costs. The length of the oxygenation season also impacts operation costs. For example, the Aurora reservoir diffuser system is typically operated at design levels for 60 to 90 d per year, while Vadnais and Pleasant are operated for up to 270 d per year, resulting in significantly higher oxygen costs. Actual annual oxygen costs averaged $66,000, $16 per hectare meter for the study site (Table 5).

In some cases, cost savings in water treatment plant chemicals can far exceed oxygen costs. At the Aurora Water’s Bimney Water Purification Facility, the installation and use of the hypolimnetic oxygen system resulted in significant savings in each year it has been in use. In the years prior to the installation of the oxygen system, water was treated from a withdrawal gate in the epilimnion to avoid high levels of dissolved manganese in the hypolimnion. During 2014 and 2015, geosmin-producing cyanobacteria were seasonally present in the epilimnion, so a major portion of the Aurora Reservoir water was treated through granular activated carbon (GAC) adsorbers to reduce geosmin taste and odor impacts. This resulted in a significant cost of approximately $350,000/yr. Since the installation of the oxygen diffuser system, this extra treatment step has not been necessary, saving the utility that expense. In addition, the oxidation of released manganese in the reservoir with oxygen, rather than in the treatment plant with potassium permanganate, has also resulted in a marked decrease in spending on that pre-oxidation chemical. Finally, withdrawing water from the hypolimnion (where algae and total organic carbon levels are lower than in the epilimnion) has allowed for a significant reduction in coagulation chemical use, as compared to their average use in the years prior to the oxygenation system installation, Considering all of these recaptured savings, the hypolimnetic oxygen system at Aurora Reservoir has provided total cost savings of more than $1.4M in 3 yr of operation, more than paying back the initial $1.0M installation cost in 8 separate reservoirs (Table 6).

Table 6. Annual oxygen costs and chemical cost savings at Aurora Reservoir (Benskin 2018).

Aurora Reservoir	2016	2017	2018
Granular activated carbon (GAC)	$350,000	$350,000	$350,000
Potassium permanganate	$15,000	$80,000	$75,000
Coagulant chemicals	$43,000	$83,000	$140,000
Chemical savings	$408,000	$513,000	$565,000
Oxygen costs	$39,000	$32,000	$40,000
Total annual savings	$369,000	$481,000	$525,000

Conclusion

The authors evaluated the benefits to the water column and treatment plant operation after installing and operating a hypolimnetic oxygenation system employing a bubble plume line diffuser. It has been observed that when designed, monitored, and operated properly, diffusers successfully add and spread dissolved oxygen to the deep waters of the hypolimnion, and over the sediments, while preserving thermal stratification.

The results obtained for the projects presented demonstrate how effective diffuser technology was able to (1) increase the DO and spread it laterally and longitudinally throughout the hypolimnion, (2) maintain elevated DO throughout the hypolimnion during successive years of operation, and (3) blanket the sediments with oxygen, even below the installation depth of the diffuser. Additionally, it was shown that with proper engineering, oxygen placement can be engineered to target specific areas or elevation layers of a reservoir.

Bubble plume line diffusers are an established, effective, and successful hypolimnetic oxygenation system design with 19 systems in operation in the United States. The system obtains high oxygen transfer efficiencies without need for a water pump or intake structure. The line diffusers are installed and maintained without divers. Diffusers provide an economical means to distribute oxygen input over large areas of the reservoir hypolimnion.

Diffusers are an attractive treatment option for water supply. Maintaining oxygen levels in the reservoir directly reduces anoxic products and nutrients and mitigates the causes of taste and odor in the reservoir before the raw water enters the treatment plant. For example, manganese that is released from bottom sediments can be almost entirely pre-oxidized by the oxygenation system in the reservoir, rather than depending on chemical oxidation at the water treatment plant. Lower algae levels during summer can allow for a significant decrease in coagulant chemical use. By addressing these problems through oxygenation in the reservoir, more of the water treatment plant capacity is available for other needs, increasing overall treatment capacity.

Installation costs vary widely depending on site specific conditions but can provide an attractive alternative to expensive water treatment plant modifications. Typical liquid oxygen supplied water supply installations run between $0.5M and $2.5M, but average about $100/kg/d. Operating costs are mainly the oxygen usage with low maintenance costs. In some conditions, it is possible to obtain substantial savings by avoiding algaecide applications and chemical treatment costs in the WTP. Several reservoirs with oxygenation systems are experiencing a reduction in oxygen use and costs over time. The results show that bubble plume line diffusers are a cost-effective treatment option for water supply reservoirs.

Acknowledgments

We thank the utilities and their staff members who provided information for this study. Also, we thank the personnel who installed, operated, and monitored the oxygenation systems.