This study was conducted at the University of Arkansas’ Savoy broiler production facility (36˚7'43.41''N, 94˚18'22.84''W) in western Washington County, northwest Arkansas (Figure 1). This facility is a small-scale, commercial production facility located in the Ozark Highlands (i.e., Major Land Resource Area 116A) [15] and consists of a set of four broiler houses. The natural soils immediately surrounding this facility, which is located in a summit landscape position at an elevation of approximately 384 m, are mapped as Nixa very gravelly silt loam (loamy-skeletal, siliceous, active, mesic Glossic Fragiudults) [16] . The Nixa series soils are very slowly permeable soils that formed in cherty limestone [16] , thus natural rainfall is prone to surface runoff from the area surrounding the broiler houses. Runoff from the area flows to nearby upland headwater tributaries that flow into the Illinois River, which is approximately 3.3 km directly west of the study site.

Rainfall-runoff simulations were conducted adjacent to house #4 at the facility (Figure 2) to evaluate the effects of fan type (i.e., sidewall ventilation fans versus tunnel ventilation fans) and fan operation time (i.e., season, warm months versus cool months) on runoff quantity and runoff water quality; thus the experiment was treated as a block design. Rainfall-runoff simulations were conducted adjacent to only one house to minimize variability in nutrient concentrations in the dust on the soil surface adjacent to each fan. Though the measured data were specific to this particular single house and this production facility, the results were assumed to be representative of other broiler houses with both sidewall and tunnel ventilation fans throughout the Ozark Highlands region.

At the study site, sidewall fans 1 (S1) and 4 (S4) operate on a timed schedule at the beginning of each flock during cool weather (Figure 2(a)). Sidewall fans S1, S3, and S4 operate on a timed schedule at the beginning of each flock during warm weather. If the in-house temperature is 1.1˚C above the desired level, all sidewall fans, including sidewall fan S2, automatically operate to maintain house temperature and humidity. Sidewall-fan use dominates during cool weather. When the in-house temperature rises to 1.6˚C, 2.2˚C, 3.1˚C, or 3.3˚C above the desired level, tunnel ventilation fans T1, T2, T3, and T4, respectively, automatically start operating (Figure 2(b)). Prior to T4 operating, airflow is directed across the house. Once T4 begins operating, airflow is pulled down the house through cooling cells at the opposite end of the house.

Tunnel fan use is greater during warm weather and each tunnel fan (34,000 m³ min⁻¹) exhausts air at an appreciably greater rate than a sidewall fan (20,400 m³∙min⁻¹), resulting in more dust being exhausted from the house via tunnel than sidewall ventilation fans. Based on the amount of electricity used by each fan to maintain optimum in-house humidity and temperate, the volume of air exhausted by each fan was determined (Figure 3).

Identically sized plots (3 m²) were installed outside broiler house #4 adjacent to each fan (Figure 4). As the grassed areas adjacent to the fans are exposed to natural rainfall, structures were built to cover the plot areas to protect deposited dust from natural rainfall (Figure 4(a)). For each fan, plots were positioned within the high-volume dust trail and in areas with consistent slopes (~3%) to obtain uniform runoff (Figure 4(b)). Dust volumes and deposition uniformity were expected to differ at least slightly among plots due to variability in fan operating time. The runoff plots were covered to avoid disturbance by natural rainfall between flocks (i.e., when there were birds in the house). Roof structures were removed prior to each rainfall simulation study, and replaced immediately following rainfall. The only disturbance to the grass at the site was cutting grass to a 5-cm height following each rainfall simulation.

Rainfall simulation plots 1.5-m wide by 2-m long with 3% slope bounded by 0.2-cm thick and 15-cm wide metal frames, were installed in the center of the dust plume adjacent to each fan (Figure 4(b)). Runoff collection gutters at the downslope edge of the plots diverted runoff to a below-ground container, from which the runoff was pumped into a 190-L barrel using a peristaltic pump. The frames were designed to fit under the rainfall simulator so the areas enclosed would receive equal amounts of rainfall with a coefficient of uniformity of >90% and followed National Phosphorus Research Project and U.S. Department of Agriculture-Natural Resources Conservation Service protocols [17] - [19] .

Rainfall simulation studies were conducted following bird removal for five consecutive flocks over a one-year period. Simulations on sidewall fan plots were conducted between April 2012 and January 2013 and on tunnel fan plots between April 2013 and January 2014. The portable rainfall frame, which was centered and leveled over the plot, was constructed of hollow metal tubing, upon which a single fixed nozzle (TeeJet^TM 1/2HH-SS-

5OWSQ), designed for a flow rate of 210 mL∙s⁻¹, was mounted central to the top of the frame and 3 m above the surface of the plot [18] . Windscreens were attached to three sides to restrict wind impact on rainfall uniformity. Water was supplied by a 25-mm diameter pipe from the farm’s well. Samples of simulated rainfall were collected prior to each rainfall simulation for chemical analysis.

A regulator maintained water pressure at 28.3 kPa (4.1 PSI) to produce raindrops of similar size to natural rainfall [20] [21] . Rainfall distribution and volume were measured prior to each simulation study using five collection pans located in each corner and the center of the plot, equidistant from each other. Pressure was maintained at 28.3 kPa (4.1 PSI) and rainfall was collected for 5 minutes prior to measurement.

Rainfall simulations were conducted at an intensity of 7 cm∙hr⁻¹, which is equivalent to a 2-year, 24-hour storm, and runoff was collected for a total of 30 minutes according to the National Phosphorus Research Project [18] . Rainfall simulations were conducted on the same day for all plots of a particular fan type.

The chemical composition of runoff from pastures adjacent to the farm, which had not received fertilizer nor had been grazed in the prior 15 years, were used as unamended control observations for comparison [22] .

Total runoff collected during the rainfall simulations was weighed to determine volume and subsampled for analyses. A 40-mL subsample was filtered (0.45 µm) immediately after collection in the field and stored at 4˚C until analyzed. The concentration of dissolved P (DP) was determined on filtered runoff by the colorimetric molybdenum-blue method of Murphy and Riley [23] and ammonia-N (NH₃-N) and nitrate-N (NO₃-N) were measured on a Lachat Instruments QuikChem 8500 flow injection analysis system (Milwaukee, WI). A 125-mL subsample was acidified with 12 drops of concentrated sulfuric acid for sample preservation. Total phosphorus (TP) and total nitrogen (TN) were determined on unfiltered runoff samples following persulfate/autoclave digestion [24] . Total solids (TS) content was determined gravimetrically on a 125-mL sample of unfiltered runoff after over-drying (105˚C) for 12 hours. Rainwater was sampled, prepared, and analyzed identically.

Flow-weighted concentrations were determined by calculating the load (i.e., event concentration multiplied by total volume from 30 minutes of runoff) for each fan from a rainfall simulation runoff event. Loads for all of the plots for the runoff event were summed and flow for all of plots and runoff events were summed. Finally, the sum of loads was divided by the sum of runoff to obtain flow-weighted concentrations for the event for sidewall and tunnel fan plots. The flow-weighted concentrations represent the worst-case scenario if a poultry house was located adjacent to a waterway.

Separate one-factor analyses of variance (ANOVA), were conducted to evaluate the effects of season and fan location/type (i.e., sidewall and tunnel fans) on runoff water quality (i.e., DP, TP, NH₃-N, NO₃-N, TN, and TS concentrations). Since sidewall and tunnel fans did not operate at the same times of the year, the interaction effects of season and fan type could not be evaluated. All relationships are reported as statistically significant at the P < 0.05 level, unless noted otherwise. Wolfram Mathematica (online beta version 2014, Wolfram, Champaign, Illinois) was used for all statistical analyses.

The volume of air exhausted from house #4 by sidewall fans S1, S2, S3, and S4 and tunnel fans T1, T2, T3, and T4 were determined from in house electricity use records over 2012 and 2013 (Figure 3). A greater volume of air was exhausted by the four sidewall (52 and 62%) than then the four tunnel (48 and 38%) fans in the mid-winter flocks of 2012 (24th February to 11th April) and 2013 (5th February to 18th March), respectively (Figure 3). In contrast, during the warmer summer months (June through August), tunnel fans exhausted a greater volume of air from house #4 (89 and 84%) than from the sidewall fans (11 and 16%) during 2012 and 2013, respectively (Figure 3).

On an annual basis, the four sidewall fans exhausted 35% and tunnel fans 65% of the total amount of air exhausted in 2012. However, in 2013 sidewall fans exhausted a greater proportion of air than in 2012, exhausting 43% of the annual volume of air with tunnel fans exhausting 57% (Figure 3). The greater tunnel fan use relative to sidewall fans in 2012 than in 2013, reflects the appreciably warmer June, July and August of 2012 (average high of 34˚C) than 2013 (average high of 29˚C from National Oceanic and Atmospheric Administration http://www.srh.noaa.gov/). The difference in air volume between fan type and over time of the year when birds are present in the house, influences the amount of dust exhausted and deposited adjacent to house ventilation fans.

The concentration of P and N forms in runoff varied with time of the year that rainfall was applied for sidewall (Figure 5 and Figure 6) and tunnel fans (Figure 7 and Figure 8). For sidewall fans, P and N concentrations were lower (P < 0.01) in runoff from fans S2, S3, and S4 for the August rainfall simulation (9.0 and 56 mg∙L⁻¹ for TP and TN, respectively) than in runoff for the other four rainfall simulation dates (2.5 and 16.4 mg∙L⁻¹). This seasonal difference reflects the fact that sidewall fans were used less than tunnel fans in warmer, more humid months, when tunnel fans operate longer to more effectively achieve desired in-house temperature and humidity.

This shift in fan use with season is reflected in greater P and N concentrations in runoff from the tunnel fans in May and August (8.9 and 86.1 mg∙L⁻¹, respectively) compared with runoff from rainfall simulations in October, January, and April (2.9 and 31 mg∙L⁻¹, respectively; Figure 7 and Figure 8). Clearly, the potential for dust emitted from fans can enrich the P and N concentration of runoff occurring immediately adjacent to the fans. This potential is directly related to fan use, which changes during the year, with seasonal fluctuations in in-house temperature and humidity triggering fan use.

The concentration of DP and TP in runoff from sidewall fan S1 tended to be greater than from S2, S3, and S4 for each simulations (Figure 5). Unlike P, N did not vary consistently among fans (Figure 6). Sidewall fan S1 operates longer than the other sidewall fans, irrespective of season, with only fan S1 operating for the first week of a new flock, when birds occupy a quarter of the house. Thus, the amount of dust emitted from fan S1 is expected to be greater than from the other three fans, which related directly to the nutrient runoff trends observed.

Concentrations of P and N were converted to a flow-weighted basis within rainfall simulation events by calculating loads at each fan, summing the loads, then dividing the total by the total flow for all runoff plots (Figure 9 and Figure 10). The average for runoff concentrations from sidewall and tunnel fans allowed a comparison

among fans and with nutrient runoff from nearby pastures that had not been fertilized or grazed in 15 years and pasture that had received poultry litter (Table 1). Fluctuations in average flow-weighted concentrations of P and N forms in runoff from plots adjacent to sidewall and tunnel fans were consistent with changes in fan use (Figure 9 and Figure 10).

The flow-weighted concentration of DP in runoff from plots adjacent to sidewall (4.9 mg∙L⁻¹) and tunnel fans (2.7 mg∙L⁻¹) was within the range of values measured in runoff from area pastures that received varying amounts of poultry litter (0.7 to 33 mg∙L⁻¹; Table 1). Total P concentration in runoff from sidewall (8.1 mg∙L⁻¹) and tunnel fans (4.7 mg∙L⁻¹) tended to be at the lower limits of the range measured in fertilized pasture runoff (3.8 to 37 mg∙L⁻¹; Table 1). As expected, runoff from plots impacted by fan dust were appreciably greater than concentrations measured in runoff from pasture that had not been fertilized or grazed for at least 15 years (0.07 and 0.10 mg∙L⁻¹; Table 1).