F.J. Sikora¹, L.L. Behrends¹, H.S. Coonrod¹, W.D. Phillips¹, and D.F. Bader²

Constructed wetlands have been successfully used to treat a wide variety of wastewaters. Municipal wastewater and acid mine drainage encompasse most of the water treated by constructed wetlands (Hedin and Narin, 1993; Kadlec and Knight, 1996). Other wastewaters treated with wetlands include petroleum industrial effluents, pulp and paper wastewater, and landfill leachates (Litchfield, 1993; Marten, et al., 1993; Thut, 1993). The main advantage of constructed wetlands is that the technology is inexpensive compared to conventional treatment options. There are capital costs associated with building the wetland, but the minor operation and maintenance costs makes constructed wetlands a cheaper alternative to conventional treatments that include yearly labor and chemical costs.

Many army ammunition plants across the country have problems with groundwater contaminated with explosives. A demonstration was initiated at the Milan Army Ammunition Plant near Milan, Tennessee early in 1996 to test the feasibility of treating contaminated groundwater with constructed wetlands.

The design for the demonstration was developed from information obtained from two preliminary microcosm studies conducted in 38 L aquaria and are reported in Sikora, et al. (1995) and Behrends, et al. (1995).

The demonstration involves a comparison of two wetland types including a lagoon system with submergent plants and a subsurface flow gravel-bed wetland with emergent plants. A lagoon system consists of two cells in series with each cell having dimensions of 24 x 9.4 x 0.6 m (L x W x H) containing 137 m³ water. A gravel-bed system has a first cell with dimensions of 32 x 11 x 1.4 m (L x W x H) containing 205 m³ water and is kept anaerobic by adding 114 kg of powdered milk to the water every two weeks. The second cell in the gravel-bed treatment system has dimensions of 11 x 11 x 1.4 m (L x W x H) containing 43 m³ water and is kept aerobic by reciprocating water back and forth from one interior partition to another (patent application filed). Gravel used in the gravel-based wetlands had 42% porosity. For the purpose of the demonstration, effluent from both wetland treatments goes through granular-activated carbon-bed canisters before being discharged to the sewer system. The lagoons were planted with sago pond weed (Potamogeton pectinatus), water stargrass (Heteranthera dubia), elodea (Elodea canadensis), and parrot feather (Myriophyllum aquaticum). The gravel-bed wetlands were planted with canarygrass (Phalaris arundinacea), woolgrass (Scirpus cyperinus), sweetflag (Acorus calamus), and parrot feather. Water began flowing to each of the wetland treatment systems at 19 L min^-1 starting on June 17, 1996. On November 21, 1996, the source of the contaminated groundwater was changed to a new well which contained the explosive contaminants at higher concentrations. The design hydraulic retention time was 10 days through the lagoon system and 9.1 days through the gravel-bed system. The retention time in each of the lagoon cells was 5 days. The retention time in the first anaerobic gravel-bed wetland was 7.5 days. The retention time in the reciprocating gravel-bed wetland was 1.6 days.

Influent and effluent water samples were collected every 2 weeks. Intensive sampling of water at interior locations in the wetlands occurred every 2 months. Interior water samples were collected at equal distances between the inlet and outlet at 4 locations in the lagoon cells and first gravel-bed wetland cell, and 2 locations in the reciprocating gravel-bed wetland. The demonstration and the scheduled sampling is planned to continue until August 1997. Data will be presented for biweeklywater samples collected from July 1, 1996, to April 30, 1997, and intensive water sampling that occurred on August 13, 1996, October 8, 1996, December 3, 1996, February 11, 1997, and April 8, 1997.

Water samples were analyzed for explosives via HPLC. A photodiode array detector with a range of 190 to 367 nm was used for analyte qualification and confirmation. A fixed-wavelength detector at 254 nm was used for analyte quanitification. Total organic C (TOC) was analyzed via Dohrmann DC 190 TOC analyzer. Total Kjeldahl N (TKN) and ammonium N (NH₄-N) were analyzed with LACHAT flow-injection analysis. Biological oxygen demand (BOD) was analyzed via standard methods (Greenberg et al., 1992). Chemical oxygen demand (COD) was determined via HACH digestion and colorimetric analysis. Dissolved oxygen (DO), temperature, and pH were determined with a hand-held YSI probe at time of sampling.

Disappearance rate of TNT, RDX, and HMX was analyzed using first-order kinetics (Kadlec and Knight, 1996). Assuming plug-flow hydraulics, the first-order equation for the reduction of a pollutant in a wetland is:

where k is the first-order rate constant with units of m/yr; q is the hydraulic loading rate with units of m/yr; y is the fractional distance from inlet to outlet (ranging from 0 to 1); Ci is the influent concentration of pollutant; and C is the concentration at y. The k value for removal of TNT, RDX, and HMX in the lagoon and gravel-bed systems was determined via linear regression of ln (C/Ci) versus -y/q where the intercept was maintained at zero. The slope from the regression was the rate constant, k.

The temperature of the water leaving the two wetland systems and the temperature of the groundwater entering the wetlands are shown in Fig. 1. The temperature of the groundwater remained relatively stable due to the insulation of the water in the ground. The temperature of the water leaving the wetlands varied considerably as a reflection of the seasonal air temperatures. The first samples collected in July reflect a high water temperature during the summer months of approximately 30^o C. During the colder winter months of January and February (days 198 to 256), the effluent water temperature decreased to approximately 5^o C.

Approximate influent concentrations into the wetland systems were 1.2, 1.8, 0.1, and 0.1 mg/L for TNT, RDX, TNB, and HMX, respectively, during the earlier period of the demonstration (Fig. 2). After 155 days, the average influent concentrations were 4.6, 4.4, 0.35, and 0.1 mg/L for TNT, RDX, TNB, and HMX, respectively. The gravel-bed wetland did a good job reducing TNT and TNB effluent concentrations below the detection level of 0.002 mg/L. The gravel-bed removed RDX and HMX below detection levels of 0.005 mg/L during the warmer periods of the year, but some release of RDX and HMX occurred during the winter. The decreased removal efficiencies for RDX and HMX could have been due to colder temperatures decreasing microbial activity. The lagoon system did nearly as well at removing TNT and TNB in the water as the gravel-bed systems but did a poorer job at reducing RDX and HMX. A seasonal effect was observed for TNT and TNB removal in the lagoons where higher concentrations of the explosives were observed in the effluent during winter.

The rate of explosives removal in the wetlands was determined by first-order kinetic analysis (Table 1). The rate of TNT removal in the anaerobic gravel-bed was rapid with k values ranging from 300 to 960 m/yr. The rate of RDX removal in the gravel-bed wetland was not as rapid as TNT but was still high with k values ranging from 41 to 347 m/yr. Removal rate constants for TNT and RDX in the lagoon cells were less than those observed in the gravel-bed wetland (Table 1). Lower-rate constants were observed during February in both the gravel-bed and lagoon wetlands which was a reflection of a decrease in the removal efficiency during colder temperatures.

The reciprocating gravel-bed wetland follows the anaerobic gravel-bed wetland in order to remove nutrients and carbon that is released in the first wetland from continual feeding with powdered milk. Reciprocating water in a gravel-bed wetland allows for aeration of microbial biofilms on the gravel which hastens ammonium and BOD removal via aerobic microbial activity. The average dissolved oxygen concentration increased from 1.6 mg/L going into the reciprocating wetland to 4.5 mg/L exiting the wetland indicating improved aeration of the water (Fig. 3). Ammonium- and BOD-removal efficiencies were greater than 90 % in the reciprocating wetland and were fairly rapid since the retention time in the reciprocating wetland was only 1.6 days.

In addition to the primary explosives, TNT by-products analyzed in the water samples were 2-amino-4,6-dinitrotoluene (2A), 4-amino-2,6-dinitrotoluene (4A), 2,4-diamino-6-nitrotoluene (2,4-A), and 2,6-diamino-4-nitrotoluene (2,6-A). 2,6-A was below the detection limit of 0.006 mg/L in all samples. The by-products of TNT were observed in the gravel-bed wetlands at the retention time of 1.5 days at concentrations approximately 10% of the influent TNT concentration (Fig. 4). The degradation of TNT in the anaerobic gravel-bed wetland is believed to have occurred via reduction of the molecule to amino derivatives to ultimately form 2,4,6-triaminotoluene (TAT) which may be further polymerized to harmless humic-like substances (Rieger and Knackmuss, 1995). The increased concentration of all the amino derivatives was about 20% of the initial TNT concentration and is comparable to results from a batch study with microcosm wetlands in aquaria where 25% of the initial TNT concentration was observed to form 4-A (Sikora, et al., 1996). Both TNT-degradation products, 2-A and 4-A, were observed to increase in the lagoon to approximately 2% of the initial TNT concentration in April (Fig. 4). No increase in degradation products was observed with a decrease in TNT concentration in August. The slight or non-existent increase in TNT-degradation products in the lagoon system was puzzling. TNT is believed to be degraded by submergent plant species via production of nitroreductase enzymes (Wolfe, et al., 1994) which should have yielded amino derivatives in the aqueous phase with a decrease in TNT concentrations as observed with parrot feather in microcosm wetlands (Sikora, et al, 1996).

By-products of RDX analyzed were mononitroso-RDX (m-RDX) and trinitroso-RDX (t-RDX). Both m-RDX and t-RDX concentrations were found to increase with concurrent decreases in RDX concentration in the gravel-bed wetlands (Fig. 5). t-RDX was more persistent in the wetland water with 0.06 and 0.7 mg/L t-RDX remaining in the effluent in August and April, respectively. The RDX by-products were not observed in the lagoon system (Fig. 5). The lack of RDX by-products was not surprising since there was only a slight decrease in RDX concentrations.

The data presented for explosives removal using wetland systems is preliminary data describing removal in newly constructed wetlands in operation for 10 months. The project is continuing for at least one full year. Longer term observation of these systems is also desirable to determine removal of explosives in matured constructed wetlands. In addition to sampling water, sediment, gravel, and plant samples are being taken and analyzed for explosives to determine if explosives and by-products are sorbing onto or into various components of the wetlands.

Influent concentrations of TNT, RDX, TNB, and HMX were respectively, 1.2, 1.8, 0.1, and 0.1 mg/L before November 21, 1996, and respectively, 4.6, 4.4, 0.35, and 0.1 mg/L after November 21, 1996. The gravel-bed wetland reduced TNT and TNB concentrations below 0.002 mg/L during the entire demonstration period. Removal of TNT and TNB was complete in the lagoons during the warmer temperatures but less complete during colder winter months. The gravel-bed wetland removed RDX and HMX during warmer temperatures with less removal efficiency than during colder winter months. The lagoon was very ineffective at removing RDX and HMX in the contaminated groundwater.

The study was funded through the U.S. Department of Defense Environmental Security Technology Certification Program administered through the U.S. Army Environmental Center in Aberdeen Proving Ground, Maryland. The project manager was Darlene F. Bader with the U.S. Army Environmental Center. In addition to the authors, many people were responsible for the success of this demonstration project. Contributors to sample collection and analyses included Cathy McDonald, Earl Bailey, Jerry Clayton, and Keith Bozeman. Contributors to wetland maintenance included Jerry Berry, Danny Williams, and Eddie White. Contributors to conceptual ideas and project direction included Richard Almond and Joseph Hoagland. Contributor to engineering design of the wetlands was Randy Summers.

Table 1. First-order rate constants (k values in units of m/yr) for removal of TNT, RDX, and HMX in wetland treatment systems sampled throughout the demonstration period.

Figure 1. Temperature changes in the influent groundwater entering the wetlands and water leaving the gravel-bed and lagoon wetland systems.

Figure 2. Concentration of four explosives, TNT, RDX, TNB, and HMX, in the inflow groundwater entering the cells and the effluent water leaving the gravel-bed and lagoon wetland systems

Figure 4. TNT and TNT by-products (2-A = 2-amino dinitrotoluene, 4-A = 4-amino dinitrotoluene, 2,4-A = 2,4-diamino nitrotoluene) observed in water samples taken from gravel-bed and lagoon wetlands during sampling events in August 1996 and April 1997. Left y-axis is for TNT concentrations and right y-axis is for concentration of by-products.

Figure 5. RDX and RDX by-products (m-RDX = mononitroso RDX, t-RDX = trinitroso RDX) observed in water samples taken from gravel-bed and lagoon wetlands during sampling events in August 1996 and April 1997. Left y-axis is for RDX concentrations and right y-axis is for concentration of byproducts.