It was specifically hypothesized that an on-farm nutrient removal system could be designed to reduce nutrients in stormwater runoff to within acceptable limits by trapping the nutrients at the source. To test this hypothesis, a rainfall simulation and a small-scale containment structure were used to test various byproduct thicknesses for nutrient removal from broiler house dust (BHD). The containment structure (i.e., leachate column) was designed by adapting 56.8-L buckets to contain the RM and WT byproducts. Dust was applied to the surface of the byproducts in each leachate column followed by simulated rainfall application. The leachate produced was then collected for chemical analysis.

Prior testing of the RM and WT byproducts [11] showed the hydraulic conductivity of 6-mm-diameter particles of RM was 123 cm 30 min⁻¹, while that of the 6-mm-diameter particles of WT was 138 cm 30 min⁻¹. These hydraulic conductivities were determined to be more than adequate to transmit the calculated volume of rainfall that would fall to the ground adjacent to broiler house fan outlets during a 25-yr, 30 min storm event (i.e., 33 cm per 30 min).

Three byproduct thicknesses (i.e., 4, 8, and 12 cm) were tested using the 56.8-L leachate columns. The columns were constructed by removing the bottoms of 27, 56.8-L buckets, leaving the remainder of the bucket for the specified byproduct thickness plus 7.6 cm of open space at the top. Welded wire mesh, with 1.27 cm openings, was attached to the bottom of the buckets and landscape fabric was placed on top of the wire mesh. The byproduct was placed on the landscape fabric. The cut buckets were placed inside uncut 56.8-L buckets, perforated on the sides to allow airflow, to capture leachate during the experiment (Figure 1). Leaching experiments were conducted in triplicate for each byproduct-thickness treatment combination, including controls for each byproduct that received no dust and dust only treatments (i.e., no byproduct).

Dust collected off fan shutters from four broiler facilities in northwest Arkansas was combined and mixed thoroughly as described by Herron et al. [12] . The dust application rate for the experiment was calculated based on direct measurements of dust thicknesses that had settled on the ground adjacent to fan outlets during one grow-out cycle (flock) of birds (approximately 6 weeks), which averaged 3.2 mm per flock. Based on the surface area of the leachate

column (638 cm²) and the estimated bulk density of the dust (0.23 g∙cm⁻³), 45.43 g of dust were used for each byproduct-thickness treatment combination. The dust was analyzed by the University of Arkansas Agricultural Diagnostic Laboratory and contained 4312 mg water extractable P kg⁻¹, 10,400 mg total P kg⁻¹, 3,785 mg NH₄-N kg⁻¹, 277 mg NO₃-N kg⁻¹, and 97,600 mg total N kg⁻¹. The dust was applied by hand uniformly to the surface of byproduct in each column. For the no-byproduct treatment, dust was applied by hand uniformly directly to the landscape fabric.

The rainfall simulator used for this experiment was based on the design of Miller (1987). The dimensions for the area under the simulator were 2.0 meters wide by 1.5 meters long. The frame was built of hollow metal tubing, on which a 25-mm-diameter, polyvinyl chloride (PVC), water-supply pipe was mounted. Polyvinyl chloride windscreens were attached to four sides to limit wind effects on rainfall distribution. A water hose supplying water from the city water supply was attached to the water-supply pipe. A single, fixed nozzle (TeeJet^TM 1/2HH-SS-5OWSQ, TeeJet, Glendale Heights, IL), designed for a flow rate of 210 mL∙sec⁻¹, was installed on the water-supply pipe and centered at the top of the frame, 3 m above the surface of the leachate columns [19] . A low-pressure regulator and an oil-filled pressure gauge were used to maintain the water pressure at 37.9 kPa (5.5 PSI). An in-line filter was placed in the supply pipe to prevent particles clogging the line and nozzle during water delivery.

Leachate columns were randomly placed under the rainfall simulator in three zones, one zone for each replicate set of treatment combinations (Figure 2, Figure 3). Rainfall intensity was verified and distribution uniformity was determined by

simulating rainfall three times onto the area where the leachate columns were to be placed and measuring the resulting rainfall volume. The exact locations of the three zones were then selected based on the results of the rainfall uniformity tests. The coefficient of uniformity was determined using the method described by Christiansen [20] , which expresses uniformity as a percentage according to the formula coefficient of uniformity = 100 * (1.0 ? Σd/m * n), where d is the standard deviation of individual observations from the mean value m, and n is the number of observations [21] .

Rainfall simulations were conducted at an intensity of 7 cm∙hr⁻¹ based on a 2-year, 24-hour storm rate and leachate was collected for a total of 30 minutes according to the National Phosphorus Research Project [19] . Leachate columns received 30 minutes of simulated rainfall prior to the first treatment with dust to equalize their water contents. After the first rainfall simulation, dust was surface applied to the byproducts in each column, which then received simulated rainfall. Six consecutive dust applications were followed by 30 minutes of simulated rainfall at a rate of 7 cm∙hr⁻¹, the combination of which occurred at one-day intervals. Dust applications approximated the amount of dust deposited on the soil outside a broiler house fan during one year of typical operation (i.e., the production of six flocks of birds; [12] ).

Following each rainfall simulation, leachate was collected from the bottom of the columns. Leachate volume was measured and a 1-L sub-sample was stored at 4˚C and analyzed within 24 hours of collection. A sample of source water used for the simulated rainfall was collected and similarly retained for analysis. Forty milliliters of leachate were filtered (0.45-µm membrane) immediately after collection and stored at 4˚C until analyzed for dissolved-reactive P (DRP) by the colorimetric molybdenum-blue method [22] . Ammonium-N (NH₄-N) and nitrate-N (NO₃-N) were analyzed colorimetrically by flow-injection analyses (Lachat Instruments QuikChem 8500, Loveland, CO). One hundred twenty-five milliliters of leachate were acidified to a pH ≤ 2 with 12 drops of concentrated sulfuric acid for sample preservation and analyzed for total P and N following persulfate-autoclave digestion [23] . Total P (TP) was determined by spectrophotometry (Beckman Coulter, Pasadena, CA) and total N (TN) was determined by flow-injection analysis (Lachat Instruments QuickChem 8500, Loveland, CO). One hundred twenty-five milliliters were analyzed gravimetrically for total solids after oven drying at 105˚C for 12 h.

The TP results were tested for and had homogeneous variances. Therefore, a two-factor analysis of variance (ANOVA) was conducted on the concentrations of TP to evaluate the effects of byproduct, byproduct thickness, and their interaction. Due to non-homogeneous variances, separate one-factor ANOVAs were conducted on the leachate concentrations of DRP, NH4-N, NO3-N, and TN to separately evaluate the effect of byproduct for each thickness and thickness for each byproduct. All relationships were reported as statistically significant at the p < 0.05 level unless noted otherwise. Wolfram Mathematica (online beta version 2014, Wolfram, Champaign, IL) was used to conduct all statistical analyses.

The average rainfall uniformity was 95%, 94%, and 87% for zones 1, 2, and 3, respectively, as determined by the coefficient of uniformity (Table 1). Additionally, the coefficient of uniformity among the three consecutive 30 min, 7 cm∙hr⁻¹

rainfalls was 93%. Thus, the configuration of leachate columns in the rainfall simulator produced uniform and repeatable rainfall both within the simulator footprint and among individual simulated rainfall events. For instance, coefficients of uniformity of 90% to 93% were considered excellent by Humphry et al. [21] and Miller [24] , and the National Phosphorus Research Project protocol [19] recommends a coefficient of uniformity of 85% or greater for standardized rainfall simulations using the same rainfall simulation equipment as used in this study.

Leachate concentrations for all thicknesses of RM ranged from 0.001 to 1.8 mg DRP L⁻¹, 0.28 to 16.7 mg TP L⁻¹, 0.6 to 442.5 mg NH₄-N L⁻¹, 0.0 to 4.8 mg NO₃-N L⁻¹, and 5.5 to 450.2 mg TN L⁻¹ (Table 2). Leachate concentrations for all thicknesses of WT ranged from 0.03 to 9.8 mg DRP L⁻¹, 0.43 to 23.3 mg TP L⁻¹, 51.4 to 307.3 mg NH₄-N L⁻¹, 0 to 114.0 mg NO₃-N L⁻¹, and 0.4 to 346.6 mg TN L⁻¹ (Table 2). Neither thickness, byproduct type, nor their interaction affected (p > 0.05) the concentrations of DRP, TP, NH₄-N, NO₃-N, or TN (Table 2), when the thicknesses and byproducts were compared together. However, when the thicknesses and type of byproduct were compared separately, the DRP concentration was lower (p < 0.05) for RM than for WT for the 4- and 12-cm byproduct thicknesses (Table 2). Similarly, the NH₄-N concentration was also lower (p = 0.02) for RM than for WT for the 8-cm byproduct thickness (Table 2). In addition, the NO₃-N concentration was lower (p < 0.002) for RM than for

^#Dissolved reactive phosphorus (DRP), total phosphorus (TP), ammonium nitrogen (NH₄-N), nitrate nitrogen (NO₃-N), and total nitrogen (TN). *Two-way ANOVA conducted on data with homogeneous variances. ^†One-way ANOVA conducted on data without homogeneous variances.

WT for the 8- and 12-cm byproduct thicknesses. Comparing the leachate analyses of dust without byproduct to the dust with byproduct treatments, the 8-cm-thick treatments of both byproducts were effective at reducing leachate TP, with RM sorbing up to 94% and WT up to 87% of P leached from dust (Table 3). In the case of N, the RM was more effective than WT in reducing leachate NH₄-N concentrations. The 8 cm-depth treatment of RM reduced leachate NH₄-N concentrations by 75%.