SENSITIVITIES OF FOUR AIR DISPERSION MODELS TO CLIMATIC PARAMETERS FOR SWINE ODOR DISPERSION

Transcription

1 SENSITIVITIES OF FOUR AIR DISPERSION MODELS TO CLIMATIC PARAMETERS FOR SWINE ODOR DISPERSION Y. Xing, H. Guo, J. Feddes, Z. Yu, S. Shewchuck, B. Predicala ABSTRACT. Using air dispersion models to predict downwind livestock odor concentrations in order to establish science-based setback distances has the potential to become a common practice for regulatory agencies. In this study, four air dispersion models, ISCST3,,, and, were used to simulate odor dispersion from a swine farm. The sensitivity analyses were conducted to develop an understanding of how model climatic parameters affect downwind odor concentrations and travel distances. Under steady-state weather conditions, the results indicated that the odor dispersion was mainly affected by atmospheric stability, wind speed, wind direction, and air temperature. Odor transport was favored by stable atmospheric conditions, low wind speed, and high ambient temperature. The four models predictions for odor concentrations varied greatly within 1 km from the source; beyond that, ISCST3 and gave similar results (within 17%), while s and s predictions were much lower than those of ISCST3 by up to 5.3%. However, when hourly averaged annual meteorological data were used, predicted the highest odor concentrations (up to 71.% higher than those of ISCST3), and s predictions were also higher than those of ISCST3, which were very different from the results obtained under steady-state weather conditions. s predictions were the lowest (up to.5% lower than those of ISCST3). When determining setback distances, it is recommended that if steady-state weather data are used, the odor concentration criterion allowed should be higher than the criterion allowed when using hourly annual or monthly weather data for the same neighboring land use. Keywords. Air dispersion model, Climatic parameter, Odor, Sensitivity analysis, Steady-state weather, Swine, Variable weather. Odor nuisance complaints against animal production farms have been increasing rapidly in the last decade and are becoming one of the major barriers for further development of the livestock industry (Guo et al., 2). Solutions for the problem frequently involve the specification of setback distances from neighboring properties; however, most of the existing setback guidelines were experience-based (Guo et al., 2). The establishment of science-based setback distances requires an accurate understanding of many factors, such as source odor emission rates, topography, and weather conditions including atmospheric stability classes, wind speed and direction, solar radiation, relative humidity, and air mixing height (Guo et al., Submitted for review in June 26 as manuscript number SE 6536; approved for publication by the Structures & Environment Division of ASABE in February 27. The authors are Yanan Xing, ASABE Member Engineer, Graduate Student, and Huiqing Guo, ASABE Member Engineer, Assistant Professor, Department of Agricultural and Bioresource Engineering, University of Saskatchewan, Saskatoon, Saskatchewan; John Feddes, ASABE Fellow Engineer, Professor, Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, Alberta; Zimu Yu, ASABE Member Engineer, Graduate Student, Environmental Engineering Division, University of Saskatchewan, Saskatoon, Saskatchewan; Stan Shewchuck, Research Scientist, Saskatchewan Research Council, Saskatoon, Saskatchewan; and Bernardo Predicala, ASABE Member Engineer, Adjunct Professor, Prairie Swine Centre, Inc., Saskatoon, Saskatchewan, Canada. Corresponding author: Huiqing Guo, Department of Agricultural and Bioresource Engineering, 57 Campus Dr., Engineering Bldg., University of Saskatchewan, Saskatoon, SK, Canada S7N 5A9; phone: ; fax: ; huiqing.guo@usask.ca. 23). Using air dispersion models to predict downwind livestock odor concentration in order to establish science-based setback distances has the potential to become a common practice for regulatory agencies. Although most commercially available air dispersion models were originally designed for industrial sources, they have been used in predicting odor concentrations downwind from agricultural sources (Zhu et al., 2; Guo et al., 21; Sheridan et al., 2; Jacques Whitford Environment, 23). Changes in the climatic parameters will affect the odor dispersion predicted by the models. The analysis of a model s sensitivity to these climatic parameters can identify the dominant parameters and their degrees of impact on downwind odor concentration and setback requirements (Smith, 1993). Smith (1993) evaluated the sensitivity of the STINK model to odor emission rate and climatic parameters for predicting odor concentration. He found that wind speed and odor emission rate were the most important factors. Atmospheric stability and surface roughness height were shown to be the next most important parameters. Wind direction in relation to the source orientation was only of moderate importance. Chastain and Wolak (1999) used a Windows-based computer program, based on Gaussian plume equations, to conduct a similar sensitivity analysis for livestock odor dispersion. The results indicated that the longest odor plume lengths occurred during stable atmospheric conditions. Neutral conditions during the day presented the next most critical period for odor dispersion. The extra vertical mixing provided by a significant increase in wind speed or the roughness associated with a forest barrier greatly reduced the odor travel distance. Transactions of the ASABE Vol. 5(3): American Society of Agricultural and Biological Engineers ISSN

2 Jacques Whitford Environment (23) used the CAL- PUFF model to conduct sensitivity analyses to develop odor dispersion factors, including climate factors, to be used in formulae for calculating minimum separation distances in Alberta, Canada. It was found impossible to develop generic values for micro-climate factors. Although individual models have been evaluated for their predictions of odor dispersion from livestock production sites, very limited work has been done to compare various air dispersion models and identify their differences in odor dispersion predictions, which is the information needed to help us understand the model differences and choose suitable models for odor dispersion applications. Zhou et al. (25) calibrated four air dispersion models, ISCST3,,, and WindTrax, using odor plume data measured within 1 km from two swine farms. They concluded that these four models performed similarly and predicted downwind odor concentrations that showed good agreement with field measured results. Considering that 58.3% of the measured odor concentrations were zero, this conclusion may need further examination. The model was compared with ISCST3 model under steady-state and variable weather conditions (U.S. EPA, 1998). The results showed that even though CAL- PUFF can be made to produce the same concentration in a steady-state environment, variable meteorological conditions can produce predictions higher than those of ISCST3. An inter-comparison of the AERMOD, ADMS, and ISCST3 models was done to assess the AERMOD model for regulatory purposes in the U.K. and its performance in relation to the other advanced dispersion models (Hall et al., 23). The comparison used four single representative boundary layer conditions, neutral (high and low wind speed), stable, and unstable boundary layers, taken from hourly meteorological data from a single year. AERMOD and ADMS generally showed a greater sensitivity to changes in atmospheric conditions than ISCST3, and the maximum concentrations and their distances from the source predicted by the different models were significantly different. In general, ISCST3 s predictions were the highest, while those of AERMOD were the lowest. It was difficult to see any consistent patterns in the differences between the models because these models reacted to a multiplicity of input parameters in complex ways that were hard to distinguish. The,, and TAPM models were evaluated against annual dispersion data sets at Anglesea and Kwinana, Australia (Hurley et al., 25). performed adequately for Anglesea, but performed poorly for Kwinana; performed marginally for Anglesea, with results worse than, and performed marginally (although better than ) for Kwinana, with overpredicting extreme concentrations. TAPM performed well for both the Anglesea and Kwinana annual data sets, and outperformed both and CAL- PUFF. The above studies indicate that these models performed differently under different simulation conditions for industrial air pollutant dispersion. In summary, very limited research has been conducted on comparing various industrial air dispersion models predictions for agricultural odors and the sensitivities of these models to various climatic parameters. OBJECTIVES The objectives of this study were to (1) conduct sensitivity analyses for four commonly used air dispersion models, i.e., ISCST3,,, and, for odor dispersion from a swine farm as affected by the primary climatic parameters, and (2) compare the predictions of these models under various climatic conditions. MATERIALS AND METHODS ISCST3 MODEL The ISCST3 (Industrial Source Complex) model was designed to support the U.S. EPA s regulatory modeling programs and is widely used in North America and worldwide (U.S. EPA, 1995a). It is a steady-state Gaussian plume dispersion model. The model can handle multiple sources, including point, volume, area, line, and open pit sources. Source emission rates can be treated as constant throughout the modeling period, or may be varied by month, season, hour of a day, or other optional periods. The user can specify multiple receptor networks in a single run, and may also mix Cartesian grid receptor networks and polar grid receptor networks in the same run. ISCST3 runs with a sequence of hourly averaged meteorological conditions to predict hourly average concentrations at receptors. Topography can be taken into account only when the deposition is considered. MODEL The model was developed by the Australian Environmental Protection Authority and it is an extension of the ISCST3 model (Australia EPA, 2). It is designed to predict ground-level concentrations or dry deposition of pollutants emitted from one or more sources, which may be stacks, area sources, volume sources, or any combination of these. Up to gridded receptors can be handled in one run. The discrete receptors can be run with gridded receptors at the same time. allows the calculation of average concentrations using a minimum average time of 3 min, even though the meteorological data is hourly. The topography may be accounted for by adjusting ground roughness height (Smith, 1972). The output data file contains gridded concentration fields for each source group and for each averaging period. MODEL The air dispersion model was a U.S. EPA regulatory model based on a Lagrangian puff model designed to simulate continuous puffs of pollutants being emitted from a source into the ambient wind flow (U.S. EPA, 1995b, 1998). It consists of three sub-systems: CALMET,, and CALPOST. CALMET is a meteorological model that combines meteorological data and geophysical data to generate a wind field. The model then combines the information provided by CALMET and source data to predict concentration, deposition flux, visibility impairment, etc., at each receptor for a specified average time. CALPOST is a post-processor for the model. can accommodate point, volume, and area source emissions. Up to 5 5 gridded receptors and discrete receptors can be handled during one run time. CAL- PUFF can use the three-dimensional meteorological fields developed by the CALMET model or the meteorological 18 TRANSACTIONS OF THE ASABE

3 Figure 1. Layout of the swine farm. files used by ISCST3. contains algorithms for near-source effects such as building downwash, transitional plume rise, partial plume penetration, and subgrid-scale terrain interactions, as well as long-range effects such as pollutant removal, chemical transformation, vertical wind shear, over-water transport, and coastal interaction effects. Most of the algorithms contain options to treat the physical processes at different levels of detail depending on the model application. Topography can be incorporated into the simulation. MODEL, a Gaussian integrated puff model, was developed by the U.S. EPA and marketed by Bee-Line Software Company (Asheville, N.C.). The Gaussian puff diffusion method is used to compute the contribution to the concentration at each receptor from each puff during every time step. It can simulate dispersion of airborne pollutants from semiinstantaneous or continuous point sources. There is no treatment of area or volume sources. It may deal with non-reactive pollutants, deposition, and sedimentation. One-hundred gridded receptors can be used, and it can use up to 1999 discrete receptors. A maximum of 1 separate meteorological periods of the same length may be used during each run. It can deal with different meteorological time intervals with a minimum of 1 s instead of the 1 h required by the other models. This makes it suitable for simulating odors measured by field odor assessors. This model includes some consideration of terrain effects through the wind field, but there is no explicit treatment of complex terrain. ODOR SOURCE (SWINE FARM) The swine farm was located in Yorkton, Saskatchewan, Canada (fig. 1). It consisted of one barn with ten rooms for 11,55 feeder pigs and an uncovered two-cell earthen manure storage basin. The barn was mechanically ventilated with wall-mounted fans. The shallow manure pits underneath the fully slatted floor were gravity-drained once every 2 to weeks. The odor emission rates measured by Guo et al. (25) were used by this study (table 1). The study area was a rural crop field with flat terrain and no obstacles. MODEL CLIMATIC SENSITIVITY ANALYSIS Odor concentration is measured by an olfactometer and a panel of panelists, and is based on the odor detection threshold (i.e., the dilution ratio of the odorous air sample when 5% of panelists have correctly identified the odorous sample from the odor-free samples) using odor units (OU) (CEN, 1999). All air dispersion models use mass concentrations in units of g m 3 and mass emission rates in g s 1. To be able to quantify odor concentration and emission rate with meaningful units and enable application of air dispersion models for odor modeling, odor researchers commonly take the odor concentration unit as OU m 3, which is considered equivalent to mass concentration in g m 3, and the odor emission rate as OU s 1, which is equivalent to the mass emission rate in g s 1 from a source (Zhu et al., 2; Guo et al., 21; Zhou et al., 25). These units will also be used in this study. Two types of meteorological conditions were considered. First, steady-state meteorological conditions were used to evaluate the sensitivities of the models as affected by each meteorological parameter and to compare the prediction differences of the four models without the bias of a varying meteorological regime (U.S. EPA, 1998). The influence of mixing height, ambient temperature, atmospheric stability, wind speed, and wind direction on the models predictions under steady-state weather conditions were evaluated. Under various steady-state weather conditions, the odor transport Table 1. Odor emission rates from the barn and manure storages. Source Total Emission Rate (OU s 1 ) Odor Emission Rate (OU m -2 s 1 ) Barn 37,928.9 Cell 1 27, Cell 2 325, Vol. 5(3):

4 distances predicted by the four models were compared using the critical detection distance (CDD), which is defined as the maximum odor travel distance from the source where odor concentration is reduced to 1 OU m 3. Because the distance of interest for setback determination is within 5 km from the source (Guo et al., 2), the odor concentrations predicted by the models were also compared within 5 km from the swine farm. Second, variable meteorological conditions were considered. A full year of hourly averaged meteorological data recorded in this study area in 23 was used in order to obtain the annual average odor concentrations in the study area; therefore, odor concentration predictions under variable meteorological conditions could be compared with those under steady-state meteorological conditions. For steady-state meteorological conditions, seven weather conditions, from stable weather (F1) to unstable weather (C5) (Pasquill, 1961), that favor odor travel and result in high odor concentrations at ground level downwind of the odor sources were chosen for this study, i.e.: F1: atmospheric stability F (moderately stable) with 1 m s 1 wind speed. F3: atmospheric stability F (moderately stable) with 3 m s 1 wind speed. E3: atmospheric stability E (slightly stable) with 3 m s 1 wind speed. E5: atmospheric stability E (slightly stable) with 5 m s 1 wind speed. D5: atmospheric stability D (neutral) with 5 m s 1 wind speed. D8: atmospheric stability D (neutral) with 8 m s 1 wind speed. C5: atmospheric stability C (slightly unstable) with 5 m s 1 wind speed. Weather conditions that are more unstable than C5 will result in stronger vertical mixing, greater air dispersion, and decreased odor transport distances as compared with the seven selected weather conditions; therefore, they are not included in this study. 3 C and 16 C, and it was assumed to be 2 C above the ambient temperature when the ambient temperature was above 16 C. The manure storage cells odor exit temperature was considered to be the same as the ambient air temperature.. As a result of considering the whole barn area as 32 point sources instead of considering each individual fan as a point source, the odor exit velocity from the barn area was considered as.5 m s 1. Odor exit velocities from the manure storage cells were also considered as.5 m s The model simulation time was set up to allow the odor to travel to the farthest distance at which maximum odor concentration was reduced to 1 OU m The receptor s detection height was considered as 1.5 m above the ground. 7. The study area was a rural crop field with flat terrain and no obstacles. Wind speed and direction were both horizontally homogeneous in the study area. 8. The wind direction was from WNW, i.e., the prevailing wind direction for this area, in studies where the wind direction did not vary except when examining the impact of various wind directions. 9. During all the simulations, neither deposition nor chemical transformation was considered. When simulating odor dispersion using hourly annual meteorological data, the same assumptions were applied except for assumptions 3, 5, and 8. The barn odor emitting temperature and the manure storage odor emitting temperature were constant at 16 C. For all steady-state CDD modeling and odor dispersion modeling using variable meteorological data, receptors were arranged in grid format and spaced 1 m from each other within 6 km from the farm. For odor concentration simulation within 5 km of the source, the receptors were placed along 72 direction radials, beginning with (north) and incrementing by 5 clockwise. In each direction, 33 receptors were placed on the centerline of the odor plume and spaced 1 m from each other within 3 km from the source, and four receptors were placed every 5 m at a distance of 3 to 5 km from the source. COMPUTATION ASSUMPTIONS The following assumptions were applied to odor dispersion simulation under the steady-state meteorological conditions: 1. All odor sources were considered as point sources in the simulation; for the other three models, the barn was considered a point source, and the two manure storage cells were considered area sources. For all models, the barn was separated into 32 point sources to best represent the shape of the barn. The odor emitting height was 1.5 m for the barn and m (ground level) for the manure storage cells. The manure storage cells were treated as single point sources for. was run in the screening mode using ISC meteorological data. 2. The odor emission rates from the barn and the manure storage cells were constant, as given in table 1, in order to exclude the effects of changing source emissions on odor dispersion. 3. For the barn, the exhaust air temperature was assumed to be 16 C when the ambient temperature was between RESULTS AND DISCUSSION MIXING HEIGHT Mixing height is defined as the depth of the surface boundary layer in which thermally generated or shear-generated turbulence is found. Under all seven weather conditions, F1 to C5, with an ambient temperature of 2 C, the simulation results indicated that the mixing height had no effect on the model predictions for all models when the mixing height was between 5 and 3 m. ATMOSPHERIC TEMPERATURE The effects of ambient temperatures between 2 C and 3 C on odor dispersion were evaluated with a mixing height of 15 m. Impact on CDD Figure 2 shows the odor CDD under E3 and D5. The odor travel distance increased with the increase of ambient temperature. The increased ambient temperature decreased the vertical dispersion and increased the molecular and eddy dif 11 TRANSACTIONS OF THE ASABE

5 CDD (km) CDD (km) E3 ISCST D5 t ( C) ISCST t ( C) Figure 2. Effect of atmospheric temperature on CDD under E3 and D5. fusion in the horizontal plane and favored horizontal odor transportation (Heinsohn and Kabel, 1999).Under E3, with the increase in the ambient temperature from 2 C to 3 C, the CDD increased 9.9%, 1.8%, 8.7%, and 7.6% as predicted by ISCST3, and, and IN- PUFF2, respectively, while the CDD under D5 increased 8.1%, 8.6%,.2%, and 3.6%, respectively, and the CDD under F3 increased 9.8%, 15.2%, 9.7%, and 52.5%, respectively. Hence, the more unstable the weather conditions, the less effect the ambient temperature had on odor dispersion. The ambient temperature had a much greater effect on than on the other models. was the least sensitive to ambient temperature. Table 2 summarizes the differences between the odor travel distances predicted by the other three models as compared to ISCST3. The difference, in percent, was calculated using the following equation: X Difference (%) = Model X X ISCST 3 ISCST 3 1 where X Model and X ISCST3 are the model-predicted and ISCST3-predicted CDD values (km), respectively. ISCST3 generally predicted the greatest odor travel distances, while had predictions very close to those of ISCST3 under E3 but slightly lower under D5. The predictions of and were 22.7% to 5.3% lower than those of ISCST3. s predictions were (1) Table 2. Difference in CDD between the other models and ISCST3. Weather Temperature Difference from ISCST3 (%) Conditions ( C) E D higher than those of when the temperature was below C but lower when the temperature was above C. Odor Concentration within 5 km The predicted odor concentrations under D5 on the plume centerline are given in table 3. At the same distance, the odor concentration increased with the increase in the ambient temperature. At 1 m, the impact of ambient temperature was significant for all models; the differences between the odor concentrations at 2 C and 2 C ranged from 65% (CAL- PUFF) to 31% (). One reason for these great differences was the higher plume rise at lower ambient temperature, which resulted in low odor concentrations close to the sources. Another reason was the inaccuracy of the models ability to predict near-source odor plumes because these models are not designed to predict air dispersion at such a close distance (U.S. EPA, 1995b). Furthermore, this distance is usually within the property line of swine farms; therefore, it is not of interest for regulatory purposes. Hence, the modelpredicted odor concentrations at 1 m will not be discussed in the remainder of this article, although model predictions may be given solely for the purpose of presenting the modeled results. At 5 m, the differences were reduced to 38% () to 123% (). At 1 km, the differences began to disappear except for (66%).These results agreed with the field measurement results, which found that the higher the temperature, the stronger the odor was at the same receptor s location (Pan et al., 25). Model Table 3. Plume centerline concentrations (OU m 3 ) downwind under D5. Temp. ( C) Distance (km) ISCST Vol. 5(3):

6 Percentage Difference (%) Temperature ( 2 C) CDD (km) ÓÔ ÓÔ ÓÔÖ ÓÔÖ ÓÔÖ ÓÔÖ ÓÔÔ ÓÔÖ ÓÔÔ Ö ÓÔÖ ÓÓÓÓ ÔÔÔÔÔ ÖÖ Ö 8 ISCST3 Ö 8 Distance (km) F1 F3 E3 E5 D5 D8 C5 Weather conditions Figure. Effect of atmospheric stability on CDD. Percentage Difference (%) Temperature (2 C) Table. Differences (%) on CDD between the other models and ISCST3 under various weather conditions. Weather Difference with ISCST3 (%) Conditions F F E E D D C Distance (km) Figure 3. Odor concentration differences (%) between other models and ISCST3 under different atmospheric temperatures. ISCST3 was also used as the benchmark to compare the models differences in odor concentration predictions, and similar results were obtained as for the CDD. As illustrated in figure 3, s predictions were within 1% of those of ISCST3 beyond 1 km, while and IN- PUFF2 s predictions were lower than those of ISCST3 by up to 7%, except at very close distances under low temperatures. ATMOSPHERIC STABILITY Impact on CDD The effect of atmospheric stability on the distance of odor dispersion is shown in figure under an ambient temperature of 2 C and mixing height of 15 m. As the atmospheric stability changed from stable (F1) to unstable (C5), the CDD decreased rapidly by 95% to 97%. Hence, atmospheric stability had a significant effect on odor dispersion, which is consistent with the finding of Chastain and Wolak (1999). Guo et al. (23) also reported that the majority of odor events were reported during either moderately or slightly stable atmospheric conditions. These four models had similar sensitivity to the atmospheric stability. It is important to point out that the great odor travel distances obtained with F1 and F3 were under the hypothetical assumption that the weather conditions would remain steady for a long enough time to allow odor to travel to such distances, i.e., 5 to 6 h with F1 or 2 h with F3. In reality, it is rare that such weather conditions would occur without any wind direction or stability shift. The 23 hourly averaged meteorological data from Yorkton, Saskatchewan, Canada, showed that the prevailing WNW wind had a total of one period of 5 h, two periods of h, three periods of 3 h, and seven periods of 2 h F1 conditions. It had a total of 62 h (annual occurrence frequency of.7%) of F1 conditions, while its annual frequency of F3 or more stable weather was 1.%. Table shows the differences between the CDD of the other models compared to those of ISCST3. predicted the lowest values under F1 to D8, which were 11.5% to 38.9% lower than those of ISCST3, followed by. Under C5, s predictions were slightly higher than those of. s predictions were close to ISCST3, with 17% to 6.1% difference. The differences between or and ISCST3 decreased as the instability of the weather increased. Odor Concentration within 5 km Modeled odor plume centerline concentrations within 5 km downwind are given in table 5. At 5 m, although the predicted odor concentrations were high under F1, this weather condition occurs with very low frequency, as discussed previously. Under F3, odor concentration gradually decreased with increased distance, and it was 27 OU m 3 or lower at 3 km (which is the maximum setback distance required by the setback guidelines of the Canadian Prairie provinces). Under E3 to C5, predicted odor concentrations at 3 km were 1 to 13 OU m 3. The odor plume centerline concentration at 3 km was reduced 98% to 99% when the weather changed from F1 to C TRANSACTIONS OF THE ASABE

7 Table 5. Plume centerline concentrations (OU m 3 ) under various weather conditions. Weather Distance (km) Conditions Model F1 ISCST F3 ISCST E3 ISCST E5 ISCST D5 ISCST D8 ISCST C5 ISCST CDD (km) CDD (km) Stability E ISCST Wind speed (m/s) Stability D ISCST Wind speed (m/s) Figure 5. Effect of wind speed on CDD. The differences between the odor concentration predictions of the four models were also analyzed. They were generally large within 1 km but were very small at 2 to 5 km. The differences were also large under stable weather conditions but were very limited under neutral and unstable conditions, as indicated in table 5. WIND SPEED Impact on CDD Because different atmospheric stabilities have different wind speed ranges, the wind speed s effect on odor dispersion was analyzed under different stabilities. The influence of wind speed on odor dispersion is shown in figure 5 under stability classes E and D, with an ambient temperature of 2 C and mixing height of 15 m. As wind speed increased, the CDD decreased. The increased turbulence associated with high wind speeds enhances air mixing and therefore decreases the horizontal odor dispersion. Under stability class E, when the wind speed increased from 2 to 5 m s 1, the CDD decreased by 5%, 53%, 5%, and 6% for ISCST3, AUS- PLUME,, and, respectively, while under stability class D, when the wind speed increased from 3 to 8 m s 1, the CDD decreased by 67%, 56%, 68%, and 5% for the four models, respectively. Comparing the four models, always predicted the lowest distances, which were 11.5% to 38.% lower than those of ISCST3; s predictions were 3.1% to 31.5% lower than those of ISCST3; and s predictions were within 1.8% to 1.1% of those of ISCST3. The higher the wind speed, the lower the differences between the models were. The possible reason is that the higher wind speeds dominated the effects of other factors accounted for by the models and became the driving force for air dispersion. Odor Concentration within 5 km Within 5 km, the odor concentration decreased with the increase of wind speed, as shown in table 6 under stability class D. These results are consistent with the observation of Guo et al. (23) that there were high occurrences of odor events during periods of low wind speeds. Pan et al. (25) also concluded that odor levels decreased faster at a high wind speeds than at low wind speeds. When the four models were compared, always predicted the lowest odor levels, which were up to % lower than those of ISCST3 (table 6). Within.5 km, predicted odor concentrations that were higher compared to ISCST3 s predictions, but beyond.5 km, they were lower compared to ISCST3 s predictions by up to 8%. gave higher values than ISCST3 within.5 km, but lower beyond.5 km by up to %. WIND DIRECTION Impact on CDD Wind directions from west-north-west (WNW), west (W), southwest (SW), and south (S) were selected to simulate the odor dispersion from the swine farm under E3 and D5 to examine the relative locations and orientations of the odor sources on odor dispersion. The ambient temperature was 2 C, and mixing height was 15 m. Vol. 5(3):

8 Model Table 6. Plume centerline concentrations (OU m 3 ) at various wind speeds under stability class D. Wind Speed (m s 1 ) Distance (km) ISCST Under E3, the CDD changed greatly with various wind directions as simulated by (fig. 6), which increased by 51.9% from 2.7 km with S wind to.1 km with SW wind. The wind directions had limited effect on the other models, with the highest effect on ISCST3 from 3.5 km with S wind to.2 km with W wind, a 2% increase. However, under D5, wind direction had a significant effect on all models except. The maximum CDD occurred with SW wind for and W wind for ISCST3, and the lowest CDD occurred with S wind for all the models except. had the largest difference between the CDDs under S and SW winds under D5, with.85 and 1.8 km, respectively, which is an increase of 112%. Hence, odor source orientations and wind direction affect downwind odor concentrations. With SW wind, the three odor sources were in a line, resulting in the highest odor concentrations downwind, while with S wind, the barn was parallel with the manure storage cells, so the centerlines of the plumes of the barn and manure storages were separated, which resulted in lower downwind odor concentrations. The differences among these four models were significant; e.g., predicted a CDD of 1.8 km with SW wind, while s prediction was 1.2 km. Odor Concentration within 5 km Table 7 gives the modeled odor concentrations within 5 km under D5. South wind was used as the reference wind direction for comparing the impact of wind direction. The Table 7. Plume centerline concentrations (OU m 3 ) with different wind directions under D5. Wind Distance (km) Model Direction ISCST3 S SW W WNW S SW W WNW S SW W WNW S SW W WNW maximum differences between the odor concentrations of the four directions at distances of.5, 1, and 2 km were 135%, 1%, and % (), while they were 27%, 16%, and 11% () at 3,, and 5 km. The results showed that the source orientation has great effect on odor dispersion, especially at distances closer to the source. s predictions for odor concentrations were similar to those of ISCST3 for all wind directions. IN- PUFF2 s results were similar to those of ISCST3 for S and SW winds but were mostly lower with W and WNW winds. The odor concentrations predicted by were higher than those of ISCST3 at close distances but lower at long distances for S and SW winds, while they were always lower than those of ISCST3 for W and WNW winds. VARIABLE METEOROLOGICAL CONDITIONS To obtain the annual average odor concentration in the nearby area of this swine farm, hourly averaged meteorological data for the year 23 from Yorkton, Saskatchewan, Canada, were used in odor dispersion modeling. Since only accepts 1 time periods in one run, the one-year hourly meteorological data were divided into numerous periods of 1 h each. The average odor concentration over each period was then calculated, and the average of all periods was the CDD (km) Ô ÓÓ ÔÔÓÔÖÖÓÔÖÖ ÓÓ ÔÔÖ ÓÔÖÖÓÔÖÖ ÓÓÔÔ ÓÓ ÔÔÖ ÓÔÖÖÓÔÖÖ ÓÓÔÔ Ö ÓÓ ÔÔÖ ÓÔÖÖÓÔÖÖ ÓÓÔÔ Ö ÓÓ ÔÔÖ ÓÔÖÖÓÔÖÖ ÓÓÔÔ Ö ÓÓ ÔÔÖ ÓÔÖÖÓÔÖÖ ÓÓÔÔ Ö E3 ISCST3 ÔÔ WNW W SW S Wind direction CDD (km) Ó Ó ÓÓ ÔÔ ÓÓ ÓÓ ÔÔ ÔÖÖÔÔ ÔÔ ÔÖÖ Ö ÔÔ ÖÓÓ Ö ÔÔÖ ÓÔÖÖ D5 Ó ISCST3 Ô WNW W SW S Wind direction Figure 6. Effect of wind direction on CDD. 11 TRANSACTIONS OF THE ASABE

9 Figure 7. Predicted annual average odor concentration (OU m 3 ) in the study area. Center of the odor sources is at (, ), and x and y axes are both in m. annual average concentration. The other three models could use this set of data directly. The receptor grid was 6 6 km, with a uniform spacing of 1 m. The results are shown in figure 7. The odor contours for various odor concentrations varied in all directions, with the maximum distances occurring leeward of the prevailing winds in the NW and SE areas. Schauberger et al. (22) calculated direction-dependent separation distance using the AODM model and found that for the area leeward of the prevailing winds, the odor occurrence frequency is higher than for areas leeward of less frequent wind directions. Guo et al. (25) studied odor occurrence in the same area as this study and found that the locations with high odor events were mostly downwind of the prevailing winds from the farm. The maximum downwind distances for 1, 2, 5, and 1 OU m 3 are presented in table 8. predicted the greatest distances for all odor concentration levels, while s predictions were the lowest. These results indicated that all four models predictions are different, and if used to determine setbacks, each will result in different setback distances, with differences up to 71.%. If annual average odor concentrations of 1 to 1 OU m 3 are used as setback criteria, then the maximum setback distance will be in the range of.3 to 2.3 km, which falls within the recommended setback distances used by some states in the U.S. and the Canadian Prairie provinces. As shown in table 8, s predictions were higher than those of ISCST3 under variable meteorological conditions, while they were lower under steady-state meteorological conditions. This was possibly caused by the difference in the minimum wind speed limitations of these two models. assumes the lowest bound wind speed of.5 m s 1, and it will overpredict when wind speed is less than.5 m s 1 (Australia EPA, 2), while the lowest wind Odor Concentration (OU m 3 ) Table 8. Maximum predicted distances for various odor levels and model differences based on ISCST3. ISCST3 Distance (km) Distance (km) Difference (%) Distance (km) Difference (%) Distance (km) Difference (%) Vol. 5(3):

10 speed for ISCST3 is m s 1. In the 23 annual meteorological data file used by this study, there were a total of 57 h during which the wind speeds were less than.5 m s 1, which caused overpredictions by. Similarly, even though s predictions were close to those of ISCST3 under steady-state weather conditions, under variable state weather conditions its predictions were much higher than those of ISCST3. U.S. EPA (1998) obtained similar results and stated that these results should come as no surprise because the meteorological assumptions used by ISCST3 and in predicting the downwind transport of the effluents and the dispersion from the respective plumes and puffs are different. The accumulation of hour-by-hour meteorological conditions on the transport of the puff is the key for the differences that are produced by these two models, and the difference is also compounded by the different treatment of dispersion during calm wind conditions (U.S. EPA, 1998). Comparing the annual average odor concentrations with the results obtained previously using steady-state weather conditions, odor can travel much farther under steady-state weather conditions using the same odor concentration criterion. The maximum distance of.7 km for achieving 1 OU m 3 obtained using annual hourly weather data is the same as the CDD for 1 OU m 3 under steady-state weather condition C5, while the other steady-state weather conditions resulted in 1 to 22 km of CDD (fig. ). Zwicke (1998) reported that ISCST3 overpredicted the hourly concentrations based on hourly averaged meteorological data by 2.5 to 1 times as compared to a series of controlled pollutant release and measurement experiments. Fritz et al. (25) also found that the appropriate time period for the Pasquill-Gifford horizontal dispersion parameter used in Gaussian-based dispersion models varied widely depending on the corresponding meteorological variations and that basing a 1 h averaged concentration on them might result in overestimated downwind concentrations. These results suggests that we may use different odor concentration criteria for steady-state weather conditions, such as F1 to C5, than we use for variable weather conditions, such as annual, seasonal, and monthly hourly weather data. The acceptable odor concentration allowed when using steady-state weather data should be higher, for example 75 OU m 3 as used in the OFFSET model (Guo et al. 25), than the acceptable average odor concentration allowed when using variable weather data over a period of time, e.g., 1 to 1 OU m 3 as for this study area. Further study is needed to quantify equivalent odor concentration criteria under various steady-state weather conditions and variable hourly annual weather conditions in the study area. CONCLUSIONS A mixing height ranging between.5 and 3 km had no effect on odor dispersion for all four air dispersion models. The ambient temperature had great influences on the critical detection distance (CDD), as predicted by, but its effect on the other models was very limited. The effect of the ambient temperature decreased with the increase of distance and disappeared at 1 km, except for. Atmospheric stability had a great impact on the CDD predicted by all models. As the weather changed from stable (F1) to unstable (C5), the CDD decreased 95% to 97% and the odor plume centerline concentration at 3 km reduced by 98% to 99%. Under the same atmospheric stability, as the wind speed increased, the CDD decreased. Under stability class E, when the wind speed increased from 2 to 5 m s 1, the CDD decreased by 53% to 6%. Under stability class D, when the wind speed increased from 3 to 8 m s 1, the CDD decreased by 5% to 67%. Wind direction had great impact on the CDD and odor concentrations near the swine farm, indicating the considerable effect of the source orientations on the odor dispersion. Comparing the model predictions under exclusively steady-state weather conditions, ISCST3 and gave similar results (within 17%), while and s predictions were much lower than those of ISCST3 (up to 5.3% beyond.5 km) for odor concentration and the CDD from the source. The differences between the model predictions generally decreased with the increase of instability and wind speed, and generally stabilized beyond 1 km from the source. Using annual hourly weather data, and predicted higher odor concentrations than ISCST3, which were very different from the results obtained under steady-state weather conditions. predicted the greatest distances for odor concentrations from 1 to 1 OU m 3 (up to 71.% higher than that of ISCST3), while predicted the shortest distances (up to.5% lower than that of ISCST3). When setting odor criteria for setback distance, it is suggested that if steady-state weather data are used, the odor concentration allowed should be set higher than that allowed when using hourly annual or monthly weather data for the same neighboring land use. Further study is needed to quantify equivalent odor concentration criteria under various steady-state weather conditions and variable hourly annual weather conditions. ACKNOWLEDGEMENTS The authors would like to acknowledge the Saskatchewan Agricultural Development Fund, Sask Pork, and the Alberta Livestock Industry Development Fund, Ltd., for funding this study. REFERENCES Australia EPA. 2. Gaussian Plume Dispersion Model: Technical User Manual. Victoria, Australia: Environment Protection Authority. CEN Determination of odor intensity using dynamic serial dilution olfactometry. Draft European Standard CEN/TC26. Chastain, J. P., and F. J. Wolak Application of a Gaussian plume model of odor dispersion to select a site for livestock facilities. Clemson, S.C.: Clemson University, Department of Agricultural and Biological Engineering. Available at: DEL.pdf. Accessed May 26. Fritz, B. K., B. W. Shaw, and C. B. Parnell, Jr. 25. Influence of meteorological time frame and variation on horizontal dispersion coefficients in Gaussian dispersion models. Trans. ASAE 8(3): Guo, H., L. D. Jacobson, D. R. Schmidt, and R. E. Nicolai. 21. Calibrating INPUFF-2 model by resident panelists for longdistance odor dispersion from animal production sites. Applied Eng. in Agric. 17(6): TRANSACTIONS OF THE ASABE

11 Guo, H., L. D. Jacobson, D. R. Schmidt, and R. E. Nicolai. 23. Evaluation of the influence of atmospheric on odor dispersion from animal production sites. Trans. ASAE 6(2): Guo, H., L. D. Jacobson, D. R. Schmidt, R. E. Nicolai, and K. A. Janni. 2. Comparison of five models for setback distance determination from livestock sites. Canada Biosystems Eng. 6: Guo, H., W. Dehod, J. Feddes, C. Lague, and I. Edeogu. 25. Monitoring odour occurrence in the vicinity of swine farms by resident observers: Part I. Odour occurrence profiles. Canada Biosystems Eng. 7: Hall, D. J., A. M. Spanton, F. Dunkerley, M. Bennett, and R. F. Griffiths. 23. An intercomparison of the AERMOD, ADMS, and ISC dispersion models for regulatory applications. Available at: Accessed May 26. Heinsohn, R. J., and R. L. Kabel Sources and Control of Air Pollution. Upper Saddle River, N.J.: Prentice Hall. Hurley, P. J., J. Hill, and A. Blockley. 25. An evaluation and inter-comparison of,, and TAPM: Part 2. Anglesea and Kwinana annual datasets. Clean Air Environ. Qual. 39(1): Jacques Whitford Environment. 23. Standard practice document for use of the dispersion factor in the calculation of minimum distance separation in the agricultural operation practices act. Project No. ABC2286. Calgary, Alberta, Canada: Jacques Whitford Environment, Ltd. Pan, L., S. X. Yang, and J. DeBruyn. 25. Measurement and analysis of downwind odors from poultry and livestock farms. In Proc. 7th International Symposium: Livestock Environment VII: St. Joseph, Mich.: ASAE. Pasquill, F The estimation of the dispersion of windborne material. Meteorol. Mag. 9(163): Schauberger, G., M. Piringer, and E. Petz. 22. Calculating direction-dependent separation distance by a dispersion model to avoid livestock odour annoyance. Biosystems Eng. 82(1): Sheridan, B. A., E. T. Hayes, T. P. Curran, and V. A. Dodd. 2. A dispersion modelling approach to determining the odour impact of intensive pig production units in Ireland. Bioresource Tech. 91(2): Smith, F. B Chapter 17: A scheme for estimating the vertical dispersion of a plume from a source near ground level. In Proc. Third Meeting of the Expert Panel on Air Pollution Modelling. Report NATO-CCMS-1. Paris, France: NATO/CCMS. Smith, R. J Dispersion of odours from ground level agricultural sources. J. Agric. Eng Res. 5(3): U.S. EPA. 1995a. User s guide for the industrial source complex (ISC3) dispersion models. Volumes I-III. EPA-5/B-95-3a-c. Washington, D.C.: U.S. Environmental Protection Agency. U.S. EPA. 1995b. A user s guide for the dispersion model. EPA-5/B Washington, D.C.: U.S. Environmental Protection Agency. U.S. EPA A comparison of with ISCST3. EPA-5/R Washington, D.C.: U.S. Environmental Protection Agency. Zwicke, G. W The dispersion modeling of particulate for point and multiple point sources in agriculture. MS thesis. College Station, Texas: Texas A&M University, Department of Biological and Agricultural Engineering,. Zhou, X. J., Q. Zhang, H. Guo, and Y. X. Li. 25. Evaluation of air dispersion models for livestock odour application. Paper presented at CSAE/SCGR Annual Meeting, Winnipeg, Manitoba, Canada. Paper No Zhu, J., L. K. Jacobson, D. R. Schmidt, and R. Nicolai. 2. Evaluation of INPUFF-2 model for predicting downwind odors from animal production facilities. Applied Eng. in Agric. 16(2): Vol. 5(3):

12 118 TRANSACTIONS OF THE ASABE