Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2012 Aug 1.
Published in final edited form as: Atmos Environ (1994). 2011 Aug 1;45(27):4617–4625. doi: 10.1016/j.atmosenv.2011.05.061

Use of AERMOD to Determine a Hydrogen Sulfide Emission Factor for Swine Operations by Inverse Modeling

Patrick T O’Shaughnessy a, Ralph Altmaier a
PMCID: PMC3144569  NIHMSID: NIHMS305620  PMID: 21804761

Abstract

This study was conducted to determine both optimal settings applied to the plume dispersion model, AERMOD, and a scalable emission factor for accurately determining the spatial distribution of hydrogen sulfide concentrations in the vicinity of swine concentrated animal feeding operations (CAFOs). These operations emit hydrogen sulfide from both housing structures and waste lagoons. With ambient measurements made at 4 stations within 1 km of large swine CAFOs in Iowa, an inverse-modeling approach applied to AERMOD was used to determine hydrogen sulfide emission rates. CAFO buildings were treated as volume sources whereas nearby lagoons were modeled as area sources. The robust highest concentration (RHC), calculated for both measured and modeled concentrations, was used as the metric for adjusting the emission rate until the ratio of the two RHC levels was unity. Utilizing this approach, an average emission flux rate of 0.57 µg/m2-s was determined for swine CAFO lagoons. Using the average total animal weight (kg) of each CAFO, an average emission factor of 6.06 × 10−7 µg/yr-m2-kg was calculated. From studies that measured either building or lagoon emission flux rates, building fluxes, on a floor area basis, were considered equal to lagoon flux rates. The emission factor was applied to all CAFOs surrounding the original 4 sites and surrounding an additional 6 sites in Iowa, producing an average modeled-to-measured RHC ratio of 1.24. When the emission factor was applied to AERMOD to simulate the spatial distribution of hydrogen sulfide around a hypothetical large swine CAFO (1M kg), concentrations 0.5 km from the CAFO were 35 ppb and dropped to 2 ppb within 6 km of the CAFO. These values compare to a level of 30 ppb that has been determined by the State of Iowa as a threshold level for ambient hydrogen sulfide levels.

Keywords: inverse modeling, AERMOD, confinement buildings, hydrogen sulfide, emission factor

1. Introduction

A large increase in the number of intensive hog production facilities in the United States has been evident over the last two decades. The United States Environmental Protection Agency (USEPA) refers to these operations as “confined animal feeding operations” (CAFOs). These facilities are operated with a high animal density in large buildings that hold manure in subfloor pits that are periodically flushed into nearby lagoons or directly sprayed on local agricultural fields. Given their potential to be sources of ambient odor and gas concentrations from the manure slurry, several research efforts have noted a relationship between living in proximity to these facilities and the prevalence of adverse health effects. (Wing and Wolf, 2000; Chrischillis et al., 2004; Merchant et al., 2005; Sigurdarson and Kline, 2006).

Given the difficulty and expense of obtaining ambient air pollutant measurements over a large region, plume dispersion modeling is an attractive alternative for the assessment of the impacts of CAFOs on air quality in surrounding areas. Plume dispersion models developed by the USEPA for Clean Air Act (CAA) compliance were initially developed for tall smoke stacks emitting a continuous pollutant stream. However, CAFOs are low-level sources occupying a large area, and may produce pollutants on an intermittent basis. Regardless, a number of attempts have been made to use air dispersion models to estimate odor emissions from CAFOs. In these studies odor was typically measured via a human odor panel and were primarily conducted for the purpose of determining setback distances from nearby residences (Heber, 1997; Piringer and Schauberger, 1999; Jacobson et al., 2001; Zhu et al., 2000; Schauberger and Piringer, 2001; Guo et al., 2004; Hoff et al., 2008). Furthermore, Wing et al. (2008) found a significant correlation between odor complaints and hydrogen sulfide concentrations at low wind speeds as well as PM10 at high wind speeds. However, only one published study described the use of a dispersion model to estimate gaseous concentrations (ammonia and hydrogen sulfide) downwind from CAFOs (MPCA, 2003), and none were found that modeled particulate contaminants.

In 2004, a Technical Workgroup organized by the Iowa Department of Natural Resources (IDNR) determined that, of all dispersion models available, AERMOD (USEPA, 2004) was the most suitable for modeling the dispersion of airborne contaminants from CAFOs (IDNR, 2004). Like all dispersion models, AERMOD requires an emission rate for each pollutant emitted at each source in the model in order to calculate downwind concentrations. However, the USEPA currently does not provide AP-42 emission factors for agricultural sources that could be used to calculate an emission rate for use in AERMOD.

The goal of this study, therefore, was to determine an emission factor for hydrogen sulfide emitted from CAFOs scalable by total operation animal weight to derive a site-specific emission rate for use in AERMOD. This study focused hydrogen sulfide because (1) it represents one of the most hazardous gases emitted from these facilities (Roth and Goodwin, 2003); (2) it can be measured accurately to the part-per-billion level by direct-reading instruments; (3) its primary source in rural areas containing swine CAFOs are the CAFOs; and (4) the Iowa State legislature adopted a limit on ambient hydrogen sulfide concentrations which are not to exceed 30 ppb more than seven times over a year (IAB, 2004).

1.1 CAFO Emission Rates

As mentioned by Bunton et al. (2007), studies involving the use of dispersion models applied to CAFOs are complicated by a lack of accurate information on the gaseous emission rates emanating from this source type. However, a number of studies have been conducted to either directly measure the emission rate of gases from swine CAFOs or estimate emission rates using an inverse-modeling approach.

Typically the direct-measurement studies we re conducted by detecting the interior concentration of a gas as well as the total exhaust flow rate exiting a CAFO from exhaust fans. The emission rate was then determined by multiplying the concentration by the flow rate (Blunden et al., 2008; Ni et al., 1999, 2000, 2002; Kim et al., 2008; Zhu et al, 2000; Heber et al., 1997; Hoff et al., 2005; Jerez et al., 2005; Jacobson et al., 2005; Koziel et al., 2005). An alternative approach involved using the decay rate of a tracer gas to estimate exhaust flow rate (Demmers et al., 1999; Gay et al., 2003). Studies of this type can be complicated when both the gas emissions from buildings and on-site waste lagoons are considered (Harper et al., 2000; Zahn et al., 2001; Aneja et al., 2001). The results from these studies vary considerably and none of the emission rates have been applied to a published study involving dispersion modeling.

The inverse-modeling method has been utilized to avoid the complications of the direct-measurement approaches, and to obtain a gas emission rate for an entire source, including both buildings and lagoons. Inverse-modeling is conducted by first applying an arbitrary emission rate to the model, Qo; running the model to determine a concentration value at a receptor location, Co; then comparing that value with a background-compensated measurement at the same location, C-Cb (Flesch, 2005). The next iteration of this procedure will result in a perfect association between modeled and measured concentration by applying a new emission rate, Q, to the model that is the result of adjusting the original rate by the following ratio:

Q=Qo(C−Cb)/Co Eq. 1

This methodology has been applied to CAFOs and other low-level emission sources utilizing a variety of dispersion model types (McCulloch et al., 1998; Rege and Tock, 1996; Hensen et al., 2009). Flesch (2005, 2007, 2009) has developed and verified the use of inverse modeling applied to a backward Lagrangian Stochastic (bLS) dispersion model to predict emission rates from agricultural facilities. This technique accurately inferred emission rates from sources that were far enough upwind of a receptor as to not be influenced by topographical features that could influence wind trajectories. Sanz et al. (2010) further verified this technique when applied to the emissions of ammonia from surface-applied swine slurry.

1.2 Study Rationale

The present study was motivated by the desire to use AERMOD to predict ambient hydrogen sulfide concentrations produced by swine CAFOs over areas large enough to encompass affected nearby residences. An inverse-modeling approach to determine CAFO emission rates was motivated by the availability of a large dataset of gas concentration measurements made with accurate instruments at point locations near large CAFOs over at least a year-long period by the IDNR (2009). This approach differs from that of other published inverse-modeling efforts in that it may not result in rates that accurately estimate the gaseous output of swine CAFOs because AERMOD has not been adequately validated for this purpose. Furthermore, AERMOD does not permit the incorporation of chemical reactions. For example, Hydrogen sulfide can react with hydroxyl radicals and ozone with a resulting residence time in the atmosphere of approximately 3 days (Warnek, 2000). However, given the large dataset available, it can be assumed that the rates determined by this method will represent values that maximize the accuracy of AERMOD to predict large-scale (>100 m), downwind concentrations given a prescribed set of input options. Furthermore, it was our intention to constrain our determination of an emission rate based on the limited information that would typically be available when modeling a swine CAFO: the location, footprint area, and permitted animal weight capacity of the facility.

1.3 Evaluation Criteria

Inverse modeling is relatively simplistic if a small set of measured downwind concentrations are to be compared to the associated modeled values. However, the comparison of modeled-to-actual concentrations at a receptor site over a time period long enough to include multiple changes in meteorological conditions can involve the assessment of thousands of observations. Therefore, an evaluation criteria is required that includes multiple pairs of modeled-versus-measured data. One approach is to perform a correlation analysis of the entire distribution of measured data obtained during the sampling period relative to the time-associated modeled values. The emission rate is then adjusted to provide the highest correlation possible. However, the USEPA decided that this approach was impracticable when using AERMOD (USEPA, 2005) because “estimates of concentrations that occur at a specific time and site are poorly correlated with actually observed concentrations.” The USEPA recognized that this poor correlation was due to a variety of uncertainties such as deviations from an exact knowledge of the true wind direction where, for example, an error of 5 to 10 degrees can lead to concentration errors of 20 to 70 percent.

The USEPA (2005) concluded that dispersion models are more reliable when estimating long time-averaged concentrations over simple topography, and when estimating the magnitude of the highest concentrations regardless of where or when they occur relative to a source location. This second criteria implies that a comparison can be made to relate the magnitude and probability of occurrence of the highest concentrations produced by a model with that of measured values as they occurred over, say, a year-long period in any radial direction downwind from the source. This approach was taken by Perry et al. (2005) when validating AERMOD using multiple field locations in which both the emission rate was known and downwind concentrations were measured. The study group relied on quantile-quantile (Q-Q) plots that pair modeled and measured values after r anking both sets of data as a visual interpretation of the accuracy of the model. The plotted points were evaluated relative to lines placed a factor of 2 above and below a 1:1 line, representing bounds of model accuracy deemed reasonable by the USEPA (USEPA, 2005).

Given that AERMOD is most accurate when predicting the likelihood of high concentration events, a comparison could be made between the highest measured and modeled concentrations to determine the accuracy of the model. However, Perry et al. (2005) used a metric known as the robust highest concentration (RHC), first developed by Cox and Tickvart (1990), for this comparison. The equation used to calculate the RHC mitigates the potential for unusual events to influence the highest concentration by incorporating a set of the largest values from a distribution of measurements to determine the highest probable concentration. The ratio of the RHC developed from modeled values to that of measured values was then assessed for its proximity to unity. A similar, but inverted, approach was used during this study whereby an accurate emission rate applied to the model was identified by adjusting it until the RHC of the modeled results was identical to the RHC of the values measured at a downwind monitoring site over the entire time period as those modeled.

2. Methods

2.1 Study sites and instrumentation

The IDNR, in cooperation with the State Hygienic Lab at the University of Iowa, conducted a sampling assessment of hydrogen sulfide and ammonia in the vicinity of CAFOs at various locations in Iowa. Data collected during 2003 and 2004 at 10 sites was used for this study. Depending on the site, between 6 and 12 months of hourly data was available for analysis. Hydrogen sulfide was measured with a continuous-pulsed fluorescent analyzer (Model 450C, Thermo Electron Corp., Waltham, MA). This instrument was operated with a range of 0–200 ppb and averaging time of 300 s, which provided a lower detection limit (LDL) of 0.06 ppb. The instrument zero level was not auto-corrected but was manually re-zeroed when the drift exceeded 2 ppb. Given this procedure for offsetting instrument drift, a true background level of hydrogen sulfide could not be obtained from the IDNR data and was therefore considered to be 0 ppb. These monitors were operated continuously to obtain hourly observations at fixed sites located on residential properties adjacent to the CAFOs. The monitors were stored in temperature-controlled trailers. A sample line with entry probe at 4 m was positioned on top of the trailer.

Of the various sites measured by the IDNR, four separate sites (Table 1, sites A – D) that were primarily influenced by swine CAFOS were chosen for determining an average hydrogen sulfide emission factor. At each site, the nearest CAFO to a measuring station that had no other CAFO within 5 km in the same line as that between the CAFO and the monitoring station was selected to apply the inverse-modeling method. Sites A–C met this criterion, whereas site D contained a second large CAFO (1.6 M kg) 4.3 km away from the station, and within 5° of the line between the station and the closest CAFO. Therefore, the emission rate for the closest CAFO was determined with the more removed CAFO added as another source. To negate influence by other CAFOs in the area, the data set of concentrations applied to the inverse-modeling approach was constrained to times when winds were blowing within +/− 22° from the selected CAFO.

Table 1.

Attributes of the Swine CAFOs Closest to the Monitoring Station at Each Site.

Site Swine Type1 Permitted Weight
kg		 Dir. & Dist.2
deg., m		 Met Data3
hours		 Met Data4
hours		 Number of
CAFOs5
A Finishing 1,123,000 185°, 975 1206 8784 10
B Finishing 1,048,000 170°, 940 973 4416 9
C Finishing 1,048,000 20°, 1095 422 4416 9
D Sow 810,000 200°, 710 1119 5880 8
E Sow 663,000 27°, 680 5880 19
F Sow 796,000 172°, 860 3768 17
G Finishing 857,000 260°, 1360 5516 4
H Finishing 1,123,000 288°, 1050 5136 22
I Finishing 1,366,000 350°, 820 8765 3
J Finishing 1,021,000 140°, 230 8760 26
Open in a new tab
1

Finishing CAFOs are used to raise pigs to market weight (100 kg); Sow CAFOs are used to hold sows prior to and during pregnancy.

2

Direction (degrees from true north) and distance (m) from monitoring station

3

Hours of meteorological data analyzed based on hours of the year during which wind was blowing from the site to the monitoring station ± 22°

4

Hours of meteorological data analyzed when applying the overall emission factor to all sites

5

CAFOs within 7 km of a monitoring site to which the overall emission factor was applied

After an average emission factor was determined by inverse-modeling of a single CAFO at sites A–D, the factor was tested by applying it to all CAFOs within 7 km of the monitoring stations at sites A–D as well as those surrounding an additional 6 sites located in other areas of northern Iowa (Table 1). Of the 10 sites, 7 utilized meteorological data from Mason City, IA. The other three sites were widely separated and used meteorological data from Waterloo, IA, Sioux City, IA, and Omaha, NE.

2.2 Meteorological data

AERMOD requires hourly surface and upper-air meteorological data as inputs. Wind speed and direction were recorded at each sampling site. Additional meteorological data needed by AERMOD such as temperature, dew point, pressure, and solar radiation were not recorded. Therefore, all necessary ground-level data was obtained from the nearest National Weather Service (NWS) station located in Mason City, Iowa, located 40 – 60 km from the original 4 study sites. Upper-air data was recorded in Omaha, Nebraska, approximately 260 km from the study region and contains information on the vertical aspects of the atmosphere including pressure, temperature, and relative humidity. Both datasets were pre-processed and made available by the IDNR Air Quality Bureau for use in AERMOD. Meteorological data between 2003 and 2004 that coincided with the time periods of the IDNR measurements at each site were utilized. As explained further in section 2.4, our goal was to develop a method for accurately modeling the highest concentrations from swine CAFOs regardless of when they occurred relative to their measured occurrence. Therefore, we assumed that the meteorological conditions actually occurring at the sampling station are represented in the meteorological data set obtained, with the understanding that they may not coincide temporally. Furthermore, in the part of the Midwestern US where the study took place there are no major terrain features to significantly change the meteorology of the sample site relative to that of the NWS station.

2.3 Modeling Options

A vendor-supplied modeling package (AERMOD View™ ver. 6.7.1, Lakes Environmental, Waterloo, Ontario) was used to execute AERMOD. All modeling was performed with the dispersion coefficient set as “rural” and terrain height option set to “elevated”. The modeled pollutant type was sulfur dioxide given no option for hydrogen sulfide. All pollutant types available in AERMOD are treated as inert compounds except sulfur dioxide when utilizing the “urban” dispersion option, and therefore produce the same model output for a given emission rate. The IDNR monitoring sites were treated as receptor locations with a height of 4 m, coinciding to the height of the inlet of the sample probe.

The adjustable options for modeling CAFOs as sources in AERMOD were source type and emission height. Source type options include point, area, line, and volume. All sites selected for initial analysis contained multiple buildings and at least one lagoon, where both structures were considered sources of hydrogen sulfide. For example, the site A CAFO consists of 10 buildings, each 88 m × 15 m, and two adjacent and adjoining lagoons with total dimensions of 130 m × 280 m (Figure 1). Given the source type options available, volume sources were chosen to represent the CAFO buildings and rectangular area sources to represent the lagoons. The building downwash option is not available for these source types. The lack of this feature is also justified given that our intent was to model the dispersion of hydrogen sulfide over a wide radius around a CAFO (>100 m) where building wake effects will little influence on ground-level concentrations.

Figure 1.

Figure 1

Open in a new tab

Site A swine operations relative to the sampling station receptor location. A detail of the large CAFO used to develop an emission factor is also provided.

Volume sources are modeled as a virtual point source with an initial lateral and vertical dimension to the plume established at the point of release of the plume. These dimensions are scaled in AERMOD relative to the physical dimensions of the volume, which results in these “initial” dimensions being smaller than the physical dimensions (USEPA, 1995). Volume sources can only be applied in AERMOD with a square base. Therefore, a long, rectangular swine building was modeled with multiple, contiguous volumes with base length and width equal to the width of the building (typically 14 m) and height equal to the height of the building roof peak (6.5 m). In some cases the volumes were not contiguous in order to have an integer number of square volumes with one exactly on either end of the building floor area. The area source type used for lagoons can be applied to AERMOD given the length and width of the rectangular area of the lagoon as well as the degrees from north that the rectangle is tilted. Building and lagoon dimensions for each site were obtained from aerial photographs available on the internet.

The emission release height of volume and area sources can be altered in AERMOD. Likewise, an initial vertical dimension can be established for an area source that accounts for an initial loft in the plume from the source. A sensitivity analysis of release height and loft was performed to determine whether there was justification for values other than zero for both. An emission factor was determined for Site A as described in section 2.4 modeled as one large area source or as a combination of area and volume sources. For each scenario, the release height was varied between 0 and 6 m. Likewise, the loft was varied from 0 to 6 m when the Site-A CAFO was modeled as an area source and an emission factor determined for each set of options.

2.4 Emission Factor Determination

The four sites, A – D, were used to obtain an average hydrogen sulfide emission factor with units of µg/s-m2-kg. To obtain this emission factor by inverse modeling, an arbitrary emission flux, Fo with units of µg/s-m2, was chosen for application to the lagoons as area sources. A review was conducted of all publications in which the hydrogen sulfide emission flux for waste storage structures (Blunden and Aneja., 2008; Gay et al., 2003; Zahn et al., 2001; Lim et al., 2003; Jacobson, 1999) or building types (Blunden et al., 2008; Gay et al., 2003; Hoff et al., 2005; Jacobson et al., 2005; Jerez et al., 2005; Kim et al., 2008; Kozial et al., 2005; Ni et al., 1999, 2000,2002; Zahn et al., 2001;, Zhu et al., 2000) was reported. These studies revealed a geometric mean (GM) of storage flux rates of 3.18 µg/s-m2 (n = 14) and a GM of building flux rates of 5.22 µg/s-m2 (n = 32). A t-test on the log-transformed rates did not demonstrate a significant difference between the two source types (p = 0.58). Therefore the flux rate per building footprint area was treated as being equivalent to the flux rate of the lagoons. Furthermore, the volume sources representing the CAFO buildings require the input of a mass emission rate. Therefore, the emission rate applied to an individual volume source, Qv, with square footprint area, Av, was calculated as:

Qv=FoAv Eq. 2

AERMOD was run with hourly meteorological data coincidental with the sampling time period to derive hourly modeled concentrations at the monitoring station location. These modeled values were ranked and the RHC was determined as given in Equation 3:

RHC=xn+(x¯n−1−xn)ln((3n−1)/2)) Eq. 3

where xn is the nth largest value, and is the average of the n-1 largest values. An n = 26 was suggested by Cox and Tikvart (1999) as being adequate for obtaining a reliable RHC value when the total number ranked is large (>100). Likewise, the measured concentrations at the monitoring station were ranked and an RHC value obtained by application of Equation 3. The ratio of the RHC value determined from the hydrogen sulfide sensor data, RHCs, to the value determined from modeled concentrations, RHCm, was used to correct the initial area flux in a way similar to that given in Eq. 1 except that the measured values were not background corrected:

F=Fo(RHCs/RHCm) Eq. 4

The resulting flux rate, F, was then scaled relative to the permitted weight of the facility to obtain an emission factor.

As explained previously, two CAFOs were added to the model as sources at Site D. This complicated the inverse modeling as a rate had to be determined for both sites during the same inverse-modeling iteration procedure. The assumption was made that the rate for the more removed CAFO was directly proportional to that of the closer CAFO and scaled relative to the ratio of their respective weights. Only the value obtained from the closest unit was used when determining an average emission factor for all sites.

2.5 Concentration Values in a CAFO Vicinity

After a hydrogen sulfide emission factor was determined it was applied to hypothetical CAFOs of varying sizes to estimate the influence of swine CAFOs on surrounding areas. The largest was given a live weight of 1 M kg, and was represented by 8 buildings, totaling 48 volume sources measuring 14 × 14 m, and two lagoons, represented by two area sources measuring 70 × 155 m and 155 × 1555 m. A 1M kg CAFO represents the average of those nearest the monitoring sites evaluated in this study, which are among the largest such facilities in the Midwestern United States. A layout typical of large swine CAFOs was chosen. The model was run to determine hydrogen sulfide concentrations at each node of a uniform polar grid with 16 rings at 500-m intervals extending to 8,000 m. The grid contained radial sections every 10° starting at 000B0;. Hourly meteorological data for an entire year (2004) was applied to the model to determine the first-highest values existing at all grid nodes for that year. Similar modeling was also performed on CAFOs with live weights of 0.25, 0.5 and 0.75 M kg, and footprint areas scaled to increase with animal weight. By comparison, for all 5,048 swine operations for which the IDNR has a live weight on record, the median weight is 0.16 M kg and the upper quartile constitute operations >0.27 M kg. For this analysis, all first-highest values at a set radial distance away from the source were averaged for a number of radial distances to determine a general relationship between concentration and downwind distance.

3. Results

As shown in Figure 2, release height with no vertical loft will affect the inverse-modeled emission flux where, in general, a greater height results in a greater emission flux. This was expected as a greater release height will cause the model to produce lower downwind receptor-level concentrations by enhanced dispersion. In order to compensate for lower concentrations, a higher emission flux is needed to match modeled with measured concentrations. This effect was more pronounced when modeling an entire CAFO as one area source than as a combination of area (lagoons) and volume (buildings) sources, and justifies the decision to model the CAFOs as a combination of area and volume sources. A similar trend is seen in Figure 3 where both release height and vertical loft are applied to CAFOs as area sources. However, loft height had very little influence on the value of the inverse-modeled emission flux between lofts of 0–3 m. Given no physical rationale for either elevating these sources or applying an initial loft, those model options were set to zero.

Figure 2.

Figure 2

Open in a new tab

The ratio of the inverse-modeled emission flux for a release height relative to the emission flux determined with a release height of 0 m for two different source configurations.

Figure 3.

Figure 3

Open in a new tab

The ratio of the inverse-modeled emission flux for a release height and loft relative to the emission flux determined with a release height of 0 m and loft of 0 m for a CAFO modeled as an area source.

An example of the general relationship between modeled and measured concentration data sets that have the same RHC value is given in Figure 4. This plot was developed from Site-A data sorted by wind direction for which there was only one isolated CAFO almost directly to the south of the monitoring station (Figure 1), hence the peak in hydrogen sulfide concentrations is associated with winds from 185°. As shown in the figure, the range and spread of concentrations modeled by AERMOD closely match the measured values.

Figure 4.

Figure 4

Open in a new tab

Measured and modeled hydrogen sulfide concentrations relative to wind direction for Site A.

Q-Q plots for all fours sites are given in Figure 5. These plots show good agreement between the highest measured and modeled concentrations. However, the measured-to-modeled relationship curves downward for lower concentration values. This pattern is explained by the relative difference between the lowest modeled values, which go to 0, and the lowest measured values which retained levels up to 2 ppb because of instrument drift.

Figure 5.

Figure 5

Open in a new tab

Q-Q plots for all monitoring sites A – D. Solid lines represent values varying by a factor of 2 to either side of the dashed 1:1 line.

The emission rates and emission factors determined by inverse modeling for each site are given in Table 2. The average emission flux for lagoons was 0.57 µg/s-m2 . The average emission factor for lagoons was 6.06 × 10−7 µg/s-m2-kg with a standard deviation of 4.38 × 10−7 µg/s-m2 -kg. The average building emission rate was 111.7 µg/s with a standard deviation of 61.6 µg/s.

Table 2.

CAFO Emission Rates and Emission Factors

Site Lagoon Emission
Flux,
µg/m2-s		 Building Emission
Rate1,
µg/s		 Lagoon Emission
Factor,
µg/m2-s-kg
A 0.54 106.0 4.82E-07
B 0.46 89.2 4.34E-07
C 0.27 53.6 2.61E-07
D 1.01 198.0 1.25E-06
Mean 0.57 111.7 6.06E-07
Open in a new tab
1

Rates applied to a volume source with 14 m × 14 m footprint

The average emission factor was applied to 10 sites in northern Iowa including the 4 original sites used to determine that factor, but with additional sources applied to the model that included all CAFOs within 7 km of the monitoring site. An RHC ratio was calculated for each site to compare modeled-to-measured concentration values. The average RHC ratio was 1.24 with a range of 0.70 – 1.89. As a measure of uncertainty in the emission factor, the coefficient of variation (CV) of the RHC ratios was computed with a result of 29.3%. A one-sample t-test demonstrated that the average ratio was not significantly different from 1.00 at the 95% confidence level (p = 0.064). The relationship between RHCs and RHCm for each site is shown in Figure 6; all values were within the EPA guideline that modeled values be within a factor of 2 of measured values (USEPA, 2005).

Figure 6.

Figure 6

Open in a new tab

Relationship between measured and corresponding modeled RHC values relative to factor-of-two ranges and the 1:1 line.

Figure 7 was created to visualize the range and extent of hydrogen sulfide concentrations downwind of hypothetical swine CAFOs of varying sizes for which the average emission factor reported in Table 2 was applied. This figure gives an average of the first-highest concentrations at all nodes of a polar grid at set distances away from the source at the origin. Figure 6 therefore demonstrates that worst-case hydrogen sulfide concentrations produced by large CAFOs (1 M kg) can extend to an average distance of 6.0 km from the CAFO before dropping below 2 ppb. Furthermore, concentrations can exceed 25 ppb within 0.5 km of a large CAFO. As when determining the emission factor, background concentrations were not included when developing Figures 6 and 7.

Figure 7.

Figure 7

Open in a new tab

Relationship between distance from CAFOs of various sizes (M-kg of pig weight) and first-highest hydrogen sulfide concentrations relative to a value (2 ppb) that represents instrument drift.

4. Discussion

The inverse-modeling approach was applied to four CAFOs in north-central Iowa represented as a combination of volume and area sources. The resulting lagoon emission flux rates and associated emission factors were within a half-order magnitude of each other. Three of the four sites produced very similar emission factors. This consistency may be attributed to their similar size and type of swine (finishing), and all were in the same meteorological region with regards to model input. The facility producing the highest emission factor was also the exception in terms of swine type (sows). Given the relative similarity of the four sites, it cannot be determined how accurately the average emission factor reported here could be applied to CAFOs of sizes or regions different from those utilized in this study. However, the CAFOs analyzed represent those that are typical of any in the United States built and operated with modern technology and management practices. Regardless, this emission factor is most suitable for modeling large CAFOs located in the Midwestern United States.

An average emission factor was developed from results obtained from four monitoring sites. When this factor was applied to all CAFOs within 7 km of the monitoring station at those sites and an additional six sites, the average RHC ratio was 1.24; a value that indicates the modeled values over estimated nearby concentrations. Regardless, we do not recommend lowering the emission factor by that amount without further research to justify that approach, especially given that the average obtained was not significantly different from 1.0.

The meta-analysis of all available papers reporting hydrogen sulfide emission rates mentioned above suggests that measured emission flux rates (GM = 3.18 µg/s-m2 for lagoon storage units) are an order of magnitude higher than the average value determined here by inverse modeling (0.57 µg/s-m2 for lagoons). Therefore, a direct application of a measured flux rate to AERMOD, with the modeling options utilized for this study, could produce extremely overestimated downwind concentrations. However, the GM of storage flux rates incorporated results from one study (Blunden & Aneja, 2008) that reported seasonal lagoon flux rates which varied between 0.0 µg/s-m2 in the winter and 0.08 µg/s-m2 in the summer. This variability in measured emission rates contributes to the uncertainty when comparing the emission factor derived here with those from other published studies. For example, an analysis performed in 2002 on the variability in ammonia emission rates from swine lagoons resulted in a CV of 25% (NRC 2002). Therefore, we cannot be certain whether the difference between the flux rate developed from this analysis and that reported in most other studies is a consequence of uncertainties inherent to AERMOD or that the CAFO types we analyzed represent those that emit less hydrogen sulfide than most of those reported elsewhere. These findings suggest that, if inverse-modeling is used to determine an emission rate, then that process should be performed with the same model to which the emission rate will be applied. However, a similar comparison of inverse-modeled e mission rates to those measured by other means could not be found in the literature to substantiate this assertion.

If the flux rate discrepancy is due to issues inherent to the AERMOD code, then there are some possible causes. Although they evaluated the ISC model rather than AERMOD, the study by Rege and Tock (1996) determined that the standard Pasquill-Gifford dispersion coefficients under-represented the lateral dispersion of contaminants from a low-level source. If applied to the model, greater dispersion would reduce the modeled concentration at any receptor point but would likewise result in a higher emission rate by inverse modeling than for the more constrained lateral dispersion inherent to the ISC model. A similar assessment utilizing AERMOD could not be found in the literature. However, a comparable effect may be causing a reduction in the emission rate by inverse modeling with that dispersion model. There are also general aspects of the AERMOD code that contribute to statistical uncertainty in modeled concentration values. Hanna (2007) provides an excellent review of this issue that includes the relative influence of model inputs such as wind speed and mixing height on model uncertainty. In general he states that the ISC model and AERMOD can produce a relative error in concentrations below 50% but not if the sources have a time-varying release, are located in areas with large elevation differences, or are located far from the nearest meteorological station.

An application of the hydrogen sulfide emission factor found for swine CAFOs to hypothetical CAFOs of various sizes by animal weight revealed that the highest levels could exceed 30 ppb but only for the largest CAFO analyzed and for distances within 0.5 km. This evaluation suggests that setback distances applied to swine CAFOs over 0.5 km will result in hydrogen sulfide concentration levels that do not exceed 30 ppb on any given hour of the year.

5. Conclusions

With the use of inverse modeling applied to AERMOD, a mean hydrogen sulfide emission factor of 6.06 × 10−7 µg/s-m2 -kg was determined for large swine CAFOs. This factor was based on emission flux rates determined from inverse-modeling utilizing AERMOD applied to four large swine CAFOs in north-central Iowa for which animal weights were supplied from a state permitting process and lagoon and building dimensions were determined by manual evaluation of aerial photographs. When these factors were applied to hypothetical swine CAFOs of varying animal weight, the highest concentrations modeled were less than 30 ppb beyond 0.5 km of the CAFO. The overall extent of hydrogen sulfide emitted from the largest CAFOs extended to 6.0 km from the source CAFO.

The emission rates determined by this method were lower than most other values reported from studies during which the emission rate was directly measured. Therefore the emission factor derived in this study represents a value that can be applied to determine ambient hydrogen sulfide concentrations produced by swine CAFOs in the Midwestern United States with the use of AERMOD, but should not be used to obtain an assessment of the actual emission rate. Furthermore, the emission factor was derived from the highest downwind concentrations and with non-local meteorological data applied to the model over a large range of climatological conditions. Therefore, the use of the emission factor in AERMOD applied to other sites should be performed with a large meteorological data set with the intent of estimating the highest concentrations at a receptor site.

Research Highlights.

  • We used inverse-modeling to determine a hydrogen sulfide emission factor for swine buildings

  • When applied to a test set of swine buildings, the emission factor accurately reproduced gas concentrations

  • Inverse modeling may result in more accurate emission rates than by direct measurement

  • Large swine operations can influence local hydrogen sulfide concentrations up to 6 km

  • For all but the largest swine operations, modeled levels were below a recommended level of 30 ppb.

Acknowledgements

This work was funded by the Environmental Health Sciences Research Center of The University of Iowa (NIEHS P30 ES05605-11). The authors are grateful for the efforts and helpfulness of the CAFO monitoring team at the Iowa Department of Natural Resources led by Sean Fitzsimmons.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Contributor Information

Patrick T. O’Shaughnessy, Email: patrick-oshaughnessy@uiowa.edu.

Ralph Altmaier, Email: ralph-altmaier@uiowa.edu.

References

  1. Aneja VP, Bunton B, Walker JT, Malik BP. Measurement and analysis of atmospheric ammonia emissions from anaerobic lagoons. Atmospheric Environment. 2001;35:1949–1958. [Google Scholar]
  2. Blunden J, Aneja VP. Characterizing ammonia and hydrogen sulfide emissions from a swine waste treatment lagoon in North Carolina. Atmospheric Environment. 2008;42(14):3277–3290. [Google Scholar]
  3. Blunden J, Aneja VP, Westerman PW. Measurement and analysis of ammonia and hydrogen sulfide emissions from a mechanically ventilated swine confinement building in North Carolina. Atmospheric Environment. 2008;42(14):3315–3331. [Google Scholar]
  4. Bunton B, O’Shaughnessy P, Fitzimmons S, Gering J, Hoff S, Lyngbye M, Thorne PS, Wasson J, Werner M. Modeling and monitoring of emissions from CAFOs. Environmental Health Perspectives. 2007;115:303–307. doi: 10.1289/ehp.8838. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Chrischilles E, Ahrens R, Kuehl A, Kelly K, Thorne P, Burmeister L, Merchant J. Asthma prevalence and morbidity among rural Iowa schoolchildren. Journal of Allergy and Clinical Immunology. 2004;113:66–71. doi: 10.1016/j.jaci.2003.09.037. [DOI] [PubMed] [Google Scholar]
  6. Cox W, Tikvart J. A statistical procedure for determining the best performing air quality simulation model. Atmospheric Environment. 1990;24A:2387–2395. [Google Scholar]
  7. Demmers TGM, Burgess LR, Short JL, Phillips VR, Clark JA, Wahthes CM. Ammonia emissions from two mechanically ventilated UK livestock buildings. Atmospheric Environment. 1999;33:217–227. [Google Scholar]
  8. Flesch TK, Wilson JD, Harper LA, Crenna BP. Estimating gas emissions from a farm with an inverse-dispersion technique. Atmospheric Environment. 2005;39:4863–4874. [Google Scholar]
  9. Flesch TK, Wilson JD, Harper LA, Todd RW, Cole NA. Determining ammonia emissions from a cattle feedlot with an inverse dispersion technique. Agricultural and Forest Meteorology. 2007;144:139–155. [Google Scholar]
  10. Flesch TK, Harper LA, Powell JM, Wilson JD. Inverse dispersion calculation of ammonia emissions from Wisconsin dairy farms. Trans. ASABE. 2009;52:253–265. [Google Scholar]
  11. Gay SW, Schmidt DR, Clanton CJ, Janni KA, Jacobson LD, Weisberg S. Odor, total reduced sulfur, and ammonia emissions from animal housing facilities and manure storage units in Minnesota. Applied Engineering in Agriculture. 2003;19:347–360. [Google Scholar]
  12. Guo H, Jacobson LD, Schmidt DR, Nicolai RE, Janni KA. Comparison of five models for setback distance determination from livestock sites. Canadian Biosystems Engineering. 2004;46:6.17–6.25. [Google Scholar]
  13. Hanna SR. Chapter 4.0: A Review of Uncertainty and Sensitivity Analyses of Atmospheric Tranport and Dispersion Models. In: Borrego C, Renner E, editors. Developments in Environmental Science. Volume 6. Atlanta: Elsevier Ltd.; 2007. pp. 331–351. [Google Scholar]
  14. Harper LA, Sharpe RR, Parkin TB. Gaseous nitrogen emissions from anaerobic swine lagoons: Ammonia, nitrous oxide, and dinitrogen gas. J. Environ. Qual. 2000;29:1356–1365. [Google Scholar]
  15. Heber AJ, Duggirala RK, Ni J–Q, Spence ML, Haymore BL, Adamchuck VI, Bundy DS, Sutton AL, Kelly DT, Keener KM. Manure treatment to reduce gas emissions from large swine houses. In: Voermans JAM, Monteny G, editors. International Symposium on Ammonia and Odour Control from Animal Production Facilities; The Netherlands: Dutch Society of Agricultural Engineering (NVTL); 1997. pp. 449–457. Rosmalen. [Google Scholar]
  16. Heber AJ. Setbacks for sufficient swine odor dispersion and dilution. Livestock and Environment Symposium; University of Nebraska Cooperative Extension Service; December 10–11, 1997; New World Inn, Columbus, Nebraska. 1997. Available at: http://pasture.ecn.purdue.edu/~heber/setba.htm. [Google Scholar]
  17. Hensen A, Loubet B, Mosquera J, van den Bulk WCM, Erisman JW, Dämmgen U, Milford C, Löpmeier FJ, Cellier P, Mikuka P, Sutton MA. Estimation of NH3 emissions from a naturally ventilated livestock farm using local-scale atmospheric dispersion modeling. Biogeosciences Discuss. 2009;6:825–862. [Google Scholar]
  18. Hoff SJ, Hornbuckle KC, Thorne PS, Bundy DS, O’Shaughnessy PT. Iowa Concentrated Animal Feeding Operations Air Quality Study. Iowa State University and The University of Iowa Study Group; 2002. Feb, Emissions and community exposures from CAFOs. 2002. Available at: http://www.public-health.uiowa.edu/ehsrc/CAFOstudy.htm. [Google Scholar]
  19. Hoff SJ, Zelle BC, Huebner MA, Gralapp AK, Bundy DS, Zhang Y, Jacobson LD, Heber AJ, Koziel JA, Beasley D. NH3, H2S, CO2, PM, and odor animal emission data from the six-state (APECAB) project: swine deep-pit finishing buildings in Iowa. Paper #648, in the proceedings of the 2005 AWMA Annual Meeting and Exhibition; June 2005; Minneapolis, MN. 2005. [Google Scholar]
  20. Hoff SJ, Bundy DS, Harmon JD. Modeling receptor odor exposure from swine production sources using CAM. Applied Engineering in Agriculture. 2008;24:821–837. [Google Scholar]
  21. IAB (Iowa Administrative Bulletin) ARC 3588B. No. 4. Vol XXVII. Iowa Legislature General Assembly; 2004. Chapter 32; p. 274. [Google Scholar]
  22. IDNR (Iowa Department of Natural Resources) Animal Feeding Operations Technical Workgroup Report on: Air Emissions Characterization, Dispersion Modeling, and Best Management Practices. 2004 Available at: http://www.iowadnr.com/air/afo/aforeport.html.
  23. IDNR (Iowa Department of Natural Resources) Iowa Dept. of Natural Resources, Air Quality Group; 2009. Animal Feeding Operations Field Study, Hydrogen Sulfide and Ammonia Results. Available at: http://www.iowadnr.gov/air/prof/monitor/monitor.html. [Google Scholar]
  24. Jacobson LD, Paszek D, Nicolai RE, Schmidt DR, Hetchler B, Zhu J. Odor and gas emission from animal manure storage units and buildings. Am. Soc. of Agric. Eng., St; ASAE Annual Int. Meeting; 18–21 July 1999; Toronto, ON, Canada. Joseph, MI: 1999. Paper no. 994004. [Google Scholar]
  25. Jacobson LD, Guo H, Schmidt DR, Nicolai RE. Influence of weather conditions on field evaluation of odor dispersions from animal production sites. Proceedings of the Sixth International Livestock Environment Symposium; 2001 May 21–23; Louisville, KY: American Society of Agricultural Engineers; 2001. pp. 679–686. [Google Scholar]
  26. Jacobson LD, Hetchler BP, Johnson VJ, Schmidt DR, Nicolai RE, Heber AJ, Ni J, Koziel JA, Hoff SJ, Zhang Y, Beasley D. Aerial pollutants emissions from confined animal buildings (APECAB) project: dry sow buildings in Minnesota. Paper #53 in the proceedings of the 2005 AWMA Annual Meeting and Exhibition; June 2005; Minneapolis, MN. 2005. [Google Scholar]
  27. Jerez SB, Zhang Y, McClure JW, Heber AJ, Ni J, Koziel JA, Hoff SJ, Jacobson LD, Beasley D. Aerial pollutant concentration and emission rate measurements from a swine farrowing building in Illinois. Paper #1026 in the proceedings of the 2005 AWMA Annual Meeting and Exhibition; June 2005; Minneapolis, MN. 2005. [Google Scholar]
  28. Kim YK, Ko HJ, Kim HT, Kim YS, Roh YM, Lee CM, Kim CN. Quantification of ammonia and hydrogen sulfide emitted from pig buildings in Korea. Journal of Environmental Management. 2008;88:195–202. doi: 10.1016/j.jenvman.2007.02.003. [DOI] [PubMed] [Google Scholar]
  29. Koziel JA, Baek B-H, Bayley cL, Bush KJ, Spinhirne JP, Balota A, Sweeten JM. NH3, H2S, CO2, PM, and odor animal building emission data from the six-state (APECAB) project: swine finisher buildings in Texas. Paper #1043, in the proceedings of the 2005 AWMA Annual Meeting and Exhibition; June 2005; Minneapolis, MN. 2005. [Google Scholar]
  30. Lim T-T, Heber AJ, Ni J-Q, Sutton AL, Shao P. Odor and gas release from anaerobic treatment lagoons for swine manure. Journal of Environmental Quality. 2003;32:406–416. doi: 10.2134/jeq2003.4060. [DOI] [PubMed] [Google Scholar]
  31. McCulloch RB, Few GS, Murray GC, Jr, Aneja VP. Analysis of ammonia, ammonium aerosols and acid gases in the atmosphere at a commercial hog farm in eastern North Carolina, USA. Environmental Pollution. 1998;102 Suppl.1:263–268. [Google Scholar]
  32. Merchant JA, Naleway AL, Svendsen ER, Kelly KM, Burmeister LF, Stromquist AM, Taylor CD, Thorne PS, Reynolds SJ, Sanderson WT. Asthma and farm exposures in a cohort of rural Iowa children. Environmental Health Perspectives. 2005;113:350–356. doi: 10.1289/ehp.7240. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. MPCA (Minnesota Pollution Control Agency) Draft Environmental Impact Statement on the Hancock Pro Pork Feedlot Project, Stevens and Pope Counties. 2003 Available at: http://www.pca.state.mn.us/news/eaw/hancock-eis.pdf.
  34. Ni J-Q, Heber AJ, Lim TT, Diehl CA. Continuous measurement of hydrogen sulfide emission from two large swine finishing buildings. Presented at the July 18–22, 1999, ASAE/CSAE-SCGR Annual International meeting; 1999. Paper # 994132. ASAE, 2950 Niles Rd. St. Joseph, MI 49085-9659 USA. [Google Scholar]
  35. Ni J-Q, Heber AJ, Diehl CA, Lim TL. Ammonia, hydrogen sulfide and carbon dioxide release from pig manure in under-floor deep pits. Journal of Agricultural Engineering Research. 2000;7(1):53–66. [Google Scholar]
  36. Ni J-Q, Heber AJ, Diehl CA, Lim TT, Duggirala RK, Haymore BL. Summertime concentrations and emissions of hydrogen sulfide at a mechanically ventilated swine finishing building. Transactions of the American Society of Agricultural Engineers. 2002;45(1):193–199. [Google Scholar]
  37. NRC (National Research Council) The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations: Interim Report. Washington DC: National Academies Press; 2002. p. 51. [Google Scholar]
  38. Perry SG, Cimorelli AJ, Paine RJ, Brode RW, Weil JC, Venkatram A, Wilson RB, Lee RF, Peters WD. AERMOD: A dispersion model for industrial source applications. Part II: Model performance against 17 field study databases. Journal of Applied Meteorology. 2005;44:694–708. [Google Scholar]
  39. Piringer M, Schauberger G. Comparison of a Gaussian diffusion model with guidelines for calculating the separation distance between livestock farming and residential areas to avoid odour annoyance. Atmospheric Environment. 1999;33:2219–2228. [Google Scholar]
  40. Rege MA, Tock RW. Estimation of point-source emissions of hydrogen sulfide and ammonia using a modified Pasquill-Gifford approach. Atmospheric Environment. 1996;30:3181–3195. [Google Scholar]
  41. Roth S, Goodwin V. Alberta: Edmonton; 2003. Health effects of hydrogen sulphide: knowledge gaps. Alberta Environment Pub No. T/696. Available at: http://www3.gov.ab.ca/env/protenf/publications/H2S_Report.pdf. [Google Scholar]
  42. Sanz A, Misselbrook T, Sanz MJ, Vallejo A. Use of an inverse dispersion technique for estimating ammonia emission from surface-applied slurry. Atmospheric Environment. 2010;44:999–1002. [Google Scholar]
  43. Schauberger G, Piringer M. Predicting odour impact using the Austrian odour dispersion model (AODM) Water Science and Technology. 2001;44:197–204. [PubMed] [Google Scholar]
  44. Sigurdarson ST, Kline JN. School proximity to concentrated animal feeding operations and prevalence of asthma in students. Chest. 2006;129:1486–1491. doi: 10.1378/chest.129.6.1486. [DOI] [PubMed] [Google Scholar]
  45. USEPA (US Environmental Protection Agency) User’s guide for the industrial source complex (ISC3) dispersion models. EPA Contract No. EPA-454/B-95-003b. 1995 Available at: http://www.epa.gov/scram001/userg/regmod/isc3v2.pdf.
  46. USEPA (US Environmental Protection Agency) AERMOD: Description of model formulation. EPA Contract No. EPA-454/R-03-004. 2004 Available at: http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf.
  47. USEPA (US Environmental Protection Agency) Revision to the Guideline on Air Quality Models: Adoption of Preferred General Purpose (Flat and Complex Terrain) Dispersion Model and Other Revisions. 40 CFR Part 51, Appendix W – Guideline on Air Quality Models. Final Rule Federal Register. 2005 Nov. 9Vol. 70(No. 216):68218–68261. 2005.
  48. Warnek P. Chemistry of the Natural Atmosphere. 2nd Ed. San Diego: Academic Press; 2000. p. 598. [Google Scholar]
  49. Wing S, Wolf S. Intensive livestock operations, health, and quality of life among eastern North Carolina residents. Environmental Health Perspectives. 2000;108:233–238. doi: 10.1289/ehp.00108233. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Wing S, Horton RA, Marshall SW, Thu K, Tajik M, Schinasi L, Schiffman SS. Air pollution and odor in communities near industrial swine operations. Environmental Health Perspectives. 2008;116:1362–1368. doi: 10.1289/ehp.11250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Zahn JA, Hatfield JL, Laird DA, Hart TT, Do YS, DiSpirito AA. Functional Classification of Swine Manure Management Systems Based on Effluent and Gas Emission Characteristics. J. Environ. Qual. 2001;30:635–647. doi: 10.2134/jeq2001.302635x. [DOI] [PubMed] [Google Scholar]
  52. Zhu J, Jacobson L, Schmidt D, Nicolai R. Evaluation of INPUFF-2 model for predicting downwind odors from animal production facilities. Applied Engineering in Agriculture. 2000;16:159–164. [Google Scholar]

RESOURCES