About Us | Help Videos | Contact Us | Subscriptions

Journal of Environmental Quality - Article



This article in JEQ

  1. Vol. 41 No. 3, p. 664-671
    unlockOPEN ACCESS
    Received: Aug 15, 2011

    * Corresponding author(s): ray.bryant@ars.usda.gov
Request Permissions


Using Flue Gas Desulfurization Gypsum to Remove Dissolved Phosphorus from Agricultural Drainage Waters

  1. Ray B. Bryant *a,
  2. Anthony R. Budaa,
  3. Peter J.A. Kleinmana,
  4. Clinton D. Churcha,
  5. Louis S. Saporitoa,
  6. Gordon J. Folmara,
  7. Salil Boseb and
  8. Arthur L. Allenc
  1. a USDA–ARS, Bldg. 3702, Curtin Rd., University Park, PA 16802
    b Constellation Power Generation, 1005 Brandon Shores Rd., FSRC, 2nd Floor, Baltimore, MD 21226
    c Univ. of Maryland Eastern Shore, 11868 Academic Oval, 3111 John T. Williams Hall, Princess Anne, MD 21853. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer. Assigned to Associate Editor Gerwin F. Koopmans


High levels of accumulated phosphorus (P) in soils of the Delmarva Peninsula are a major source of dissolved P entering drainage ditches that empty into the Chesapeake Bay. The objective of this study was to design, construct, and monitor a within-ditch filter to remove dissolved P, thereby protecting receiving waters against P losses from upstream areas. In April 2007, 110 Mg of flue gas desulfurization (FGD) gypsum, a low-cost coal combustion product, was used as the reactive ingredient in a ditch filter. The ditch filter was monitored from 2007 to 2010, during which time 29 storm-induced flow events were characterized. For storm-induced flow, the event mean concentration efficiency for total dissolved P (TDP) removal for water passing through the gypsum bed was 73 ± 27% confidence interval (α = 0.05). The removal efficiency for storm-induced flow by the summation of load method was 65 ± 27% confidence interval (α = 0.05). Although chemically effective, the maximum observed hydraulic conductivity of FGD gypsum was 4 L s−1, but it decreased over time to <1 L s−1. When bypass flow and base flow were taken into consideration, the ditch filter removed approximately 22% of the TDP load over the 3.6-yr monitoring period. Due to maintenance and clean-out requirements, we conclude that ditch filtration using FGD gypsum is not practical at a farm scale. However, we propose an alternate design consisting of FGD gypsum-filled trenches parallel to the ditch to intercept and treat groundwater before it enters the ditch.


    FGD, flue gas desulfurization; ICP–OES, inductively coupled plasma–optical emission spectroscopy; TDP, total dissolved phosphorus; UMES, University of Maryland Eastern Shore

A large area of intensive poultry production on the Delmarva Peninsula has been scrutinized for its contributions of nutrients to the Chesapeake Bay. Decades of chicken litter applications have led to excessive soil phosphorus (P) levels and are of concern for dissolved P losses (Pautler and Sims, 2000; Sims et al., 2000). This legacy P can be a significant source of P entering drainage ditches that empty into streams and rivers that eventually flow to the Chesapeake Bay. University of Maryland Eastern Shore (UMES) and USDA–ARS researchers have documented substantial P concentrations in agricultural drainage waters derived from these high-P soils (∼450 mg kg-1 Mehlich-3 P). Even when fields receive no further P additions, total P concentrations in ditch drainage are 1 to 3 mg L−1, with dissolved P accounting for approximately 70% of the load (Kleinman et al., 2010). Loads can vary widely in response to precipitation patterns, but annual P losses of 25 kg ha−1 are common (Kleinman et al., 2007). Existing conservation practices, such as minimum tillage and edge-of-field grass filter strips, are designed to reduce sediment-bound particulate P in runoff and offer no control over dissolved P losses (Sharpley et al., 2004). Eliminating P applications during future cropping cycles (i.e., phytomining) will do little to reduce P losses in the near term because, in extremely high-P soils, phytomining is a decadal process that primarily removes P from the topsoil (van der Salm et al., 2009; Kleinman et al., 2011). The importance of agricultural drainage to nonpoint-source pollution of surface waters and the need for soil and water conservation practices that can minimize P losses in subsurface flow in the Atlantic Coastal Plain of the United States and other areas were emphasized by Sims et al. (1998). However, over the last decade, there has been little success in developing viable strategies for reducing these losses.

In this study, our strategy for controlling dissolved P losses was to intercept the flow path and sorb dissolved P, thereby removing it from agricultural drainage waters leaving the farm. The drainage ditch itself is the concentrated flow path for water and dissolved P originating from upstream nonpoint source areas. By removing dissolved P from drainage ditch water by filtration at one location within a ditch, we can protect against downstream environmental impacts due to P losses from all upstream, nonpoint source areas. Because this previously proposed strategy (Penn et al., 2007; Penn et al., 2010) affords no agronomic benefit that could translate to profit, the cost of establishing and maintaining a treatment system must be kept to a minimum. In recognition of the cost constraints and in search of beneficial uses for waste materials, considerable work has been done to characterize the composition and P sorption characteristics of industrial byproducts (Leader et al., 2008; Penn et al., 2011). For our purpose, we used flue gas desulfurization (FGD) gypsum, also referred to as synthetic gypsum, which has sorption properties that are effective in removing or immobilizing P in water and soil and in poultry, dairy, and swine manures (Moore and Miller, 1994; Stout et al., 1998; Stout et al., 2000; Clark et al., 2001; Dou et al., 2003; Zhang et al., 2004; Penn and Bryant, 2006; Dick et al., 2006; Penn et al., 2007; Leader et al., 2008; Penn et al., 2010; Penn et al., 2011). Perhaps more importantly, FGD gypsum has a low metal content (Kost et al., 2005), and studies have shown that concentrations of metals in leachate are below levels of concern (Ghosh and Subbarao, 1998; Punshon et al., 2001; Ishak et al., 2002; Kost et al., 2005). Other studies have used calcium (Ca) to effectively precipitate dissolved P from runoff, but lime filters (Kirkkala et al., 2011) and crushed concrete (Egemose et al., 2012) result in high pH values (>11) in the effluent. Flue gas desulfurization gypsum was expected to be pH neutral. Additionally, several studies have shown the beneficial effects of FGD gypsum use as an agricultural amendment (Chen et al., 2001; Clark et al., 2001; Kukier et al., 2001), and a research and demonstration effort is underway in the United States to promote acceptance in the agricultural community for using FGD products (Dick et al., 2006).

The specific objectives of this study were to design, build, and monitor the effectiveness of an in-ditch filter to remove dissolved P, thereby reducing nonpoint source P losses from upstream agricultural fields. Key considerations included effective sorption of dissolved P, filtration of large flow volumes, and minimal adverse environmental impacts due to the release of heavy metals or toxics.

Materials and Methods

Study Site

This study was conducted on the UMES Research and Teaching Farm located near the city of Princess Anne in Somerset County, Maryland (Fig. 1). The farm, which was formerly a commercial poultry farm, was acquired by UMES in 1993. It lies in the heart of the poultry producing area and has some of the highest soil P values on the Eastern Shore in close proximity to the Chesapeake Bay. The fertility index values shown in Fig. 1 are equivalent to soil test values by the Mehlich-3 method for available P. Soil test values by other methods were converted to Mehlich-3 equivalents. Mean values for each county represent the mean for only those fertility index values that exceed 150. The farm typifies ditch-drained agriculture in the area; a corn and soybean rotation with a winter wheat cover crop is grown under no-till. Poultry litter, which typically has a N:P ratio of 2:1, is usually spring applied on fields to be planted with corn at rates equivalent to crop removal of P in accordance with guidelines set forth in the Maryland P Site Index.

Fig. 1.
Fig. 1.

Location of study area and mean Maryland P fertility index values (equivalent to Mehlich-3 P values) by county. UMES, University of Maryland Eastern Shore.


In April 2007, one of the larger collection ditches on the farm, bounded on both sides by poultry houses and soils with Mehlich-3 P values averaging approximately 450 mg kg−1, was selected as the construction site for the ditch filter (Fig. 2). Within the approximately 17-ha area drained by the ditch, the only litter application during the study period (2007–2010) occurred in 2009. A poultry litter application at the rate of 0.5 Mg ha−1 that was applied and incorporated before planting had no discernible effect on P loss. During the study period, high soil P values controlled dissolved P loss in drainage waters.

Fig. 2.
Fig. 2.

The flue gas desulfurization gypsum ditch filter was installed on a major collection ditch draining 17 ha. The average Mehlich-3 soil test value is approximately 450 mg kg-1. The “gypsum curtain” is a second-generation design.


Soils in the drainage area have silt loam surface horizons and well structured, silty clay loam, argillic, subsurface horizons that favor preferential flow. Below a depth of approximately 50 cm, the argillic horizon transitions to highly permeable, medium, and coarse sands containing discontinuous clay lenses. Soils are predominantly poorly drained Typic Endoaquults and Umbraquults (Quindocqua, Othello, and Kentuck), with small areas of Aquic Hapludults (Manokin, Glassboro and Woodstown) (Soil Survey Staff, 2010). Slopes are 0 to 2%, and elevation is approximately 2.5 m above sea level.

Ditch Filter Construction

Flue gas desulfurization gypsum (110 Mg) and sand (5 Mg) were used in the construction of the filter bed (Fig. 3). To avoid flooding during large storm events, the gypsum ditch filter was designed to allow excess flow to spill over and bypass the filter. To measure bypass flow during large flow events, a compound V-notch, straight-walled weir was used to block flow and to establish a hydrologic head above the filtration bed. After the weir was installed, gabions (rock-filled wire cages) were placed behind the weir to hold it in place and to prevent erosion of the ditch side walls. A manifold, placed in front of the weir, was connected to a drain pipe, which routed the filtered effluent underground around the weir and through a partially buried metal box that provided access to flow monitoring and sampling equipment. The filter bed consists of six 30-m-long, 10-cm-diameter tile lines that were encased in cylindrically shaped filter fabric, attached to the manifold, and sandwiched within a layer of sand with FGD gypsum above (25 cm thick) and below (10 cm thick). A coconut fiber erosion control mat stabilized the surface of the bed until vegetation established by natural succession.

Fig. 3.
Fig. 3.

Clockwise from upper left: 110 Mg of flue gas desulfurization (FGD) gypsum was used in construction; a weir impedes ditch flow and measures overflow (white arrows show direction of flow); six 10-cm tile lines (30 m long) attach to the manifold in front of the weir; tile lines embedded in sand are sandwiched in FGD gypsum above and below; and filtered effluent is routed underground around the weir and through flow and concentration monitoring instruments before release downstream.



The source of the FGD gypsum used in this study was the Conemaugh Generating Station, a coal-fired power station owned and operated by Reliant Energy near New Florence, Pennsylvania. As received from the power station, the FGD gypsum was comprised of uniform silt-sized particles, and it was used in the construction of the filter without physical modification. Samples of “fresh” FGD gypsum that was used to construct the filter in April 2007 and samples of the “spent” gypsum that were taken from the filter bed in January 2011 were digested following USEPA standard method 3050b (Kimbrough and Wakakuwa, 1989). The digests were analyzed by inductively coupled plasma–optical emission spectroscopy (ICP–OES) (Varian 730-ES Axial ICP Spectrometer) for the following elements and detection limits (mg L−1): Al (0.01), As (0.01), Ca (0.1), Cd (0.01), Cu (0.01), Fe (0.01), Hg (0.01), K (0.1), Mg (0.01), Mn (0.01), Mo (0.01), Na (0.1), Ni (0.01), P (0.01), Pb (0.1), Na (0.1), Ni (0.01), S (0.1), and Zn (0.01). American Sigma 900Max automated samplers were used to take water samples from the ditch upstream of the filter and from the collection pipe that routed the filtered effluent around the weir and discharged it to the ditch downstream of the filter. A shaft encoder was used to detect a rise in water level during a flow event and trigger the automated samplers to begin drawing samples at 2-h intervals during and after the event. Water samples were filtered (0.45 μm) immediately after collection, and pH was measured at the UMES Nutrient Analysis Laboratory. Samples were stored in a cold room and shipped on ice to the USDA–ARS Water Quality Laboratory at University Park, Pennsylvania for ICP–OES analysis for the same elements and detection limits reported above. Phosphorus measured by ICP–OES in the filtered water samples is hereafter referred to as total dissolved P (TDP). Filtered samples were also analyzed using the ICP–OES equipped with a hydride generator to provide detection limits for determinations of mercury (Hg) and arsenic (As) of 0.001 mg L−1, suitably below the drinking water standards of 0.002 and 0.01 mg L−1, respectively. Dilutions of 1000 mg L−1 National Institute of Standards and Technology traceable standards for Hg and As were used to develop standard curves for these analyses. Analyses of long-term performance were calculated by the event mean concentration efficiency method and the summation of load method (USEPA, 2002). Erickson et al. (2010) describe both methods in detail and provide examples.

Storm Flow Separation

Data were summarized by individual storm events, which necessitated separating storm flow from base flow. For longer duration storms (rainfall lasting at least several hours), a semi-log separation technique (Hall, 1968) was used. This approach assumes that the point at which storm flow stops appears as the beginning of a straight line when the hydrograph is plotted on a semi-log scale. This straight line was then projected back to the time of the hydrograph peak. Another straight line was used to connect the projected line with the stream flow hydrograph at the time the stream flow began to increase. The area below the two lines was assumed to be base flow. The area above the projected lines and below the stream flow hydrograph was assumed to be storm flow.

For storms of shorter duration (those in which rainfall lasted no more than a couple of hours and in which stream flow rates returned to prestorm values 1 or 2 h after rainfall ceased), base flow was assumed not to increase during the storm. Consequently, the base flow was given a constant value during the storm and was then subtracted from the stream flow to produce storm flow.

Results and Discussion

The success of the gypsum filter as a strategy for removing TDP from agricultural ditch drainage waters was assessed in terms of “chemical efficiency,” “system efficiency,” and “environmental impact.” Chemical efficiency refers to the amount of P removed from water that passed through the FGD gypsum filter bed. System efficiency is defined as TDP removed as a proportion of TDP that entered the ditch and either passed through the FGD gypsum filter bed or over the weir in bypass flow. System efficiency is therefore dependent on the hydraulic conductivity of the gypsum bed because it determines the ability of the filter to process large volumes of water during storm events. System efficiency is assessed in terms of efficiency of P removal during storm events and in terms of overall efficiency of P removal, including intervals of base flow between storm events.

Expressed as a percentage of the total cations (by weight) that were determined by ICP–OES as listed above, a digested sample of the “fresh” FGD gypsum used to construct the filter contained 55.3% Ca and 38.2% sulfur (S) (total = 93.5%). Impurities include Na (3.55%), Fe (1.32%), Mg (0.77%), Al (0.54%), K (0.26%), and P (0.03%). All other impurities were present at <0.01%. Although precipitation as CaPO4 is the mechanism for P removal by gypsum (Penn et al., 2011; Stoner et al., 2012), Fe and Al oxides present as impurities in the FGD gypsum may remove phosphate by adsorption. In this discussion, we use the term “P removal” to include both mechanisms of removing TDP from solution. Before decommissioning the ditch filter, we present a preliminary assessment of the environmental impacts of using FGD gypsum for filtration in terms of the fate of Hg (2.7 × 10−5%) and As (1.9 × 10−3%), which are the elements of environmental concern that were originally present as impurities in the FGD gypsum. A more thorough analysis of these and other environmental impacts is planned when the ditch filter is decommissioned and we can destructively sample the gypsum bed.

Chemical Efficiency during Storm Flow

A total of 31 storm-induced flow events occurred from April 2007 to December 2010. With the exception of two events on 23 Apr. 2009 and 29 May 2009 when we experienced equipment failure, water chemistry and flow data were collected for all storm-induced flow events. Storm event date, volume of flow passing through the filter bed, flow-weighted mean concentrations, loads of TDP in the influent and effluent, and percent removal of TDP for storm events are reported in Table 1. Assessing the efficiency of the ditch filter during storm events is important because a large proportion of TDP loss occurs during storm events. For storm-induced flow, the event mean concentration efficiency for TDP capture for water passing through the gypsum bed was 73 ± 27% confidence interval (α = 0.05). By the summation of load method, we calculated a removal of 65 ± 27% confidence interval (α = 0.05). Erickson et al. (2010) provide a rationale for choosing one method over the other depending on assumptions and the desired interpretation of the results. In our case, the results obtained by the two methods are in close agreement and serve to reinforce the outcome.

View Full Table | Close Full ViewTable 1.

Flow volumes of water passing through the gypsum filter bed and total dissolved phosphorus removal (chemical efficiency) for individual storm events.

Storm event date Filtered flow Chemical efficiency based on concentration
Chemical efficiency based on load
Influent TDP† Effluent TDP Removal Influent TDP Effluent TDP Removal
L × 103 mg L−1 % kg %
13 Apr. 2007 16 0.056 0.032 42 0.001 0.001 54
18 Apr. 2007 571 1.576 1.073 32 0.811 0.503 38
6 June 2007 2 0.171 0.012 93 0.001 0.001 100
30 July 2007 57 0.879 0.289 67 0.129 0.027 79
22 Aug. 2007 97 1.000 0.700 30 0.098 0.064 35
8 Feb. 2008 158 0.200 0.136 32 0.031 0.025 21
20 Feb. 2008 366 0.365 0.330 10 0.220 0.210 42
28 Apr. 2008 340 0.578 0.266 54 0.229 0.116 49
9 May 2008 148 1.581 0.405 74 0.319 0.104 67
16 May 2008 87 0.576 0.241 58 0.154 0.028 82
5 June 2008 315 1.721 0.310 82 0.540 0.106 80
5 July 2008 170 1.843 0.255 86 0.314 0.050 84
11 Dec. 2008 484 1.111 0.759 32 0.635 0.393 38
15 Mar. 2009 124 0.616 0.281 54 0.077 0.034 56
6 Apr. 2009 230 0.722 0.548 24 0.166 0.093 44
11 Apr. 2009 201 0.810 0.734 9 0.104 0.076 27
14 Apr. 2009 318 0.526 0.607 −15 0.081 0.090 −11
18 June 2009 225 1.909 0.078 96 0.014 0.001 96
3 Aug. 2009 68 1.871 0.128 93 0.127 0.009 93
12 Aug. 2009 146 1.207 0.057 95 0.157 0.009 95
24 Aug. 2009 256 2.231 0.065 97 0.571 0.016 97
10 Sept. 2009 255 0.985 0.064 93 0.274 0.014 95
17 Oct. 2009 1 1.196 0.049 96 0.316 0.013 96
12 Nov. 2009 263 1.033 0.040 96 0.396 0.015 96
9 Dec. 2009 73 0.234 0.035 85 0.060 0.005 91
23 Jan. 2010 118 0.729 0.021 97 0.086 0.002 97
29 Mar. 2010 16 0.821 0.013 98 0.013 0.001 98
9 Apr. 10 27 1.860 0.011 99 0.050 0.001 99
1 Oct. 10 285 1.433 0.577 60 1.314 0.545 59
Mean 1.029 0.280 0.251 0.088
Total dissolved phosphorus.
Dates in italics indicate that storm flow did not breach the weir. All flow passed through the gypsum bed.

At a head of 50 cm above the surface of the gypsum bed, water passed through the gypsum bed at a rate of approximately 2.5 L s−1 during the first year of operation. The filtration rate was approximately 2 L s−1 in the second year, but in Year 3 the filtration rate dropped to slightly less than 1 L s−1. In Year 3, storms with low flow breached the weir. We observed this decrease in filtration rate over time and suspected sealing due to deposition of silt on the surface of the filter bed. In August 2010, we tilled the surface of the gypsum bed, and flow rate through the filter increased to 2.7 L s−1, approximately the same as during the first year of operation.

Flow rate determines the residence time of water passing through the gypsum bed (i.e., the time available for P removal by precipitation or adsorption to occur). After June 2009, removal of TDP was consistently above 90%, corresponding to the reduced filtration rate and longer residence time as described above. There was a weak inverse correlation between percent removal of TDP and flow rate through the gypsum bed (R2 = 0.36; p = 0.005), indicating that other factors, such as antecedent moisture conditions in the gypsum bed, play a role in determining the efficiency of P removal. Total dissolved P in the effluent was high during the December 2008 event after a 6-mo dry spell. Under initially dry conditions, some time is required for the gypsum to become wet and dissolve sufficiently to support a high concentration of dissolved Ca to react with the dissolved P. Furthermore, animal burrows and root channels can provide conduits for flow that bypasses the filter bed and routes water directly to the buried drain tiles. Although the surface of the filter bed was obscured by vegetation, we speculate that high levels of TDP in the effluent during three successive storms in April 2009 may have been due to the presence of burrows that allowed bypass flow through the filter bed but which later became clogged. The October 2010 event occurred after tilling the surface of the filter bed in August, and the level of TDP in the effluent from that event was similar to levels observed during 2007 when flow-through rates were greater.

System Efficiency during Storm Flow

After construction of the ditch filter, the area received 84 mm of precipitation within 30 h (18 Apr. 2007 event). This was the second largest ditch flow event, with 7879 × 103 L recorded during the monitoring period (Table 2). Although this event proved that the design of the ditch filter could withstand extreme flow events, 93% of the storm flow topped the weir and bypassed the gypsum bed. Only 7% of the storm flow passed through the gypsum bed. The maximum ditch flow rate during this event was 215 L s−1, whereas the maximum filtration rate of the gypsum bed was 4 L s−1. The remainder of 2007 was relatively dry, and 100% of storm flow in the June, July, and August events passed through the filter bed. Beginning in 2008, precipitation events returned to normal, and storm events produced bypass flow over the weir in all but two of the remaining storm events. The durations of the 29 storm-induced flow events ranged from 10 to 130 h, with an average duration of 50 h.

View Full Table | Close Full ViewTable 2.

Total ditch flow volumes, percent of total flow passing through the filter bed, influent and effluent total dissolved phosphorus, and total dissolved phosphorus removal as a proportion of total phosphorus load entering the ditch filter system (system efficiency) for individual storm events.

Storm event date Total ditch flow Influent TDP† Effluent TDP Removal
L × 103 (% of total flow) kg %
13 Apr. 2007 16 (100) 0.001 0.000 54
18 Apr. 2007 7879 (7) 13.807 13.498 2
6 June 2007 2 (100) 0.000 0.000 100
30 July 2007 57 (100) 0.129 0.027 79
22 July 07 97 (100) 0.098 0.064 35
8 Feb. 2008 454 (35) 0.113 0.084 25
20 Feb. 2008 1328 (28) 0.730 0.520 29
28 Apr. 2008 406 (84) 0.312 0.200 36
9 May 2008 183 (81) 0.464 0.249 46
16 May 2008 87 (100) 0.154 0.028 82
5 June 2008 2326 (14) 3.900 3.466 11
5 July 2008 250 (68) 0.516 0.252 51
11 Dec. 2008 2336 (21) 3.412 3.169 7
15 Mar. 2009 124 (100) 0.077 0.034 56
6 Apr. 2009 592 (39) 0.681 0.609 11
11 Apr. 2009 800 (25) 0.599 0.571 5
14 Apr. 2009 1289 (25) 0.299 0.299 0
18 June 2009 233 (97) 0.082 0.068 17
3 Aug. 2009 148 (46) 0.127 0.009 46
12 Aug. 2009 439 (33) 0.157 0.009 33
24 Aug. 2009 1876 (14) 0.571 0.016 16
10 Sept. 2009 2953 (9) 3.453 3.193 8
17 Oct. 2009 3 (9) 3.703 3.399 8
12 Nov. 2009 9237 (3) 10.974 10.593 3
9 Dec. 2009 3416 (2) 1.542 1.487 4
23 Jan. 2010 2135 (6) 1.460 1.376 6
29 Mar. 2010 848 (2) 0.657 0.644 2
9 Apr. 2010 82 (33) 0.150 0.100 33
1 Oct. 2010 1202 (24) 1.720 0.951 45
Total 51.87 47.11
Total dissolved phosphorus.
Dates in italics indicate when storm flow did not breach the weir. All flow passed through the gypsum bed.

By the summation of load method, system removal of TDP was 9.2 ± 3.7% confidence interval (α = 0.05) during storm-flow events. Obviously, the hydraulic conductivity of the gypsum bed limits its ability to filter drainage during higher-flow storm events.

Overall System Efficiency

Continuous flow monitoring throughout the study allowed calculation of the total amount of TDP removed during storm flow and base flow. After storm flow, the ditch filter functioned as a drainage control structure and slowly released filtered water over a period of several days to weeks. Although P concentrations during base flow conditions were considerably lower than during storm flow, the additional TDP in base flow that was treated by the filter was substantial. Ditch flow, TDP loads, and overall P removal efficiencies are summarized by year in Table 3. The trends are similar to those observed in storm flow. Although total flow was highly variable from year to year, the percent of total ditch flow that passed through the filter declined in 2009 and 2010 as the hydraulic conductivity of the gypsum bed decreased. As expected, the total load of dissolved P entering the ditch filter within a yearly period is highly correlated to the total ditch flow for the year (R2 = 0.98; p = 0.008). The overall system efficiency ranged from 17 to 30% removal of TDP per year (Table 3), which is considerably better than the overall 9.2% removal of TDP in storm flow alone. The lower overall system efficiencies occurred in years 2009 and 2010 in spite of higher chemical efficiencies during these years. The ability to filter larger flow volumes is more important to the overall efficiency of the system than is the chemical efficiency of the filter medium (FGD gypsum).

View Full Table | Close Full ViewTable 3.

Ditch flow, total dissolved phosphorus loads, and phosphorus removal efficiencies by year of operation.

Year Apr.–Dec. 2007 Jan.–Dec. 2008 Jan.–Dec. 2009 Jan.–Dec. 2010
Total ditch flow, L × 103 11.6 12.8 75.9 27.3
Ditch flow passing through the filter, L × 103 4.3 7.0 13.2 4.6
Ditch flow passing through the filter, % 37 55 17 17
TDP† in ditch flow, kg 8.6 13.3 48.2 22.8
TDP removed, kg 2.2 4.4 9.8 4.0
System efficiency based on load, % 25 30 20 17
Chemical efficiency based on load, % 68 57 84 89
Total dissolved phosphorus.

Environmental Impact

Gypsum is a neutral salt and should not strongly affect water pH. The pH of ditch flow in our study typically ranged from 6.0 to 6.5; the effluent that passed through the gypsum filter had a slightly higher pH, typically ranging from 6.5 to 7.0. Measured concentrations of Ca and S in the effluent were used to calculate that approximately 40 Mg of gypsum (36% of the amount used in construction) was dissolved over the course of the study. The fate and environmental impact of elevated concentrations of Ca and S as sulfate were not determined because neither element is regulated. However, ditches in the study area flow to the brackish and marine waters of the Chesapeake Bay where gypsum is a naturally occurring mineral deposit.

Mercury can volatilize after a photoreduction reaction that occurs on exposure of Hg to light (Nriagu, 1994). After installation, the gypsum filter bed was covered with an erosion mat, and vegetation soon established on the surface of the filter. Therefore, Hg losses from FGD gypsum due to Hg volatilization via this photoreduction reaction are expected to be minimal, although we did not attempt to measure volatilization losses. We did detect Hg in both the influent and the effluent during storm flow at concentrations ranging from 0.001 to 0.004 mg L−1 (USEPA drinking water standard = 0.002 mg L−1), indicating that Hg is not removed from solution by the gypsum filter, nor does it leach from the gypsum at concentrations that might cause concern. The farming history includes a period of tomato production. Mercury may have been introduced to these soils through fungicide applications.

Arsenic-containing roxarsone (4-hydroxy-3-nitrobenzenearsonic acid) that was routinely fed to chickens to prevent coccidiosis is the source of arsenic in soils of the study area. Microorganisms of the genus Clostridium in chicken litter rapidly transform roxarsone to inorganic arsenate under anaerobic conditions. Dissolved As (assumed to be in the form of arsenate) was present in the influent during storm flow at concentrations ranging from 0.001 to 0.004 mg L−1 (USEPA drinking water standard = 0.010 mg L−1). Dissolved As in the effluent was below detection limits, suggesting that dissolved arsenate is precipitated as calcium arsenate as it passes through the gypsum filter (Bothe and Brown, 1999). The solubilities of gypsum, As in calcium arsenate, and P in calcium phosphate are 2.1, 0.1, and 0.005 g L−1, respectively. Therefore, As and P should remain insoluble as long as gypsum is present. To prevent the subsequent dissolution and loss of As to receiving waters, the filter bed will have to be removed before complete dissolution of the gypsum. Incubation studies designed to evaluate the effects of land applying “spent” gypsum from a decommissioned filter bed concluded that there will be little effect on dissolved P concentrations in amended soils (Grubb et al., 2011a, 2011b). In these studies, spent gypsum did not appear to provide additional P fertilizer value.

The Future of Flue Gas Desulfurization Gypsum Use in Filtration

For a period of more than 3 yr, 110 Mg of FGD gypsum was used to remove approximately 20 kg of TDP from ditch drainage waters. During that period, approximately 40 Mg of FGD gypsum was dissolved. The data provide evidence that the FGD gypsum filter is chemically effective at reducing TDP and that there is no Hg or As loss that would have a negative environmental impact. However, the system efficiency of the gypsum filter is disappointingly low, and large P loads that move during large storm events mostly bypass the filter and flow to the receiving water body. Perhaps the system efficiency could be improved by mixing sand or a coarser material with the gypsum to increase the flow rate through the gypsum bed. However, the decline in flow rate through the gypsum bed that we observed over time, and as early as the second year of operation, was apparently due to sediment additions that caused surface sealing. The condition was remediated by tilling the surface of the gypsum bed in August 2010. Any increased flow rate through a modified gypsum bed will be defeated by surface sealing due to sedimentation, thus requiring annual removal of vegetation and tillage to maintain the higher flow rate. This filter protected a 17-ha area. To completely treat drainage from an area the size of the average farm would require several filters treating multiple ditches. Annual maintenance requirements and the need to periodically replace the gypsum bed in multiple filters on a single farm are not likely to be acceptable to farm managers.

Subsequent to the construction of the ditch filter, research on the fate and transport of P in soils at this site showed that 90% of the P that reaches the drainage ditch moves laterally in groundwater when water tables are high; only 10% moves overland in runoff (Vadas et al., 2007). Based on this knowledge, we propose trenching adjacent to drainage ditches and filling the trenches with FGD gypsum to intercept and treat P-laden groundwater before it enters the ditch. This alternative design, referred to as the “gypsum curtain,” is currently being tested at the UMES Research and Teaching Farm (Fig. 2). Lateral groundwater movement is much slower than ditch flow, meaning that the hydraulic conductivity of the gypsum should be adequate to accommodate the lower groundwater flow rate. Also, sediment additions leading to sealing and a lower filtration rate of the gypsum should not be an issue. By installing gypsum curtains on all ditches, effluent from an entire farm can be treated before the water reaches the ditch using a system that requires no maintenance. When the treatment begins to fail as gypsum is dissolved, we propose installing a second curtain parallel to the first. There should be no need to remove spent gypsum from the first curtain as long as the groundwater entering the first curtain is saturated with Ca. This new, low-maintenance design would be expected to maximize the chemical efficiency of using FGD gypsum to filter drainage water while minimizing the limiting aspect of a slow through flow rate.


The authors thank the staff and students of the University of Maryland Eastern Shore's Nutrient Management Laboratory for their valuable contributions to this study. Don Mahan collected samples and maintained the study site. Janice Donohoe and Leonard Kibet led and supervised a dedicated group of undergraduate and graduate students who processed and assisted in analyzing samples. This project was funded by the University of Maryland Eastern Shore and the USDA Agricultural Research Service.




Be the first to comment.

Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.