Recycling Vertical-Flow Biofilter: A Treatment System for Agricultural Subsurface Tile Water

Agricultural runoff and similar nonpoint sources of pollution are responsible for widespread degradation of surface water quality in the U.S. (Hall and Killen, 2005; Hardy and Koontz, 2008). In almost three-quarters of the rivers studied in the National Water Quality Survey, nonpoint discharges were major contributors to water quality impairment (US EPA, 1992). Nonpoint source discharges resulting from agricultural runoff add large amounts of inorganic nitrogen and phosphorus to surface water (Goolsby and Battaglin, 2001; Powers, 2007; US EPA, 1992). In the Chesapeake Bay Region (US), nonpoint source discharges contribute about two-thirds of the nitrogen and one-quarter of the phosphorus inputs (Correll et al., 1995). In the 1200 km2 Conestoga River watershed in Pennsylvania, 47.2 kg/ha/yr total nitrogen and 44.7 kg/ha/yr nitrate-nitrogen are discharged from nonpoint sources adding, ultimately, to the nutrient load of the Chesapeake Bay (Woltenmade, 2005). The addition of excessive inorganic nutrients to surface waters leads to eutrophication, which, in turn, is associated with the development of hypoxic zones such as those in the Gulf of Mexico, the Chesapeake Bay, and similar areas (Alexander et al., 2008; Boesch et al., 2001; Mitsch et al., 1999; Wang et al., 2001). Subsurface tile drainage is a common agricultural water management practice used in regions with a seasonally high water table. By taking advantage of this system, farmers are able to extend their growing season by allowing for earlier spring planting and later harvest dates. The use of subsurface tile drainage has been shown to significantly improve crop production (Kladivko et al., 2005). Skaggs et al. (1994) noted that subsurface artificial drainage has improved agricultural production on nearly one-fifth of U.S. soils. In the intensively cropped watersheds of the Midwest United States, the use of subsurface tile drainage has allowed one of the highest agricultural productivities in the world. Approximately 30% of all agricultural lands in the upper Midwest are artificially drained (Zucker and Brown, 1998). Despite all of the benefits to crop production, tile drain lines can have a negative environmental impact. Tile drain lines can act as conduits for contaminants, promoting the rapid movement of these substances to surface waters (Fleming and Ford, 2004; Gentry et al.,

100 2000). A study of tile drain outlets in southwestern Ohio found an average concentration of nitrate-N of 17 mg L -1 was discharged to receiving waters (Fleming et al., 1998). The crop production system employed, the amount, rate, and timing of fertilizer application, the size and arrangement of drainage tiles, and the presence of cover crops are all known to influence nitrogen inputs to surface waters from tile drainage systems (Kaspar et al., 2007;Kladivko et al., 2004;Nangia et al., 2008;deVos et al., 2000;Domagalski et al., 2008;Dinnes et al., 2002). Agronomic controls such as crop and fertilizer management, however, are not usually sufficient to rectify nutrient pollution resulting from tile drainage systems (Jaynes et al., 2008;Madramootro et al., 2007). Therefore, additional methods for nutrient removal and control are needed where subsurface tile drainage is common. Passive treatment systems such as vegetated riparian zones and biofilters have been shown to be effective in controlling nutrient inputs from surface runoff (Cors and Tychon, 2007;Dodds and Oakes, 2006;Mankin et al., 2007;Mayer et al., 2007;Spruill, 2004;Yamada et al., 2007). The preponderance of water flow in tile drainage systems though, is subsurface, within the vadose zone. Therefore, the efficiency of surface systems for treatment may be reduced because substantial amounts of contaminated water may bypass the active treatment zone. To address this limitation, subsurface systems such as in-situ bioreactors, permeable reactive barriers, biofilters, and subsurface flow constructed wetlands have been investigated (Bezbaruh and Zhang, 2003;Darbi et al., 2003;Greenan et al., 2006;Robertson et al., 2007;Schipper and Vojvodic-Vakovic, 2000;Schipper and Vojvodic-Vakovic, 2001;Schipper et al., 2004;Su and Puls, 2007;van Driel et al., 2006). These subsurface systems generally depend on microbial denitrification to mineralize and remove nitrate. Denitrification is an anaerobic process. As such, it requires anaerobic conditions in the subsurface as well as an adequate supply of electron donors and available carbon. Thus, an exogenous source of carbon such as wood chips or sawdust is usually required for these systems to function properly (Greenan et al., 2006;Lin et al., 2002;Vymazal, 2007). We recently reported on a recycling vertical-flow bioreactor (RVFB) for the treatment of household greywater (Gross et al., 2007). This system intercepts and re-circulates contaminated water to a vegetated soil biofilter for aerobic treatment. We report here on the potential use of this system for the removal of excess nutrients from tile water.

Materials and methods
The mesocosm scale RVFBs used in this study have been described elsewhere (Gross et al., 2007). Briefly, each unit consisted of two tiers, each made of a 55 x 40 x 30 cm plastic container. The top container functioned as a soil based treatment unit, while the bottom container served as a reservoir from which water was recycled continuously to the treatment unit ( Figure 1). The RVFB units were run with a total of 40 L of synthetic tile water (STW: [g L -1 ] CaCl 2 , 1.7; NaSO 4 , 1.8; NaHCO 3 , 0.1; KNO 3 , 0.1; K 2 HPO 4 , 0.0004; Humic acid, 0.003) in a semi-batch mode. After initial loading of each unit, water from the reservoir was recirculated to the treatment unit at a rate of 0.41 L min -1 using a 4.6 L min -1 (5 watt) submersible pump. Milk tubing (0.635 cm) modified into a drip line provided uniform water distribution over the surface of the treatment unit. The recirculation rate was set to prevent ponding of water on the soil surface.  Evaluation of the RVFB was conducted under conditions typical of a temperate climate such as that found in the Mid-Atlantic region of the Eastern US. Experimental systems contained a mixed plant community of emergent plants common to southeastern Pennsylvania. An initial period of 4 weeks (designated weeks -4 to 0; data not shown) was allocated for the establishment of the plant community before STW was added for treatment. Plants were excluded from the control system by weeding twice per week. In addition, plants were harvested from one of the experimental treatments at day 30 of the growing season to assess the importance of vegetation in nutrient removal. The systems were maintained in a greenhouse at ambient temperatures for the duration of the study, one growing season (May -October). After allowing for the initial plant establishment and system acclimation, samples were collected twice weekly by draining 20 L of effluent from the reservoir container of each RVFB and replacing it with 20 L of freshly prepared STW. An aliquot (1 L) of the drained effluent was transported on ice immediately to the laboratory for analysis. Samples were stored at 4 o C and analyzed within 24 hours of collection. Nitrate-nitrogen (NO 3 -N), nitrite-nitrogen (NO 2 -N), and dissolved reactive (ortho) phosphate were assayed using standard chemical test kits (HACH Test-N-Tube Plus Methods 835/836, 839 and 834 respectively. All kits follow USEPA approved methods SM 4500. Hach Company, Loveland, CO.). Data was analyzed using the statistical program Prism 4.0 (GraphPad, Inc). Treatments were compared using paired T-tests at a level of significance ( = 0.05) Figure 2 summarizes the removal of nitrate in the RVFB systems. There was no noticeable removal of NO 3 -N from the influent tile water in the control system, indicating that passive removal via adsorption to the soil or microbial transformation was not a significant factor. Effluent nitrate concentrations in the vegetated systems were consistently below the EPA guidelines of 10 mg L -1 . In fact, effluent concentrations of NO 3 -N in these systems rarely exceeded 2 mg L -1 , corresponding to a removal of > 90% of the influent NO 3 -N. Harvesting of the plant community (day 30) resulted in a rapid increase in the effluent NO 3 -N concentration. Within two weeks of the vegetation removal, the concentration of nitrate discharged by the harvested unit approached the concentration discharged in the control system indicating that the bulk of the nitrogen removal in the RVFB was the result of plant uptake and assimilation rather than of denitrification or other soil microbial processes. Nitrate levels in the effluent from the harvested unit remained significantly elevated, in excess of discharge limits, for the remainder of the study. The rapid increase in NO 3 -N seen upon the removal of vegetation indicates that possible harvesting of the plants in a functioning RFVB should be limited to times when tile water discharge is minimal.

Results
Ammonia-N in the effluent of all of the planted systems was consistently below 0.1 mg L -1 (data not shown). The concentration of NO 2 -N was significantly (p = 0.05) lower in the vegetated units than in the control unit. (data not shown). Removal of vegetation from one of the vegetated units (Harvested; day 30) did not have a clear impact on the discharge of nitrite by that unit. While there appeared to be a slight elevation in the concentration of NO 2 -N in the Harvested unit, this increase was transient and may not have been significantly different from the discharge of the remaining experimental units.
Mean +/-S.D. of effluent nitrite-N concentration in RVFB treating synthetic tile water. Plants were harvested in the 30 th day of the experiment (arrow).

Fig. 3. Nitrite (NO 2 -N) in RVFB Effluent
Removal of reactive phosphate from influent tile water was negligible in the RVFB without vegetation (Control; Figure 4). In fact, at times, the concentration of phosphate in the effluent was higher than that in the influent indicating that phosphate was being leached from the soil. In the presence of vegetation, the concentration of phosphate in the effluent was reduced to 20% of the control effluent. Removal of vegetation had no apparent impact on the concentration of PO 4 in the effluent. Retention of phosphate in the remaining root biomass as well as uptake by plant re-growth may account for the low concentrations of PO 4 in the post-harvest effluent, however, the specific cause of this pattern was not established.

Conclusions and future recommendations
Nutrient enrichment from non-point source runoff is a major factor in the degradation of surface water quality (Hardy andKoontz 2008, Ribaudo et al. 2001). The sources of nutrient runoff vary in scale from small individual households to large regional agricultural activities. Similarly, the options for prevention and remediation available for this type of pollution vary widely. Ultimately, multiple technologies at multiple scales must be available to address this issue (Ribaudo et al. 2001).
Mean +/-S.D. of effluent phosphate (PO4) concentration in RVFB treating synthetic tile water. Plants were harvested in the 30 th day of the experiment (arrow).

Fig. 4. Reactive (ortho) Phosphate in RVFB Effluent
The efficacy of management and control procedures in minimizing the runoff of nutrients has been documented widely (Mitsch et al., 1999). However, agronomic practices alone are not enough to eliminate all agriculture-related nutrient runoff. In order to adequately reduce the impact of non-point source runoff, a combination of agronomic practices and treatment techniques is required. A wide variety of treatment options for nutrient runoff have been developed. These include riparian buffer zones, biofilters, denitrification walls and constructed wetlands (Bezbaruah and Zhang, 2003;Darbi et al., 2003;Jaynes et al., 2008;Kelly et al., 2007;Lin et al., 2007;Su and Puls, 2007;Vymazal, 2007;Yamada et al., 2007). Although each of these systems has been shown to reduce the NO 3 -N concentration in groundwater, there is no single ideal system appropriate for use under all circumstances. Our research demonstrates the potential use of a recycling vertical-flow biofilter (RVFB) as an alternative treatment option for the removal of nutrients from contaminated tile water. Using this relatively simple system we were able to achieve a > 95% removal of reactive phosphate and a > 90% removal of nitrate-nitrogen from STW. The RVFB combines characteristics of a constructed wetland, a riparian buffer zone and a trickling filter for aeration. Subsurface flow intercepted by the RVFB is recycled to the soil surface. From there, it flows through a vegetated soil bed treatment system where combined biotic and abiotic processes remove excess nutrients and then flows through a layer of hollow plastic spheres, finally trickling into a reservoir. Movement of the water through the hollow spheres re-aerates the water and prevents the development of anaerobic conditions. Thus the RVFB is a hybrid treatment system combining the advantages of several existing treatment processes into one system capable of treating both surface and subsurface runoff.
For example, reactive phosphate is most likely removed by a combination of plant uptake and soil sorption. Removal of nitrogen compounds, on the other hand, is most likely the result of plant uptake as well as limited nitrification and denitrification by the soil microbial community. The importance of plant uptake in the removal of nutrients is reflected in the increase in the concentration of both nitrate and phosphate in the harvested unit after the removal of vegetation from the system. No such increase in the concentration of nutrients was seen in units from which the vegetation was not removed. Because of the dominant role of vegetation in the RVFB, application of this technology to field situations must consider the management and ultimate use of the vegetation. The RVFB has several advantages over other systems for the treatment of agricultural runoff. Since it is based on modular components -the upper soil treatment module, functionally similar to a riparian buffer, the plastic spheres, functionally similar to a trickling filter and the lower recirculation and reservoir module -there is flexibility in the design, allowing complete units to be tailored to a specific site. In addition, the use of separate modules should reduce maintenance costs since repair to a single component can be done by simply replacing the module without the need to disassemble the entire system and disrupting its operation. In exploiting the ability of vegetated soil systems to sequester and transform inorganic nutrients, the RVFB reflects the advantages of a riparian buffer, a well-established treatment modality for the prevention of surface water pollution from agricultural runoff (Cors and Tychon, 2007;Mankin et al., 2007;Mayer et al., 2007;Schoonover et al., 2005;Yamada et al., 2007). By adding a subsurface recirculating reservoir, the RVFB also is capable of intercepting and treating subsurface runoff, particularly tile water that normally bypasses riparian buffers. Unlike constructed riparian wetlands, however, the RVFB operates in a primarily aerobic mode: water trickling through the soil and into the reservoir is aerated. Thus, the generation and release of nitrogenous greenhouse and ozone depleting gases and precursors of acidic deposition (e.g. N 2 O) associated with denitrification-based systems may be avoided (Magner et al., 2004;David et al., 2009) Our results provide a proof-of-concept only for the RVFB. Additional research is needed to demonstrate how well the system functions under realistic field conditions as well as the costs of this system compared to alternatives. We believe, however, that the RVFB has the potential to be a useful addition to the armamentarium in the fight against non-point source pollution.