Diurnal and seasonal variations of odor and gas emissions from a naturally ventilated free-stall dairy barn on the Canadian prairies

ABSTRACT

This study characterized the seasonal concentration (C) and emission (E) patterns of odor, ammonia (NH₃), and hydrogen sulfide (H₂S) over the course of a whole year and their diurnal patterns in cold, warm, and mild seasons for a naturally ventilated free-stall dairy barn. It was found that seasonal odor and NH₃ and H₂S emissions varied greatly: from 17.2 to 84.4 odor units (OU) sec⁻¹ AU⁻¹ (AU: animal unit, 500 kg of animal body mass), from 0.27 to 0.92 mg sec⁻¹ AU⁻¹, and from 3 to 105 μg sec⁻¹ AU⁻¹, respectively. The overall concentrations of odor and NH₃ were higher in the winter, whereas the emissions were higher in the mild and warm seasons. Diurnal variation was most significant for odor emission (OE) in the mild season when the ratio of maximum (279.2 OU sec⁻¹ AU⁻¹) to minimum value (60.5 OU sec⁻¹ AU⁻¹) was up to 4.6. The indoor air quality was also evaluated by considering not only the health effect of individual gases, but also the additive effect of NH₃ and H₂S. Results showed that the indoor air quality was poorest in cold seasons when NH₃ C could exceed the threshold limit set out in occupational health regulation, and in fact could worsen due to the additive effect of the two gases. Further, it was suggested NH₃ was a good indicator for predicting odor concentration (OC) or OE. The impact of climatic parameters on odor and gases were also examined, and it was found ventilation rate (VR) negatively affected OC and NH₃ C, but positively impacted OE and NH₃ E. Using 70% of the total data, a multilinear model for OE was developed as a function of VR and indoor relative humidity and was validated to be acceptable using the rest of the data.

Implications: Diurnal and seasonal variations of odor, NH₃, and H₂S concentrations and emissions were monitored for a naturally ventilated dairy barn in a cold region. The emission factors were calculated and indoor air quality was evaluated. The overall odor and NH₃ concentrations were higher in winter, whereas emissions were higher in the mild and warm seasons. Diurnal variation was most significant for odor emission in the mild season, when the ratio of maximum to minimum value was up to 4.6. The results can be used to estimate odor and gas emissions from other dairy barns in Canada and other cold regions.

Introduction

Intensive animal housing and feeding operations around the world have been rapidly developed and have raised public concern about their adverse impact on human health and the environment. Intensive animal production is associated with various air emissions, including NH₃ and H₂S, as well as being a general odor nuisance, leading to complaints of eye, nose, and throat irritation, headache, and drowsiness (Schiffman, 1998). At low concentrations, NH₃ and H₂S are both potential threats for respiratory irritation, neurological effects, and immunological effects, and can even result in death at very high concentrations (Copeland, 2014). Additionally, NH₃ not only contributes to eutrophication of surface water and nitrate contamination of groundwater, but also impairs atmospheric visibility through forming an aerosol (U.S. EPA, 2004).

Agriculture is the most important NH₃ emission source, with the majority of NH₃ emissions being attributed to livestock production (Webb et al., 2005). In Canada, animal agriculture contributed 64% of total NH₃ emissions as shown by a national inventory in 2002 (Carew, 2010). Several studies have been carried out to quantify air emissions from different animal sectors (Gay et al., 2003; Hayes et al., 2006); however, odor data from dairy operations are still limited (Zhu et al., 2000; Gay et al., 2003; Zhao et al., 2007; Akdeniz et al., 2012a; Mosquera et al., 2006; Rzeźnik et al., 2014). For dairy barns in Canada specifically, there are no published odor emission data, and NH₃ and H₂S emission data are also insufficient. It was indicated that odor and gas emissions varied by animal species, climate, region, ventilation rate, and so on (Gay et al., 2003; Ngwabie et al., 2009), and odor and gas concentrations and emissions from animal barns would also vary diurnally and seasonally (Sun et al., 2010; Wang, 2007). Therefore, long-term measurements for different animal species, different regions (climates), and different seasons are necessary to improve the air emission database involving livestock production. This first step is necessary for establishing appropriate control and mitigation strategies for odor and gas emissions. Diurnal and seasonal variations of gas emissions, including NH₃ from dairy barns, have been characterized (Saha et al., 2014); however, diurnal and seasonal odor emission patterns from dairy buildings have not been well studied. Additionally, given that livestock room air is composed of hundreds of components (Ni et al., 2012), the indoor air quality of dairy barns has not been thoroughly evaluated in research studies considering the combined health effect of the multiple components, as well as their possible diurnal and seasonal concentration variabilities.

As odor measuring is time-consuming and costly, there is often a desire to relate odor to an odorous component such as NH₃ or H₂S, components that are relatively easy to measure. Amon et al. (1997) reported a good relationship between OC and NH₃ C in a clinoptilolite-treated broiler room. Blanes-Vidal et al. (2012) found seasonal patterns of odor perception were associated with seasonal variations in NH₃ C in nonurban residential communities. In another study, Blanes-Vidal et al. (2009) observed OC was most strongly related to the sulfur-containing compounds, including H₂S, from agitated swine slurry. Fewer studies considered multiple factors for predicting livestock odor, such as various odorants, volatile organic compounds (Hobbs et al., 2000; Akdeniz et al., 2012b), or climatic parameters (Wang, 2007).

Hence, this study was conducted at a naturally ventilated free-stall dairy barn in a cold region climate (the Canadian Prairies), aiming to (1) reveal seasonal variations of odor, NH₃, and H₂S concentrations and emissions and their diurnal variations during the cold, mild, and warm seasons; (2) evaluate indoor air quality in different seasons; (3) investigate the correlations between odor and odorous gases (NH₃ and H₂S) and the impact of climatic parameters on odor and gases; and (4) develop a prediction model for odor emission (OE) and validate the model.

Material and methods

Description of the dairy barn

The dairy barn chosen for this study is a research dairy barn located in Saskatoon, Saskatchewan, Canada (106.6º W and 52.1º N). Most of the facility is a free-stall area where approximately 112 milking cows are housed. The dairy barn has a smaller average herd size compared to the provincial average (178 in Saskatchewan), but is a middle-scale barn across the whole country (Government of Canada, 2016). A photograph of the inside of the barn is provided in Figure 1. The dairy breed is Holstein, and the cows are routinely fed a mixed ration of barley silage and alfalfa hay, as well as concentrates with barley grain, canola meal, soybean meal, distiller’s dried grains, and also a mineral–vitamin supplement. The milk production is 38 L per cow daily. The floor area is 3, 230 m². On the south side, 4 pens of 12 cows each are housed and milked in the parlor, while on the north side there are 52 stalls where cows are milked in an automatic milking unit, or optionally in the parlor. The automatic manure scraper is set to clean the alleyways four times daily. The manure and all wash water are pumped to a covered slurry tank with a capacity of 2.52 million L outside the barn.

Figure 1. Inside view and brief schematic plan of the dairy barn.

In the mild and warm seasons, ventilation of the free-stall area is controlled by sliding window panels on the side walls, while in winter all the windows are closed and six chimney fans provide ventilation. On hot summer days, the end-wall door is also opened to increase ventilation. Three large-volume recirculation fans are used to keep the air temperature uniform, and a few radiant natural gas heaters are used to keep the temperatures above freezing in winter.

Measurement schedule

Diurnal measurements were performed for two days, respectively, in February (February 9 and 12), July (July 21 and 23), and October (October 13 and 15) of 2015, which represent the cold, warm, and mild seasons in Saskatoon. The two days were not consecutive due to the difficulty in setting up the odor laboratory sessions. Only daytime was considered to do measurements, when outdoor activities mainly occurred and air quality was a concern to the neighboring residents. The on-site gas sampling station was continuously measuring NH₃, H₂S, and carbon dioxide (CO₂) from 6:00 a.m. to 9:00 p.m. on each measurement day. Two air samples were collected simultaneously every 3 hr for odor concentration (OC) measurement (from 6:00 to 9:00 a.m., from 9:00 a.m. to 12:00 p.m., from 12:00 to 3:00 p.m., from 3:00 to 6:00 p.m., and from 6:00 to 9:00 p.m.).

The seasonal sampling and measurements were conducted on one picked day each month (when the weather was typical) from February 2015 to January 2016. On each measuring day, the concentrations of NH₃ and H₂S were continuously monitored for 2 hr in the early morning (6:00 to 8:00 a.m.), and for another 2 hr in the early evening (6:00 to 8:00 p.m.). During each morning or evening sampling period, two air samples were collected for OC measurement. When the diurnal measurements were performed on February 9, July 21, and October 13, gas concentrations during the same morning and evening periods and OC from 6:00 to 9:00 a.m. and from 6:00 to 9:00 p.m. were extracted to represent the monthly results.

Measurement methods

The room air was continuously drawn by an air pump from a fixed sampling point. A short piece of Teflon tubing was used to fix the sampling height at around 1.8 m above the center area (Figure 1). The air was then sent to the gas analyzers for concentration measurements. The equipment included a CO₂ sensor (K30 CO₂ sensor, CO₂ Meter, USA), an NH₃ sensor (C21 NH₃ transmitter, GFG Instrumentation, USA), and an H₂S analyzer (JEROME 631-X, Arizona Instrument Corporation, Arizona Instrument LLC, USA), which were all located on the overhead walkway, as shown in Figure 1. The measurement ranges and accuracies were 0–10,000 ppm and ±3% (±30 ppm) for the CO₂ sensor, 0–100 ppm and ±5% for the NH₃ sensor, and 0.003–50 ppm and ±0.003 ppm at 0.05 ppm and ±0.03 ppm at 0.50 ppm for the H₂S analyzer. Every 5 min, one measurement for each of the three gas concentrations would be recorded by a data logger (CR10X, Campbell Scientific Corporation, Canada). Air samples were collected using 10-L Tedlar air bags and were analyzed for OC in the Olfactometry Laboratory at the University of Saskatchewan. There were usually eight or at least six trained panelists participating in each odor session. The screening of panelists and measurements of OC were conducted in compliance with the CEN (2003) standard. Additionally, the barn’s indoor temperature (T_in) and relative humidity (RH_in) were also monitored continuously by two wireless T/RH data loggers (OM-EL-USB-2, Omega, Canada) with –35°C to 80°C and 0 to 100% for T and RH measurement ranges and 0.5°C and 3.5% for T and RH accuracies, and were recorded every 5 min. The ambient hourly temperature (T_out) and relative humidity (RH_out) were acquired from Environment Canada (the department of the government of Canada with responsibility for coordinating environmental policies and programs).

The maintenance and calibration of the olfactometer and gas analyzers were all performed according to their operational requirements.

Ventilation rate and emission rate calculation

Hourly ventilation rate (VR) was estimated using a CO₂ mass balance method, which is a widely used method for estimating VR for naturally ventilated dairy barns (Ngwabie et al., 2011; Wu et al., 2012). The equation used to calculate VR is based on per heat production unit (HPU) (1000 W total heat produced by the livestock), which is as follows (CIGR [International Commission of Agricultural Engineering], 2002):

(1)

(2)

(3)

where VR is ventilation rate (m³ hr⁻¹); (CO₂)_P is CO₂ production per HPU based on a 24-hr period (0.185 m³ hr⁻¹); (CO₂)_i is indoor CO₂ concentration and (CO₂)_o is outdoor CO₂ concentration in parts per million (ppm); a is a constant of 0.22 for dairy cows in free-stalls; h is the hours after midnight; h_min is hours after midnight with minimum cow activity, which is 2.9 (2:55 a.m.); Φ_tot is total heat production (W); m is body mass, which is 755 kg on average for the barn; Y₁ is milk production (kg day⁻¹); and P is days of pregnancy, the effect of which was neglected in this study. To modify Φ_tot per HPU outside the thermoneutral zone, the following equation was used for cattle (CIGR, 2002):

(4)

where T_in is indoor T (°C). The actual (CO₂)_o on the measurement days was measured in the mild and warm seasons. To reduce the possible pollution from the open windows, (CO₂)_o was measured 5 m away from the southeast side of the dairy barn for 30 min at the end of the day. In the cold season, (CO₂)_o was assumed to be 390 ppm (International Panel on Climate Change [IPCC], 2013) when the ambient measurement was not applicable.

Knowing the odor and gas concentrations and VR, the odor and gas emissions were calculated as follows:

(5)

where E is odor emission rate in odor units (OU) per second per animal unit (AU), odor units per second per square meter of the floor area (OU sec⁻¹ m⁻²), or odor units per second per cow (OU sec⁻¹ cow⁻¹). Gas emission E is in units of mg sec⁻¹ AU⁻¹, mg sec⁻¹ m⁻², or mg sec⁻¹ cow⁻¹; VR is ventilation rate of the room (m³ sec⁻¹); and ∆C is the difference of odor and gas concentrations between the room incoming air and exhaust air in units of OU m⁻³ or ppm. The concentrations of odor, NH₃, and H₂S of inlet air (ambient air) were negligible compared to the indoor concentrations and were treated as zero.

Gas emissions were calculated on an hourly basis for both diurnal and seasonal results. Therefore, each data point of gas concentrations and emissions in the figures that follow was the average of hourly results within the 3 hr, or the average of hourly results from the morning and evening periods. For odor emissions, 3-hr average of the diurnal results and 2-hr average from both morning and evening periods of the seasonal results were calculated for a basis.

Statistical data analysis

The statistical evaluation of data was performed by SPSS software 22. “Daily” effect combined the results of ambient weather, ventilation rate, animal management, and so on, while “diurnal” effect was a function of the time of day. There were two days chosen in each season with each having five diurnal levels: early morning (EM) from 6:00 to 9:00 a.m., late morning (LM) from 9:00 a.m. to 12:00 p.m., early afternoon (EA) from 12:00 to 3:00 p.m., late afternoon (LA) from 3:00 to 6:00 p.m., and early evening (EE) from 6:00 to 9:00 p.m. To examine the “diurnal” effect in each season, a General Linear Model (GLM) was performed where “daily” and “diurnal” were the two factors. The main effects of the two factors and the interaction between “daily” and “diurnal” were examined. If the interaction effect was significant, “diurnal” effect was then examined separately for each day. To do multiple comparisons, the Duncan test was selected in GLM, or a nonparametric test (Kruskal–Wallis one-way analysis of variance [ANOVA]) was selected when error variances of the data were unequal. For seasonal results, only the factor of “seasonal” was considered, and the same methods were used to examine the “seasonal” effect and perform multiple comparisons of the monthly results. The significance was indicated by p-value (0.05 level). A p-value less than or equal to 0.05 indicated a significant effect or variance. Multiple comparisons for seasonal OC and OE were not performed.

The relationships among odor, gases, and climatic parameters were indicated by Pearson correlation coefficients (r). The outcome of the results from SPSS showed two different significance levels (0.05 level and 0.01 level) for which p ≤ 0.05 indicated significant correlation and p ≤ 0.01 indicated very significant correlation. Using 70% of the data, randomly selected, the regression model for OE was derived as follows:

(6)

where Y is the dependent factor OE, A₀ is constant, A₁, A₂, A₃, …, A_p are coefficients, and X₁, X₂, X₃, …, X_p are independent factors including VR, T, and RH. As there may exist significant correlations among climatic parameters (e.g., VR and T_out), multicollinearity among the independent variables probably would occur when input all the factors to regress the model. To solve the problem, a stepwise regression method was used to remove some of the highly correlated factors. The paired-samples t-test was performed to compare the difference between modeled results and observed results.

Results and discussion

Diurnal odor and gas profiles

Climatic parameters

The T_out varied greatly among the three seasons. The T_in of this dairy barn was controlled within a narrow range of 5.1–5.9°C during the cold season, and was more affected by outdoor weather in the warm and mild seasons with ranges of 19.3–28.0°C and 8.6–16.0°C, respectively. Figure 2 gives the diurnal profile of VR. No significant daily difference in VR between the two days was observed in any season (p > 0.05). Diurnal VR was significantly lower in the EM than in the other periods in the cold season and was significantly higher in the afternoon during the mild season (p < 0.05), but showed no significant difference in the warm season (p > 0.05). There was a noticeable amount of difference for VR in the afternoon between the mild season and warm season, which was explained by the significantly positive correlation between VR and wind speed (r = 0.63, p < 0.01) along with higher wind speeds during afternoons of October. Wind direction also affected where the fresh air was coming from the windows and how it mixed with the room air. The orientation of the dairy barn is northeast to southwest. Wind in October came from the northwest-west (299º), which was more favorable for ventilation than the wind direction in July (from the south-southwest, 200º).

Figure 2. 3-Hour average VR, OC, and OE in the cold (a), warm (b), and mild (c) seasons (error bar was plotted by standard deviation).

Odor concentrations and emissions

Diurnal 3-hour average OC and OE are presented in Figure 2. The daily difference of OC or OE was not significant in any season (p > 0.05). The diurnal effect was significant for OC and OE in both cold and mild seasons (p < 0.05), which contradicts the conclusion of Zhu et al. (2000), who found OC and OE were both constant for a naturally ventilated dairy barn in October. This could be due to the smaller variations of VR in their study. During the cold season, diurnal OC varied differently on the two days, while OE showed highly consistent patterns on both days with being obviously low in the EM and high in the evening. No significant diurnal variance (p > 0.05) was found for OC or OE during the warm season when analyzing the results from the two days together; however, great diurnal variations of OC and OE were observed on July 21 when OC and OE were as low as 153 OU m⁻³ and 32.5 OU sec⁻¹ AU⁻¹, respectively, in the EM, and as high as 549 OU m⁻³ and 109 OU sec⁻¹ AU⁻¹, respectively, in the LA. A higher impact of VR on OE was observed during the mild season than in the cold and warm seasons, with the maximum and minimum ratio of OE being up to 4.6. Because of consistent VR variations, the OC and OE of the two days both presented highly consistent diurnal patterns with lower OC, but apparently higher OE during the entirety of the afternoon.

Ammonia concentrations and emissions

Diurnal NH₃ C and NH₃ emissions (NH₃ E) can be found in Figure 3. Zhao et al. (2007) observed a significant daily difference for NH₃ C from a naturally ventilated dairy barn in summer, which was observed for both NH₃ C and NH₃ E in the cold and warm seasons in this study. The T_in, T_out, and VR were all slightly lower, while the RH_in and RH_out were slightly higher on February 9 than on February 12. This together may result in significantly higher levels of NH₃ C and NH₃ E (p < 0.05). Slightly higher VR and thus lower NH₃ and NH₃ E on July 21 than on July 23 were observed as well. Zhang et al. (2005) and Saha et al. (2014) indicated considerable diurnal variations of NH₃ E for naturally ventilated dairy buildings, which was also found to be true in this study. In the cold season, there were significant differences in both diurnal NH₃ C and NH₃ E on February 9 (p < 0.05). Yang et al. (2016) observed larger diurnal variations of NH₃ E in the summer and fall at two dairy feedlots, and this pattern was confirmed in this study. In the warm season, NH₃ C was significantly lower in the EM on July 21 (p < 0.05), but showed no significant diurnal effect on July 23 (p > 0.05). Significant diurnal variances (p < 0.05) were found in NH₃ E for both days with lower values in the entire morning.

Figure 3. 3-Hour average NH₃ C and NH₃ E in the cold (a), warm (b), and mild (c) season.

In the mild season, very high similarities were observed from the two curves of both NH₃ C and NH₃ E, with the trend of diurnal NH₃ C being opposite to that of NH₃ E. Saha et al. (2014) observed the highest NH₃ E between 1:00 p.m. and 3:00 p.m. and the lowest between 4:00 a.m. and 6:00 a.m. Similarly, this study also found NH₃ E tended to be highest from 12:00 to 3:00 p.m. and lowest in the EM.

Hydrogen sulfide concentrations and emissions

Diurnal H₂S C and H₂S emissions (H₂S E) are plotted in Figure 4. Due to the malfunction of the H₂S analyzer, the data for H₂S C for the whole day of February 12 and from 6:00 to 9:00 a.m. on February 9 are missing. In the cold season, a significant diurnal effect was indicated by the obvious difference between H₂S C in the LM and in the EE (p < 0.05). In the warm season, significant variance (p < 0.05) was observed in diurnal H₂S C on July 21 with lower H₂S C in the EM and EE but this was not found on July 23. The diurnal variations of H₂S E showed similar patterns to diurnal VR on both days, with the maximum in the EA being significantly higher than the minimum in the EE (p < 0.05). In the mild season, H₂S C and H₂S E were very low (less than 0.03 ppm and 0.01 mg sec⁻¹ AU⁻¹, respectively) and showed highly consistent diurnal patterns for the two days. Significant diurnal variance was concluded from both H₂S C and H₂S E (p < 0.05), which simultaneously showed a peak in the LM and a low in the LA.

Figure 4. 3-Hour average H₂S C and H₂S E in the cold (a), warm (b), and mild (c) seasons.

Table 1 summarizes the minimum, maximum, and average of the diurnal results. It was found that the average OC in February (cold season) was obviously higher than in July (warm season) and October (mild season), while OE was the opposite. Due to the fluctuations of VR, seasonal variations of OE could be extremely large, with the average OE in October and July being 5.4 and 2.7 times that in February. Ammonia concentration was obviously high in winter and was low and similar in the warm and mild seasons. However, no such great difference was observed for NH₃ E (p > 0.05). Similar to OC and NH₃ C, H₂S C also showed obvious seasonality. The highest H₂S C occurred in the cold season and the lowest in the mild season. The seasonal variations of H₂S E were even greater than that of OE, with a ratio of maximum (in July) to minimum (in October) being up to 13.8. The possible great seasonal variations of odor and gas emissions suggest it is necessary to obtain detailed seasonal (such as monthly) concentration and emission profiles of odor and gases from naturally ventilated dairy barns.

Table 1. Descriptive statistics of 3-hour average odor, NH₃, and H₂S concentrations and emissions in the cold, warm, and mild seasons.

	Cold			Warm			Mild
Items	Minimum	Maximum	Mean (SD)	Minimum	Maximum	Mean (SD)	Minimum	Maximum	Mean (SD)
VR (m^3 sec^−1)	5.0	8.5	6.7 (1.1)^b	19.1	33.5	27.9 (5.3)^a	24.2	135.5	60.3 (45.8)^a
OC (OU m^−3)	381	897	630 (153)^a	153	549	370 (125)^b	311	483	403 (64)^b
OE (OU sec^−1 AU^−1)	11.7	35.0	24.3 (6.9)^b	32.5	109	66.2 (25.7)^ab	60.5	279.2	130.1 (86)^a
OE (OU sec^−1 m^−2)	0.6	1.9	1.3 (0.4)	1.6	5.2	3.2 (1.2)	3.2	14.8	6.9 (4.5)
OE (OU sec^−1 cow^−1)	17.6	52.8	36.7 (10.5)	49	165	100 (38.8)	91.4	421.5	196.5 (130)
NH3 C (ppm)	16	31	23 (6)^a	2	8	5 (2)^b	1	6	4 (2)^b
NH3 E (mg sec^−1 AU^−1)	0.4	0.9	0.6 (0.2)^a	0.3	0.8	0.6 (0.2)^a	0.5	0.9	0.6 (0.1)^a
NH3 E (μg sec^−1 m^−2)	21.7	48.3	35.1 (8.2)	12.6	38.9	27.9 (7.7)	24.1	47.8	33.1 (7.6)
NH3 E (mg sec^−1 cow^−1)	0.60	1.34	0.98 (0.23)	0.40	1.23	0.88 (0.25)	0.69	1.37	0.95 (0.22)
H2S C (ppb)	186	275	217 (42)^a	211	193	176 (29)^b	0	28	13 (9)^c
H2S E (μg sec^−1 AU^−1)	9.4	15.1	12.2 (2.4)^b	35.2	68.3	48.2 (11.4)^a	0.1	7	3.5 (2.1)^b
H2S E (μg sec^−1 m^−2)	0.5	0.8	0.7 (0.1)	1.7	3.3	2.3 (0.5)	0.01	0.4	0.2 (0.1)
H2S E (μg sec^−1 cow^−1)	14.2	22.9	18.5 (3.6)	53.2	103	72.8 (17.3)	0.2	10.6	5.3 (3.2)

Note. C is concentration and E is emission; SD is standard deviation. The letters a, b, and c are used to indicate the significance of the difference in the measured items among the three seasons at the 0.05 level; the same letters show the difference is not significant. The minimum, maximum, and mean were calculated from the 3-hr diurnal averages of the two days in each season.

Seasonal odor and gas profiles

Climatic parameters

From February 2015 to January 2016, the number of cows, T_in, RH_in, and RH_out changed from 102 to 116, from 6°C to 24°C, from 49% to 91%, and from 35% to 88%, respectively. The T_out and VR are plotted in Figure 5. It shows that seasonal VR ranged from 5.0 to 69.2 m³ sec⁻¹ with significant variances (p < 0.05). The overall VR in winter was obviously lower than during the mild and warm seasons, excluding March, when the windows were occasionally partially opened. In April, VR presented great diurnal variance, which was because the slightly higher wind speed (10 km hr⁻¹) and more favorable wind direction (317º) in the early evening resulted in much higher VR than in the morning (7 km hr⁻¹ and 290º). It was also found that VR in August and September was significantly lower than in the other warm months (p < 0.05), which may be attributed to the relatively lower wind speed in August and lower T_out and T_in in September.

Figure 5. Seasonal VR and concentrations and emissions of odor, NH₃, and H₂S.

Odor concentrations and emissions

Figure 5 also gives seasonal OC and OE variations. The average OC ranged from 203 to 639 OU m⁻³, which was within the ranges reported by Rzeźnik et al. (2014) and Mosquera et al. (2006), but higher than the 103–312 OU m⁻³ reported by Akdeniz et al. (2012a) for dairy barns with mechanical ventilation. The average OC of March, June, and August was 374 OU m⁻³ and was higher than the 100 OU m⁻³ measured by Zhao et al. (2007). This can be attributed to the fact that the dairy barns Zhao et al. (2007) studied had additional wide-open ridge and cooling fans that were likely to provide more fresh air and further decrease odor and gas concentrations inside. Zhao et al. (2007) found no significant seasonal variations for OC, while the overall OC in the cold season, including November to March, was 37% higher than that in the mild and warm seasons from April to October. No obvious difference was observed between the results of the mild and warm seasons, which is consistent with the previous discussion that no significant difference existed between OC in July and October (p > 0.05).

As shown in Figure 5, OE tended to be low and stable in the cold season (excluding March), but fluctuated greatly in the mild and warm seasons. Greater variations were presented in the seasonal OE than for OC. The highest OE was estimated to be 84.4 OU sec⁻¹ AU⁻¹, which was almost 5 times the minimum OE at 17.2 OU sec⁻¹ AU⁻¹. When comparing OE on a per floor area basis, the average OE was 2.38 OU sec⁻¹ m⁻², which was close to 3.30 OU sec⁻¹ m⁻² reported by Maasikmets et al. (2015) for a naturally ventilated dairy barn with loose housing system in Estonia, and was higher than the results of Zhu et al. (2000) due to their high VR, and thus very low OC, but was within the range summarized by Gay et al. (2003) for 13 dairy barns in Minnesota. The average OE on a per cow basis was 69.3 OU sec⁻¹ cow⁻¹, which was comparable to the results of Mosquera et al. (2006). Seasonal odor and gas concentrations and emissions are summarized in Table 2.

Table 2. Descriptive statistics of seasonal odor, NH₃, and H₂S concentrations and emissions.

Items	Minimum	Maximum	Mean ± SD
OC (OU m^−3)	203	639	437 ± 134
OE (OU sec^−1 AU^−1)	17.2	84.4	45.9 ± 24.2
OE (OU sec^−1 cow^−1)	26	127.5	69.3 ± 36.6
OE (OU sec^−1 m^−2)	0.89	4.22	2.38 ± 1.24
NH3 C (ppm)	2	29	9 ± 7.3
NH3 E (mg sec^−1 AU^−1)	0.27	0.92	0.53 ± 0.18
NH3 E (mg sec^−1 cow^−1)	0.41	1.39	0.80 ± 0.27
NH3 E (μg sec^−1 m^−2)	14	50	28 ± 10
H2S C (ppm)	0.01	0.20	0.13 ± 0.06
H2S E (μg sec^−1 AU^−1)	3	105	28 ± 32
H2S E (μg sec^−1 cow^−1)	4.7	157.8	41.5 ± 48.2
H2S E (μg sec^−1 m^−2)	0.16	5.23	1.42 ± 1.63

Note. C, E, and SD are described in Table 1. The minimum, maximum, and mean were calculated from the 12 daily averages for the 12 months.

Ammonia concentrations and emissions

Significant variances (p < 0.05) existed in the seasonal NH₃ C as plotted in Figure 5. Similar to OC, NH₃ C was significantly higher in the cold season excluding August, when NH₃ C (11 ppm) was significantly higher than in the other warm and mild months. In addition to the relatively lower VR in August compared to the overall VR over the summer, it was found that both RH_out and RH_in were higher, and increasing RH_in and RH_out proved to be associated with increasing NH₃ C by their positive correlations (p < 0.01). From November 2015 to January 2016 when the windows were closed, NH₃ C increased to above 11 ppm, but was considerably lower than the previous winter (February 2015), which may be explained by a relatively warmer winter when the windows were partially opened at times. The average NH₃ C was 9.1 ppm in this study, which was higher than the 2.1 ppm from Samer et al. (2012) and 6.6 ppm from Ngwabie et al. (2009); their studies had higher ventilation. On the contrary, Arcidiacono et al. (2015) reported much higher NH₃ C (averaged 16.6 ppm) for naturally ventilated free-stall dairy buildings in Italy, which may be explained by their lower manure-removing frequency (one or two times daily), as well as the low sampling height of 10 cm above the floor. Ngwabie et al. (2009) found NH₃ C presented little variation during the winter months, which was also observed in this study from November 2015 to January 2016. Additionally, it was reported NH₃ C in May was 60% of the winter values (Ngwabie et al., 2009), while in this study the average NH₃ C over the summer only amounted to 39% of the winter results. The greater difference of seasonal NH₃ C in this study can probably be attributed to greater outdoor weather differences between Canadian Prairies summers and winters. The release of NH₃ is also directly related to the conversion of feed nitrogen to animal product (Maasikmets et al., 2015); however, no discussion could be carried out since the feed composition for this dairy barn was not analyzed.

As reported by Amon et al. (2001) and Yang et al. (2016), NH₃ E from dairy barns varied over the course of the year, which was also observed in this study when seasonal NH₃ E changed from 0.27 to 0.92 mg sec⁻¹ AU⁻¹. This finding is different from the conclusion of diurnally measured results (the variations of NH₃ E among February, July, and October were not obvious). Seasonal NH₃ E averaged 0.53 mg sec⁻¹ AU⁻¹, which is higher than 0.21 mg sec⁻¹ AU⁻¹ from the study of Maasikmets et al. (2015). Wu et al. (2012) found that NH₃ E varied from 32 to 77 g day⁻¹ HPU⁻¹ for a naturally ventilated free-stall dairy building in Jutland. Using the same unit, seasonal NH₃ E was in the range of 20–69 g day⁻¹ HPU⁻¹, which is comparable to the result of Wu et al. (2012). If one only compared the summer results, the average NH₃ E would be 0.53 mg sec⁻¹ AU⁻¹, which is lower than the 2.33 mg sec⁻¹ AU⁻¹ reported by Samer et al. (2012), but is close to the lower end of the range of 0.56 to 1.11 mg sec⁻¹ AU⁻¹ indicated by Fiedler and Müller (2011) (for two naturally ventilated dairy barns in Germany with more openings) and in the range of 0.36–0.78 mg sec⁻¹ AU⁻¹ reported by Schrade et al. (2012) (for a naturally ventilated dairy barn in Switzerland). In winter, NH₃ E averaged 0.44 mg sec⁻¹ AU⁻¹ and is lower than the 1.50 and 0.99 mg sec⁻¹ AU⁻¹ found by Samer et al. (2012) and Snell et al. (2003). When calculated by floor area, NH₃ E averaged 28 μg sec⁻¹ m⁻², which is lower than the 43 μg sec⁻¹ m⁻² that was discussed by Gay et al. (2003). Besides the ventilation rate, other factors could also be the reasons for the different NH₃ E found in previous studies, including floor type, feeding routine, management, and so on.

Hydrogen sulfide concentrations and emissions

As can be seen from Figure 5, H₂S C fluctuated drastically (p < 0.05) over the year (from 0.01 to 0.20 ppm), and is higher than the results found by Zhao et al. (2007) (from 2 to 31 ppb). Different from OC and NH₃ C, H₂S C tended to be high in both cold and warm seasons (excluding September), but low in the mild season, which confirmed the conclusion from diurnal results. As for H₂S E, it was significantly higher from April to July (p < 0.05), but remained below 20 μg sec⁻¹ AU⁻¹ during the other months. The range was from 3 to 105 μg sec⁻¹ AU⁻¹. The average H₂S E was 1.42 μg sec⁻¹ m^−2, which is higher than 0.29 μg sec⁻¹ m⁻² found by Maasikmets et al. (2015), but within the range of 1.04 to 2.89 μg sec⁻¹ m⁻² as reported by Gay et al. (2003).

Indoor air quality evaluation

As most Canadian provinces’ occupational health and safety regulations use the indoor threshold limits of NH₃ and H₂S recommended by the U.S. Occupational Safety and Health Administration (OSHA), the indoor air quality in the study barn was evaluated against OSHA threshold limits. It regulates indoor threshold limit values (TLV) of NH₃ C and H₂S C: They can be no more than 25 ppm and 1 ppm, respectively, for an 8-hr time-weighted average (T-TLV), and 35 ppm and 5 ppm, respectively, for short-term exposure limits (S-TLV) of less than 15 min (American Conference of Government Industrial Hygienists [ACGIH], 2010). The U.S. OSHA also suggests the effect of the mixtures should be considered as additive where the health effect and target organ or system are the same when none of the components have a value exceeding the TLV (ACGIH, 2010). That is, if the sum of

(7)

exceeds unity, the threshold limit of the mixture should be considered exceeded (where C₁ is the observed concentration and T₁ is the corresponding threshold limit). Since NH₃ and H₂S both cause upper respiratory tract irritation, their combined effect was examined in addition to their individual impact and was described by an indicator as given in Table 3.

Table 3. Indicator (the ratio of 8-hr average concentration to T-TLV) of individual gas concentration levels and their combined indicator.

	Cold		Warm		Mild
	February 9	February 12	July 21	July 23	October 13	October 15
NH3 C	1.09	0.68	0.16	0.21	0.11	0.11
H2S C	0.22	0.16	0.18	0.21	0.01	0.01
Combined	1.31	0.84	0.34	0.42	0.12	0.12

Note. T-TLV is time-weighted average (8-hr) threshold limit value and C is concentration.

The concentrations of NH₃ and H₂S remained below the S-TLV in all seasons. From 9:00 a.m. to 5:00 p.m. in the warm and mild seasons with good ventilation, the T-TLV was not exceeded for NH₃ or H₂S, and the additive indicator was also below 1. However, the February results showed the T-TLV for NH₃ C could be exceeded in the cold season. The NH₃ C was below the T-TLV on February12, but exceeded the T-TLV on February 9, which suggests respiratory threat when an occupant is exposed to the room air for 8 hr or longer. The combined indicator rose from 0.68 to 0.84 on February 12 if using the average H₂S C of 155 ppb during the winter (November to March) to fill in the missing data, and rose from 1.09 to 1.31 on February 9, which has a substantial impact on indoor air quality as opposed to only considering the individual components.

Overall, the indoor air quality reduced from November to February when VR was small, which suggests the indoor air quality was poor during the cold season and sometimes could exceed TWA and cause respiratory irritation. Although for this dairy barn, it’s not usual for workers to stay in the barn for a continuous 8 hr, the results of the potential health risk in the cold season could still provide reference for the building design and operation plan of the other dairy barns; for example, for dairy barns where long-time stay inside the barn is a possibility, it might be necessary to install an NH₃ sensor to continuously monitor the indoor concentration during the cold season and increase ventilation when necessary. Besides, indoor air quality could differ greatly among the winter or summer months, which is explained by the great variations of NH₃ C. The best indoor air quality was observed during the mild season, followed by the warm season; however, it should be noted that higher odor and H₂S emissions occurred during the warm and mild seasons, which would have more outdoor impact.

Impact of climatic parameters

A significant effect was observed for all five climatic parameters on OC (p < 0.05), with a negative effect of T_in, T_out, and VR, and a positive effect of RH_in and RH_out. Excluding T_in, the other four climatic parameters all had significant impact on OE (p < 0.01), but with an effect opposite to that on OC, suggesting the change of climatic parameters would increase one, but decrease the other.

The most related climatic parameter to OC and OE was T_out (r = –0.55, p < 0.01) and VR (r = 0.92, p < 0.01), respectively. Using curve estimation in SPSS, the best single linear regression model for OE was developed in cubic function with VR being the independent factor: OE (OU sec⁻¹ AU⁻¹) = –0.10 + 3.25 VR – 0.03 VR² + 1.10 × 10⁻⁴ × VR³ (R² = 0.89), where VR is in m³ sec⁻¹, or OE (OU sec⁻¹ AU⁻¹) = 1.51 + 519 VR – 813 VR² + 493 VR³ (R² = 0.88), where VR is in m³ sec⁻¹ AU⁻¹.

Climatic parameters proved to have significant impact on gases as well. Amon et al. (2001), Zhang et al. (2005), and Ngwabie et al. (2011) suggested NH₃ E correlated with T_in, and Li et al.

(2014) revealed NH₃ E was strongly related to T_out and RH_in. Similar to our findings for OC, T_in, T_out, and VR all had significantly negative effects on NH₃ C, and RH_in and RH_out both had positive impacts on NH₃ C (p < 0.01), whereas NH₃ E was only positively correlated with VR (p < 0.01) and negatively related to RH_out (p < 0.05). The only parameter that was correlated with H₂S C was VR with a negative relationship (p < 0.05). For H₂S E, a positive correlation between H₂S E and T and a negative relationship between H₂S E and RH were found, while no relationship between H₂S E and VR was observed. Moreover, significant correlations were also revealed among those climatic parameters, (e.g., T_in versus T_out (r = 0.94, p < 0.01), RH_in versus RH_out (r = 0.79, p < 0.01)), which strongly suggested the indoor thermal environment of the dairy barn was highly dependent on the outdoor weather. Also, we tried to develop a prediction model for VR based on the independent factors of T_in, T_out, RH_in, RH_out, difference between T_in and T_out (T_in – T_out), and difference between RH_in and RH_out (RH_in – RH_out). The best prediction model for VR was an exponential function based on the difference between T_in and T_out: VR (m³ sec⁻¹) = 36.43 × e ^{−0.084 (Tin – Tout)} (R² = 0.65), or VR (m³ sec⁻¹ AU⁻¹) = 0.22 × e ^{−0.087 (Tin – Tout)} (R² = 0.67), where T_in and T_out are in °C.

Relationship between odor and odorous gases

In line with previous studies (Blanes-Vidal et al., 2012; Akdeniz et al., 2012b), a significantly positive correlation was found between OC and NH₃ C (OC, OU m⁻³ = 141.3 ln(NH₃ C, ppm) + 179.5, R² = 0.51). On the contrary, no significant relationship was revealed between OC and H₂S C (p > 0.05). Gay et al. (2003) reported a moderate correlation between OE and total reduced sulfur emissions (r = 0.51, p < 0.05) and between OE and NH₃ E (r = 0.48, p < 0.05) for dairy housing facilities. In this study, it was found that OE was negatively related to NH₃ E (r = 0.46, p < 0.01) but not to H₂S E (p > 0.05). The results confirmed that NH₃ C plays a vital role in predicting OC or OE; however, H₂S cannot be an indicator in the absence of other factors.

Multilinear regression model for OE and validation

The multilinear model for OE was regressed as follows:

(8)

or

(9)

The modeled OE (OE_m) from eq 8 and the corresponding observed OE (OE_o) are plotted in Figure 6a. The remaining 30% of the data was used for validation. The comparison of predicted OE (OE_p) from eq 8 and OE_o is presented by Figure 6b. No significant difference in the average results was observed for either group of comparisons (p > 0.05).

Figure 6. Comparison of modeled results and observed results: (a) comparison of OE_m and OE_o using 70% of the data, and (b) validation using the remaining 30% of the data.

We also used fractional bias (FB) and Pearson correlation, which are two of the general performance measures for paired statistical comparison (ASTM, 2014). Significant correlations between the paired results were found for both groups of comparisons (r > 0.88, p < 0.01). The FB was 0.03 for the comparison of OE_m and OE_o, and –0.03 for the comparison of OE_p and OE_o, both of which fell into the range of –0.25 < FB < 0.25 (FB = 0 indicates an ideal model) suggested by ASTM (2014). The preceding results indicate the source odor emission model performs satisfactorily.

Conclusion

Diurnal and seasonal variations of odor, NH₃, and H₂S concentrations were measured for a naturally ventilated free-stall dairy barn in a cold region climate (the Canadian Prairies). With this knowledge, we evaluated indoor air quality and acquired diurnal and seasonal variations of odor, NH₃, and H₂S emissions, which could be further employed to study the outdoor impact of odor and gases by dispersion modeling and provide relative references for policymakers. Our findings show:

1. Great diurnal variations of odor and gas concentrations and emissions were observed in all seasons, but especially in the mild season. Overall, higher OC and NH₃ C were likely to occur in the early evening from 6:00 to 9:00 p.m., while higher OE and NH₃ E were likely to occur in the afternoon when VR was high. Odor, NH₃, and H₂S concentrations and emissions also showed great seasonal variations, with relatively higher odor and NH₃ concentrations in the winter, and higher emissions during the warm and mild seasons.

2. Considering the additive health effect of NH₃ and H₂S made a big difference for describing the indoor air quality than only considering individual gases. Based on the health effect of respiratory irritation, indoor air quality was poor and could exceed the exposure limit during the cold season.

3. Positive relationships were revealed between OC and NH₃ C and between OE and NH₃ E, which were in agreement with previous studies. Conversely, no relationship between odor and H₂S was found. Climatic parameters presented significant, but opposite, effects on OC and OE. The climatic parameters most relevant to OC and OE were T_out with negative influence and VR with positive influence. The prediction model of OE was developed as a multilinear function of VR and RH_in, which explained 87% of the data and was validated to be acceptable. Additionally, significant effects of climatic parameters on gas concentrations and emissions were also examined.

Acknowledgment

Special thanks are given to Zhu Gao, Jingjing Han, Shamim Ahamed, and Ali Motalebi for their assistance in field sampling. The authors also appreciate the great help from the workers of the dairy barn.

Funding

The authors thank the Natural Sciences and Engineering Research Council of Canada (NSERC) and the China Scholarship Council (CSC) for their financial support.

Additional information

Funding

The authors thank the Natural Sciences and Engineering Research Council of Canada (NSERC) and the China Scholarship Council (CSC) for their financial support.

Notes on contributors

Dandan Huang

Dandan Huang is a Ph.D. student in the Department of Biological Engineering, University of Saskatchewan, Saskatoon, Canada.

Huiqing Guo

Huiqing Guo is a professor in the Department of Mechanical Engineering, University of Saskatchewan, Saskatoon, Canada.