2
Determining Emission Factors

INTRODUCTION

The U.S. Environmental Protection Agency (EPA) has asked the committee to address a number of specific questions (see Executive Summary) relative to characterizing emissions from animal feeding operations (AFOs). The committee has addressed these questions based on the following assumptions developed in earlier sections of this report: (1) emissions estimates are needed at the individual AFO level (Finding 2); (2) it is not practical to measure emissions at all individual AFOs (Finding 3); (3) therefore a modeling approach to predict emissions at the individual AFO level has to be considered; and (4) it is necessary to establish the set of independent variables that are required to characterize AFO emissions at the individual AFO level (Finding 4).

Most local, state, and federal agencies rely on emission factors to develop emission inventories for various substances released to the atmosphere. As defined by the Emission Factor and Inventory Group in the EPA Office of Air Quality Planning and Standards, an emission factor is (EPA, 1995b):

A representative value that attempts to relate the quantity of a pollutant released to the atmosphere with an activity associated with the release of the pollutant.

Emission factors are generally expressed as mass per unit of activity related to generating the emission per unit time or instance of occurrence. EPA (2001a) proposed defining emission factor as the mass of the substance emitted



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations 2 Determining Emission Factors INTRODUCTION The U.S. Environmental Protection Agency (EPA) has asked the committee to address a number of specific questions (see Executive Summary) relative to characterizing emissions from animal feeding operations (AFOs). The committee has addressed these questions based on the following assumptions developed in earlier sections of this report: (1) emissions estimates are needed at the individual AFO level (Finding 2); (2) it is not practical to measure emissions at all individual AFOs (Finding 3); (3) therefore a modeling approach to predict emissions at the individual AFO level has to be considered; and (4) it is necessary to establish the set of independent variables that are required to characterize AFO emissions at the individual AFO level (Finding 4). Most local, state, and federal agencies rely on emission factors to develop emission inventories for various substances released to the atmosphere. As defined by the Emission Factor and Inventory Group in the EPA Office of Air Quality Planning and Standards, an emission factor is (EPA, 1995b): A representative value that attempts to relate the quantity of a pollutant released to the atmosphere with an activity associated with the release of the pollutant. Emission factors are generally expressed as mass per unit of activity related to generating the emission per unit time or instance of occurrence. EPA (2001a) proposed defining emission factor as the mass of the substance emitted

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations per animal unit (AU) per year. EPA and the USDA have different definitions of AU (see Appendix B). Throughout this report, the EPA definition is used. Emission factors are usually derived from calculations based on measured data. Actual measurements of concentrations and flow rates yield a value for an emission rate, the mass of a substance emitted per unit time (e.g., kilograms of ammonia [NH3] per year). Sometimes it is more appropriate to measure the flux of an emitted substance, the mass emitted per unit area of the source per unit time (e.g., kilograms of NH3 per hectare-year). An emission rate can be estimated from flux measurements by integrating emissions over the whole area of the emitting source. Emission rates for an AFO can be estimated from emission factors through the simple expression in Equation 2-1: ER = AU × EF, (Eq. 2-1) where ER is the emission rate, AU is the number of animal units associated with the source, and EF is the emission factor in units of mass per AU per unit time. Equation 2-1 illustrates that the uncertainty contained in the numerical values selected for AU and EF are also present in the derived values for ER. The main goal of the approach outlined by EPA (2001a) is to develop a method for estimating emissions at the individual AFO level that reflects the different kinds of animal production units commonly used in commercial-scale animal production facilities. Specifically, the approach attempts to subdivide the populations of AFOs according to the different production or manure management systems that are commonly used and to develop emission factors for model farms characterized by the processing steps. Assignment of emission factors to each of the individual processing steps within a model farm leads to an estimate of the annual mass of emissions. An estimate of the emissions from an individual AFO can then be made by associating it with the proper model farm, accounting for the AUs housed there, and adding the contributions from the processing steps (housing, manure storage, and land application). The central assumption of this approach is that the individual processing steps within each identified manure management system are the principal factors that influence emissions. In other words, although there is inherent variability in emissions within each processing step that constitutes a manure management system, the act of subdividing the AFO population into model farms succeeds in decreasing this inherent variability to the point that single emission factors for individual processing steps, when combined, can adequately describe emissions from a model farm and thus from individual AFOs that are assigned to a given model farm category. It is further implied in this approach that the dominant factor controlling the magnitude of the calculated emissions is the number of AUs housed and not other unaccounted-for or unknown factors. This also explains the emphasis on finding the correct

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations emission factors for the individual processing steps since there is an implied supposition that such unique values must exist (EPA, 2001a). The data quality objectives (defined as the quality of data that will be necessary to solve a problem or provide useful information; Kateman and Pijers, 1981) required to meet the needs of the EPA Office of Air and Radiation are not specified by EPA (2001a). Whatever method is eventually selected to estimate emissions from individual AFOs, the derived estimate will contain some degree of uncertainty. Here the committee emphasizes the data quality that can be assigned to measurements of emissions, and to subsequently derived emission rates and emission factors. This discussion is placed in the context of the five specific questions from the EPA. SCIENTIFIC CRITERIA What are the scientific criteria needed to ensure that reasonable and appropriate estimates of emissions are obtained? In this report, “reasonable and appropriate estimates of emissions” is taken to mean emission estimates with acceptable estimates of uncertainty. For emission rates from AFOs—as with all numerical measurements and numerical calculations based on them—uncertainty can be described in terms of accuracy and precision (Taylor, 1987). Accuracy In this report “accuracy” is taken to mean the measure of systematic bias in the average of a set of measurements or estimates, and “precision” is taken as the measure of overall reproducibility. Systematic bias can arise from the measurement technology selected to characterize concentrations or from the selection of AFOs that are not representative of the larger population. Typically, concerns about accuracy are limited to the calibration of the analytical instrumentation used. While accurate calibration is an important component of the measurement process, it does not address the possibility that the analytical instrumentation selected may be ill-suited for the task or that bias may be introduced by the experimental design. Possible sources of systematic bias that should be considered include a predominance of daytime sampling when emissions are often higher; ignoring times during the year when buildings are empty; sampling locations that are not representative of exhaust air composition; odor panel sensitivities; and lack of adequate background sampling, especially at larger facilities with multiple housing units in close proximity. The representativeness of the emission factors reported in the scientific literature and used by EPA (2001a) is a major concern since the EPA’s Office of Air and

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations Radiation has no criteria for how to select the AFOs whose measurements are to be used (e.g., whether the AFO was being operated optimally or not), nor have AFOs been chosen at random. Management of an AFO can have a significant impact on its emissions. AFOs at which individual emission measurements have been made have been selected largely based on access (finding operators willing to allow access to their facilities) and the physical characteristics of the sites (as required by criteria associated with the emission measurement technique selected). Thus, calculating a mean emission factor from screened published data by no means guarantees that the calculated value is representative of the AFO population. Because there are no universally accepted analysis methods, the presence of systematic bias in emission measurements is best evaluated via intercomparison studies in which emissions are determined by two or more separate analytical techniques with differing overall experimental designs. An assessment of accuracy can also be made through the use of elemental (nitrogen [N], carbon [C], or sulfur [S]) mass balances. Nutrient excretion factors (see Appendix B) offer an independent means to set upper limits on possible emission rates. Reported emission rates in excess of nutrient excretion rates should be viewed with suspicion; they may indicate measurement conditions atypical of normal operation, or a fatal flaw in the overall experimental design or instrumentation used in the study. Precision Assigning an estimate of precision to measurements of concentrations emitted from different components in a manure management system is not a simple task. One method is to make paired observations with similar instrumentation over the same space and time (Cochran, 1977). The variance is then obtained as follows: (Eq. 2-2) where Ai and Bi represent the ith pair of observations and n represents the number of pairs (Cochran, 1977). This approach often requires duplication of equipment that may not be possible. Spatial variations in emissions may also become important for area sources such as lagoons or cropland receiving manure or lagoon water. Robarge et al. (2002) applied Equation 2-2 (with n = 90 paired observations) to estimate precision, expressed as percent coefficient of variation (CV) associated with ambient atmospheric concentrations of gaseous and particulate species measured using annular denuder technology (Purdue,

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations 1992). For ammonia (NH3) and sulfur dioxide (SO2), the calculated CV was <10 percent. For nitrous (HONO) and nitric (HNO3) acids, the CV values were 17.5 and 31 percent, respectively; for particulate ammonium (NH4 +), sulfate (SO4 2-), and nitrate (NO3 -), CVs were 13, 18, and 25 percent, respectively. Determining the precision of emission concentration measurements is also complicated by the fact that such measurements are actually part of a time series with a substantial degree of covariance between measurements. Emissions of gaseous chemical species are highly dependent on microbial decomposition and conversion processes and on physical transport across air-liquid or air-solid interfaces. These processes are in turn dependent on temperature, and variations in temperature are not random but are autocorrelated. The presence of a significant degree of positive autocorrelation in data requires corrections of the standard error of the mean. The variance is underestimated if it is calculated using standard statistical formulas (Code of Federal Regulations, 2001). The presence of autocorrelation in emissions data also suggests reconsideration of the sampling frequency in order to characterize emissions. Limiting sampling to one or several short series of sequential measurements (as is often done to reduce cost) may in fact be an inefficient and possibly ineffective way to determine actual diurnal or seasonal variations of emissions with time. Assigning an estimate of precision to an emission factor for an individual AFO is more challenging than assigning it to a set of concentration and airflow measurements. The relative uncertainty associated with emission factors from individual AFOs can be obtained by remembering that emission factors are an estimate of emissions of particulate matter (PM) or a chemical species from a source. According to Equation 2-1, multiplying an emission factor by the AU, yields an emission rate. Integration of the emission rate over time (e.g., one year) yields the total mass emission from the source. For AFOs the total mass emission for a gaseous species containing nitrogen, carbon or sulfur must be a percentage of the total amount of that element excreted. If the individual AFO is in a steady state with regard to the excreted elements nitrogen, carbon, and sulfur, then the percent emissions of these elements should be relatively constant when averaged across several years. A certain percentage is retained for periods longer than one year (e.g., sludge accumulated at the bottom of treatment lagoons), but most of the elements excreted are applied to agricultural land for row crops and grasses, with the remainder emitted as gases or lost in leachate. The percentage of an excreted element lost as air emissions must fall between 0 and 100 percent, and it is highly unlikely to be at either extreme. Adoption of nutrient management plans further decreases the range of potential emission, since a certain percentage of the excreted nutrients will be used to support crop growth. The problem of determining the relative uncertainty associated with emissions from an individual AFO, then reduces to determining

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations the variation in the percentages of nitrogen, carbon, and sulfur lost from year to year. By way of example, if 60 percent of the excreted nitrogen on a swine AFO is assumed to be emitted to the atmosphere as NH3 (the value of 60 percent is selected for illustration purposes only and is not a value endorsed by the committee to be used to characterize AFOs), a 1 percent CV associated with this number would mean an uncertainty of ±0.6 percent, while a 10 percent CV would mean an uncertainty of ±6 percent. Given the dependence of NH3 volatilization on ambient air temperature, it is highly unrealistic to expect uncertainties of 1 percent CV; such uncertainties can be approached only in a laboratory environment. Values of CV of 10 percent or greater are probably much more realistic for real AFOs. Continuing with the example of 60 percent of the excreted nitrogen emitted as NH3, the range in uncertainty in emissions, and therefore calculated emission factors, associated with a 10 percent CV can be calculated directly based on the amount of nitrogen excreted and the number of animal units housed. For a finisher swine operation housing 10,000 head (4,000 AUs; 2.5 head per AU), the annual amount of nitrogen excreted is 1.37 x 105 kg using a nitrogen excretion factor of 13.7 kg N/yr per head (Doorn et al., 2002). (This nitrogen excretion factor assumes that 70 percent of nitrogen intake is excreted.) If 60 percent of excreted nitrogen is emitted as NH3, these numbers translate into an emission factor of 20.6 kg N/AU per year. Although the actual variation is not known, for the purpose of this example, a CV of 10 percent will be assigned, yielding a standard deviation of ±2.1 kg N/AU per year. Given a normal distribution in the percentage of excreted nitrogen lost as NH3, 95 percent (approximately two times the standard deviation) of the derived emission factors for this single AFO fall in the range of 16.4 to 24.8 kg N/AU per year. Carrying through the same calculations, and assuming instead that 80 percent of excreted nitrogen is released as ammonia, yields emission factors ranging from 21.9 to 32.9 kg N/AU per year. As noted above, these calculations are for illustration purposes only to demonstrate how a relatively modest variation in emissions from a single AFO (10 percent CV) translates into a range of potential emission factors. Yearly variations in emissions are to be expected and cannot be ignored. After careful evaluation of ammonia emissions from swine houses by various methods, Doorn et al. (2002) recommended a general emission factor for houses of 3.7 ± 1.0 kg NH3/yr per finished hog, which is a 27 percent CV. Groot Koerkamp et al. (1998) reported CVs ranging from 17 to 49 percent for different livestock and housing systems in England, the Netherlands, Denmark, and Germany, with between-season CVs ranging from 24 to 57 percent. Although the yearly variation in emissions from single AFOs is not well characterized, the assumed value of 10 percent CV used in the above calculations appears quite conservative compared to these measures of precision reported.

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations Viewing emissions as a percentage of an element excreted offers a means of estimating the relative uncertainty associated with emissions from individual AFOs. The approach will be most successful for those gaseous species (NH3, CH4 [methane], or H2S [hydrogen sulfide]) whose emissions comprise a substantial portion of the element (nitrogen, carbon or sulfur) excreted. For gaseous species whose emissions represent relatively minor fractions of these excreted elements (e.g., volatile organic compounds [VOCs]), the percent emission becomes less certain, but the approach still makes it possible to set an upper limit on emissions, and the use of percent CV values to estimate relative uncertainty still applies. This approach cannot be used for PM, whose emissions are not a direct function of the amount of a given element excreted, nor can it be applied to odors. In summary, to ensure that reasonable and appropriate estimates of emissions are obtained from AFOs, the measured and derived emission values must have accompanying measures of uncertainty, including accuracy and precision. Accuracy does not depend simply on instrument calibration; representativeness must be considered since AFOs may not be selected at random and there are no standard methods for measuring emissions. All measurements of emissions should be assumed to have systematic bias and should be compared to other measurements or derived data, such as excretion factors and mass balances. Methods to obtain an estimate of precision do exist and should be included in experimental designs. Short-term sequential measurements will undoubtedly be autocorrelated, and deriving estimates of precision by applying normal statistical techniques to such data will underestimate uncertainties. There are methods for deriving estimates of variance from highly autocorrelated data (Code of Federal Regulations, 2001). PUBLISHED LITERATURE What are the strengths, weaknesses and gaps of published methods to measure specific emissions and develop emission factors that are published in the scientific literature? Ammonia Several well-designed research studies have been published establishing some of the factors that contribute to variations in NH3 emissions. For example, Groot Koerkamp et al. (1998) reported wide variations in emissions for different species (cattle, sows, and poultry) measured in different European countries, across facilities within a country, and between summer and fall. Amon et al. (1997) demonstrated that emissions increase as animals age.

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations Differences due to the manure storage system have been demonstrated (Hoeksma et al. 1982). Climate, including temperature and moisture, also affects NH3 emissions (Hutchinson et al., 1982; Aneja et al., 2000). Zhu et al. (2000) reported diurnal variation in emission measurements. With so many sources of variation in NH3 emissions, it is unreasonable to apply a factor determined in one system, over a short period of time, to all AFOs within a broad classification. Although NH3 emissions have been reported under different conditions, there are few reliable data to estimate total NH3 emissions from all AFO components for all seasons of the year. Twenty-seven articles were used for NH3 emission factors by EPA (2001a); of these, only eleven with original measurements were from peer-reviewed sources. Additional data were taken from six progress reports from contract research. Two of these (Kroodsma et al., 1988; North Carolina Department of Environmental and Natural Resources, 1999) were identified as “preliminary,” and in one case (Kroodsma et al., 1988), the airflow measurement equipment was not calibrated. Emission factors for NH3 were also taken from nine review articles (EPA, 2001a); three of these modeled or interpreted previously reported information with the objective of determining emission factors (Battye et al., 1994; Grelinger, 1998; Grelinger and Page, 1999). Several of the reviews reported factors used in other countries, but not the original research used to develop them. Other reviews summarized data from primary sources that were already considered. Thus, the review articles may not provide new information. Most measurements and estimates reported did not represent a full life cycle of animal production. As animals grow or change physiological state, their nutrient excretion patterns vary, altering the NH3 volatilization patterns (Amon et al., 1997). A single measurement over a short period of time will not capture the total emission for the entire life cycle of the animal. In addition, most measurements for manure storage represent only part of the storage period. The emissions from storage vary depending on length of storage, changing input from the animal system, and seasonal effects such as wind, precipitation (Hutchinson et al., 1982), and temperature (Andersson, 1998). Only one article reported measurements over an entire year (Aneja et al., 2000), although the measurements may not have been continuous. In this case, NH3 emissions were measured from an anaerobic lagoon using dynamic flow-through chambers during four seasons. Summer emissions were 13 times greater than those in winter, and the total for the year was 2.2 kg NH3-N per animal (mean live weight = 68 kg) per year. Expressing NH3 emission factors on a per annum and per AU basis facilitates calculation of total air emissions and accounts for variation due to size of AFOs, but it does not account for some of the largest sources of variation in emissions. Clearly, there is a great deal of variation in reported measurements among AFOs represented by a single model. For example, only two references

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations were provided for beef drylot NH3 emission factors, but the values reported were 4.4 and 18.8 kg N/yr per animal (See Table 8-11, EPA, 2001a). For swine operations with pit storage, mean values reported in eight studies ranged from 0.03 to 2.0 kg/yr per pig of less than 25-kg body weight (See Table 8-17, EPA 2001a). This higher rate represents 66 percent of the nitrogen estimated to be excreted by feeder pigs per year (See Table 8-10, EPA, 2001a). The actual variation among AFOs represented by a single model cannot be determined without data representing the entire population of AFOs to be modeled. This would require greater replication and geographic diversity. Much of the variation among studies within a single type of model farm can be attributed to different geographic locations or seasons and the different methods and time frames used to measure the emission factors. The approach in EPA (2001a) was to average all reported values in selected publications—both refereed and non-refereed—giving equal weight to each article. Emission factors reported in some represented a single 24-hour sample, while in others, means of several samples were used. Emission factors from review articles were averaged along with the others. Properly using available data to determine emission factors, if it could be done, would require considering the uniqueness and quality of the data in each study for the intended purpose and weighting it appropriately. The causes of the discrepancies among studies would also have to be investigated. Adding emissions from housing, manure storage, and field application, or using emission factors determined without considering the interactions of these subsystems, can easily provide faulty estimates of total emissions of NH3. If emissions from a subsystem are increased, those from other subsystems must be decreased. For example, most of the excreted nitrogen is emitted from housing, much of the most readily available nitrogen will not be transferred to manure storage. If emissions occur in storage, there will be less nitrogen for land application. The current approach ignores these mass balance considerations, and simply adds the emissions using emission factors determined separately for each subsystem. Dividing the total manure nitrogen that leaves the farm by the total nitrogen excreted can identify some potential overestimation of emission factors. For example, using emission factors in Table 8-21 of EPA (2001a) for swine model farms, the total ammonia nitrogen emissions for 500 AUs in Model S2 can be estimated to be 1.12 x 104 kg/yr. (Three significant digits are carried for numerical accuracy from the original reference and may not be representative of the precision of the data.) The total nitrogen excreted by 500 AUs of growing hogs is 1.27 x 10 4 kg/yr (EPA, 2001a). Thus, one calculates that 90 percent of estimated manure nitrogen is volatilized to ammonia, leaving only 10 percent to be accumulated in sludge, applied to crops, and released as other forms of nitrogen NO [nitric oxide], N2O [nitrous oxide], and N2). Thus, these emission

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations factors suggest that almost all excreted nitrogen is lost as NH3, which seems unlikely. Nitric Oxide Although nitric oxide was not specifically mentioned in the request from the EPA, the committee believes that it should be included in this report because of its close relationship to ammonia. An appreciable fraction of manure nitrogen is converted to NO by microbial action in soils and released into the atmosphere. NO participates in a number of processes important to human health and the environment. The rate of emission has been widely studied but is highly variable, and emissions estimates are uncertain. Attempts to quantify emissions of NOx from fertilized fields show great variability. Emissions can be estimated from the fraction of the applied fertilizer nitrogen emitted as NOx, but the flux varies strongly with land use and temperature. Vegetation cover greatly decreases NOx emissions (Civerolo and Dickerson, 1998); undisturbed areas such as grasslands tend to have low emission rates, while croplands can have high rates. The release rate increases rapidly with soil temperature—emissions at 30°C are roughly twice emissions at 20°C. The fraction of applied nitrogen lost as NO emissions depends on the form of fertilizer. For example, Slemr and Seiler (1984) showed a range from 0.1 percent for NaNO3 (sodium nitrate) to 5.4 percent for urea. Paul and Beauchamp (1993) measured 0.026 to 0.85 percent loss in the first 6 days from manure nitrogen. Estimated globally averaged fractional applied nitrogen loss as NO varies from 0.3 percent (Skiba et al., 1997) to 2.5 percent (Yienger and Levy, 1995). For the United States, where 5 Tg of manure nitrogen is produced annually, NOx emissions directly from manure applied to soil are roughly 1 percent or 0.05 Tg/yr, neglecting emissions from crops used as animal feed. Williams et al. (1992) developed a simplified model of emissions based on fertilizer application and soil temperature. They estimated that soils accounted for a total of 0.3 Tg or 6 percent of all US NOx emissions for 1980. Natural variability of emissions dominates the uncertainty in the estimates. In order of increasing importance, errors in land use data are about 10-20 percent, and experimental uncertainty in direct NO flux measurements is estimated at about ±30 percent. The contribution of soil temperature to uncertainty in emissions estimates stems from uncertainty in inferring soil temperature from air temperature and from variability in soil moisture. Williams et al. (1992) show that their algorithm can reproduce the observations to within 50 percent. A review of existing literature indicates that agricultural practices (such as the fraction of manure applied as fertilizer, application rates used, and tillage) introduce variability in NO emissions of about a factor of two.

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations Variability of biomes to which manure is applied (such as short grass versus tallgrass prairie) accounts for an additional factor of three (Williams et al., 1992; Yienger and Levy, 1995; Davidson and Klingerlee, 1997). Future research may have to focus on determining the variability of emissions, measured as a fraction of the applied manure nitrogen, with agricultural practices, type of vegetative cover, and meteorological conditions. Hydrogen Sulfide Most of the studies on hydrogen sulfide emissions from livestock facilities were conducted recently and included current animal housing and manure management practices. Several recent publications from Purdue University document H2S emissions from mechanically ventilated swine buildings (Ni et al., 2002a, 2002b, 2002c, 2002d). A pulsed fluorescence SO2 analyzer with an H2S converter was used to measure H2S concentrations in the air, and a high-frequency (16 or 24 sampling cycles each day) measurement protocol was used for continuous monitoring. In one of the studies reported, H2S emission from two 1,000-head finishing swine buildings with under-floor manure pits in Illinois was monitored continuously for a six-month period from March to September 1997. Mean H2S emission was determined to be 0.59 kg per day, or 6.3 g per day per 500-kg animal weight. Based on emission data analysis and field observation, researchers noticed that different gases had different gas release mechanisms. Release of H2S from the stored manure, similar to carbon dioxide and sulfur dioxide, was through both convective mass transfer and bubble release mechanisms. In comparison, the emission of NH3 was controlled mainly by convective mass transfer. Bubble release is an especially important mechanism controlling H2S emission from stirred manure. The differences in release mechanisms for different gases are caused mainly by differences in solubility and gas production rates in the manure. Some measurements from swine buildings were also conducted in Minnesota (Jacobson, 1999; Wood et al., 2001). Very few data are available on H2S emission from other types of livestock facilities, such as dairy, cattle, and poultry. Using emission data from swine operations to estimate emission factors for other species such as dairy and poultry is not scientifically sound. Outside manure storage, such as storage in tanks or anaerobic lagoons, can be important sources of H2S emissions. Emission data for such sources are lacking in the literature. EPA (2001a) stated that H2S emissions from solid manure systems— such as beef and veal feedlots, manure stockpiles, and broiler and turkey buildings—were insignificant, based on the assumption that these systems are mostly aerobic. Such an assumption is not valid because it is not based on scientific information. Published data indicate that a significant amount of H2S

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations TABLE 2-1. Odor Emission Rates from Animal Housing as Reported in the Literature Animal Type Location Odor Emission (OU/s-m2)a Flux Rate Reference Nursery pigs (deep pit) Indiana 1.8a Lim et al., 2001 Nursery pigsb Netherlands 6.7 Ogink et al., 1997; Verdoes and Ogink, 1997 Nursery pigs Minnesota 7.3-47.7 Zhu et al., 1999 Finishing pigs Minnesota 3.4-11.9 Zhu et al., 1999 Finishing pigsc Netherlands 19.2 Ogink et al., 1997; Verdoes and Ogink, 1997 Finishing pigsd Netherlands 13.7 Ogink et al., 1997; Verdoes and Ogink, 1997 Finishing pigs (daily flush)e Indiana 2.1 Heber et al., 2001 Finishing pigs (pull-plug)e Indiana 3.5 Heber et al., 2001 Finishing pigs (deep pit) Illinois 5.0 Heber et al., 1998 Farrowing sows Minnesota 3.2-7.9 Zhu et al., 1999 Farrowing sows Netherlands 47.7 Ogink et al., 1997; Verdoes and Ogink, 1997 Gestating sows Minnesota 4.8-21.3 Zhu et al., 1999 Gestating sows Netherlands 14.8 Ogink et al., 1997; Verdoes and Ogink, 1997 Broilers Australia 3.1-9.6 Jiang & Sands, 1998 Broilers Minnesota 0.1-0.3 Zhu et al., 1999 Dairy cattle Minnesota 0.3-1.8 Zhu et al., 1999 Note: Rates have been converted to units of OU/s-m2 for comparison purposes. a Net odor emission rate (inlet concentration was subtracted from outlet concentration). b Number of animals calculated from average animal space allowance. c Pigs were fed acid salts. d Multiphase feeding. e Odor units normalized to European Odor Units based on n-butanol. SOURCE: Adapted from Sweeten et al. (2001). CHARACTERIZING VARIABILITY How should the variability in emissions be characterized that is due to regional differences, daily and seasonal changes, animal life stage, and different management approaches? Each model farm proposed by EPA (2001a; Appendix D) includes three variable elements: a confinement area,

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations manure management system, and land application method. The manure management system was subdivided into solid separation and manure storage activities. The model farm assumes that emissions depend primarily on the category identified for each individual element. The potential influences of regional differences, hourly, daily and seasonal changes, animal life stages, and different management approaches are not explicitly considered. Climatic and Geographic Differences Differences in climate will influence emissions from AFOs because of differences in temperature, rainfall frequency and intensity, wind speed, topography, and soils. EPA (2001a) notes several possible influences of climatic differences by acknowledging the influence of air temperature on gaseous emissions and the effect of rainfall frequency on stocking densities at cattle and dairy feedlots. Climatic differences per se were excluded from the criteria used to select emission factors from the scientific literature; however the van’t Hoff-Arrhenius equation was used to adjust CH4 conversion factors for mean temperature differences (See Chapter 8; EPA, 2001a). Increases in mean ambient temperature are expected to increase gaseous emission rates from several components of the model farms, including manure storage and land to which manure has been applied. It is unclear how averaging reported emission factors would remove this influence of temperature, especially if the selected emission factors used were mostly determined in climatic region of the country. The same logic applies to estimates of emissions from housing units or land. Depending on one or two published emission factors from one region of the country results in a possible systematic bias because of climatic differences. This bias is still present when emission factors for one species are applied to others by adjusting them to reflect differences in excretion rates, or by assuming that emissions from an anaerobic poultry lagoon are similar to those from an anaerobic swine lagoon (See Chapter 8, EPA, 2001a). Differences in emissions from AFOs may also arise because of other geographic differences such as availability of land for manure or lagoon effluent disposal, rates of evapo-transpiration, and differences in soil texture and drainage that can impact application rates of lagoon water, or differences in soil microenvironments that affect microbial action and the resulting gaseous emissions. The breed of a given animal species (e.g., selection for cold or heat tolerance) and feed formulations (due to changes in animal maintenance requirements) may also vary in response to geographic and climatic differences. It is difficult to project how these various sources of uncertainty will combine to influence gaseous emissions and whether these factors will have significant impact on total percentages of nitrogen, carbon or sulfur lost in

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations gaseous species, when averaged over a year’s time. Climatic differences do not negate the mass balance flow of elements through AFOs, so that, unless there is a significant change in storage of an element within the manure management system, changes in total emissions (air and water) can come about only because of changes in excretion (resulting from changes in feed formulation or efficiency of animal nutrient utilization). Differences may not be as important for annual emissions of major gaseous species (such as NH3 and CH4) as for VOCs and PM. Hourly, Daily, and Seasonal Changes Changes in emissions from individual AFOs due to hourly, daily, and seasonal variations are discussed here because measurements to characterize emissions are usually conducted for short periods of time, preferably during different seasons of the year. Failure to account for short-term cycles in an experimental design used to characterize emissions could result in significant systematic error in a derived emission factor, when extrapolated to a one-year time period. Individual AFOs are essentially a collection of different biological systems each operating with its own hourly, daily, and seasonal cycles. At the scale of the individual animal, there are daily cycles in activity related to eating, defecating, and moving about (the latter being particularly important for generating PM from cattle feedlots). Microbial cycles that produce emissions may be closely tied to animal activity through the amount and frequency of defecation. As an animal grows, the amount and composition of its feed intake change, as does the amount and composition of its manure (National Research Council, 1994, 1998, 2000, 2001). This gives rise to corresponding changes in total microbial activity and emissions. Lactating animals experience changes in productivity throughout their natural cycle, with changes in feed consumed and nutrients excreted (National Research Council 1998, 2000, 2001). Although the capacity within an AFO remains essentially constant, a number of different animals may occupy this space during the year, depending on the production cycle used. Thus, the cycling of animals through an AFO is another source of variation in emissions. Upsets in daily rhythms of animals must also be considered, because they may result in changes in feed uptake and nutrients excreted for a period of several days. Such upsets may occur due to illness, drastic short-term changes in weather, or breakdowns of farm equipment. Depending on the manure management system being employed, such event-driven processes may not be significant in terms of emissions of NH3 or CH4 but may have a major impact on other emitted species such as VOCs and PM. Other event-driven processes that can occur include lagoon turnover, flush cycles for housing units, and manure

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations scraping at feedlots. As noted by EPA (2001a), these events can result in enhanced emissions. The impact of daily cycles on emissions is not important when averaged over a yearly time scale, provided a sufficient number of observations are made to account for such cycles. However, given the paucity of emissions data deemed valid for the development of emission factors to characterize the model farms, it is not possible to determine to what extent such cycles may have impacted published emission measurements. As noted earlier, averaging published emission factors does not compensate for the presence of systematic bias that may be present as a result of a failure of the experimental design to account adequately for such cycles. Animal Life Stage Reference has already been made to differences in feed formulations that occur during the life cycles of most animals produced at AFOs, and the subsequent effects on the amount and composition of fecal matter and urine excreted. In this section, a specific example is provided (Figure 2-1) of changes in the rate of nitrogen excreted for “grow-finish” swine produced at AFOs in the southeastern United States. The data are based on a growth model (Agricultural Research Council, 1981) used by a commercial swine producer to adjust feed formulations. To prevent the disclosure of proprietary information, data have been normalized to 100 percent for the highest rate of nitrogen excretion per day. As expected, the relative amount of nitrogen excreted daily tends to increase as the pig grows, reflecting changes in the daily total nitrogen consumed. The actual feed formulation is changed four times during the growth cycle of the hog (not twice as assumed by EPA, 2001a) to account for changes in nitrogen required for maintenance and growth. The changes in the relative amount of nitrogen excreted per day with changes in formulation are not simply an artifact of the model but reflect periods of adjustment by the animal to the changes in feed composition. Overall there is a series of curvilinear increases in the amount of nitrogen excreted per day for finishing swine under this model, with nitrogen excretion nearly doubling during the latter half of the animal’s growth period. The emphasis in Figure 2-1 is on total nitrogen excreted. Expressed as a percentage of body weight, the nitrogen excreted would actually be decreasing throughout the growth cycle. Figure 2-1 illustrates that if daily housing emissions of NH3 are directly related to daily nitrogen excretion and the model is an accurate representation of nitrogen excretion, then there will not be a simple increase in emissions from the

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations FIGURE 2-1. Relative excretion rate of nitrogen versus day in the life cycle of a grow-finish hog at a commercial swine production facility in the southeastern United States. Animals attain the designation of grow-finish hog at approximately day 40 in their life cycle and are finished at about day 174. Note: Relative excretion rates refer to kilograms of nitrogen per day excreted on day n relative to day 174. confinement unit with time. Thus, averaging together emission measurements made from several different housing units with different age animals, or from the same housing unit during different times during one growth cycle, may significantly under- or overestimate emissions, depending on the age of the animals when sampled. Actual emissions, however, will also depend on the manure collection practices (flush frequency, pit recharge, pull plug, or pit storage) associated with the confinement unit. A manure collection practice that accumulates manure for relatively long periods of time, such as pit storage, may act to smooth the variations in emissions due to variations in daily excretion of nitrogen. At a minimum the data displayed in Figure 2-1 demonstrate that the same sampling scheme may not be applicable to all swine confinement units and that measurements of emissions may have to be weighted to account for differences in animal age. Management Optimal management is vital to the success of individual AFOs for the production of quality animals, and should also result in decreased emissions. Appropriate drainage and manure removal minimizes PM generation from cattle

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations feedlots (Sweeten et al., 1998). Effects of animal health on feeding habits is important to maintain consistent nutrient uptake efficiency and prevent feed spoilage. This attention includes maintenance of proper ventilation for animals in confined housing units, maintenance of drainage systems to remove wastes from housing units on a frequent basis, and regular (perhaps daily) visual inspection of animals and their daily routines. Adherence to nutrient management plans will reduce the potential for of excessive air emissions or surface runoff resulting from overapplication of nutrients to crops. Anaerobic lagoons should not exceed design-loading rates and should be maintained at the proper pH range for waste stabilization. Assessing the overall quantitative impact of effective management on decreasing emissions is currently not possible due to the paucity of emissions data. However, management practices should not be excluded in assessing emissions from individual AFOs. One way to achieve this goal would be to determine whether managers at AFOs where measurements of emissions are scheduled are in compliance with animal industry guidelines for decreasing emissions, including odors. An illustration of one such program is the America's Clean Water Foundation’s (ACWF) On Farm Assessment and Environmental Review (OFAER) project (2002), which reportedly provides livestock producers a confidential, comprehensive, and objective assessment of water quality, odor, and pest risk factors at their operations. (Reference to the OFAER program is for illustration purposes only and should not be construed as an endorsement of this program by the committee or the National Research Council.) The OFAER project currently has the participation of approximately 3,200 AFOs nationwide. Using voluntarily provided emission factors from individual AFOs may produce the database necessary to assess the impact of management on emissions. In summary, the answer to the question of how the variability in emissions due to regional differences, hourly, daily, and seasonal changes, animal life stage, and different management approaches should be characterized is through consideration of these factors in experimental designs for measuring emissions and deriving emission factors. Average ambient temperatures are the main differences among different regions of the country. Selecting an emission factor based on data from one region (e.g., the southeastern United States) and extrapolating it to other regions or even to other animal types is questionable at best, and must necessarily introduce systematic bias into the derived emission rates for individual AFOs. Because of the importance of temperature effects on microbial activity and gas exchange across different interfaces, accounting for regional differences must include actual measurements of emissions at AFOs across the United States. Consideration of daily and seasonal changes and animal life stages speaks to the need to consider variations in emissions that occur on the same time scale as most field measurements of emissions at AFOs. Proper characterization of these variations will require experimental designs that

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations encompass the full life cycle of the animals under production and consider whether measured emission rates are nonlinear during the typical animal life cycle. If emissions are in fact nonlinear, then observations of emission rates have to be weighed accordingly when extrapolating to a one-year time frame. Since AFOs will probably never be chosen at random for field measurements of emissions, selection criteria should be developed for what constitutes an acceptable AFO for field measurements. These criteria should include an evaluation of management and reflect the growing volunteer effort to address water quality and odor and pest issues, for example the ACWF OFAER project. STATISTICAL UNCERTAINTY How should the statistical uncertainty in emissions measurements and emissions factors be characterized in the scientific literature? As noted earlier in this chapter, uncertainty can be described in terms of accuracy and precision. Deviations from accuracy (systematic bias) for individual measurement technologies will be addressed in more detail in the final report. This section addresses the broader issue of uncertainty associated with published emissions data and their use in deriving emission factors. An example of the uncertainty associated with published emission rates from AFOs is illustrated in Table 2-2 adapted from Tables 9 and 10 in a recent review paper (Arogo et al., 2001) summarizing recently published measurements of NH3 flux (kilograms of NH3-N per hectare per day) from primary anaerobic swine lagoons. Multiplying the fluxes by the lagoon surface areas gave the daily emission rates for various seasons. The majority of observations listed in Table 2-2 were from “farrow-finish” AFOs, with the remainder from “farrow-wean,” “grow-finish,” and “breed-wean” facilities. The range in lagoon pH values was 6.8-8.3, but the majority were between 7.4 and 8.2. The variability in the calculated emission rates in the table is evident in the range of values listed for each combination of measurement method and measurement period, with typical factors of 3 to 7. Seasonal differences in emission rates are also evident, with the ratio of summer to winter rates being as large as 10 or more. Within-lagoon variation in total ammoniacal nitrogen (TAN) is much less, but between lagoons the values vary by factors as high as 10. There is also no obvious association between TAN concentrations in the lagoons and calculated emission rates. The range of rates for individual lagoons is evidence of the uncertainty that must be associated with emission factors derived from published emission rates. Failure to document this uncertainty in tabulated values of emission factors can lead to unrealistic expectations

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations TABLE 2-2. Calculated Emission Rates of Ammonia from Primary Anaerobic Swine Lagoons as a Function of Measurement Method and Measurement Period Measurement Methoda Period TANb mg/L Emission Rate (kg NH3-N/d) Reference Micromet. Aug-Oct 917-935 29-51 Zahn et al. (2001) Micromet. Summer 230-238 11.2-140 Harper et al. (2000) Micromet. Winter 239-269 4.6-6.7 Harper et al. (2000) Micromet. Spring 278-298 11-34 Harper et al. (2000) Micromet. Summer 574 42-59 Harper and Sharpe (1998) Micromet. Winter 538 14-33 Harper and Sharpe (1998) Micromet. Spring 741 14-42 Harper and Sharpe (1998) Micromet. Summer 193 7.0-20 Harper and Sharpe (1998) Micromet. Winter 183 14-22 Harper and Sharpe (1998) Micromet. Spring 227 7.2-16 Harper and Sharpe (1998) Chamber Summer 587-695 145 Aneja et al. (2000) Chamber Fall 599-715 30 Aneja et al. (2000) Chamber Winter 580-727 11 Aneja et al. (2000) Chamber Spring 540-720 63 Aneja et al. (2000) TG OP-FTIR May - 93-305 Todd et al. (2001) TG OP-FTIR November - 20-169 Todd et al. (2001) Chamber September 101-110 0.44-2.7 Aneja et al. (2001) Chamber November 288-311 0.04-0.14 Aneja et al. (2001) Chamber November 350 0.17-0.62 Aneja et al. (2001) Chamber Feb/March 543-560 0.35-2.6 Aneja et al. (2001) Chamber March 709-909 0.32-1.2 Aneja et al. (2001) Chamber April-July 978-1143 319 Heber et al. (2001) Chamber May-July 326-387 48 Heber et al. (2001) a Micromet. = micrometeorological; TG OP-FTIR = tracer gas open path fourier transform infrared spectroscopy; Chamber = dynamic flow through chamber. b TAN = total ammoniacal nitrogen. SOURCE: Data derived from Tables 9 and 10, Arogo et al., 2001 regarding the accuracy of emissions calculated for individual AFOs. In addition, large uncertainties associated with emission rates for the principal components of a manure management system reduce the probability of documenting success in the application of emission reduction technologies. As a first approximation, estimates of the variance associated with emission rates, such as those in Table 2-2 can be obtained using normal

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations statistical procedures. If estimates of the variance are included in published reports, then the variance associated with the derived emission factor can be calculated by using well-known formulas for the propagation of error (Beers, 1957), and assuming no significant autocorrelation between sequential observations. As noted earlier, emissions from AFOs are most likely parts of time series with autocorrelation between observations, especially those taken over relatively short periods of time (hours or days). The presence of autocorrelation within a data set means that calculated values for the variance of the sample mean using standard statistical procedures will be biased low, and that the overall uncertainty for a derived emission factor will be underestimated. When values of the variance associated with emission rates are not included in the published literature, very rough approximations of the population variance can be obtained from the range of reported values (Natrella, 1963; Deming, 1966). For example, if it is assumed that the data follow a normal distribution and the reported range in emission rates encompasses 95 percent of the sample population, the estimate of the population standard deviation (σ ) is (Eq. 2-3) Values for the denominator in Equation 2-3 range from 3.5 (random) to 4.9 (triangular) for other assumed shapes of the data distribution (Natrella, 1963). For the purposes of this report, the data are assumed to follow a normal distribution. Applying Equation 2-3 to the data in Table 2-2, and assuming that each population mean is equal to the average of the minimum and maximum values, we find percent CV values ranging from 8.4 to 42.6 for the individual combinations of measurement method and measurement period, with a mean (for 17 entries) of about 25 percent. This is similar to values noted earlier for field measurements (Groot Koerkamp et al., 1998; Doorn et al., 2002), and reinforces the argument that the uncertainty associated with published values of emission rates (or flux) cannot be ignored when deriving emission factors. These calculations illustrate that at a minimum, a derived emission factor for NH3 emissions for a single AFO based on the data in Table 2-2 will probably have an associated CV of at least 25 percent. This is a minimum estimate because our calculations using the data in Table 2-2 are based only on the within-study variance. The approach for estimating uncertainty represented by Equation 2-3 can provide only a rough estimate of the standard deviation of the sample population. If the reported range in emission rates represents a limited number of observations, then the assumption that the range encompasses 95 percent of the possible observable values is less likely to be true. Proper characterization of

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations the uncertainty associated with emissions in the published literature, therefore, also requires knowledge of the number of observations. This is especially important when averaging values for derived emission factors as is done by EPA (2001a). Simple averaging implies equality in the uncertainties associated with the emission rates used to determine emission factors. In reality, the actual numbers of observations associated with reported values in the published literature vary substantially among investigators, requiring serious consideration of weighted averaging as a more valid means of calculating emission factors. Developing a weighting protocol will require examination of the experimental design employed for each set of emissions data considered, determining the most likely sources of variation in the reported values, and considering whether the experimental design gathered sufficient data to obtain realistic estimates of this variation. Weighted averaging is not considered by EPA (2001a). The model farm construct proposed by EPA (2001a) attempts to reduce the uncertainty in deriving emission factors for individual AFOs by subdividing the overall AFO population according to the manure management systems used. Subdivision of large sample populations into smaller subsets is an acceptable procedure to reduce uncertainty (i.e., improve sample quality). The measurement of emissions from an individual AFO (or component of an individual AFO) will necessarily be interpreted as being representative of all AFOs in a defined subset of the larger sample population. However, further subdivision of the sample population also increases the need for data in terms of emission rates and emission factors. This approach must necessarily reach a point of diminishing returns. Emission rate measurements obtained on two AFOs using the same management schemes for animal housing and manure handling will likely not be the same. To include both operations in the same sample AFO population will therefore require the overall uncertainty in the emission factor to be increased to allow both to be part of the same statistical population. Attempting to use only a mean value for a sample population to characterize an individual member of that population must necessarily have a large degree of uncertainty associated with it. To decrease this uncertainty, specific information concerning the individual member of the sample population to be characterized must be included in deriving the estimated value. This necessarily will increase the complexity of the model used to describe individual members of the population and therefore the size of the database required to accomplish the desired goal. In summary, an example has been given of how the statistical uncertainty in emissions measurements and emissions factors can be characterized in the scientific literature, provided sufficient information is available in published reports. The example speaks solely to the issue of precision and cannot address the question of accuracy (systematic bias) of the reported values. However, issues concerning systematic bias have been addressed elsewhere in this chapter. Failure of investigators to note the degree

OCR for page 23
The Scientific Basis for Estimating Air Emissions from Animal Feeding Operations of uncertainty associated with their reported values for emission rates may be a reflection of the limited number of observations upon which their reported values are based. Equal weighting should not be given to reported emission rates and derived emission factors when the actual number of observations on which these reported values are based differs significantly among investigators. All other things being equal, reported values for emissions based on a relatively large number of observations should be given greater weight than those derived from relatively few observations. As presented in this chapter, a wide range of factors can influence air emissions of gases, PM, and other substances from AFOs. Combinations of these factors that will be most useful in pursuing regulatory goals will depend on research-based information about the strength of the relationship between each combination of factors and the rate of emission of a particular pollutant. Finding 5: Reasonably accurate estimates of air emissions from AFOs at the individual farm level require defined relationships between air emissions and various factors. Depending on the character of the AFOs in question, these factors may include animal types, nutrient inputs, manure handling practices, output of animal products, management of feeding operations, confinement conditions, physical characteristics of the site, and climate and weather conditions.