The Canadian Society for Bioengineering The Canadian society for engineering in agricultural, food, environmental, and biological systems.

La Société Canadienne de Génie Agroalimentaire et de Bioingénierie La société canadienne de génie agroalimentaire, de la bioingénierie et de l’environnement

Paper No. 06-170

Livestock Odour Dispersion Modeling: A Review

H. Guo Department of Agricultural and Bioresource Engineering, University of Saskatchewan, Saskatoon, S7N 5A9

Z. Yu Division of Environmental Engineering, University of Saskatchewan, Saskatoon, S7N 5A9

C. Lague Department of Agricultural and Bioresource Engineering, University of Saskatchewan, Saskatoon, S7N 5A9

Written for presentation at the CSBE/SCGAB 2006 Annual Conference

Edmonton Alberta July 16 - 19, 2006

Abstract This paper reviewed the studies of livestock odour dispersion modeling including odor sources, odor emission rate, odor characteristics and measurements, odor dispersion modeling methods, and methods and results of adapting industrial air dispersion models for livestock odor dispersion simulations. The Gaussian plume models, Puff models, Fluctuating models and other models that have been used in livestock odour dispersion modeling were explicitly reviewed. Research work on validation or evaluation of industrial air dispersion models using various field odour plume measurement results and the various relationships between odour concentration and odour intensity generated by different studies were also discussed. Although many Gaussian models have been applied in livestock odour dispersion and achieved relatively satisfactory results, there are many differences between industrial gas and livestock odour, including emission sources, travel distance, measurement methods, that can challenge the use of Gaussian models. Since the existing commercial models are originally designed for industrial gas dispersion, it is necessary to develop a new model that can accurately predict livestock odour concentrations downwind the sources. Finally, a fluctuating model that can deal with multiple sources and variable emission rates was suggested for livestock odour dispersion with proper diffusion coefficients within the interest short distance. Papers presented before CSBE/SCGAB meetings are considered the property of the Society. In general, the Society reserves the right of first publication of such papers, in complete form; however, CSBE/SCGAB has no objections to publication, in condensed form, with credit to the Society and the author, in other publications prior to use in Society publications. Permission to publish a paper in full may be requested from the CSBE/SCGAB Secretary, PO Box 23101, RPO McGillivray, Winnipeg MB R3T 5S3 or contact bioeng@shaw.ca. The Society is not responsible for statements or opinions advanced in papers or discussions at its meetings.

Introduction Researchers have been using industrial air dispersion models to predict livestock odours downwind of livestock operations since 1980's. However, most of these models are designed for air contaminantes from industrial sources and studies indicated that they cannot be directly used for livestock odour dispersion predictions (Zhu et al. 2000a; Guo et al. 2001; Zhang et al. 2005). The industrial and agricultural sources are different regarding source type, emitting height, emitting temperature, etc. Unlike specific air contaminants from industrial sources, livestock odour consists of more than 300 compounds and chemical or biological reactions may occur during its transportation (Schiffman et al. 2001). Odour is measured differently from the other air contaminants. Odour concentration, in detection threshold or Odour Unit, is measured using olfactometer and human panelists. Field odour plume can only be measured for odour intensity by using human sniffers (Li et al., 1994; Hartung and Jungbluth, 1997; Zhu et al., 2000a; Guo et al., 2001; Zhang et al., 2005). Furthermore, most of the industrial air dispersion models calculate hourly average concentration, whereas the interval of one field odour plume measurement is between 10 s to 1 min and with the total measurement duration of 10 min. Odours can generate community complaints from a series of short detectable exposures even though the hourly-averaged concentration is very low or undetectable.

Objectives

The objectives of this study are to review the differences between livestock odour dispersion modeling and conventional industrial air dispersion modeling regarding odour emission sources and the difference of odour from gas, review various approaches for adapting the commercial air dipersion models for livestock odour dispersion, and identify the requirement for development of livestock odour dispersion models.

Livestock Odour And Its Measurement Livestock Odour Odor is the human olfactory response to many discrete odorous gases. Odour from livestock operations has many different compounds and most of the compounds have very low concentrations. A total of 321 different compounds in livestock buildings have been identified by Schiffman et al. (2001). The compounds are often listed in groups based on their chemical structure. Some of the principal odorous compounds, individual and as group, are ammonia, amines, hydrogen sulphide, volatile fatty acids, indoles, skatoles, phenols, mercaptans, alcohols, and carbonyls (Curtis, 1983). Yu et al. (1991) reported that indole, p−cresol, phenol, skatole, volatile carboxylic acids, and ammonia appeared to be the most important constituents of odor from swine waste. Zahn et al. (1997) indicated that the C2 through C9 carboxylic acids represented the greatest threat to air quality because of their high transport efficiencies and high airborne concentrations. The concentrations of some major compounds (H2S, NH3, VOCs) of livestock odour had been measured by different researchers. The average H2S concentrations measured in swine facilities were less then 2 ppm (Donham and Popendorf, 1985; Schiffman et al., 2001), although the peak concentrations were up to 100 ppm during agitation of manure and up to 220 ppm from exhaust air from the pit fan of a deep-pitted swine facility (Patni and Clarke, 1991). The average Ammonia concentrations measured in cattle housing, swine housing and poultry buildings were less than 8 ppm, 5-18 ppm and 5- 30 ppm, respectively (Koerkamp et al., 1998). Donham and Popendorf (1985) measured higher mean NH3 concentration of 33.8 ppm in 21 randomly

2

selected swine producing farms in Iowa. The concentrations of VOCs in swine buildings ranged from 0.62 mg/m3 to 11.72 mg/m3 (Schiffman et al., 2001; Hartung and Phillips, 1994) and were much higher (up to 108.7 mg/m3) from slurry storages (Hobbs et al., 1997; Zahn et al.,1997). Although the concentrations of many individual compounds of the livestock odour are below the standardized odor thresholds, the intensity of the total mixture may be very strong and irritating. This indicates that the intensity of the odorous emissions result not only from individual compounds that are present at concentrations above their odour thresholds, but from the aggregate effect of hundreds of odorous chemicals present at subthreshold concentrations (Schiffman et al., 2001). It was found that concentrations of the most common specific odorous gases, such as hydrogen sulfide and ammonia, are not correlated to livestock odour concentration (Spoelstra, 1980; Pain and Misselbrook, 1990; Jacobson et al., 1997; Zahn et al., 1997). Therefore, no individual compound can be used to quantify the livestock odor intensity, rather, overall odor strength or odor intensity of the air emitted from the livestock facilities is measured. Odour Characteristics And Measurement There are various techniques for measuring and describing odours. Odours can be characterized by five attributes: concentration, intensity, persistence, hedonic tone and character descriptor. Concentration and intensity are the most widely accepted and used characteristics of odour. Sensory methods using dynamic triangular olfactometry are most widely used to evaluate odor concentration. There are two odor concentrations (thresholds) that can be measured: detection threshold and recognition threshold. They are measured by an olfactometer (a dilution apparatus to dilute the odorous air sample by fresh air) and a panel of trained odor assessors and usually reported as odor units (OU). Odour unit is defined as the dilution ratio of the odorous air sample by fresh air when 50% of an odor panel can detect or recognize the odor. If not specifically pointed out, the detection threshold is usually used as the odor concentration in all odor research such as air quality, odor emission, and dispersion. Odor intensity describes the strength of an odour sample. It is measured at concentrations above the detection threshold. Intensity changes with odor concentration. It can be measured against a five-step scale using n-butanol, a standard reference chemical (ASTM, 1998). Different n-butanol scales have been used by different researchers (Jacobson et al., 2005, Zhang et al. 2003 and 2005; Feddes et al. 2005). Jacobson et al. (2005) used a 5 point scale while Zhang et al. (2003, 2005) and Feddes et al (2005) used an 8 point scale. The comparison of the two scales is given in Table 1. For the same n-butanol concentration-in-water, the intensity interpretation was not the same. For example, 240 ppm n-Butanol in water was considered a little annoying in 8 point referencing scale, while it was considered very faint in 5 point scale. Therefore, for the same n-butanol concentration in water, the 5 point scale will achieve high odour concentration (25 OU/m3) than 8 point scale (0.2 or 5 OU/m3). This will result in the different relationship between odor concentration and intensity. Persistence indicates how easily the full-strength odorous air is diluted to below the detection threshold. It is a calculated value based on the full-strength intensity and the detection threshold concentration. Hedonic tone describes the unpleasantness or pleasantness of an odor (ASCE. 1995). It is typically rated using a scale that ranges from -10 to +10 indicating unpleasant to pleasant. Character descriptors are used to describe the character of the odor. For livestock odour, character descriptors can be used by the panelists to identify the different odour sources

3

as indicated by such as swine manure odour or odour from cattle. Character descriptors are used when the samples’ concentrations are at or above the recognition threshold concentration.

Table 1. Odour intensity referencing scale Odour concentration

（OU/m3） Odor Scale Odour

intensity Odour Strength

n-Butanol

in water (ppm)

by Zhang et al. (2005)

University of Manitoba

by Feddes et al. (2005)

University of Alberta

0 No odour 0 0 1 1 Not annoying 120 0 2 2 A little annoying 240 0.2 5 3 A little annoying 480 3 12 4 Annoying 960 24 26 5 Annoying 1940 116 58 6 Very annoying 3880 412 128

Very annoying 7750 1204 287

0 to 8

7 8 Extremely

annoying 15500 3051 640

0 No odour 0 0 1 Very faint 250 25 2 Faint 750 72 3 Moderate 2250 212 4 Strong 6750 624

0 to 5 (Guo et al.,

2001)

5 Very strong 20250 1834 Livestock Odour Sources And Emission Rates Odor emissions from an animal production site originate from three primary sources: manure storage unit, animal housing, and land application of manure. Animal mortalities/carcasses and feedstuffs, storage sites and silage piles/pits are also the sources of livestock odours. Most of odorous gases from livestock operations are by-products of anaerobic decomposition/ transformation of livestock wastes by microorganisms. Livestock wastes include manure, spilled feed wash water, bedding materials, wash water, and other wastes. The by-products of microbial transformations mainly depend on the ways they are done (aerobic or anaerobic). Microbial transformations done under aerobic conditions generally produce fewer odorous by-products than those done under anaerobic conditions. Moisture content and temperature affect the rate of microbial decomposition. Hence, the odor production from the odour sources varies diurnally and seasonally with the changing indoor and outdoor climatic conditions and animal conditions. The exhaust air emitting height is at or lower than the ground level for outdoor earthen manure storages while for above ground concrete or steel storages the emitting height will be up to 5 m above ground. For animal buildings, exhaust air outlets are the openings on the walls and roofs of the building and the wall or ceiling mounted fans, the emitting height varies between 1 m on the wall to up to 6 m for ridge vent and ceiling mounted fans. There are often a number of vertical or horizontal openings or fans for one building. For manure storages, the exhaust air temperature is the same or close to the ambient air

4

temperature. In winter, if the manure storage is frozen, there will be no odor emission. For livestock buildings, depending on the housing system, the exhaust air temperature is kept at required temperature range for animals, which may be as low as above 0oC for dairy in winter, as high as 32oC for young chicks, and normally 2 to 4oC higher than ambient temperature in summer. Odour concentration is an important air quality parameter for the workers and animals in the livestock buildings. Odour emission rate from livestock sources measures the size of an air pollution source for the surrounding area and can be used as standard of evaluation of odour impact. As a necessary input of odour dispersion model, it is important for the study and application of the odour dispersion model as well as the determination of the setback distance. For a building source, odour emission rate is the product of the odour concentration of the exhaust air and ventilation rate of the building. For manure storages, a wind tunnel or ventilated chamber with a certain air flow rate will be placed on the manure surface and the odor emission rate will be the product of the odor concentration of the exhaust air from this device and the air flow rate of the device (Ormerod, 1990; Pain et al., 1988; Pain and Misselbrook, 1990). When dealing with odour dispersion model, confusion results from the unit of the emission rate because the odour concentration, i.e. odour detection threshold, is expressed as OU rather than g/m3. Some researchers used OU as the odor concentration unit (Smith, 1993; Mahin, 1998), this caused confusion because using OU as the odor concentration unit, ventilation rate has a unit of m3/s, then the odor emission rate will have a unit of OU m3/s, which is not consistent with the mass concentration and emission rate accepted by air dispersion models. If odor is treated as a specific mater in air dispersion modeling, the “mass” of odour should be expressed as OU which is equivalent to unit of mass “g”. Then, the odor concentration takes the unit of OU/m3 and odour emission rate OU/s, the same format as mass concentration and emission rate. This is employed by most odor researchers in odour dispersion models (Williams, 1985; Carney and Dodd, 1989; Pain et al., 1991). Another potential approach for dealing with the “mass” of odor is to use the equivalent n-butanol mass concentration. CEN (2002) considered that n-butanol in air with mass concentration of 123 µg/m3 has odor detection threshold of 1 European Odour Unit (OUE), therefore, 123 µg/m3 is considered as the mass concentration of 1 OUE odorous air. In air dispersion modeling, the odor concentration is converted to the equivalent n-butanol mass concentration by multiplying the odor detection threshold with 123 µg/m3 n-butanol, then the odor dispersion modeling is virtually changed to normal air dispersion modeling of n-butanol vapor. However, this approach has no advantage comparing to the method that directly uses Odour Unit, because the model output should be converted back to Odour Unit in order to validate the model, evaluate the odour impact or make other applications. Odor emission rates change constantly with changing animal mass and number and outside weather conditions. None of the existing setback models or odor dispersion models considers diurnal and seasonal variation in odor emission rates. Odor emission rates have been measured more or less randomly during specific time periods (Heber et al., 1998; Lim et al., 2001; Jacobson et al., 2000, Verdoes and Ogink, 1997; Klarenbeek, 1985; Wood et al., 2001, Zhou and Zhang, 2003; Zhang et al., 2005). Great variations in odor concentrations and emission rates have been measured in each study and among different studies (Wood et al., 2001). The means or geometric means of the limited measured odor emission rates for each type of odor sources were used as representative values in odor dispersion and setback modeling without considering the diurnal and seasonal variations (Zhu et al., 2000b; Jacobson et al., 2000; Lim et al., 2000). Guo et al. (2006) found large variations in annual and diurnal odor emission rates

5

from swine barns but no specific seasonal or diurnal patterns were observed. It was suggested that multiple measurements should be taken from an odor source to obtain the mean, maximum, and minimum odor emission rates for air dispersion modeling purposes. Odour Plume Measurement The accuracy of odor dispersion models needs to be evaluated by field odor plume measurement data. Challenges exist in air dispersion model evaluation using odor plume measurement. Although odor concentrations, i.e., odor detection thresholds (OU/m3), are used as inputs to dispersion models, air samples taken in the odor plume downwind of a source are generally below the sensitivity of olfactometry panels (Zhang et al., 2003), which excludes the use of an olfactometer for odor plume determination. Instead, odor intensity which measures odor strength by using number and word categories to describe an odor is widely accepted to measure downwind odor plumes (Li et al., 1994; Hartung and Jungbluth, 1997; Zhu et al., 2000a; Guo et al., 2001; Zhang et al., 2003). There are two methods of measuring odour plume dispersion, i.e., odor intensity at the observer’s location. The first method is to measure the odour plume using a panel of trained odour observers. The second method is to monitor odour occurrence at neighbouring residences using trained resident odour observers. For the first method, groups ranging from 5 to 15 trained odour observers, are brought leeward of an odour source and the odour intensity of the odour plumes are measured. Several studies have used this method to measure odour plumes (Li et al. 1994; Hartung and Jungbluth 1997; Kaye and Jiang 1999; Zhu et al. 2000a; Jacobson et al. 2000; Zhang et al. 2003). Usually, one measurement only takes 10 min and the downwind distance from the odour source is less than 1 km and most often less than 0.5 km. Beyond this distance, little odour could be detected. There are two reasons for the inability to detect odour at a greater distance: a) due to the changing wind direction, it is difficult to position the odour observers in the right place on time at such a long distance to catch the odour plume, and b) the measurement takes place mostly during the daytime when unstable or neutral atmospheric stability may not allow odours to travel for a longer distance. Hence, although rather costly, this method is only practical for short distance measurement and the results are obtained under specific weather conditions and topography, which may not be replicated under other conditions. This method is used because the quality of the data is relatively easy to control when compared with the other method using resident observers. However, the setback distances in most setback guidelines are greater than 0.5 km, and livestock odours have been detected up to 6 km away from the odour source (Guo et al. 2004), so this method will not be helpful for odour dispersion model validation for long distances. Besides, most industrial air dispersion models are intended for long distance predictions up to 50 km rather than 1 km or less (EPA US 2004). Ideally, an alternative way for using trained assessors to monitor odours in a certain area over a long period of time would be to arrange odour observers on a grid area to live at the monitoring locations and work full time as assessors (VDI 1993). The high cost would make it impractical. The second method, i.e., using trained voluntary resident odour observers to monitor odour, has it merits and demerits. First, it is very useful for long term odour monitoring at the resident’s location considering that the resident is at home for a relatively long period of time. Residents are normally at home and available to observe odours during the stable atmospheric weather conditions from the late afternoon, throughout the night, and to the early morning, and some rural residents may be available to monitor odours at home all the time. Therefore, odour occurrence can be observed under various weather conditions and seasonal and diurnal odour

6

occurrence profiles can be obtained. The cost is relatively low because the assessors are voluntary. There has been very limited research done with this method. Jacobson et al. (2000) and Guo et al. (2001, 2003) monitored odour in a 4.8 x 4.8 km grid of farmland that had 20 livestock farms within or adjacent to it in Minnesota, U.S.A. Nineteen trained resident odour observers monitored odour at their residences from late June to mid-November for five months during their normal daily activities. Nimmermark et al. (2003) also used a similar method and measured odours in five areas of Minnesota. Guo et al. (2005b, 2006) monitored odor occurrences around three swine farms using 39 families living within 8.6 km (5 miles) from the swine farms for two separate years. These studies have proved that using resident odour observers for long term and long distance odour dispersion measurement to be practical and effective. However, measures need to be taken to increase the accuracy and credibility of the data. The possible options include implementing periodic nose calibration, screening the observers for bias for or against the intensive livestock operations, and taking measurements at designated times. Another drawback of this method is that odour monitoring can only be done at the volunteers’ residence locations, which might not cover all desired locations (Guo et al., 2005b). An alternative method by combining the above two was used by Guo et al. (2005b) that used hired trained odor observers to travel on designated locations at designated time in a study area to monitor odors downwind of swine farms for the warm season (May to October). This method has proved to ensure the quality of data for short and long distances, and long term observations.

Odour Dispersion Modeling

Atmospheric dispersion models have been proved to be a very powerful tool to predict the odour concentration downwind the agricultural source, so as to determine where odour nuisance is likely to occur near livestock production facilities. There are several models (e.g. ISC3, ADMS3, AUSPLUME, INPUFF and CALPUFF) that are commercially available and often particular models are favoured in different parts of the world. However, most of these models are originally designed for industrial pollution sources and may not suitable for livestock odour dispersion modeling. Only a few models (AODM, ODODIS) were developed specifically for odour dispersion from agricultural sources, however these models still use the same air dispersion theories regardless of the differences between gas and odour or need more validations. Gaussian Plume Model Traditionally, the Gaussian plume model is the most common air pollution model. It is the cornerstone of most dispersion calculations in regulatory applications, which are based on this Gaussian dispersion model for a continuous point source in a uniform flow with homogeneous turbulence. In a general reference system, the Gaussian plume formula is expressed as (Arya, 1999):

⎪⎭

⎪⎬⎫

⎪⎩

⎪⎨⎧

⎟⎟⎠

⎞⎜⎜⎝

⎛ −−+⎟⎟

⎠

⎞⎜⎜⎝

⎛ +−⎟

⎟⎠

⎞⎜⎜⎝

⎛−= 2

2

2

2

2

2

2)(exp

2)(exp

2exp

2 zzyzy

HzHzyU

QCσσσσπσ

[1]

where: Q is the source emission rate, g/s; C is the downwind concentration at the receptor location (x, y, z); U is the average horizontal wind speed, m/s; H is the effective emission height,

7

m; and yσ and zσ are the standard deviations (horizontal and vertical) of the plume concentration, m. ISCST3 and AUSPLUME are the most commonly used air dispersion model based on the Gaussian dispersion theory. ISCST3 (Industrial Source Complex) model is designed to support the US EPA's regulatory modeling programs and is widely used in North America and worldwide (US 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. AUSPLUME model was developed by Australian Environmental Protection Authority and it is an extension of the ISCST3 model (US EPA, 2000). 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. Some work has been done in using the air dispersion models for predicting the odours from industrial or urban sources, especially from composting facilities. Engel et al. (1997) utilized ISCST model to simulate the odours from a composting facility, the model output 1-hour concentrations were converted to 30 seconds peak concentrations with a peak to mean ratio of 1.97. This methodology appears to generate reasonable results in terms of predicting the frequency of nuisance conditions and the model results correlate with quantitative measurements obtained from field sampling studies. Heinemann and Wahanik. (1998) applied a Gaussian plume model in predicting the local distribution of odours emanating from mushroom composting facilities. A composter survey was conducted to verify the model and the results appeared to be reasonable, however only one odorous gas (dimethy1 disulfide) was modeled to indicate odour. Mahin (1997) summarized the selected case studies relative to use of dispersion model to predict odours from composting facilities and wastewater treatment facilities. ISCST model and SCREEN (An EPA Gaussian plume model) models were used in most of the selected cases and different adjustment factors (peak-to-mean ratios) were applied to convert long time (1-hour) average concentration to short term peak concentration (2 min, 5 min or 10 min). Many researchers simplified and modified the Gaussian plume model and applied in agricultural odour dispersion. Stoke (1977) applied the Pasquill’s equation (equation [2]) to predict odour dispersion from 10 pig barns. Equation 2 was used to calculate the distance for 1 OU/m3 with measured emission rate and odour panels were employed to determine the downwind distance where odour concentration was equal to 1 OU/m3 which meant that 50% of the panel can detect odour. Agreement between predicted and measured distance for 1 OU/m3 was reasonable.

⎥⎦

⎤⎢⎣

⎡+−=

zyzy

HyQ

UHyxC2

2

2

2

21exp

21),0,,(

σσσπσ [2]

Janni (1982) evaluated the importance of various parameters in the PTDIS model developed by the US EPA with regards to their potential for reducing odour problems from agricultural sources. The model was a simplified Gaussian model and was used to calculated ground-level concentrations between 0.5 and 30 km downwind of a point source. He concluded that downwind distance, stability class and wind speed, source emission rate are important factors in odour dispersion, and the emission height has little effect on the odor dispersion. He also pointed out that field measurement data is needed to validate the results obtained.

8

Majer and Krouse (1985) conducted research on using modified Gaussian dispersion model to predict emissions from agricultural sources. They considered the difference between the gas dispersion from industrial sources and odour dispersion from agricultural sources and convinced that the Gaussian plume formula should be used only for those downwind distances for which the empirical diffusion coefficients have been determined by standard diffusion experiments. They developed a modified Gaussian plume model suitable for predicting the concentration of pollutants on the centerline in the downwind direction (Equation [3]), although the validation of this modified model was not conducted.

)2

exp()2

1( 2

2

22

22

zz

x

yx

sc

Hx

HU

vCCσσ

σσσπ

−−= , [3]

where: Cc is the downwind concentration on the centerline, g/m3; v is the volume emission rate at the source, m3/s; Cs is the source concentration g/m3; xσ is the standard deviations downwind of the plume concentration, m. Similarly, Williams (1985) simplified the Gaussian plume model to estimate the odour concentration emanating from animal buildings and land application of manure. This model (Equation [4]) described the downwind ground level concentration on the plume centerline for a ground level source (H=0).

U

DvCCzy

c σπσ0= [4]

Where: D is the number of dilutions to detection threshold; and C0 is the detection threshold concentration, g/m3. The distance that odour complaints might be expected was calculated by equation [5]. Comparison with empirical formulae relating distance of complaint to odour emission obtained from a large number of experimental studies showed that dispersion modeling approach provided reasonably accurate results.

U

DFSR

zy πσσ )(= [5]

where: R is the peak to mean ratio; S is the factor by which recognition or annoyance threshold causing odour complaints is larger than the detection threshold. Carney and Dodd (1989) also applied modified Gaussian plume models to evaluate the ground level odour concentration downwind a point source by using the same equation as equation [2] and a linear source by equation [6]. Comparisons were made between the model predicted concentrations with actual odor emission rate and odor plume measurement data for various agricultural sites including a point source (a slurry tank), a line source (a strip of land spread with slurry), an areal source (field spread with slurry), and a 450-sow swine production unit. Since there was no equation for the areal source, the model for the linear source was used and the answer multiplied by the width of the area. Results showed that the model had a good indicator of how an odour dispersed from a point source and a linear source. When odours were dispersed from an area source, the linear model gives good agreement with experimental results if an equivalent width of 10 m was used.

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛−⎟

⎠⎞

⎜⎝⎛=

25.0

5.0exp2),0,,(zz

HU

QHyxCσσπ

[6]

9

Gassman (1993) reviewed the Gaussian plume methodology, its inherent problems in estimating odour dispersion, and previous attempts to model the dispersion of agricultural odours. He discussed the shortcomings of the commonly used Gaussian method of modeling dispersion, which are that the Gaussian method does not take into account non-steady-state flow, topography, inconsistent wind velocities or the fact that odour moves in puffs rather than in continuous flow. Gassman (1993) stated that the Gaussian method was adequate when comparing differences between different scenarios, but was not recommended for finding absolute odour concentrations. Smith (1993) developed a Gaussian plume model (STINK) which can predict the dispersion of odours downwind of areal sources, such as feedlots, of any shape and any orientation with respect to the wind direction. The program was based on the equation 7. Concentration from areal sources was given by the numerical integration of the concentrations from the strips that the areal source had been divided into. The advantage of STINK model is to predict the dispersion considering an accurate description of the source geometry which is important for the dispersion relatively close to large area sources. The sensitivity of the concentration predictions to variation in the model parameter was assessed. The wind speed and odour emission rate are shown to be the most important variables in the prediction of downwind concentration. Other parameters of importance are the atmospheric stability and the roughness height of the ground surface.

∑⎪⎭

⎪⎬⎫

⎪⎩

⎪⎨⎧

⎥⎥⎦

⎤

⎢⎢⎣

⎡

⎟⎟⎠

⎞⎜⎜⎝

⎛ −−⎟⎟⎠

⎞⎜⎜⎝

⎛ +⎟⎟⎠

⎞⎜⎜⎝

⎛−=Φ

n

yiyizizi

XYyerfYyerfzzyx1

2

2

22

22

2exp

21),,( δ

σσσσπ [7]

where:

QUzyxCzyx ),,(),,( =Φ is a normalized concentration. The areal source is divided into n strips.

yiσ and ziσ refer to the ith strip of the source, m, distance xi from the receptor, X and Y are the width and length of the areal source, m. Smith (1995) modifed STINK model to predict the spatial average odour emissions from relatively large area sources given simultaneous point measurements of odour concentration and wind speed at a location immediately downwind of the source. The comparison with the TPS (Theoretical Profile Shape) method showed that STINK was suitable for predicting the rate of odour emissions. Comparison between calculated spatial average emission rates for a small feedlot pen 25 m square and the point measurements of emission rate obtained from wind tunnel samples also demonstrated the ability of this model to predict emission rate for small sources (Smith and Hancock, 1992). Smith and Kelly (1996) compared the predicted odor emission rates from the STINK model to the predicted from a Lindvall hood for a number of different sources, including a feedlot pad, an artificial feedlot pad and an anaerobic feedlot pond. The results were of a similar order of magnitude but only a poor relationship between the back-calculated values and the wind tunnel values. The model predicted odor emission rates were higher than samples collected directly from the emitting surface using a wind tunnel, which was similar to that observed by Koppolu (2002). Koppolu et al. (2002) compared the STINK model and the AERMOD (An EPA regulatory model) to assess the emission rate and near-source-dispersion of low-weight volatile fatty acids (VFAs)

10

from an artificial source under controlled conditions. VFAs were sampled at six receptors using thermal desorption tubes in one experiment and solid phase microextraction (SPME) fibers in a second experiment. Comparisons between measured concentrations and predicted concentration by STINK and AERMOD showed good agreement. However, better prediction by AERMOD, compared to STINK, was observed for the experiments involving SPME sampling. The performance of AERMOD was also assessed using meteorological conditions averaged over various time intervals. It was assumed that the average conditions observed during 1, 5, 15 and 30 minute intervals prevailed for the entire 1-hr modeling period. Koppplu et al. (2004) compared an Odor Footprint Tool (OFT) through the use of EPA regulatory model, AERMOD with Minnesota’s Odor From Feedlot Setback Estimation Tool (OFFSET) based on meteorological data from six regional locations in or near Minnesota. Comparison of measured odor intensities from livestock facilities to predicted ambient odor levels from AERMOD had been done to validate the OFT. Scaling factors in the range of 0.2 to 3900 were suggested to adjust AERMOD predictions to short-term odor measurements depending on the source type (point, area, volume) and the type of facility being modeled. Stowell, et al. (2005) described in detail the Odor Footprint Tool that utilizes AERMOD, weather databases, new source code, and user input to generate regional odor roses, odor footprints, and directional setback distance curves. Significant differences in predicted odour concentrations have been shown in comparisons between two widely used models ISCST3 and ADMS 3.1 (Atmospheric Dispersion Modeling Systemn) (Curran et al., 2002). Sheridan et al. (2004) selected ISCST3 as the most appropriate model to use in predicting where odour nuisance is likely to occur near pig units in Ireland because of the previous validation study. A Gaussian plume model, the Austrian odour dispersion model (AODM), was developed by Schauberger et al. (2000). The AODM consists of three modules: the first calculates the odour emission of the livestock building, the second estimates mean ambient concentrations using the Austrian Gaussian regulatory dispersion model, and the last transforms the mean odour concentration of the dispersion model to instantaneous values depending on wind velocity and atmospheric stability. The resultant distances for different odour thresholds and their dependence on meteorological parameters are investigated, focusing on the distance for the detection limit, 1 OU m-3, the so-called sensation distance. The results indicated a stronger dependence of the distances from meteorological conditions than from odour emission parameters and was judged to have given satisfactory results. Using the same model, Schauberger et al. (2001) calculated the separation distance between livestock buildings and residential areas and concluded that the AODM model is appropriate for regulatory purpose. Puff Model The Gaussian puff model has been developed to model the dispersion of a cloud formed by an instantaneous release of light toxic gas. The theoretical basis of this model is the same as the standard Gaussian model with the difference that longitudinal dispersion is considered. In the center of the traveling puff, concentrations are maximum and then represent a concentration peak. The general expression of puff model is expressed as (Arya, 1999):

⎭⎬⎫

⎩⎨⎧

⎥⎦

⎤⎢⎣

⎡ +−+⎥

⎦

⎤⎢⎣

⎡ −−

⎥⎥⎦

⎤

⎢⎢⎣

⎡−−= 2

2

2

2

2

2

2

2

23 2)(exp

2)(exp

22exp

)2( zzyxzyx

ip HzHzyxQC

σσσσσσσπ [8]

11

where: Qip is the instantaneous point source emission rate; xσ , yσ and zσ are the standard deviations of the puff concentration. Some models have been developed based on the puff theory, in which INPUFF2 model (Petersen and Lavdas, 1986) and CALPUFF model (USEPA 1995b) are the most representative and most commonly used. INPUFF2, a Gaussian integrated puff model, was developed by the US 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 every time step. It can simulate dispersion of airborne pollutants from semi-instantaneous or continuous point sources. There is no treatment of area or volume sources. It may deal with non-reactive pollutants, deposition, and sedimentation. It can deal with different time intervals with minimum of 1 s instead of 1 h required by the other models. This makes it suitable for simulating odors as measured by field odor assessors. This model has some consideration of terrain effects through the wind field but there is no explicit treatment of complex terrain. CALPUFF air dispersion model was an US EPA regulatory model based on Lagrangian puff model designed to simulate continuous puffs of pollutants being emitted from a source into the ambient wind flow (US EPA, 1995b; US EPA, 1998). It consists of three sub-systems: CALMET, CALPUFF, and CALPOST. CALMET is a meteorological model that combines meteorological data and geophysical data to generate a wind field. CALPUFF model then combines the information provided by CALMET and source data to predict concentration, deposition flux, visibility impairment, etc., at each receptor for specified averaging time. CALPOST is a post-processor for the model. CALPUFF can accommodate point, volume, and area source emissions. CALPUFF can use the three dimensional meteorological fields developed by the CALMET model or the meteorological files used by the ISCST3. CALPUFF contains algorithms for near-source effects such as building downwash, transitional plume rise, partial plume penetration, sub-grid 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. Diosey et al. (2000) compared the ISCST3 and CALPUFF models for odour emission from a wastewater treatment plant with area and point sources, and found that, in general, the two models give results in the same range. However, there were cases when the impacts predicted by the CALPUFF modeling system were either significantly lower or higher. Bedogni et al. (2004) applied CALPUFF model in investigating the odour impact of a solid waste landfill in Italy and satisfactory results had been gain by comparing the measured and predicted concentration of methane which was used as an odour indicator. McPhail (1991) stated that puff models should be used to predict agricultural odours because odour moves as a series of puffs rather than flowing as a continuous stream. Gassman (1993) stated that puff models might yield higher odour perception since the cycling back and forth of the odour presence actually increase odour to an observer, and as a result, the threshold and peak concentrations will greatly exceed the average concentrations for a short time period. Gaussian dispersion models are unable to account for this puff phenomenon and the corresponding peak concentrations (Gassman, 1993). Zhu (1999) used INPUFF2 model to theoretically evaluate the influence of stability class on downwind odour concentration. The downwind odour concentrations at different distances from the odour source were investigated with respect to different stability classes and wind speed.

12

According to his study, the unstable and neutral stability categories tended to govern the odour levels within 200 m range, while the stable categories yielded higher odor levels beyond that range. The odour plume width varies with the stability classes. He also pointed out that Gaussian models can not predict odour for distances less than 100 m from the source and stability class E and F are not suitable for use in Gaussian models to predict agricultural odour dispersions. Field data had been used to evaluate the INPUFF2 model for predicting downwind odours from animal production facilities (Zhu, et al., 2000a). Results showed that INPUFF2 model can predict downwind odour concentrations generated from animal production facilities within 300 m satisfactorily. At further distance, the accuracy of prediction by the model is significantly reduced. It was found that INPUFF2 cannot be directly applied for odor dispersion because the results obtained were much lower than the field measured results. Scaling factors of 35 for animal building sources and 10 for surface sources such as manure storage unit are used to amplify the odor emission rates by this study to adjust the model predicted concentration to the same numerical range as the filed odour measurement (Zhu, et al., 2000a). Following this work, Guo et al. (2001) calibrated the INPUFF2 model using odor monitoring data by resident odor observers for long-distance (up to 4.8 km) from animal production sites using the same scaling factors. Comparison between the modeled and measured odour intensity indicated that the model successfully estimated odour intensity 1 (faint odour) traveling up to 3.2 km under stable atmospheric conditions. However, the model underestimated moderate to strong or very strong odours and odours that occurred during neutral and unstable weather. The results indicate that the model could serve as an effective tool for agriculture odour dispersion estimation from animal production sources. Based on these two studies, OFFSET model, a setback distance model for determination of odour-annoyance setback distances form animal production sites, was developed based on the INPUFF2 and the scaling factors (Jacobson et al., 2005; Guo et al. 2005a). Meandering/Fluctuating Model It is well known that changes in meteorological conditions also change air pollution levels. But even under apparently constant meteorological conditions still fast and random concentration fluctuations exist due to the turbulent motion of the atmosphere. The currently existing dispersion models can estimate reasonable accurate one-hour average concentrations for regulatory use. Models describing concentrations on shorter averaging times are needed for assessing odour problem that can be caused by a series of seconds’ detectable exposure even though the hourly mean concentration is below the detection threshold. The first model that considers explicitly concentration fluctuations related to plume meandering was developed by Gifford (1959). The model is based on the hypothesis that the absolute or total plume dispersion can be separated into two components, the spreading component that represents the relative dispersion about the center of gravity or local axis of the instantaneous plume, and the meandering component representing the variance of the displacement of the fluctuating plume axis from the fixed axis of the steady plume. Mathematically, it can be expressed as (Arya, 1999):

222ymyiy σσσ +=

222zmziz σσσ += [9]

where: the subscript i denotes instantaneous or short term-averaged plume and m denotes the meandering component.

13

Gifford (1959) stated that it is reasonable to assume that the concentration distribution in the instantaneous plume is also Gaussian, that is:

⎥⎥⎦

⎤

⎢⎢⎣

⎡ −−

−−= 2

2

2

2

2)(

2)(exp

2 zi

m

yi

m

ziyi

zzyyu

Qcσσσσπ

[10]

This implies that instantaneous concentration in the fluctuating plume model is a random function of time, as a consequence of the random variability of ym and zm. ym and zm are also assumed to have Gaussian probability density functions. When the model was proposed by Gifford (1959), some statistical properties (mean, variance, and frequency distribution) of this model for a point concentration had been described. Some qualitatively comparisons with measured concentration data had been made to confirm the prediction of the fluctuating model. Data on peak to mean ratio had also been shown to have the expected behaviors from fluctuating model. Comparison of variance of point concentration as a function of distance between observed data and the model calculation had certain agreement. They were in the same order of magnitude and the largest values occur in agreement with model calculation. Although, more observed data of concentration should be used to verify and validate this model, the theory of fluctuating model itself is very attractive to odour study because of its ability of predicting instantaneous concentration. Högström model (Högström, 1972) was proposed to predict odour with the capability of describing plume meandering based on the principles proposed by Gifford (1959). The model yields estimates of the odour frequency at a certain point downwind of the source. The odour frequency is defined as the percentage of the time within the observation period that the odour exceeds the detection threshold concentration. The main idea of Högström’s model is the determination of the weighted odorous width of the plume at ground level and its variation due to the meandering of the plume. The odorous width at a certain distance from the source is that part of the instantaneous plume within which the odour threshold is exceeded at ground level. The odorous width can be calculated by the fluctuating plume model for certain threshold concentration, then the frequencies of the odour threshold are calculated based on this odorous width. From the comparison between calculated data and observed data from two field odour experiments, Högström found that it was possible to make quite realistic estimates of odour frequencies near a point source with this method. In the distance range of 2 to 5 km from the source no systematic difference is found between the calculated and the observed frequencies. For the distance range of 10 to 20 km the calculated frequencies are significantly lower than the observed ones, but it was likely that even the results obtained for that distance may be acceptable for many practical purposes. However, the ability of this model to estimate the odour frequencies for different strength of odour hadn’t been tested, and more field experiment work should be done in order to validate this method for more application. A modified Högström model was described by de Bree and Harssema (1987), the computation proceeded by calculating the probability-weighted width of the detectable odour plume at ground level due to vertical meanderings, and integrating the predicted transverse distribution of the plume centreline over the interval of the weighted width, centred laterally at the receptor location. A field odour measurement was conducted by eight trained panel members who were placed at a line downwind of the source, perpendicular to the expected wind direction. The panellist made instantaneous observations of the detectability of the odour every 10 seconds during 10 minutes. Over all, the calculated and measured odour frequencies were strongly related (R = 0.90) under unstable and near neutral atmospheric conditions However, this modified Högström model was tested for relatively short traveling times only (10 minutes), for large distances the applicability

14

was not tested and theoretically assumed to be low due to the less effects of the plume meandering on the concentration fluctuation. Thierry and Christophe (1997) discussed the micrometeorology of the odour transport and stated that the approach of combining the Gifford meandering model and the Gaussian model is quite interesting because it considers the evolution of concentration fluctuations and can be used with classically available/standard hourly meteorological data. A fluctuating plume dispersion model based on the theory proposed by Gifford (1959) had been developed by Mussio et al. (2001) to facilitate the prediction of odor-impact frequencies in the communities surrounding elevated point sources. This model was tested by the field odor data collected from the residential areas surrounding the paint shop of an automotive assembly plant. The field odour measurements were conducted by observers every 12 seconds. Results showed that the simulation of the total frequency of occurrence where the odour was frequent (i.e. readily detectable more than 30% of the time) was good because the ratio of modeled-to-observed frequencies approached the ideal value of one as the magnitude of the odour frequency increased. At low frequencies of occurrence the model prediction was poor. The stability class did not seem to affect the model’s ability to predict filed frequency values. Meanwhile, the model provided good predictions of the maximum odour levels without being sensitive to either stability class or distance from the source. However the testing field odour data was very limit, further calibration of the model is necessary to increase its accuracy and widen its applicability, especially for the long distance and wider variety of atmospheric conditions. De Melo Lisboa et al. (2005) developed a model based on the theory established by Högström (1972) on odor dispersion of puff emissions to estimate the odour impact. Nine approaches that explore several solutions within the Gaussian domain for the atmospheric dispersion problem are proposed in software named ODODIS ( ODOur DISpersion software). This model needs more validations with the field data, although a comparison with an existing database (the Prairie Grass database), which is a set of field experiments carried out in 1956 and was an important database for testing the Gaussian models, showed good agreement. Other Models There are some other models that have been used in odour dispersion prediction, such as numerical models. The characteristic of numerical models is that the area to be examined is divided into individual cells. Instead of a closed solution, the model equations are solved for each cell and time interval. With regard to the numerical models, a distinction is made between Euler model and Lagrangian model. Boeker et al. (2000) used MISKM and NaSt3D models as representatives of Euler and Lagrange models respectively to discuss the possibilities and limitations of such models. The modification and new development regarding NaSt3D can carry out the time-resolved simulation of odorant dispersion. An Eulerian-Lagrangian model was used by Schiffman et al. (2005) to simulate the dispersion of odour from a confined animal feeding operation (CAFO) under different meteorological conditions. The predicted odour dispersion distance was found to be greater at night-time than during daytime and was consistent with field reports from individuals living near the CAFO. This model needs more validation to ensure its application in different sources and varied meteorological conditions. A Lagrangian particle model-AUSTAL2000G has been developed in Germany as a regulatory dispersion model to deal with odour dispersion problems (VDI 2000). However, there is little information on validation and evaluation of this model by field measurement. Another Lagrangian particle model-WinTrax, developed by Thunderbeach Scientific (Nanaimo, BC, Canada) has been evaluated by Zhou et al. (2005). They concluded that the WindTrax dispersion model can predict odour concentrations

15

with good agreement for distances of 500 m and 1000 m. Bjerg et al. (2004) conducted a study of using CFD (Computational Fluid Dynamics) model to investigate the possibilities to reduce odour concentrations by optimising the location and design of exhausts. The CFD model was validated against full scale tracer gas (SF6) measurements around a commercial growing-finishing pig building and showed that it was a suitable technique to predict the spreading of exhausted air 50 to 150 m from a livestock building.

Issues on Adapting and Validating Of Odour Dispersion Models When either adapting existing industrial air dispersion model for odor dispersion modeling or using dispersion models developed specifically for odors, several most important issues have to be addresses, including the difference of odor and specific air contaminant/gas, the instantaneous nature of odor, and the relationship between odor concentration and intensities in validating of odor dispersion models. First, in evaluation of the accuracy of odor dispersion models, the ultimate method is to compare the model predictions with the field plume measurement data. As discussed previously, field odor plume is measured by estimating odor intensities at desired locations in the odor plume. To compare the field measured odor intensity with the model predicted odor concentration, the relationship between the odor concentration and intensity has to be known. Second, due to the inherent difference between odor and gas measurements, directly applying industrial air dispersion models for odor dispersion has been proved incorrect and inappropriate (Zhu et al. 2000a; Guo et al. 2001; Zhang et al. 2005). Measures have to be taken to make the industrial air dispersion model usable for livestock odor dispersion. Third, odor plume measurement interval is between 10 s to 1 min and duration of each plume measurement is generally 10 min whereas most air dispersion models use standard meteorological data with interval of 1 hr. Researchers have taken the average meteorological data during that 10 min period as 1 h values in the odor dispersion modeling and then compare the “hourly average” predictions with the 10 min average measured odor concentration (Zhu et al. 2000a; Zhou et al. 2005; Xing, 2006). Under the same meteorological condition, this will cause over prediction due to the longer odor transportation period, especially for puff models. The second issue may be dealt with by using the “black box” concept and combined with the third issue by applying appropriate scaling factors or peak-to-mean ratios to adjust the emission rate or the modeled odor concentration. The third issue of time interval may be dealt with individually by peak to mean ratios that transform the long term average concentration to short term concentration. Adapting Air Dispersion Models For Odor Dispersion As discussed previously, Zhu et al. (2000a) found that INPUFF2 cannot be directly applied for odor dispersion because the results obtained were much lower than the field measured results and scaling factors have to be used to amplify the odor emission rates in order to adjust the model predicted concentration to the same numerical range as the filed odour measurement. Guo et al. (2001) confirmed this finding using downwind odor measurement data from trained resident sniffers for long distance odour dispersion from animal production sites. Koppplu et al. (2004) reported that scaling factors in the range of 0.2 to 3900 may be needed to adjust AERMOD predictions to short-term odor measurements depending on the source type (point, area, volume) and the type of facility being modeled after comparison of measured odor intensities from livestock facilities to predicted ambient odor levels from AERMOD. Zhou et al. (2005) calibrated four air dispersion models, ISCST3, AUSPLUME, INPUFF2, and WindTrax using odor plume measurement data 100 to 1000 m from two swine farms. They concluded that these four models performed similarly and predicted downwind odor concentrations with good agreement with field measured results. Considering that 58.3% the measured odor

16

concentrations were zero, this set data was re-examined by Xing et al. (2006) for ISCST3, AUSPLUME, INPUFF2, and CALPUFF models. It was found that although the agreement between the model predictions and measured odor intensities was between 37% and 50% for the four models considering all the measurements, however, if the measurements with intensity zero (no odour) were excluded, the agreement reduced to between 28 and 35%. No scaling factors were used by Zhou et al. (2005) and Xing et al. (2006). Xing et al. (2006) found the scaling factors could not improve the models’ performances significantly because some model predictions were lower than measured values and some were higher, which is different than the findings by Zhu et al. (2000a) and Guo et al. (2001) that the model predictions of INPUFF2 were always lower than the measured values which made the scaling factors useful. The main reason for the difference between these studies might be that the odor intensity and concentration conversion equations are very different from each other (Xing et al. 2006). The odour intensity and concentration conversion equation is very important to ensure the accuracy of the comparison of the modeled and measured odour intensities as well as the effectiveness of improving model performance using scaling factors. This will be further discussed later. Peak To Mean Ratio Filed odour intensities are measured at 10 s to 2 min interval rather than 1 hour interval as used in most dispersion models. This will reveal the actual odour nuisance caused by short term odours that excess the odour detection threshold. However, most of the Gaussian models consider an average concentration for a time period ranging from 10 min to 1 h, according to the level of standard deviation in the concentrations that are used (Beaman, 1998). This implies that the short term fluctuations are ignored, as schematically shown in Fig.2 (Högström, 1972). The irregular curve presents the odorous gas concentration downwind of a chimney. The figure shows that after 1h, a concentration above the odour threshold has been detected several times, despite the fact that the hourly mean concentration is lower than this index.

Fig. 1. The instantaneous concentrations and hourly mean concentration (Högström, 1972) The peak-to-mean ratio can be used as a correction factor to make up this disadvantage of dispersion model and estimate the maximum values in the field. Smith (1973) gives the following relationship to transform the regulatory model calculated half hour mean concentrations to the momentary odour concentration which sensation of odour depends on.

u

p

m

m

p

tt

CC

⎟⎟⎠

⎞⎜⎜⎝

⎛= [11]

With the mean concentration calculated for an integration time of and the peak concentration for a integration time of . u is the stability dependant power law exponent. Smith (1973) suggested the following values of the exponent u depending on the stability of the

mC mt

pC pt

17

atmosphere: 0.65 (SC=2), 0.52 (SC=3) and 0.35 (SC=4) (Schaugerger et al. 2000). However, the exponent from Duffee, O‘Brien and Ostojic (1991) was: 0.5 (SC=1 0r 2), 0.33 (SC=3), 0.20 (SC=4) and 0.167 (SC = 5 or 6). Mahin (1997,1998) stated that there was no agreement on the appropriate power law exponent (u) for different stability class. One approach to convert averaging times to a shorter time is to assume a power exponent of 0.2 for all stability classes. An odour dispersion model (AODM) developed by Schaugerger et al. (2000, 2001) transformed the half-hour average concentrations calculated by a Gaussian dispersion model to instantaneous values by an attenuation function decreasing the peak-to-mean ratio with increasing wind velocity, stability, and distance from the source. The peak-to-mean ratio in equation [11] was modified by an exponential attenuation function of T/tL:

)7317.0exp()1(1 0Lt

T−−Ψ+=Ψ [12]

where is the peak-to-mean factor calculated in equation [11]; T=x/u is the time of travel with the distance x and the mean wind speed u, and t

0ΨL is Lagrangian time scale.

The Lagrangian time scale is equal to Lt εσ where )(31 222

wvu σσσσ ++= is the variance of

the wind speed as the mean of the three wind componentsu , v and , respectively, and w ε is the rate of dissipation of turbulent energy using the following approximation:

3)3.1

(1 w

KZσ

ε = [13]

where k=0.4 is the von Karman constant and Z = 2m is the height of receptor, the human nose. The peak-to-mean ratio calculated by this method was considered to be reasonable (Schaugerger et al. 2000). Relationship Between Odour Concentration And Intensity Most of the odour dispersion models can predict odour concentrations downwind the sources, while the odour intensities are measured in the field plume measurement. This results in another problem to be solved in order to validate odour dispersion models, i.e., the odour detection threshold needs to be converted to the odour intensity in order to compare the field odour plume measurement to the result calculated by an air dispersion model. There are three kinds of relationship between odour intensity and odour concentration that have been found by researchers including Weber-Fechner law, Stevens power law and Baidler model. Fechner (1860) formulated the principle that the intensity of a sensation increases as the logarithmic of the concentration increase ,i.e. the Weber-Fechner law:

2101 )(log ww KCKI += [14] where I is the increment scale number of the intensity, C is the detection threshold concentration, and Kw1 and Kw2 are constants (Misselbrook et al., 1993). Stevens (1957) proposed that the relationship between concentration and intensity is a power function expressed as:

[15] 21

sKs CKI =

where Ks1 and Ks2 are constants. Wegenaar (1975) concluded that the Weber-Fechner model and Stevens power model produce practically indistinguishable predictions. Beidler (1954) proposed a fundamental taste equation to account for the relationship of neurophysiological responses from tastes receptors.

18

CKCKK

B

BBI2

211+= [16]

where KB1 and KB2 are constants. Cain and Moskowiz (1974) and Chen et al., (1999) applied this model to odour evaluation. Although all the researchers have agreements on the presence of the relationship between odour intensity and concentration, the best fit model is different and the constants in each model are different among the researchers. Misselbrook et al., (1993) related the odor detection threshold and odour intensity for emissions following land spreading of pig slurry and also emissions from broiler houses. The Weber-Fechner logarithmic model was applied to the both emissions and the relationships between odour concentration and odour intensity obtained were expressed as:

45.0)(log61.1 10 += CI for odours from pig slurries [17] 30.0)(log35.2 10 += CI for broiler house odours [18]

Chen et al., (1999) compared intensity and threshold from four different swine facilities (gestation, farrowing, nursery, and finishing). They concluded that the widely used Weber-Fechner model did not adequately fit the data as well as the Stevens power model. Nicolai et al., (2000) investigated the Weber-Fechner, Stevens, and Beidler models to determine which best described the relationship between detection threshold and intensity. In their study odour detection threshold and intensity were measured from swine buildings and manure storages. They found that the Weber- Fechner logarithmic model provided the best form to describe the relationship between odour detection threshold and intensity from the swine buildings:

[19] 424.0)(log57.1 10 −= CIFor odour from manure storage, all three models were similar with the Weber- Fechner model indicating a slightly better fit:

519.0)(log61.1 10 −= CI [20] Zhang et al., (2003) investigated the odour data collected on four swine farms to determine the relationship between odour intensity and concentration. Odour intensity of bagged samples measured in laboratory correlated well with the odour concentration measured with olfactometers and the relationship could be adequately predicted by the Weber-Fechner model and the Stevens model. The relationship they found can be expressed as:

36.0)(log82.0 10 += CI [21] Odour samples collected in Tedlar bags from swine farms and manure storages were measured in the Olfactometry lab for both odour intensity and concentration in order to establish the relationship between odour intensity and concentration. Zhang et al. (2005) from the University of Manitoba and Feddes et al. (2005) from the University of Alberta obtained different relationships from the same odour data. The conversion equation from University of Manitoba takes the form of

78.0)(log43.1 10 += CI [22]

while the University of Alberta got the equation as

046.0)(log245.1 10 −= CI [23]

19

The concentrations for different intensity scales of these two equations were listed in table 1.When generating the equation [22] by the University of Manitoba, only 20 odour samples were collected in Tedlar bags from the farms and presented to trained human panel for odour intensity and odour concentration measurement in the olfactometory lab. However, there were over 100 odour samples used to generate the equation [23] by the University of Alberta. From this point of view, equation [23] is more reliable than equation [22]. Furthermore, the equation [22] is not reasonable because intensities 1 to 3 have virtually the same odour concentrations as seen in table 1. There are also two problems with the equation [23]: a) odour concentrations for high intensities levels seem too low, b) low intensity levels have too low odour concentrations. For example, odour concentration of 640 OU/m3 is at the moderate or low end of odour concentrations measured in swine barns and manure storages in warm seasons, and may not be considered strong comparing with odour measured in the manure storage or from the barns in winter.

Since both equation [22] and [23] are based on the same scale of n-butanol concentration-in-water, Xing et al. (2006) also used equation [23] (Feddes 2005) to evaluate four air dispersion models. For all measurements without zero odors, the agreement between measured and model predicted odor intensity was between 13 to 23%. If all measurements included, the agreement was between 56 and 62%. Hence, the agreement was not improved by using equation [23]. Guo et al. (2001), the University of Minnesota, also obtained the relationships between odour concentration and intensity on a 5 point n-butanol scale. This is based on odor intensity and concentration measurements of 124 odor samples collected from 60 swine buildings and 66 swine manure storage facilities, and 55 odor samples collected at 10 dairy and beef farms in Minnesota during 1998 and 1999. The Weber-Fechner model was the best fit for both swine and cattle data. The relationships between odor intensity on the 0 to 5 scale and concentration are expressed as:

97.1)(log928.0 10 −= CI for swine odours [24] 068.2)(log922.0 10 −= CI for cattle odours [25]

This conversion relationship for swine odours (equation [24]) is very different from the equations [22] and [23]. This results from the different intensity interpretation for the same n-butanol concentration-in-water for different odour scale as shown in table 1. For example, intensity 1 on this scale is perceived as very faint odour and it is equivalent to intensity 2 for n-butanol concentration-in-water on the 8-point scale, but its swine odor concentration 25 OU/m3 is equivalent to intensity 4 on the 8-point scale represented by equations [22] and [23]. Regarding swine odor concentration, intensity 2 on the 5-point scale is between intensities 5 and 6 in equation [24]; intensity 3 is between intensities 6 and 7 in equation [24], and intensity 4 is equivalent to intensity 8 in equation [24]. The constants in Weber-Fechner models and odour scales used by different researchers were summarized in table 2. The odour intensity and concentration conversion equation is very important to ensure the accuracy of the comparison of the modeled and measured odour intensities as well as the effectiveness of improving model performance using scaling factors.

Table 2. Constants in weber-Fechner models and odour scales used by different researchers

Constants Odour Source

Kw1 Kw2Odour Scale Reference

Pig slurries 1.61 0.45 0-6 Misselbrook et al., 1993Broiler house 2.35 0.30 0-6 Misselbrook et al., 1993

20

Swine building 1.57 -0.424 0-5 Nicolai et al., 2000 Swine manure storage 1.61 -0.519 0-5 Nicolai et al., 2000

Swine farms 0.82 0.36 0-8 Zhang et al., 2003 Swine farms and manure

storages 1.43 0.78 0-8 Feddes et al., 2005

Swine farms and manure storages 1.245 -0.046 0-8 Zhang et al., 2005

Swine buildings and manure storage facilities 0.928 -1.97 0-5 Guo et al., 2001

Dairy and beef farms 0.922 -2.068 0-5 Guo et al. 2001

Research Gaps in Odour Dispersion Modeling

Although many air dispersion models have been used in predicting downwind odour concentrations from agricultural sources, the model performance has not been satisfactory. Several important factors which may forfeit the use of air dispersion models in the agricultural area appeared ignored. Differences Between Agricultural And Industrial Air Dispersion Mejer and Krause (1985) compared the industrial and agricultural air pollution in the process of emission, transmission and immission. They pointed out that the large scale values of industrial air pollutions, on which the established dispersion models based, are too different from those in agriculture. Normally, the industrial gas can travel the distance farther than 10 km, sometimes even farther than 100 km. However livestock odour can only travel less than 8-10 km distance. The interested dispersion distance for odour is much less than gas. Therefore, the attempt to apply the established dispersion models to agricultural emission sources may lead to unreasonable results. Smith (1993) pointed out that the emphasis in Gaussian models is the prediction of dispersion over relatively large distances due to strong industrial pollutants. For agricultural odour sources, some features that differ from sources of industrial pollutants have to be taken into account. These features may include: (1) the odour source is at or near ground level; (2) there is insignificant plume rise due to the vertical momentum or lower intensity of a mass flow of warm gas;(3) the source may be of relatively large areal extent (such as aerobic lagoons); (4) the important receptor zone may be relatively close to the source of emissions; (5) the difficulty in measuring the odor emission rate; (6) the spatial and temporal variability in emission rates; and (7) the relatively low intensity of emissions.

Inherent Problems With Gaussian Model There are also a number of assumptions that come with the Gaussian model (Turner, 1994) such as constant emission rate and meteorological condition, conservation of pollutants and normal distribution of concentration. These assumptions plus the above special situations raised by Mejer and Krause (1985) and Smith (1993) undoubtedly challenge the concept of using Gaussian models in the agricultural field. Since Pasquill-Gifford dispersion parameters were developed based on concentrations that were taken over smaller time increments, Fritz et al. (1997) stated that many models developed by US EPA could be misapplied to the estimation of odours downwind for one-hour or longer time

21

period when using Pasquill-Gifford dispersion parameters. Some dispersion coefficients for Gaussian models are for long distance, e.g. BNL (The Brookhaven National Laboratory) (Singer and Smith, 1966) and TVA (The Tennessee Valley Authority) (Carpenter et al., 1971) dispersion parameters were obtained from the diffusion data of high stack release and long distance measurement of gas concentration. These may not suitable for short distance odour dispersion modeling. Using four air dispersion models for livestock odor dispersion simulation, Xing et al. (2006) found that livestock odor can travel much farther under steady state meteorological conditions than using variable hourly annual meteorological data and concluded different odor concentration criterion should be used for setback distance determination for using steady-state or variable hourly meteorological data. Guo et al. (2005b, 2006) studied the impact of weather conditions including wind speed and atmospheric stability on odour occurrence with the data reported by the resident observers living within 8.6 km from three intensive swine farms in eastern Saskatchewan, Canada. From their study, most odour events (61.7%) were detected under neutral atmospheric stability class D while only 15% were detected under stable atmospheric conditions. Stable atmospheric conditions occurred the least in the period from May to August, yet this period had the highest number of odour events. They concluded that atmospheric stability class has little effect on odour occurrences in the vicinity of swine farms. This finding is important because it would be contrary to the basic air dispersion principle that stable weather stations would allow air to travel for farther distances than unstable conditions. This will result in two problems a) the P-G stability class can not reflect the impact of weather stability conditions on odour dispersion at close distance from the source; b) the basic air dispersion principle is not suitable to odour dispersion for short distance within 8 km. Thus, in the future study of odour dispersion modeling, the stability class should be further evaluated. There are some models (such as AERMOD and ADMS) that treat PBL properties in a different way from Pasquill-Gifford stability class that had been used in many traditional models like ISCST and AUSPLUME. ADMS described the boundary layers by two parameters: the boundary layer depth and Monin-Obukhvo length (CERC, 2004). AERMOD represented turbulence based on PBL similarity theory and defined the stability by heat flux, Monin-Obukhov leangth, and mixing height (USEPA, 2004). The future development of odour dispersion model may consider these methods for treatment of boundary layer. Drawbacks Of Existing Models A main drawback of the Gaussian model is that they can only predict average concentrations for the time scale as 10-60 min, which may also be responsible for its inaccurate for short distance odor dispersion from livestock sources. Peaks of concentrations are then neglected. Although the peak to mean ratio or scaling factor can partly solve this problem, the proper averaging time of odour is not clearly defined. Some puff models can estimate the short term odour concentration downwind the sources (such as INPUFF model), however, the input meteorological data should be at the same time scale, which is difficult or expensive to obtain comparing to the regular 1-hr meteorological data. The numerical model may be a good approach for odour dispersion modeling, but it is complex and inefficient. The fluctuating model seems to be the most suitable models for odor dispersion, because they consider specifically the evolution of concentration fluctuations and can be used with standard meteorological data. Unfortunately, limited research has been done regarding such models, and the models are not validated to be applied for livestock operation sources. Beside this, odour concentration is measured using a totally different method from gas, thus, the air dispersion models are not applicable for odour dispersion predictions. To validate odour

22

dispersion model using field odour plume measurement data, accurate relationship between odour intensity and concentration is needed to convert the modeled concentration to odor intensity. In summary, 1) the Gaussian model is inaccurate for odour dispersion modeling due to the difference between the industrial sources and agricultural sources and the discrepancy between the original assumptions and the actual situations of the agricultural sources. 2) Stability class should be used in the odour dispersion models with more cautions 3) Gaussian models predict the mean concentration and ignore the fluctuation of concentration. 4) Some dispersion parameters are not suitable for odour dispersion model due to the time and distance scales. 5) Some puff models calculate instantaneous concentration only with instantaneous input. 6) Numerical models are complex, inefficient and demanding high request of computers. 7) The fluctuating models are seldom studied and not validated to be applied for livestock operation sources. 8) The air dispersion models are not applicable for odour dispersion predictions due to the different measurement between gas and odour. And 9) the conversion of odour intensity and concentration has to be considered by the model in order to validate it by field odour plume measurement data.

Conclusions and Recommendations Livestock odours come from three primary sources: manure storage unit, animal housing, and land application of manure. There are more than 200 compounds in the mixture of livestock odour. Odour concentration and odour intensity are the two most widely used measurement of livestock odour. Three main methods have been used to measure odour concentration and intensity including sensory methods, measurement of specific odorous gases and electronic nose. The sensory methods using olfactometry are most widely used to evaluate odour concentration. Odour emission rate can be determined by odour concentration and airflow rate. Many commercial Gaussian models including Gaussian plume model (ISCST3, ADM3), Gaussian puff model (INPUFF2 and CALPUFF) that have been originally designed for industrial gas dispersion have been used in odour dispersion predicting, while most researchers simplified and modified the Gaussian models to predict the odour dispersion from livestock operations. A fluctuating model that can predict the instantaneous odour concentration with hourly meteorological input data seems to be suitable in odour dispersion modeling. In order to test and validating the odour dispersion model, the field odour plume measurement should be taken. The field odour plume is measured by intensity rather than concentration. Thus the conversion between odour intensity and odour concentration is critical in validating odour dispersion model. The differences between industrial sources and livestock sources and the inherent problems of Gaussian model can question the application of Gaussian models in livestock odour dispersion modeling. Furthermore, the ability of Gaussian models that can only predict mean concentration and the differences of the measurement ways between gas and odour require the model that can accurately predict livestock odour dispersion to be carried out. Due to the research gaps in odour dispersion modeling, a model that can accurately predict the livestock odour dispersion is urgent. The model for the livestock odour should be designed for the near ground sources with large area, insignificant plume rise and variable emission rates, etc and for the interest short distance that the livestock odour can travel. The proper dispersion coefficients that are suitable to livestock odour should be found out, and the P-G stability class should be used in caution according to Guo et al. (2005b, 2006). A fluctuating model that can deal with the multiple sources and variable odour emissions is a good suggestion for livestock modeling in view of its ability to predict instantaneous odour concentration (a serial of seconds)

23

that can cause odour nuisance. And the conversion between odour concentration and intensity should be included in this model.

References

Arya, S. 1999. Air pollution meteorology and dispersion. Oxford University Press, New York. ASCE. 1995. odour control in wastewater treatment plants. American Society of Civil Engineers,

Report 82. New York. ASTM. 1998. E 544-88 Standard practices for referencing suprathreshold odour intensity.

American Society for Testing Materials, Philadelphia, PA. Beaman, AL. 1988. A novel approach to estimating the odour concentration distribution in the

community. Atmos Eviron, 22:561-567. Bedogni M., Sergio R. 2004. An applied methodology to evaluate odour impact. 9th int. Conf. on

Harmonisation within Atmospheric Dispersion modeling for Regulatory Purposes. Beidler, LM. 1954.A theory of taste stimulation. J Gen Physiol. 38(2):133–139 Bjerg B., P. Kai, S. Morsing, and H. Takai. 2004 .“CFD Analysis to Predict Close Range

Spreading of Ventilation Air from Livestock Buildings”. Agricultural Engineering International: the CIGR Journal of Scientific Research and Development. Manuscript BC 03 014. Vol. VI. August, 2004.

Boeker, P., O. Wallenfang, F. Koster, R. Croce, B. Diekmann, M Griebel and P.S. Lammers.

2000. The modeling of odour dispersion with time-resolved models. Agrartechnische Forschung, 6 heft 4, S.E 84- E 89.

Cain, W.S., and H.R. Moskowitz. 1974. Psychophysical scaling of odour. In Human Responses

to environmental odours. A Tuck, J.W. Hohnston, Jr., and D.G. Moulton, eds. Academic Press, NY.

Carney, P.G. and V.A. Dodd. 1989. A comparison between predicted and measured values for

the dispersion of malodours from slurry. J. agric. Engng Res. 44:67-76. Carpenter, S.B., Montgomery, T.L., Leavitt, J.M., Colbaugh, W.C., and Thomas, F.W. 1971.

Principle plume dispersion models: TVA power plant. J. Air Pollut. Cont. Assoc., 21: 491-495.

CEN. 2002. EN13725. Determination of odour concentration by dynamic olfactometry. European

Committee for Standardization. CEN Central Secretariat: rue de Stassart, 36 B-1050 Brussels.

CERC.2004. ADMS: User Guide. Version 3.2, July 2004. Cambridge Environmental Research

Consultants Ltd., 3 King’s Parade, Cambridge, CB2 1SJ, UK. Chen, Y., D.S. Bundy, S.J. Hoff. 1999. Development of the relationship between odour intensity

and threshold dilution ratio for swine units. J. of the Air & Waste Management Association. 49: 1082-1088

24

Curran, T.P., Sheridan, B.A., Hayes, E.T., Dodd, V.A., 2002. Comparison of two dispersion

models for odour nuisance prediction from pig units. ASAE/CIGR Annual International Meeting, Chicago, Paper No. 024014, July 28-31, 2002.

Curtis, S.E, 1983. Environmental management in animal agriculture. Iowa State University

Press, Ames, IA. De Bree, F., H. Harssema. 1987. Field evaluation of a fluctuating plume model for odours with

sniffing teams. In: Grefen, K., Löbel, J.(Eds.), Environmental Meteorology, Proceedings of an International Symposium in Wurzburg, FRG. Kluwer Academic Publishers, Dordrecht, Netherlands. 473-486.

De Melo Lisboa, H., J.M. Guillot, J.L. Fanlo, P.L. Cloirec. 2005. Dispersion of odorous gases in

the atmosphere—Part I: Modeling approaches to the phenomenon. Science of the Total Environment.

Diosey, P.G., E.H. Maureen, J.P. Richard. 2000. Modeling odour and VOCs: a comparison of

models and modeling techniques. Water Environment Federation Specialty Conference: Odors and VOC Emissions 2000, Cincinnati,OH.16-18 April 2000.

Donham, K.J., W.J. Popendorf. 1985. Ambient levels of selected gases inside swine

confinement buildings. Am Ind Hyg Assoc J. 46(11): 658-61. Duffee, O’Brien and Ostojic, 1991,'Odor Modeling - Why and How', Recent Developments and

Current Practices in Odor Regulations, Controls and Technology, Air & Waste Management Association, 1991.

Engel, P.L.. 1997. Utilizing ISCST to model composting facility odours. Presentation at the Air &

Waste Management Association’s 90th Annual Meeting & Exhibition, June 8-13, 1997, Toronto, Ontario, Canada.

EPA. 2004. Guideline on Air Quality Models. Appendix W to Part 51, 467. Umited States

Environmental Protection Agency. http://www.epa.gov/scram001/guidance/guide/ appw_03.pdf (2005/10/12)

Fechner, G.T. 1860. Elemente der psychophsik. Leipzig: Breitkopf and Hartel. English

translation of Vol 1 by Helmut E. Adler. 1966. Elements of psychophysics. (D. H. Howes and E.G.Boring, Eds.). Holt, Rinehart and Winston, New York.

Fritz, B.K., G.W. Zwicke, B.W. Shaw, and C.B. Parnell. 1997. A re-examination of the

assumptions and parameters of the Gaussian dispersion model. ASAE Paper #:974098. St. Joseph, MI. USA.

Gassman, P.W. 1993. Simulation of odour transport: a review. ASAE paper#: 924517. St.

Joseph, MI. USA. Gifford, F.A. 1959. Statistical properties of a fluctuating plume dispersion model. Advances in

Geophysics,6:117-137. Guo, H., L.D. Jacobson, D.R. Schmidt, and R.E. Nicolai. 2001. Calibrating INPUFF-2 model by

25

resident-panelists for long-distance odour dispersion from animal production sites. Applied Eng. In Agric. 17(6): 859-868

Guo, H., L.D. Jacobson, D.R. Schmidt and R.E. Nicolai. 2003. Evaluation of the influence of

atmospheric conditions on odour dispersion from animal production sites. Transactions of the ASAE 46(2): 461-466.

Guo, H., L.D. Jacobson, D.R. Schmidt, R.E. Nicolai, J. Zhu, K.A. Janni. 2005a. Development of

the OFFSET model for determination of odour-annoyance-free setback distances from animal production sites: Part II. Model development and evaluations. Transactions of the ASAE, 48(6).

Guo, H., W. Dehod, J. Feddes and C. Laguë. 2004b. Odour monitoring by trained resident-

observers in the neighbouring area of swine production operations. ASAE Paper No. 044122. St. Joseph, MI: ASABE.

Guo, H. , W. Dehod, J. Feddes, C. Laguë and I. Edeogu. 2005b. Monitoring odour occurrence in

the vicinity of swine farms by resident observers - Part I: Odour occurrence profiles, Canadian biosystems engineering. 47: 57-65

Guo, H. , W. Dehod, J. Feddes, C. Laguë and I. Edeogu. 2006. Monitoring odour occurrence in

the vicinity of swine farms by resident observers- Part II: Impact of weather conditions on odour occurrence. Canadian biosystems engineering. 48: 23-29

Heber, A.J., D.S. Bundy, T.T. Lim, J. Ni, B.L. Haymore, C.A. Diehl and R.K. Duggirala. 1998.

Odour emission rates from swine finishing buildings. In: Animal Production Systems and the Environment: An International Conference on Odor, Water Quality, Nutrient Management and Socio-economic Issues. Des Moines, Iowa, USA, July 19-22, pp. 305- 310.

Hartung, E., and T. Jungbluth, 1997. Determination of the odour plume boundaries from animal

houses. In Proc. 5th Int. Symp. on Livestock Environment, Bloomington, Minn., 1, 163-139. St. Joseph, Mich. ASAE.

Hartung, J and V.R. Phillips. 1994. Control of gaseous emissions from livestock buildings and

manure stores. J Agric Engng Res. 57:173-189. Heinemann, P., D. Wahanik. 1998. Modeling the generation and dispersion of odours from

mushroom composting facilities. Transactions of ASAE. 41(2):437-446. Hobbs P.J., T.H. Misselbrook, B.F. Pain. 1997. Characterisation of odorous compounds and

emissions from slurries produced from weaner pigs fed dry feed and liquid diets. J. Sci. Food Agric. 73:437-445.

Högström, U. 1972. A method for predicting odour frequencies from a point source. Atmos

Environ, 6: 103-121. Jacobson, L.D., C.J. Clanton, D.R. Schmidt, C. Radman, R.E. Nicolai, K.A. Janni. 1997.

Comparison of hydrogen sulfide and odour emissions from animal manure storages. Proceedings of the Ammonia and Odour Control from Animal production Facilities. (Oct 6-10, 1997) Vinkeloord, The Netherlands, Pages 405-412.

26

Jacobson, L.D., H. Guo, D.R. Schmidt, R.E. Nicolai, J. Zhu, K.A. Janni. 2005. Development of

the OFFSET model for determination of odour-annoyance-free setback distances from animal production sites: Part I. Review and experiment. Transactions of the ASAE, 48(6)

Jacobson, L. D., H. Guo, D. R. Schmidt, R. E. Nicolai, K. A. Janni and J. Zhu. 2000.

Development of an odor rating system to estimate setback distances from animal feedlots: Odor from feedlots - setback estimation tool (OFFSET). ASAE Paper No. 004044. St. Joseph, MI: ASABE.

Janni, K.A. 1982. Modeling dispersion of odorous gases from agricultural sources. Transaction

of ASAE. 25:1721-1723. Kaye, R. and J. Jiang. 1999. Integration of odour emission measurements from area sources

with dispersion modeling analysis of environmental impacts. Research Report No. 136. Sydney, NSW, Australia: Urban Water Research Association of Australia.

Klarenbeek, J.V. 1985. Odor emissions of Dutch agriculture. In Agricultural Waste Utilization and

Management, Proc. 5th International Symposium on Agricultural Wastes, 439-445, Chicago, IL, 16-17 December.

Koerkamp, P.W.G., J.H.M. Metz, G.H. Uenk, V.R. Phillips, M.R. Holden, R.W. Sneath, J.L.

Short, R.P. White, J. Hartung, J. Seedorf. 1998. Concentrations and emissions of ammonia in livestock buildings in Northern Europe. J Agric Engng Res. 70(1):79-95.

Koppolu, L., D.D. Schulte, S. Lin, M.J. Rinkol, D.P. Billesbach, and S.B. Verma. 2002.

Comparison of AERMOD and STINK for Dispersion Modeling of Odorous Compounds. Paper No. 024015. ASAE Annual International Meeting, Chicago, Illinois, July, 2002.

Koppolu, L., D. Schmidt, D.D. Schulte, L. Jacobson. 2004. Development of Scaling Factors

(Peak-to-Mean Ratios) through Dispersion Modeling with AERMOD and Field-based Odor Measurements for Livestock Facilities. ASAE# 044196. August 1-4, 2004 ASAE/CSAE Annual International Meeting Fairmont Chateau Laurier, The Westin, Government Centre, Ottawa, Ontario, Canada

Li, J., D. S. Bundy, S.J. Hoff, and Q. Liu, 1994. Field odour measurement and applications of

Gaussian plume model. ASAE Paper No. 94-4054. St. Joseph, Mich.ASAE. Lim, T. T., A. J. Heber, and J. Ni. 2000. Odor setback guideline for swine production systems. In

Odors and VOC Emissions 2000. Alexandria, Va.: Water Environment Federation Lim, T.T., Heber, A.J., J.Q. NI., A.L. Sutton, D.T. Kelly. 2001. Characteristics and emission rates

of odour from commercial swine nurseries. Transactions of ASAE. 44(5): 1275-1282. Mahin T.D.1997. Using dispersion modeling of dilutions to threshold (D/T) odour levels to meet

regulatory requirements for composting facilities. For presentation at the Air & Waste Management association’s 90th annual meeting & exhibition, June 8-13, 1997, Toronto, Ontario, Canada.

Mahin, T.D. 1998. Odour modeling using D/T levels. Water Environment & Technology. 2: 49-53.

27

Majer, G.J., K.H. Krause. 1985. Dispersion models for emissions from agricultural sources. In:

Nielson, V.C., J.H. Voorburg; P.L’ Hermite (eds). Odour Prevention and Control of Organic Sludge and Livestock Farming, London, Elsevier.

McPhail, S. 1991. Modeling the dispersion of agricultural odours. Proceedings of a workshop on

agricultural odours, Toowoomba, Queensland, Australia. AMLRDC Report No. DAQ 64/7. Toowoomba, Queensland, Australia: Feedlot Services Group, Queensland Department of Primary Industries.

. Mejer, G.J., K.H. Krause. 1985. Dispersion models for emissions from agricultural sources. In:

Nielson, V.C.; J.H. Voorburg; P.L. Hermite (eds). Odour prevention and control of organic sludge and livestock farming. London, Elsevier.

Misselbrook, T.H., C.R. Clarkon, B.F. Pain. 1993. Relationship between concentration and

intensity of odours for pig slurry and broiler houses. J. Agric. Eng. Res. (55): 163-169 Mussio, P., A.W. Gnyp, P.F. Henshaw. 2001. A fluctuating plume dispersion model for the

prediction of odour-impact frequencies from continuous stationary sources. Atmospheric Environment, 35:2955-2962.

Nicolai, R.E., C.J. Clanton and H. Guo. 2000. Modeling the relationship between detection

threshold and intensity of swine odours. Proceedings of the Second International Conference: Air Pollution from Agricultural Operations, ASAE, St. Joseph, MI, p296-304

Nimmermark S., L.D. Jacobson, S.W. Gay and D.R. Schmidt. 2003. Evaluation of OFFSET

(Odour Setback Model) using neighborhood monitors. ASAE Paper No. 034024. St. Joseph, MI: ASABE.

Ormerod, R.J. 1990. Odours from cattle feedlots – A review of Australian experience.

Proceedings of the International Clean Air Conference, Auckland, NZ, 39-42 Pain.B.F., and T.H. Misselbrook. 1990. Relationships between odour and ammonia emissions

during and following application of slurries to land. In: Proc. Odour and Ammonia Emissions from Livestock Farming. Ed: V.V. Nielsen, J.H. Voorburg and P. L’Hermite. Elsevier Applied Science, Newwork.

Pain, B.F., C.R. Clarkson, V.R. Phillips, J.V. Klarenbeek, T.H. Misselbrook, M. Bruins. 1991.

Odour emission arising from application of livestock slurries on land: measurements following spreading using a micrometeorological technique and olfactometry. J. agri. Engng. Res. 48: 101-110.

Pain, B.F., Rees, Y.J. and D.R. Lockyer. 1988. odour and ammonia emissions following the

application of pig or cattle slurry to land. In Nelson, V.V., ed, Volatile Emissions from Livestock Farming and Sewage Operations, Elsevier, 2-11

Pain, B.F., Y.J. Rees, and D.R. Lockyer. 1988. Odour and ammonia emissions following the

application of pig or cattle slurry to land. In Volatile Emissions from Livestock Farming and Sewage Operations, eds. V.C. Nielsen, J.H. Voorburg, and P. L'Hermite, 2-11. New York: Elsevier Science Publishing Co., Inc.

28

Patni N.K., S.P. Clarke. 1991. Transient hazardous conditions in animal buildings due to manure gas released during slurry mixing. Appl Eng Agric. 53:23-50.

Petersen, B.W., and L.G. Lavdas. 1986. Inpuff 2.0 – A multiple source Gaussian puff dispersion

algorithm- User’s guide. Research Triangle Park, N.C.: Atmospheric Sciences Research Laboratory. Office of Research and Development, USEPA.

Schauberger G., M. Piringer, E. Petz. 2000. Diunal and annual variation of the sensation

distance of odour emitted by livestock buildings calculated by the Austrian odour dispersion model (AODM). Atmos Environ, 34:18-28.

Schauberger G., M. Piringer, 2001, E. Petz. Separation distance to avoid odour nuisance due to

livestock calculated by the Austrian odour dispersion model (AODM). Agriculture, Ecosystems and Environment, 87, 13–28.

Schiffman, S.S., B. McLaughlin, G.G. Katul, H.T. Nagle. 2005. Eulerian-Lagrangian model for

predicting odour dispersion using instrumental and human measurements. Sensors and Actuators, B 106:122-127

Schiffman, S. S., J. L. Bennett, and J. H. Raymer. 2001. Quantification of odors and odorants

from swine operations in North Carolina. Agric. Forest Meteor. 108(3): 213−240. Sheridan B.A., E.T. Hayes, T.P. Curran, V.A. Dodd. 2004. A dispersion modelling approach to

determining the odour impact of intensive pig production units in Ireland. Bioresource Technology, 91:145-152

Singer, I.A, and Smith, M.E. 1996. Atmospheric dispersion at Brookhaven National Laboratory,

Int. J. Air Water Pollut., 10, 125-135. Smith, R. J. 1993. Dispersion of odours from ground level agricultural sources. J. agric. Engng

Res. 54(3):187-200. Smith, R. J. 1995. A Gaussian model for estimating odour emissions from area

sources. Mathematical and Computer Modelling, 21( 9):23-29 Smith, M.E., 1973. Recommended Guide for the Prediction of the Dispersion of Airborne

E%uents. ASME, New York. Smith, R.J. and J.P. Kelly. 1996. Comparison of two methods for estimating odour emissions

from area sources. In Proc. International Conference on Air Pollution from Agricultural Operations, 263-269. St. Joseph, Mich.: ASAE.

Smith, R. J. and N. Hancock. 1992. The prediction of feedlot odour emissions from downwind

measurements of odour concentration. Paper #: 920503, International conference on agricultural engineering.

Spoelstra, S.F. 1980. Origin of objectionable odourous components in piggery wastes and the

possibility of applying indicator components for studying odour development. In: Agriculture and Environment, Elsevier Science, Amsterdam. The Netherlands. pp. 241-260.

Stevens, S.S. 1957. On the Psychophysical Law. The Psychological Review. 64(3):153-181.

29

Stoke, B. 1977. Dispersion of odours in the neighbourhood of pig houses. Agriculture and

Environment. 3:287-295. Stowell, R.R., L. Koppolu, D. D. Schulte, and R. K. Koelsch1.2005. APPLICATIONS OF USING

THE ODOR FOOTPRINT TOOL. Proceedings of the Seventh International Symposium, 18-20 May 2005 (Beijing, China). Ed. T. Brown-Brandl. Copyright 2005 American Society of Agricultural Engineers, St. Joseph, Michigan USA.

Thierry, P., Christophe, G. 1997. Odour dispersion modeling for odour impact study. For

presentation at the air & waste management association’s 90th annual meeting & exhibition, June 8-13, 1997, Toronto Ontario, Canada.

Turner, D.B. 1994. Workbook of atmospheric dispersion estimates: an introduction to dispersion

modeling. 2nd edition. Lewis Publishers. Ann Harbor, MI, USA. USEPA. 1995a. User’s Guide for the Industrial Source Complex (ISC3) Dispersion Models.

Volumes I - III. EPA-454/B-95-003a-c. USEPA. 1995b. A User’s Guide for the CALPUFF Dispersion Model. EPA-454/B-95-006. USEPA. 1998, A Comparison CALPUFF with ISCST3. EPA-454/R-98-020 USEPA.2000. AUSPLUME GAUSSIAN PLUME DISPERSION MODEL USER MANUAL. USEPA.2004. AERMOD: DESCRIPTION OF MODEL FORMULATION. EPA-454/R-03-004.

September 2004 VDI (Verein Deutscher Ingenieure). 1993. VDI 3940: Determination of odorants in ambient air by

field inspections. Berlin, Germany: Beuth. VDI. 2000. Environmental meteorology atmospheric dispersion models, particle model. Verlag,

Berlin Verdoes, N. and N.W.M. Ogink. 1997. Odour emission from pig houses with low ammonia

emission. In Proc. International Symposium on Ammonia and Odour Control from Animal Production Facilities, eds. J.A.M. Voermans and G. Monteny, 317-25. Vinkeloord, The Netherlands, Rosmalen, The Netherlands: NVTL. 2.

Wegenaar, W.A. 1975. Stevens vs. Fechner: A plea for dismissal of the case. Acta

Psychologica. 39: 225-235 Williams, M.L. 1985. The effects of weather on odour dispersion from livestock buildings and

from fields. In: Nielson, V.C., J.H. Voorburg; P.L’ Hermite (eds). Odour Prevention and Control of Organic Sludge and Livestock Farming, London, Elsevier.

Wood, S. L., D. R. Schmidt, K. A. Janni, L. D. Jacobson, C. J. Clanton, and S. Weisberg. 2001.

Odor and air emissions from animal production systems. ASAE Paper No. 014043. St. Joseph Mich.: ASAE.

Xing Y., H. Guo, J. Feddes, S. Shewchuck. 2006. Evaluation of Air Dispersion Models Using

30

Swine Odour Plume Measurement Data. CSBE paper No. 06-172. Yu, J. C., C. E. Isaac, R. N. Coleman, J. J. R. Feddes, and B. S. West. 1991. Odorous

compounds from treated pig manure. Canadian Agric. Eng. 33(1): 131−136. Zahn, J. A., J. L. Hatfield, Y. S. Do, A. A. DiSpirito, D. A. Laird, and R. L. Pfeiffer. 1997.

Characterization of volatile organic emissions and wastes from a swine production facility. J. Envir. Qual. 26(6): 1687–1696.

Zhang, Q., J.J.R. Feddes, I.K. Edeogu, and X. J. Zhou. 2003. Correlation between odour

intensity assessed by human assessors and odour concentration measured with olfactometers. Canadian Biosystems Eng. 44(6): 27-32.

Zhang, Q., X.J. Zhou, H.Q. Guo, Y.X. Li, and N. Cicek. 2005. Odour and greenhouse gas

emissions from hog operations. Final Report Submitted to Manitoba Livestock Manure Management Initiative Inc.

Zhou, X., and Q. Zhang. 2003. Measurement of odour and hydrogen sulfide emissions from

swine barns. Canadian Biosystems Eng. 45: 6.13-6.18. Zhou, X.J., Q. Zhang, H. Guo, Y.X. Li. 2005. Evaluation of air dispersion models for livestock

odour application. CSAE/SCGR 2005 Meeting, Winnipeg, Manitoba, Paper No. 05-026 Zhu, J. 1999. The influence of stability class on downwind odour concentration predicted by air

dispersion models. J. ENVIRON. SCI. HEALTH. A34(10):1919-1931. Zhu, J., L. Jacobson, D. Schmidt, and R. Nicolai. 2000b. Daily variations in odor and gas

emissions from animal facilities. Transactions of the ASAE 16(2):153-58. Zhu, J., L.D. Jacobson, D. R. Schmidt, and R. Nicolai. 2000a. Evaluation of INPUFF-2 model for

predicting downwind odours from animal production facilities. Applied Engineering in Agriculture. 16(2): 159-164.

31