# Journal of Environmental Quality - Article

1. OPEN ACCESS

Received: Apr 09, 2013
Published: October 4, 2013

* Corresponding author(s):
View
Permissions

doi:10.2134/jeq2013.04.0121

# Ammonia Emission Model for Whole Farm Evaluation of Dairy Production Systems

1. C. Alan Rotz *a,
2. Felipe Montesb,
3. Sasha D. Hafnerc,
4. Albert J. Heberd and
5. Richard H. Grante
1. a USDA-ARS, Bldg. 3702, Curtin Rd., University Park, PA 16802
b Research Associate, The Pennsylvania State Univ., University Park, PA 16802
c Hafner Consulting LLC, Washington DC 20011
d Professor, Dep. of Agricultural & Biological Engineering, Purdue Univ., 225 South University St., West Lafayette, IN 47907-2093
e Professor, Dep. of Agronomy, 215 Plant and Soils, Purdue Univ., West Lafayette, IN 47907-2093. USDA is an equal opportunity provider and employer

## Abstract

Ammonia (NH3) emissions vary considerably among farms as influenced by climate and management. Because emission measurement is difficult and expensive, process-based models provide an alternative for estimating whole farm emissions. A model that simulates the processes of NH3 formation, speciation, aqueous-gas partitioning, and mass transfer was developed and incorporated in a whole farm simulation model (the Integrated Farm System Model). Farm sources included manure on the floor of the housing facility, manure in storage (if used), field-applied manure, and deposits on pasture (if grazing is used). In a comprehensive evaluation of the model, simulated daily, seasonal, and annual emissions compared well with data measured over 2 yr for five free stall barns and two manure storages on dairy farms in the eastern United States. In a further comparison with published data, simulated and measured barn emissions were similar over differing barn designs, protein feeding levels, and seasons of the year. Simulated emissions from manure storage were also highly correlated with published emission data across locations, seasons, and different storage covers. For field applied manure, the range in simulated annual emissions normally bounded reported mean values for different manure dry matter contents and application methods. Emissions from pastures measured in northern Europe across seasons and fertilization levels were also represented well by the model. After this evaluation, simulations of a representative dairy farm in Pennsylvania illustrated the effects of animal housing and manure management on whole farm emissions and their interactions with greenhouse gas emissions, nitrate leaching, production costs, and farm profitability.

### Abbreviations

bLS, backward Lagrangian Stochastic; DM, dry matter; IFSM, Integrated Farm System Model; RPM, Radial Plume Mapping; TAN, total ammoniacal nitrogen

Gaseous emissions from animal agriculture have become an important issue in the United States and in many other countries. Emissions include greenhouse gases, volatile organic compounds, and specific toxic compounds, of which ammonia (NH3) is the most important. Ammonia can create human and animal health hazards when concentrations reach critical levels in confined spaces, such as enclosed manure storages or poorly ventilated barns (NRC, 2003). When large quantities of these compounds are released to the atmosphere, they contribute to poor air quality and ecosystem degradation. Atmospheric NH3 contributes to the formation of fine particulate matter, which can cause or exacerbate human respiratory problems (Kampa and Castanas, 2008). Ammonia emission can also cause regional degradation of terrestrial and aquatic ecosystems (NRC, 2003).

Ammonia emission is regulated by the USEPA in response to the Clean Air Act (USEPA, 1990). Additionally, the Emergency Planning and Community Right-To-Know Act requires the reporting of a release that exceeds 45 kg within any 24-h period (USEPA, 2010), and daily emissions from large animal feeding operations can exceed this level. Accurate measurement and monitoring of animal feeding operations is very difficult and costly due to the many sources on farms, the large areas of exposure, and the relatively low concentrations in air. This has led to the recommendation that a process-based modeling approach be used to estimate emissions from animal feeding operations (NRC, 2003). By simulating farm systems with a comprehensive model that includes components representing the formation and emission processes and their interactions with other farm processes, individual source and total farm emissions can be estimated.

As discussed in other reviews (Ni, 1999; Montes et al., 2009), much work has been completed over the past 30 yr to model NH3 emission processes from manure sources on farms. Most models are applied to specific emission sources, such as the housing facility, manure storage, or field-applied manure. A few models have integrated the various sources to predict emissions at the farm level (Hutchings et al., 1996; Pinder et al., 2004; Rotz and Oenema, 2006; Li et al., 2012).

Despite the work that has been conducted, a need exists for a comprehensive farm-scale model that represents all of the major sources of NH3 emission and their interaction with other farm processes. Some of the earliest models of NH3 emission from manure were based on an empirical mass transfer coefficient for NH3 derived by Haslam et al. (1924), and much of the work since has followed this approach. Although this mass transfer model did not properly represent emissions from a flat surface of manure, parameters such as the Henry’s constant and dissociation coefficient were adjusted to allow models to fit empirical data. This has led to confusion, with many diverse models representing the same processes (Ni, 1999; Montes et al., 2009). Recent work has identified more fundamental and consistent relationships for representing the principal NH3 emission processes (Montes et al., 2009; Hafner et al., 2013).

Our objective was to develop and evaluate a process-based model that predicts NH3 emissions from all important sources in cattle production systems for use in whole farm simulation. This was accomplished by further developing and applying relationships for predicting emission processes as affected by the important farm components of animal housing, manure storage, land application of manure, and deposits by grazing animals.

### Model Description

The emission model was developed for incorporation into the Integrated Farm System Model (IFSM) for use in simulating and evaluating dairy and beef production systems (USDA-ARS, 2012). This comprehensive farm model provides a framework for linking process-level simulation of emissions for each of the major farm sources. This emission component can be adapted to other farm-scale models and animal species.

### Integrated Farm System Model

The IFSM is a research tool used to assess and compare the environmental and economic sustainability of farming systems. Crop production, feed use, and the return of manure nutrients back to the land are simulated for many years of weather on a crop, beef, or dairy farm (Rotz et al., 2012). Growth and development of crops are predicted for each day based on soil water and N availability, ambient temperature, and solar radiation. Simulated tillage, planting, harvest, storage, and feeding operations predict resource use, timeliness of operations, crop losses, and nutritive quality of feeds as influenced by weather. Feed allocation and animal responses are related to the nutrient contents of available feeds and the nutrient requirements of the animal groups making up the herd. The quantity and nutrient contents of the manure produced are a function of the feeds consumed and herd characteristics.

Nutrient flows are tracked through the farm to predict nutrient losses to the environment and potential accumulation in the soil (Rotz et al., 2012). Environmental losses include NH3 volatilization, denitrification and leaching losses of N from soil, erosion of sediment across the farm boundaries, and the runoff of sediment-bound and dissolved phosphorus. Carbon dioxide, methane, and nitrous oxide emissions are tracked from crop, animal, and manure sources and sinks to predict net greenhouse gas emission. Whole-farm mass balances of N, P, K, and C are determined as the sum of nutrient imports in feed, fertilizer, deposition, and legume fixation minus the nutrient exports in milk, excess feed, animals, manure, and losses leaving the farm.

Simulated performance is used to determine production costs, incomes, and economic return for each year of weather. A whole-farm budget includes fixed and variable production costs (Rotz et al., 2012). All important production costs are subtracted from the total income received for milk, animal, and feed sales to determine a net return to management. By comparing simulation results, differences among production systems are determined, including annual resource use, production efficiency, environmental impacts, production costs, and farm profit. Simulations are conducted over a 25-yr sample of recent historical weather, so the resulting distribution of annual predictions represents the effects of varying weather.

### Emission Processes

Emissions are predicted through the simulation of NH3 formation and emission processes as influenced by farm processes (Fig. 1). Manure is a primary source of NH3 emissions on livestock farms. Major manure sources include the housing facility, the storage (if one is used), field-applied manure, and deposits on pasture if grazing is used. The IFSM predicts the quantity and nutrient content of excreted feces and urine, which are used to establish the initial characteristics of the manure. In previous work, the IFSM-simulated dry matter (DM) and N excretions were shown to accurately represent individual cow and herd excretions (Rotz et al., 2006; Rotz et al., 1999). Losses and transformations in each stage are used to predict the manure characteristics at the succeeding stage. These conditions are then used to drive the simulation of the emission processes at each source, normally on an hourly time step throughout each day. Emission processes include the formation of ammoniacal N, the mass transfer within the manure, the speciation of NH3, and the mass transfer to the surrounding air. Relationships used to represent these processes are provided in Supplemental Table S1. These processes are simulated for the conditions of each day, and daily emissions from all sources are summed to give annual emissions.

Fig. 1.

Flow of information between farm and emission processes and their link to the Integrated Farm System Model (IFSM).

Total manure N consists of organic and ammoniacal N where only the ammoniacal form is readily volatilized. Depending on how the cattle are fed, 40 to 50% of this organic N is in the form of urea excreted in urine (Rotz, 2004). Excess protein N fed to cattle ends up in the urine, increasing the concentration of urea in the mixture of feces and urine (Rotz et al., 2012). After excretion, enzymatic hydrolysis quickly decomposes the urea to aqueous unionized NH3, which exists in equilibrium with ammonium (NH4+), and the sum of the concentrations of NH3 and NH4+ give the total ammoniacal N (TAN) content. The transformation of urea to TAN is modeled on an hourly time step as a function of temperature and the urea concentration in the manure (Supplemental Table S1). Because the pH of fresh cattle manure normally falls within an optimum range for urease activity (Sommer et al., 2006), pH effects on hydrolysis are ignored.

The distribution of TAN between NH3 and NH4+ in the manure solution (i.e., TAN speciation) is modeled using thermodynamic equilibrium principles (Supplemental Table S1) (Stumm and Morgan, 1996). Below pH 8, a one-unit increase in pH increases the NH3 fraction by about an order of magnitude, and this fraction approximately doubles with each 10°C increase in temperature. As the NH3 fraction in a solution increases, the potential emission rate increases.

Manure is not a dilute solution, so solute interactions are considered when predicting speciation. A fixed ionic strength is assumed and the Davies equation is used to incorporate nonideal behavior. To represent typical cattle manure, an average ionic strength of 0.35 is used (Chaoui et al., 2009), which gives an activity coefficient of 0.74 for NH4+ (Montes et al., 2009). Because NH3 has no charge, its activity coefficient is close to 1.0. To include activity corrections, the dissociation constant is multiplied by 0.74 (Supplemental Table S1).

Because NH3 concentration is sensitive to pH, knowing the pH at the surface of the manure is critical for an accurate prediction of emission rate. When manure is exposed to air, dissolved CO2 is released more rapidly than NH3 due to its lower solubility. The rapid loss of CO2 leads to a 0.5- to 1.0-unit increase in manure surface pH above the manure bulk pH (Chaoui et al., 2009; Ni et al., 2009; Blanes-Vidal et al., 2009; Montes et al., 2009). The magnitude of the pH increase depends on solution chemistry, manure depth, and environmental properties (Hafner et al., 2013). On a barn floor with constant animal movement, there is continuous mixing of the manure, so the surface pH likely varies spatially across a manure-covered surface. Due to the lack of a process model and the expected variability, surface pH is modeled as a fixed amount greater than the bulk manure pH, and this amount varies among emission sources.

The movement of NH3 away from the manure surface into the surrounding atmosphere is modeled using a mass transfer coefficient (Supplemental Table S1). A two-film interface mass transfer model is used to simulate the transfer of ammonia from the soil or manure solution through the liquid–air interface into the atmosphere. The model assumes the presence of quiescent thin films on both sides of the interface where mass transfer is dominated by molecular diffusion. Because NH3 is very soluble, the liquid coefficient has little effect on overall mass transfer.

In a large volume of manure such as in a storage tank, aqueous phase mass transfer can become limiting. As the NH3 is emitted, there is a drop in the concentration of TAN at the surface. This forms a gradient in concentration from the bulk material to the surface, and the TAN migrates from the high concentration at lower depths toward the lower concentration at the surface. The rate of this migration is controlled by the resistance to mass transfer (Supplemental Table S1).

The movement of NH3 away from the manure surface into the surrounding atmosphere is modeled using a mass transfer coefficient. The hourly rate of emission is a function of this overall mass transfer coefficient and the difference in NH3 concentration between the manure surface and surrounding atmosphere (Supplemental Table S1). Ammonia concentration in the ambient air is negligible and is thus set to zero. The NH3 concentration in the manure is calculated from the bulk TAN concentration.

### Farm Processes

By linking models for the emission processes, emission rates are predicted for each of the major farm sources. Farm processes affecting each source are modeled to determine the appropriate manure and environment characteristics, and the emission process routine described above is used to predict emissions. Farm processes are linked by accounting for losses and transformations as the manure is moved to subsequent farm processes. For example, the total quantity of manure and nutrient constituents removed from the housing facility provide the amounts placed into storage or that applied daily to fields when storage is not used.

### Animal Housing

Housing facilities include free stall, tie stall, and bedded pack barns and open lots. Exposed manure surface areas in each are established based on typical designs for cattle housing (Penn State Extension, 2013). The soiled areas assigned to tie stall, free stall, bedded pack, and open lot facilities are 1.2, 3.5, 3.0, and 5.0 m2 per cow, respectively. For growing animals, the areas are 1.0, 2.5, 2.0 and 3.2 m2 per animal. Manure is represented as a thin layer with a uniform concentration of TAN below the liquid film. Urea hydrolysis is the primary source of TAN in the manure deposited over the barn floor. Diffusion processes are neglected because the layer of manure is thin.

The mass of TAN on the floor of the housing facility is a function of the time animals spend in the facility, amount excreted, manure removal rate, and rates of urea hydrolysis and NH3 emission. These processes occur simultaneously in our model on a 1-h time step. During the day, urea N accumulates in proportion to time and excretion rate. Urine and fecal production and N excretion are functions of animal size, feed intake, protein intake, and milk production (Rotz et al., 2012). The amount of urea N excreted is set at 70% of the total urine N, with 9% of the fecal N and 1% of the urine N excreted as ammoniacal N (Bristow et al., 1992; Rotz, 2004). All remaining N is in more stable organic forms that do not contribute to emissions from the housing facility.

Manure removal rate is a function of housing type. The fraction of the manure removed each day by scraping, flushing, or infiltration into the manure pack is set to 0.98, 0.98, 0.90, 0.30, and 0.20 for tie stall, flushed free stall, scraped free stall, bedded pack, and open lot facilities, respectively. The portion not removed remains on the exposed surface where emission can continue. As the urea accumulates, the rate of urea N conversion to TAN is determined, and the ammonia emission rate is predicted. Emission rates are determined separately for the lactating cow and growing animal facilities due to differences in manure excretion, composition, and management.

When manure is removed by flushing, three changes are made to account for differences compared with scraped manure. After a scraping operation, a thin layer of manure is spread over the surface. This causes increased CO2 loss, which increases surface pH and the resulting emission rate. After flushing, a cleaner and wetter floor surface follows, which removes this effect on surface pH. This response is modeled by setting the surface pH equal to that of the bulk manure pH for the first hour after flushing. For manure accumulated after the first hour, the surface pH is modeled the same as after scraping. The removal factor is also increased to 0.98 to represent a cleaner floor immediately after removal. The third change is that the urinary N deposited after flushing is diluted by increasing the volume of solution on the floor by 20%.

Low-emission barn floors are being developed, evaluated, and used to a limited extent to reduce NH3 emission. Different floor designs are used where urine flows into channels and is carried to a holding area while feces remain on the barn floor. Partial separation of the urine and feces reduces hydrolysis of the urea in the barn. The urine and feces are typically combined for storage where the emission processes continue but at a much lower rate due to less exposed surface area per unit volume. This housing option is modeled by removing half of the urine and the associated TAN from the barn floor and placing it into the manure storage. The remainder enters the storage along with the feces after removal from the barn. This has been shown to appropriately represent the effect of this housing system (Rotz and Oenema, 2006).

Important parameters for predicting housing emissions are temperature, air velocity, and manure pH. For open facilities, temperature is set to that of ambient air. For enclosed, mechanically ventilated barns, air temperature in the barn is set as a function of the outside air temperature to account for increasing ventilation rates with increasing temperature (Supplemental Table S2). Air velocity is assumed to be equal to the ambient wind speed for open lots and at half the ambient wind speed for open, naturally ventilated barns. For mechanically ventilated barns, the air velocity in the barn is determined as a function of ambient temperature (Supplemental Table S2). The pH of excreted cattle manure is about 7 for feces and 8 for urine. The mixture is assumed to have a bulk pH of 7.3 to 7.5. Surface pH is most important because the NH3 concentration at the surface controls emission rate. For manure on the floor of the housing facility, the surface pH is fixed at 0.7 units above the bulk pH of the mixture (Chaoui et al., 2009; Ni et al., 2009; Blanes-Vidal et al., 2009).

### Manure Storage

The storage facility is another important emission source when long-term storage of manure is used on livestock farms. By the time manure is placed into storage, hydrolysis is assumed to be complete, and all urea is converted to TAN. On a given day, the amount of TAN in storage is that accumulated up to that day minus that lost from the storage since loading began. The incoming TAN is that removed from the barn plus a portion of the organic N that mineralizes to an ammoniacal form during long-term storage (Supplemental Table S2).

Manure is stored in a liquid, slurry, or solid form depending on the manure management strategy used. Bedding and manure solids can be separated from manure to form liquid manure (<5% DM). This liquid portion, containing most of the TAN, is typically stored in an earthen basin or tank. Due to wind-induced mixing and the mixing created when manure is pumped into the storage, this liquid portion remains relatively well mixed. When manure is stored as slurry (7–12% DM), less mixing occurs within the storage structure, so diffusion is more important. If the slurry is pumped into the bottom of the storage tank or basin, a crust can form on the manure surface. This crust provides additional resistance, further reducing the rate of migration to the surface. Manure mixed with bedding material may also be stored as semi-solid or solid manure (>12% DM). In this form, diffusion through the manure becomes a major constraint to the emission rate.

Stored manure emissions are modeled on an hourly time step as a function of exposed surface area, temperature, and pH. Slurry and liquid manures are assumed to spread across the exposed surface of the storage where the surface area is determined by the storage dimensions. Thus, in the early stages of loading, manure is in a relatively thin layer, with a large surface area per unit volume stored. This surface area-to-volume ratio decreases as the storage fills. The filling process is repeated after each emptying of the storage. Daily loss of NH3–N is determined such that the cumulative loss up to a given date cannot exceed the accumulated TAN in the storage. This is particularly important in the early stages of loading when a thin layer of manure on the bottom of the storage creates maximum exposure for emission. Manure temperature in the storage is set to the average ambient temperature over the previous 10 d.

Manure pH is modeled as a function of the solids content of the manure. A relationship is used to vary the bulk pH of stored cattle manure from 7 with no manure solids (pure water) to 8.5 with a relatively high solids content (Supplemental Table S2). Carbon dioxide production in the manure is assumed to be due to decomposition of organic matter through microbial activity and therefore is associated with the solids content. With low solids, CO2 production is low, and the effect of CO2 emission on manure surface pH is negligible (i.e., the surface pH is the same as the bulk manure pH). With increasing solids, there is greater opportunity for microbial decomposition, formation, and emission of CO2, and thus a greater increase in surface pH (Supplemental Table S2). For bottom-loaded storages, this surface pH effect is not included because fresh manure is not exposed at the surface.

A major difference among storage types is in the diffusion properties of the manure and the constraint they place on the movement of TAN to the surface. The overall resistance is the sum of the resistances to transport through the bulk manure and from the surface to the free atmosphere including any cover that may be used on the storage (Supplemental Table S1). The effective resistance of the manure was set using the approach of Hutchings et al. (1996) as implemented by Rotz and Oenema (2006). Assigned values are 3 × 105, 2 × 105, 33 × 103, and 0 s m-1 for solid, semisolid, slurry, and liquid manure types, respectively. Additional resistances for covered and enclosed manure storages are 2 × 105 and 2 × 106 s m-1, respectively.

### Field Application

Manure is applied to fields through daily hauling from the barn or through periodic emptying of a long-term storage. With daily hauling, manure is applied the day it is produced. With 6-mo storage, about half of the annual manure produced and stored on the farm is applied to cropland over 10-d periods in early- to mid-April and early- to mid-October. For 12-mo storage systems, all manure for the year is applied during a 10-d period in April.

Four manure application methods are modeled: broadcast spreading, spray irrigation, band spreading, and direct injection into the soil (Rotz et al., 2011). Some TAN is lost as the manure moves through the air in the application process. This loss is fixed at 1 and 10% of the applied TAN for broadcast spreading and spray irrigation with no loss for band spreading and injection (Rotz and Oenema, 2006). The manure TAN reaching the field surface is that hauled from the barn or manure storage on a given day minus this loss.

When applied to a soil surface, the manure is placed in a thin layer where remaining TAN can readily volatilize as NH3. Emission from the manure applied on a given day is determined by integrating over the period until the manure is incorporated by a tillage operation. This provides an exponential decline in the emission rate through time as influenced by changes in manure TAN content, infiltration rate, and DM content along with the effects of rainfall and ambient air temperature (Supplemental Table S2). When manure is incorporated on the day of application, an average exposure time of 8 h is assumed. When manure is not incorporated, the remaining TAN becomes negligible after several days, and the emission rate approaches zero. Manure pH is set to increase to 8.6 immediately after application due to the rapid release of CO2 (Sommer et al., 1991). As the manure lies in the field, the pH decreases at a rate of 0.3 units per day until it stops at a neutral pH of 7.0 (Sommer et al., 1991).

The mass of water contained in the manure on the field surface varies through time. The initial amount after application is determined from the application rate and the manure DM content. The remaining manure moisture is adjusted during each time step for infiltration, evaporation, and rainfall (Supplemental Table S2). Manure TAN on the soil surface also varies through time. The initial TAN is that reaching the soil after the application process. During each time step, NH3 loss occurs to the atmosphere and TAN moves into the soil in proportion to the infiltration of manure solution (Supplemental Table S2).

To predict loss from manure directly injected into the soil, the process level simulation of surface emission is bypassed. Ammonia N loss is set at 5% of the TAN in manure applied through deep injection into cropland and 8% of the TAN in manure applied through shallow injection to grassland. This provides relatively small losses, similar to those measured in field experiments (Rotz et al., 2011).

### Grazing Animals

The approach to modeling NH3 emission from pastures is similar to that used for field application of manure, except for some simplifying assumptions. The N available for volatilization is the urea N and TAN excreted by grazing animals. Excreted N is determined by how animals are fed, and the portion applied is proportional to the time each animal group spends on pasture. The N in feces is primarily organic, so about 90% of the NH3 emission occurs from the N in urine (Rotz, 2004). Some of the urine infiltrates into the soil where the urea is hydrolyzed and the resulting TAN binds to the soil. The remaining portion settles on plant and soil surfaces where it comes in contact with urease. Although hydrolysis must occur to transform the urea to TAN, this process is relatively fast compared with the time animals are on pasture. Thus, the hydrolysis process is ignored, and all urea is converted to TAN for the hourly simulation.

The daily urine mass applied is the sum of that from all animals on the pasture. A portion is assumed to immediately infiltrate into the soil and the remainder infiltrates at a slower rate (Supplemental Table S2). If rainfall occurs on a given day, the urine is diluted by the rain. This dilution reduces the concentration of the remaining TAN in the solution and increases infiltration (Supplemental Table S2).

Hourly emission rates are determined based on ambient temperature, air velocity (average daily wind speed), and pH (Supplemental Table S1). The pH is set at 8.5 to reflect the increase that occurs in urine patches after deposition (Haynes and Williams, 1992). The N available for fertilization of the pasture is the N excreted on pasture minus N losses.

### Model Evaluation

The assessment of the accuracy of a model is often called validation. However, the term “validation” implies an extensive statistical evaluation to compare model predicted and measured data to determine if the model prediction is an adequate representation of the measured data. A true validation of a farm-scale NH3 emission model would require a massive dataset across a full range of environmental and management conditions. Because this type of dataset does not exist, a less formal evaluation is used representing the more important conditions for cattle production systems used in the eastern United States.

### Animal Housing

Common cattle housing facilities in the eastern United States are tie stall, free stall, and bedded pack barns. Most reported emission measurements have been made from free stall barns with limited data from tie stall and bedded pack facilities. In the National Air Emissions Monitoring Study, five free stall dairy barns were monitored in the eastern United States, providing extensive data for model evaluation (Bogan et al., 2010; Cortus et al., 2010; Lim et al., 2010). Other published studies present more general data representing a range in barn design, animal type, feeding practices, and measurement methods. Data from all of these sources were used to evaluate model predictions of barn emissions.

### National Air Emissions Monitoring Study Data

Aerial emissions from each of the five free stall barns were measured using the same equipment and operating procedure (Bogan et al., 2010; Cortus et al., 2010; Lim et al., 2010). The barns were continuously monitored over a 2-yr period beginning in the autumn of 2007. Ammonia concentrations in air entering and exiting the facilities were measured using an INNOVA 1412 Multi-gas analyzer, and barn ventilation rate was determined by monitoring speed and static pressure of fans along the barn walls. Reported uncertainties in the measured data varied among the locations from 6.6 to 11.6%.

A six-row free stall barn was monitored in Onondaga County New York (Lim et al., 2010). At the beginning of the study, the barn contained 442 stalls and a bedded pack area, but the bedded pack was replaced by free stalls about 6 mo into the monitoring period. The barn was 97.5 m long and 33.5 m wide with 4.3-m-high sidewalls and housed about 470 lactating cows. Animals were fed a total mixed ration consisting of 54% forage (corn and alfalfa silages), ground corn grain, and additional supplemental feeds. Manure was removed using automatic alley scrapers. The manure was cycled through a manure digester, and solids were separated after digestion. Some of the manure solids were used as bedding material along with wheat straw harvested on the farm.

Two barns were monitored in Saint Croix County, Wisconsin (Cortus et al., 2010). The first was 93 m long by 28 m wide with 4-m-high sidewalls containing four rows of free stalls that housed 211 cows. The second was 107 m by 30 m with five rows of free stalls housing 355 cows. Thus, the design of the second barn used about 25% less floor area per cow. Animals were fed a total mixed ration consisting of corn silage, finely ground shelled corn, ground hay, cottonseed, and other supplemental feeds. During the first year of monitoring, a flush system was used to remove manure from the barns three times per day while the cows were being milked. The manure moved through a solids separator, and the liquid effluent was recycled through the flush system. For the second year, the manure flush system was replaced with a tractor scrape system.

The final two barns were in Jasper County, Indiana (Bogan et al., 2010). Each of these was 472 m long by 29 m wide with 4.3-m-high sidewalls and housed 1600 cows. Animals were fed a total mixed ration of about 50% forage and 50% grain. Manure was removed from the barns by scraping where a tractor-mounted scraper was replaced by an automatic floor scraper in late summer of 2009. Manure passed through an anaerobic digester, and solids separated after digestion were used as bedding material.

Each barn was simulated over the monitoring period using continuous weather data collected at the nearest weather station (NOAA, 2013). These weather data were compared with those measured at the site to assure similar values. Characteristics of the herd, feeding strategy, barn, and manure storage were set to represent those of the actual facilities. In our model, parameters such as animal numbers and feeding strategy were constant throughout the simulation, whereas minor variations occurred on the actual farms. Daily emissions per cow were compared over the monitoring period using only those days when measured emissions were reported. Daily values were summarized and reported by season designated as winter (Dec.–Feb.), spring (Mar.–May), summer (June–Aug.), and autumn (Sept.–Nov.). Measured and simulated daily emissions were compared using the mean absolute error, RMSE, systematic RMSE, unsystematic RMSE, and an index of agreement determined using the relationships documented by Waldrip et al. (2013).

For most conditions, the model worked well in predicting NH3 emissions from these barns (Table 1). The best agreement between measured and simulated data was found for the two barns in Indiana. Mean simulated emissions for each season were close to those measured, and a correlation coefficient of 0.9 and a mean absolute error of 25% were found comparing daily measured and simulated emissions across the 2-yr period. The RMSE of daily values was about 14 g d-1 cow-1 with low systematic error (Table 1). Measured and simulated daily highs and lows showed agreement (Fig. 2). For each of the barns, there were some outlying measured data. These outliers may have been due to subtle temporary measurement issues or to processes and conditions not represented by the model. Over all years and all barns using manure removal by scraping, the model predicted seasonal emissions similar to those measured with a small overprediction in the summer period and underprediction in the winter (Fig. 3). Simulated and measured daily data were highly correlated, with a mean absolute error of 35%, RMSE of 18.7 g d-1 cow-1, and very low systematic error (Table 1).

View Full Table | Close Full ViewTable 1.

Comparison of measured and simulated ammonia emissions from dairy barns for four seasons from autumn 2007 to autumn 2009.

 Season† New York Wisconsin Indiana Scraped barns, average Flushed barns, average Barn 1, yr 1‡ Barn 1, yr 2 Barn2, yr 1 Barn 2, yr 2 Barn 1 Barn 2 g d-1 cow-1 Winter Measured§ 30.3 33.7 23.7 24.3 15.3 21.4 19.1 22.0 29.0 Simulated 14.9 5.2 7.2 3.7 5.0 14.2 14.6 11.2 4.5 Spring Measured 42.5 35.3 38.4 27.9 36.0 42.7 42.9 40.5 31.6 Simulated 40.8 35.0 37.7 32.0 36.5 46.5 49.2 42.1 33.5 Summer Measured 65.7 41.9 64.0 33.3 49.4 83.9 82.4 69.1 32.6 Simulated 82.1 77.9 78.5 73.6 69.0 82.0 82.0 78.7 75.8 Autumn Measured 48.6 33.8 49.2 26.5 34.9 50.3 47.4 46.1 30.2 Simulated 59.5 44.9 50.6 40.2 45.0 51.8 52.9 52.0 42.6 Annual Measured 43.2 36.3 45.3 28.1 33.2 46.9 44.7 42.7 32.2 Simulated 47.6 39.6 45.5 37.4 38.4 44.9 46.2 44.5 39.1 Correlation¶ 0.68 0.44 0.88 0.60 0.92 0.88 0.90 0.79 0.46 Mean absolute error 19.1 19.8 19.5 11.5 10.2 15.0 25.4 RMSE# 22.8 23.1 22.9 14.6 12.8 18.7 28.3 Systematic error 6.7 6.4 12.6 3.1 1.5 1.0 8.2 Unsystematic error 21.8 22.2 19.2 14.2 12.7 18.7 27.1 Index of agreement†† 0.75 0.73 0.68 0.93 0.95 0.87 0.45
Winter is Dec. through Feb. spring is Mar. through May; summer is June through Aug.; autumn is Sept. through Nov.
In Wisconsin during year 1, a flush system using recycled manure liquid was used to remove manure; during year 2 and in all other barns, a scrape system was used.
§Measured data obtained from the National Air Emissions Monitoring Study (Bogan et al., 2010; Cortus et al., 2010; Lim et al., 2010).
Pearson’s correlation coefficient calculated using daily values.
#Root mean square errors are for simulated vs. measured daily values (Waldrip et al., 2013).
††A value of 1 indicates complete agreement, and a value of 0 indicates complete disagreement (Waldrip et al., 2013).
Fig. 2.

Measured (Lim et al., 2010) and simulated daily ammonia emissions for a dairy barn in Indiana from September 2007 to September 2009.

Fig. 3.

Simulated ammonia emission rates from dairy barns compared with the mean and reported uncertainty of measured data for each season from autumn 2007 to autumn 2009. For the first year, two barns in Wisconsin used a flush system for manure removal; all others used a scrape system.

The model did not predict daily emission rates similar to those measured during the first year for the Wisconsin barns (Table 1), and this difference cannot be fully explained. Measured data showed a similar emission rate across all seasons of the year, whereas the model predicted lower emissions in the winter and high emissions in the summer (Fig. 3). Throughout the second year and for all other barns, the simulated and measured data clearly showed seasonal changes in emission rate (Fig. 2 and 3). During this first year, a flush manure removal system was used in these barns. Compared with a scrape system, flushing is expected to reduce emissions (Ndegwa et al., 2008), but this does not explain the lack of variation in seasonal emission rate. The only other change noted was a change in bedding material in both barns. During the winter and spring, manure solids were used for bedding and during the summer and autumn wood shavings were used. Compared with the use of manure solids, the use of wood shaving bedding has been shown to reduce NH3 emissions in a tie stall barn by 25% (Powell et al., 2008b). In a free stall barn where less bedding material is used, any effect obtained would not be expected to offset the increase in NH3 emissions that occurs during the hot summer months. The correlation between simulated and measured daily values was low in these flushed barns, with a mean absolute error over 50%, RMSE of 28.3 g d-1 cow-1, and greater systematic and unsystematic error (Table 1). Despite the differences in daily and seasonal values, predicted annual emissions were just 9 and 33% greater than measured over this first year for these two barns.

### Other Published Data

Two studies have reported NH3 emissions from tie stall facilities in Wisconsin. Powell et al. (2008a, 2008b) used four chambers constructed within a tie stall barn to measure the effect of protein feeding and bedding material on NH3 emissions from bred dairy heifers and lactating dairy cows. Each chamber contained four animals where air flow rates and inlet and outlet gas concentrations were monitored. Emissions were measured during different seasons of the year, resulting in a wide range in barn temperatures for each treatment (Table 2). Although pine shaving bedding was found to provide a small reduction in emission rate, only the overall seasonal emission data from this study (Powell et al., 2008a) were used in model evaluation. Protein feeding and seasonal effects were evaluated using the data from Powell et al. (2008b) (Table 2). For these two studies, average daily N excretion rates of 238 g per bred heifer and 420 g per cow were reported. Simulated daily N excretions compared well with 262 g per bred heifer and 469 g per cow.

View Full Table | Close Full ViewTable 2.

A comparison of measured ammonia emissions reported in the literature to those simulated using similar animal and feed characteristics and climate conditions.

 Reference Location Barn type Animals Diet protein Season Avg. inside air temp. Emission Measured Simulated % of DM °C g d-1 AU-1† Powell et al., 2008a central Wisconsin tie stall, gutter 16 heifers 15.6 winter 5.3 12.4 1.7 17.1 summer 26.1 27.9 24.5 16.2 autumn 10.1 28.9 4.3 Powell et al., 2008b central Wisconsin tie stall, gutter 16 cows 15.7–17.3 winter 8.7 6.1 0.9 17.0–21.5 spring 17.5 17.6 18.5 15.7–17.3 early autumn 21.4 8.1 27.0 Adviento-Borbe et al., 2010 central Pennsylvania free stall, scrape 120 cows 16.4–17.3 spring 7.5 18.7 21.0 summer 21.3 30.1 36.6 Harper et al., 2009 south-central Wisconsin free stall, scrape 900 cows 16.5–17.8 winter na‡ 14.0 12.9 summer na 27.7 46.4 autumn na 31.1 26.0 northeast Wisconsin free stall, flush 1700 cows 17.3–17.8 winter na 6.6 6.9 summer na 32.0 32.3 autumn na 19.7 14.7 Smits et al., 1995 The Netherlands free stall, low emission floor 34 cows 14.7 – 10–24 20.2–26.0 23.5 20.0 – 10–24 26.0–52.0 32.8 Overall mean 21.4 20.6 Mean absolute error 6.9
An animal unit (AU) is 500 kg of body weight.
Temperatures were not reported by Harper et al. (2009).

Harper et al. (2009) used open path lasers and inverse-dispersion analyses to measure NH3 emissions from commercial dairy facilities in Wisconsin. For model evaluation, two of the facilities with complete data were used representing scrape and flush manure removal systems (Table 2). These were large free stall barns housing 900 to 1700 cows. Measurements were made throughout the year, giving a wide range in ambient and barn temperatures, but actual temperatures were not reported.

Adviento-Borbe et al. (2010) measured NH3 emissions from a naturally ventilated free stall barn in Pennsylvania. The barn housed 120 dairy cows fed precision diets at two protein levels, and manure was removed twice per day by scraping. Ammonia emission measurements were made during 18 12-h-long sampling events during spring and summer using a static flux chamber technique where 64 locations on the barn floor were sampled during each sampling day. Although the flux chamber technique was well suited for collecting many samples to evaluate differences in diet, the procedure did not represent NH3 emission from a barn floor exposed to ambient air movement. Recognizing this limitation, the authors developed a calibration procedure to adjust NH3 emissions measured with the static flux chamber to better represent emissions from the naturally ventilated barn. For this procedure, NH3 emissions from manure inside a mechanically ventilated test room were measured using the flux chamber and compared with emissions determined from measured air flow and NH3 concentrations at the inlet and outlet of the test room. Their data were then adjusted using the ratio of emission rates from the test room to those measured with the flux chamber. No effect on NH3 emissions was found between diets in this study; therefore, the average over both diets was used in model evaluation (Table 2).

A final study was used where emissions were measured from a low-emission barn floor in a free stall barn housing 34 dairy cows in The Netherlands (Smits et al., 1995). To represent this barn, the option of free stalls with a low-emission floor was selected in the model as the housing type. This low-emission floor was modeled using partial separation of urine and feces on the barn floor as described above. Two animal diets were fed with very different crude protein concentrations (Table 2). Emission measurements were made over 126 d, providing a range in ambient and barn temperatures.

Each of the five systems was simulated using the facility, manure handling strategy, and type and size of animals used in the experimental study. Feed types and characteristics were set to provide similar diet protein contents as those fed. Each facility was simulated over 1 yr using historical weather data for the region of the study with temperatures similar to those measured during the measurement period. Simulated emission rates were averaged over the same dates of the experiments to compare measured and simulated daily emissions. For consistency, all emissions were expressed per animal unit, where 1 animal unit was 500 kg of body mass.

The model appropriately represented measured emissions and the influences of temperature, diet protein, and manure management strategy (Table 2). There were a few inconsistencies, primarily related to the effect of temperature or season. The largest differences between measured and simulated emissions were for the autumn measurements of the two studies by Powell et al. (2008a, 2008b). For the study using bred heifers (Powell et al., 2008a), the highest emission rate was measured in the autumn, even though the reported temperature during this period was considerably lower than that measured in the summer and only a little greater than that for the winter period (Table 2). The protein content of the diet was also less than that fed in the summer, so this high emission rate cannot be explained with the available information. For the study using lactating cows (Powell et al., 2008b), measured emissions in early autumn were unexpectedly low given that the reported temperatures during this period were greater than that of the summer period. Their measurement may have been confounded by the ventilation rate used or by an error in airflow measurement. The reported airflow rate for this measurement period was relatively low considering the high air temperature reported. Simulated emissions for these tie stall barns also showed an apparent underprediction of emissions during the winter. This underprediction may be due to the barn temperatures being greater than what our model predicted for this type of barn because of the effects of the enclosed chambers within the barn.

For the free stall barn in Pennsylvania, measured and simulated emissions responded similarly to temperature, with a small overprediction by the model (Table 2). Simulated emissions from the two larger dairy operations in Wisconsin were similar to those measured across all seasons. The model also represented a reduction in emission through the use of a flush manure removal system when compared with the scrape system (Table 2). Simulation of the facility in the Netherlands that used a low-emission barn floor produced average emissions near the median of the range measured over similar weather. For the measured and simulated data, emissions from the high-protein diet were about 40% greater than those of the low-protein diet (Table 2).

These comparisons support the effectiveness of the model in predicting NH3 emissions from various types of dairy barns. Averaged over all comparisons, simulated NH3 emissions were 4% less than the reported measured values, and simulated values correlated well with measured values (r = 0.64) despite the few inconsistencies (Table 2). The mean absolute error over all data comparisons was 33% of the predicted emission.

## Manure Storage

### National Air Emissions Monitoring Study Data

Emissions from manure storages in Indiana and Wisconsin were monitored over 2 yr using open-path techniques (Grant and Boehm, 2010a, 2010b). Ammonia concentrations were measured using scanning tunable diode laser absorption spectrometry along with meteorological measurements including air flow using 3-dimensional sonic anemometers. From concentration and air flow data, Radial Plume Mapping (RPM) and backward Lagrangian Stochastic (bLS) models were used to calculate emission rates. Reported uncertainties of NH3 emission measurements were 18 and 24%, respectively, using the RPM and bLS methods. Manure pH, temperature at 0.3 m depth, and local weather conditions were also recorded. Measurements made by the two methods were found to be similar (Grant et al., 2013) and were thus combined to get the maximum number of daily measurements where a daily value required at least 25 valid 30-min measurements within the 24 h.

The manure storage in Indiana was located in Jasper County but at a different farm than the barn-monitoring site. The clay-lined lagoon was 85 m wide and 116 m long (Grant and Boehm, 2010a). At maximum capacity, the liquid depth was 5 m, with a volume of 48,000 m3 and surface area of 9900 m2. The lagoon received effluent from the milking parlor and holding area, which provided a more dilute solution than normally obtained from a free stall barn. The daily manure input was set to that of 161 cows, which represented the 2580-cow herd spending 1.5 h per day in the parlor and holding pen area. This storage was not emptied during the year. Emissions were monitored continuously from September 2008 through August 2009. Valid daily measurements were made for 71 d using the RPM method and for 131 d using the bLS method, giving a total of 148 valid daily emission measurements. Simulated emissions for only these days were compared with the measured values.

The manure storage in Wisconsin was on the same farm as the monitored barn (Grant and Boehm, 2010b). Emissions were measured for up to 21 d during each of four seasons over 2 yr. Valid daily NH3 measurements were obtained for 1 d using the RPM method and for 19 d using the bLS method, for a total of 19 daily values. Most of those days were in the autumn period, providing a relatively limited and unbalanced data set. The first and second stages of a three-stage lagoon system were monitored. The first lagoon had a surface area of 4264 m2 and a volume of 10,600 m3 at maximum capacity. The second had a surface area of 2900 m2 and a volume of 6420 m3 at maximum capacity. The lagoons were emptied twice during the year.

For these manure storages, the model predicted daily NH3 emission rates representative of those measured (Table 3). For the Indiana storage, a weak correlation was found between simulated and measured daily values, with a high mean absolute error of 69%. For the limited data of the Wisconsin site, a very high correlation was obtained, but the error was also high. Averaged over days, simulated NH3 emissions were similar to those measured during each season for both manure storages. Summer emissions in Wisconsin were greater than those measured in Indiana due to a greater N concentration in this stored manure. The average simulated emission over all measurement dates was equal to that measured at the Indiana site and 14% greater at the Wisconsin site. At the Indiana site, most of the RMSE among daily values were unsystematic, but in Wisconsin the systematic and unsystematic errors were similar (Table 3).

View Full Table | Close Full ViewTable 3.

Measured and simulated ammonia emission rates from manure storages at two locations and for four seasons during the years 2008 and 2009.

 Measurement period, statistical measure Indiana Wisconsin Valid days† Measured‡ Simulated Valid days Measured Simulated Mean SD Mean SD Mean SD Mean SD µg s-1 m-2 µg s-1 m-2 Spring§ 47 27.1 14.9 24.1 26.7 0 Summer 24 50.5 21.4 47.2 43.2 2 170.0 56.4 236.0 55.9 Autumn 33 22.4 9.4 21.0 15.4 14 16.8 14.6 12.4 12.2 Winter 44 2.3 6.5 8.9 15.1 3 0.0 2.3 0.6 0.3 Annual average 148 22.5 20.7 22.6 28.1 19 29.8 53.2 34.1 73.2 Daily correlation¶ 0.56 0.98 Mean absolute error 15.5 13.2 RMSE# 23.7 25.2 Systematic error 5.0 18.6 Unsystematic error 23.2 17.0 Index of agreement†† 0.72 0.96
Emission rates were determined from open-path measurements using the radial plume mapping or backward Lagrangian Stochastic emission models. Valid daily data from each method were averaged or used where a valid daily measurement was obtained when there were at least 25 of 48 valid half-hour measurements throughout the day.
Measured data were obtained from the National Air Emissions Monitoring Study (Grant and Boehm, 2010a, 2010b).
§Winter is Dec. through Feb., spring is Mar. through May, summer is June through Aug.; autumn is Sept. through Nov.
Pearson’s correlation coefficient calculated using daily values.
#Root mean square errors are for simulated vs. measured daily values (Waldrip et al., 2013).
††A value of 1 indicates complete agreement, and a value of 0 indicates complete disagreement (Waldrip et al., 2013).

### Other Published Data

Data from four studies were used to further evaluate the manure storage component. In the first study, NH3 emissions were measured from an earthen manure storage pond in Ohio (Zhao et al., 2007). A convective flux chamber technique was used to measure emissions for 1 d per month from April through November. The storage pond was 61 m wide, 128 m long, and 4.6 m deep with a capacity of 34,000 t of liquid manure. A settling basin was used to partially remove sand bedding and manure solids. The storage was emptied twice per year in April and September.

A dairy farm was simulated using the animal and manure storage characteristics of the monitored farm. Model inputs for feed characteristics and manure solids separation were selected to provide manure entering the storage consistent with that of the monitored farm. For a simulated year, days during each month were selected where the ambient temperature and the amount of manure in the storage were similar to that reported on the day measurements were made. Simulated daily emissions for each month were similar to those measured (Table 4). Over all measurements, the correlation between simulated and measured daily emissions was high (r = 0.97), with a relatively low error and with the mean simulated emission 10% greater than the measured emission.

View Full Table | Close Full ViewTable 4.

A comparison of measured ammonia emissions from a dairy manure storage pond in Ohio with emissions simulated by the Integrated Farm System Model under similar climatic conditions.†

 Month Daily ambient temp. Ammonia emission Measured Simulated Measured Simulated °C g d-1 m-2 Apr. 11 12.5 4.1 3.3 May 21 14.5 7.8 8.8 June 20 23.2 6.1 6.3 July 26 24.8 14.6 14.5 Aug. 26 24.5 9.9 13.4 Sept. 15 20.9 2.9 3.2 Oct. 4 12.8 0.8 1.1 Nov. 0 7.6 0.5 0.9 8-mo avg. – – 5.8 6.4 Correlation 0.97 Mean absolute error 2.4
Measured data from Zhao et al. (2007).

A second study was used to evaluate the model’s ability to represent surface covers. In an experiment conducted in Denmark, NH3 emissions were measured over an 8-mo period (Feb.–Sept. 1990) from several small manure storage tanks with different storage covers (Sommer et al., 1993). Although the experiment included data for swine manure, only the emissions from tanks storing dairy cattle slurry manure were used for model evaluation. Cover treatments included stirred slurry with no cover, a natural crust cover, and a lid providing an enclosed structure. Other covers included oil, peat, straw, foil, and expanded clay aggregate. As long as an adequate cover was maintained, these other covers had similar effects in reducing emissions and thus were grouped for model evaluation.

A dairy farm was simulated using historical weather for this region of Denmark over the period of the study. A dairy herd and manure storage were defined in the model to represent the manure and storage characteristics of the experimental study but at a larger scale suitable for a small dairy farm in this region. Simulated average daily area-specific emissions were similar to measured emissions over the measurement period (Table 5). Seasonal effects were represented well by the model with emissions during the summer period being about double those during the winter and spring periods. The effect of the cover material also provided similar simulated emission reductions as those measured for the natural crust (62 vs. 66%), other cover (76 vs. 77%), and enclosed storage types (94 vs. 91%) (Table 5).

View Full Table | Close Full ViewTable 5.

A comparison of measured ammonia emissions from cattle manure in different storages in Denmark and Wisconsin to that simulated with the Integrated Farm System Model using similar weather and manure storage conditions.

 Storage type Period of year Emission Emission reduction Measured Simulated Measured Simulated g d-1 m-2 % Slurry manure† Open winter–spring 4.4 4.1 – – summer 6.4 8.8 – – both periods 4.9 5.5 – – Crust summer 2.2 3.3 66 62 Cover winter-spring 0.9 0.9 80 78 summer 1.5 2.1 77 76 both periods 1.1 1.3 79 76 Enclosed winter–spring 0.5 0.2 88 95 summer 0.2 0.5 97 94 both periods 0.4 0.3 91 94 Digested slurry‡ Open winter 4.0 4.4 – – spring 5.6 8.8 – – summer 26.6 24.0 – – autumn 8.5 11.1 – – annual 4.0 kg m-2 4.4 kg m-2 – – Cover/enclosed annual 0.1–0.27 kg m-2 0.24 kg m-2 93–97% 95 Liquid manure§ winter 0.0¶ 0.43 – – summer 4.2 5.0 – – autumn 2.9 1.9 – – Overall mean 4.5 5.0 84.1 83.8 Overall correlation 0.98 0.96 Overall mean absolute error 1.0 3.2
Measured data from Sommer et al. (1993).
Measured data from Sommer (1997).
§Measured data from Harper et al. (2009).
Below the detection limit of the measurement procedure (Harper et al., 2009).

In the third comparison, measured NH3 emission data were used from a study of farm-scale tanks containing anaerobically digested manure from a biogas plant with and without surface covers (Sommer, 1997). Emissions were continuously monitored over a full year using a micrometeorological mass balance technique. The input to the digester included 75% animal slurry and 25% slaughter house and fish processing waste. A manure storage of the actual size (surface area of 519 m2) was simulated with and without an enclosed cover using weather for this region of Denmark for the year of the study (1994). The material entering the storage was simulated as digestate, with N and DM contents similar to that of the measured material. Without a cover, simulated and measured NH3 emissions were well correlated across seasons (r = 0.98), and the simulated total emission for the year was 10% greater than that measured (Table 5). Simulated emissions were a little greater than that measured in the spring and autumn and a little less during the summer period. Both the deep straw and clay bead covers used on the actual storage were very effective in reducing NH3 emission (Sommer, 1997). This reduction was represented by the model using an enclosed manure storage, which provided a 95% reduction in NH3 loss (Table 5).

In the final comparison, the manure storage on farm 2 of the study by Harper et al. (2009) was simulated. The manure produced was that of the northeast Wisconsin dairy farm with 1700 cows (Table 1). The manure storage consisted of two adjacent lagoons and a sand/solids separation channel, with exposed surface areas totaling 24,400 m2. The lagoons stored cattle manure along with runoff from the barns and parlor wash water. Ammonia emissions were measured using the same open-path lasers and inverse–dispersion analysis used to measure the barn emissions described above for this farm (Harper et al., 2009). The storage was modeled as reported with liquid manure stored at a DM content of 5%. The storage was simulated over a full year, and average emissions for December to February, June to July, and October to November were compared with those reported for winter, summer, and autumn, respectively. Simulated NH3 emissions were well correlated with those reported across the three seasons (r = 0.91) (Table 5). During the winter period, emissions were below the minimum quantification limits of the monitoring equipment (Harper et al., 2009), and simulated emissions were also very low. During the summer period, simulated emissions were 19% greater than measured values; this result was within the uncertainty of the measurement procedure (Harper et al., 2009). During the autumn period, simulated emissions were slightly less than the measured value.

These comparisons across different published studies further support that the model can appropriately represent NH3 emissions from different types of manure storages. Over all comparisons of different storage conditions, simulated emissions were well correlated (r = 0.97) with measured values. The average of simulated values was 10% greater than the average of measured values, with a mean absolute error of 22% (Table 5).

### Field Application

Ammonia emissions from field-applied manure are related to the DM content of the manure, with greater emissions from drier manure (Sommer and Olesen, 1991). The time the manure remains on the soil surface is also a major factor, with 50 to 75% of the potential emission lost during the first day. Emissions are influenced by the weather, with greater emissions in warmer temperatures and suppressed emissions after rain. Through a comprehensive review and summary of field measured data, Meisinger and Jokela (2000) reported NH3 losses as a function of manure DM content and time to incorporation (Table 6). For model evaluation, field application of dairy manure was simulated for 25 yr of Pennsylvania weather, and simulated annual percentages of TAN loss were compared with the reported values. The model was able to represent these major effects. The simulated 25-yr mean emissions were similar to those reported for different manure DM contents, and the range in annual predictions normally encompassed the reported mean values (Table 6). The model overpredicted the first-day emission from solid (>20% DM) manure, but the agreement was very good for all other forms of manure and periods of manure exposure.

View Full Table | Close Full ViewTable 6.

Reported and simulated total ammonium nitrogen losses after field application of various forms of cattle manure.†

 Manure Dry matter content Reported losses Simulated annual losses‡ First day Not incorporated First day incorporation§ Not incorporated¶ Mean Mean Mean Range Mean Range % Solid >20 40 90 71 57–86 83 71–93 Semi-solid 12–20 60 80 54 38–67 70 56–82 Slurry 8–11 45 60 38 28–53 56 43–68 Liquid slurry <8 30 40 25 16–36 43 31–54
Reported losses from Meisinger and Jokela (2000).
Average annual losses of ammonia-N as a fraction of available total ammoniacal N. The mean is a 25-yr average, and “range” is the range in annual values predicted for individual years.
§Manure is incorporated the same day it is applied (within 8 h), and ammonia emission ceases after incorporation. These simulated data can be generally compared with the reported losses during the first day after field application.
Manure remains on the surface until all total ammoniacal N is lost or incorporated by infiltration.

For a second evaluation, NH3 emissions measured over 4 yr after various application methods were compared with those simulated using the same conditions in central Pennsylvania. Emissions were measured using ventilated chambers and passive diffusion samplers (Dell et al., 2012). Two trials were conducted each year after dairy manure applications in 2006 and 2007 and swine manure applications in 2008 and 2009. Application methods included broadcast without incorporation, broadcast with tillage incorporation within an hour of application, and subsurface injection. Each application method was simulated using the actual weather data for the experimental periods from State College, Pennsylvania, about 30 km from the field site. Simulated manure DM and N contents were set to those of the manure applied in the field experiment (Table 7) (Dell et al., 2012).

View Full Table | Close Full ViewTable 7.

A comparison of measured and simulated ammonia emissions from dairy cattle and swine manure applied to crop land using different application methods.†

 Application date and method Manure DM‡ content Manure TAN§ content Emission Emission reduction Measured Simulated Measured Simulated % g L-1 kg NH3–N ha-1 % 2006 trials, dairy manure Broadcast 9.6 1.3 63 48.1 – – Broadcast and incorporated 9.6 1.3 12 12.2 82 75 Injected 9.6 1.3 5 3.6 91 93 2007 trials, dairy manure Broadcast 9.1 1.5 62 68.4 – – Broadcast and incorporated 9.1 1.5 11 33.8 82 51 Injected 9.1 1.5 1 4.2 99 94 2008 trials, swine manure Broadcast 3.4 1.5 51 54.4 – – Injected 3.4 1.5 4 4.2 93 92 2009 trials, swine manure Broadcast 1.7 1.7 35 35.6 – – Injected 1.7 1.7 2 4.8 93 87 Overall mean 24.6 26.9 90 82 Overall correlation 0.93 0.83 Overall mean absolute error 5.6 8.7
Measured data from Dell et al. (2012).
Dry matter.
§Total ammoniacal N.

For most trials, predicted total emissions after application were very similar to measured emissions (Table 7). Exceptions were that the simulated emissions after broadcast application in the 2006 trials were about 25% less than that measured, and the emission after one of the 2007 trials using incorporation was high. These differences appear to be associated with wind speed because the historical weather data indicate a relatively low wind speed during the 2006 trials, whereas, on the first day of the 2007 trial, the average wind speed was much greater. Because wind speeds at the field site were not measured, differences in reported wind conditions cannot be evaluated. The chamber method used to measure NH3 emissions may have also protected the sampling area from the wind’s effect on emission rate. Overall, predicted total emissions after the application methods and the reduced emission obtained from rapid incorporation and subsurface injection were normally within 20% and often within 10% of measured values (Table 7). Predicted emissions averaged 9% greater than measured values, with a mean absolute error of 23%.

### Grazing

To evaluate NH3 emissions from grazing animals, simulated emissions were compared with data measured over three grazing seasons in The Netherlands using a micrometeorological mass balance method (Bussink, 1992, 1994). Reported data were averaged for three periods defined as spring, summer, and autumn (Table 8). A Dutch dairy farm, similar to that modeled by Rotz et al. (2006), was simulated over 25 yr of weather for The Netherlands using similar stocking rates and the same fertilizer application as that used in the original study. Predicted daily emissions per cow were averaged over the same periods defined by the measured data for comparison. With a moderate level of N fertilization, the model predicted greater emission rates than those measured, particularly in the spring and autumn periods (Table 8). Under heavy use of fertilizer, simulated and measured emissions were similar (mean absolute error, 5%).

View Full Table | Close Full ViewTable 8.

Measured and simulated ammonia emissions from grazing Holstein dairy cattle in The Netherlands.†

 Season‡ Average ammonia emission Moderate fertilization§ Heavy fertilization Measured Simulated Measured Simulated kg d-1 cow-1 Spring 6.3 34.7 39.0 39.7 Summer 32.5 44.4 71.7 68.9 Autumn 11.8 30.7 40.4 37.6 Correlation 0.72 0.91 Mean absolute error 19.7 2.8
Data from Bussink (1992) and Bussink (1994).
Spring represents early May to mid-June, summer is mid-June to late Aug., and autumn is Sept. and Oct.
§Moderate fertilization is 250 kg N ha-1 per year of inorganic nitrate fertilizer plus manure N deposited on pasture. Heavy fertilization is 400–550 kg N ha-1 per year plus manure N deposited on pasture.

A second evaluation of NH3 emission from grazing cattle was conducted using data from a study by Lockyer and Whitehead (1990). Measurements made during different seasons of the year were compared with values predicted by the model simulating emissions from dairy cattle on pasture in the United Kingdom under similar temperature conditions. Comparisons were based on the percentage of total urine N deposited that was lost by volatilization. Simulated losses were well correlated (r = 0.91) with measured losses across seasons (Table 9). During the winter, simulated losses were considerably less than those measured. The N loss measured in the winter appears high compared with those measured in the spring and autumn considering the relatively cold winter temperature. This response in the measured data cannot be explained with the available information. The wind tunnel procedure used may have affected measured emission rates, but the wind velocity used in the simulation was that produced in the wind tunnel (normally 1 m s-1). Simulated losses in the spring were a little less than that measured, but there was good agreement during the summer and autumn periods (Table 9).

View Full Table | Close Full ViewTable 9.

Measured and simulated ammonia nitrogen loss from grazing dairy cattle in England.†

 Season‡ Air temp. Ammonia N loss Measured Simulated °C % applied N Winter 4.0 11.8 2.2 Spring 8.9 9.6 5.5 Summer 19.8 24.7 24.4 Autumn 12.7 12.0 12.1 Correlation 0.91 Mean absolute error 3.5
Measured data from Lockyer and Whitehead (1990).
Measurements made in Jan., Mar., July, and Sept., respectively.

Although there were some differences between measured and simulated data, these two evaluations confirm that the model can represent NH3 emissions from grazing cattle. Over all comparisons made, simulated emissions were well correlated with measured values (r = 0.83), with simulated values averaging 15% greater than those measured (mean absolute error of 30%).

### Potential Model Improvements

Although the model provides a useful tool in its current form, there are potential improvements that could be made to better represent system processes. The proportion of urinary N in the form of urea varies with diet. As new information comes available, this will be made to vary with the amount and form of protein intake. Better understanding is needed to predict the surface pH of manure. An improved model of the processes controlling manure pH, particularly at the surface, would be useful (Hafner et al., 2013). Considering the continuous mixing of the manure and application of fresh urine and feces on the floor of a housing facility, a process-based model of this component may be impractical. A model predicting surface pH may be more practical for simulating the effect on emissions when the manure remains undisturbed during storage and after field application.

For manure storage, a mechanistic model of the kinetics and mass transfer of the compounds within the stored manure would be useful. Our simplified approach of using a mass transfer resistance provides a model that captures the average transfer, but it does not consider effects from daily or seasonal changes in manure and environmental conditions. Despite these shortcomings, model predictions compare well with measured NH3 emission rates from dairy farms, suggesting that the model is a reasonable representation of these processes.

### Model Application

To demonstrate the use of the farm model, simulations were conducted to compare tie stall and free stall housing facilities, grazing of animals, and different manure handling strategies on a representative dairy farm in Pennsylvania. Parameters were set to represent a typical farm and the animal and manure management strategies commonly used in this region. This simulation analysis is provided as an example of further evaluation of the NH3 emission model within a whole farm simulation and to illustrate the interactions and overall impacts of management changes. A comprehensive evaluation with complete documentation of all model parameters and assumptions is beyond the scope of the present analysis. Only important parameters are provided to support the systems compared, including the type of equipment and facilities used and their initial costs. Prices of all farm inputs and outputs were set to reflect long-term relative values in current dollars. Additional details on the farm model are provided by Rotz et al. (2012).

The farm was on 100 ha of a medium silt loam soil. Crops grown included 20 ha of alfalfa, 50 ha of corn, and 30 ha of perennial grassland. Crops were fertilized with dairy manure produced on the farm along with additional inorganic N fertilizer applied to corn and grassland at a rate of 60 kg N ha-1. Alfalfa and grass were produced and fed primarily as silage and secondarily as dry hay. Corn was harvested and fed as silage and high-moisture grain. The dairy herd consisted of 100 Holstein cows plus 80 replacement heifers. Animals not on pasture were fed total mixed rations to maintain a target annual milk production of 10,000 kg per cow. Rations were prepared using the feeds produced on the farm plus purchased supplemental feeds required to meet energy, protein, and mineral requirements.

Four production strategies were compared on the representative farm. For the first two strategies, cows were housed in a tie stall barn and heifers were housed in a bedded pack barn. Manure was removed from gutters in the tie stall barn each day and broadcast onto crop or grassland without incorporation. The initial cost of the housing facility was $285,000 plus$160,000 in milking and manure handling equipment. For the first strategy, animals were housed and fed all year in barns. For the second strategy, the 30 ha of grassland was converted to pasture, and heifers and nonlactating cows were rotationally grazed from April to October. Initial costs for creating the pasture were $20,000 for the perimeter fence,$6000 for electric fencing, and $5000 for watering equipment. Labor for pasture management was set at 5 h per week. The grassland received the manure from the grazing animals plus 30% of the manure removed from the barns. For the third and fourth strategies, all animals were housed and fed all year in free stall barns. Manure was scraped from the barn floors twice daily and pumped into the bottom of a slurry storage tank, allowing a natural crust to form on the storage. Manure was removed from the tank each April and October for broadcast application to farm land. Initial costs for the housing and manure storage facilities were$247,000 and $68,000, respectively. Investments in milking and manure handling equipment were$120,000 and $113,000, respectively. The fourth strategy was similar to the third except that an unsealed cover (initial cost of$18,500) was used on the manure storage, and injection equipment was added to the manure applicator (added initial cost of $17,000) to place the manure under the soil surface. The use of the tie stall barn with daily hauling of manure gave an annual NH3 emission of 56 kg per cow, with most of this emission occurring after field application (Table 10). By grazing the heifers and nonlactating cows during the growing season, NH3 emissions from the pasture were offset by reduced emissions from the barn and field applied manure, giving an 8% reduction from the farm (Table 10). Other benefits included a 10% reduction in the carbon footprint of the milk produced and an increase in farm profit through reduced costs for bedding, labor, and fuel. The high concentration of N deposited in urine and fecal spots on pasture increased nitrate leaching and denitrification losses from the farm by 8 and 5%, respectively. View Full Table | Close Full ViewTable 10. Simulated annual environmental impacts and economics of using different housing facilities and manure handling methods on a representative Pennsylvania dairy farm.†  Tie stall barn Free stall barn Daily haul‡ Daily haul, grazing§ Slurry storage¶ Covered storage, injection# Ammonia emission, kg cow−1 Barn 19.3 15.1 18.2 18.5 Manure storage 0.0 0.0 9.3 6.0 Field application 36.8 31.7 38.7 5.8 Pasture 0.0 5.0 0.0 0.0 Total 56.1 51.8 66.1 30.3 Net greenhouse gas emission, kg CO2e cow−1 Housing and animal 6643 5626 4668 4674 Manure storage 0 0 3085 1328 All other 3914 3878 3470 3492 Carbon footprint, kg CO2e kg−1 milk 1.06 0.95 1.12 0.95 Nitrogen loss, kg ha-1 Volatilization 46.2 42.6 54.4 24.9 Leaching 25.9 28.0 22.3 31.5 Denitrification 21.8 22.9 19.0 25.8 Production costs,$ cow−1 3649 3546 3517 3565 Income, $cow−1 4042 4029 4033 4054 Net return to management,$ cow−1 393 483 516 489
Herd of 100 cows plus 80 replacement heifers on 100 ha of cropland (20 ha alfalfa, 50 ha corn, and 30 ha grass) annually producing 10,000 kg of milk (3.5% fat) per cow.
Manure was removed from gutters each day and applied to crop or grass land.
§Same as daily haul, except that 30 ha of grassland on the farm were converted to pasture (including fencing and watering costs) and heifers and dry cows were rotationally grazed during the growing season.
Manure was scraped from barn floor twice daily and pumped into the bottom of a slurry storage tank that was emptied in April and October with broadcast application to farm land.
#Same as slurry storage except that an unsealed cover was used on the manure storage and manure was injected into the soil during field application.

The use of free stall barns for all animals gave similar barn emissions as the combined tie stall and bedded pack barns used in the first strategy (Table 10). Combined emissions from manure storage and field application increased NH3 emission from the farm by 18%. With greater NH3 emission to the atmosphere, nitrate leaching and denitrification losses were reduced by 14%. Other effects included a small increase in the carbon footprint of the milk produced and an increase in farm profit. The use of a cover on the manure storage and injection of the manure during field application reduced storage emissions by about 35%, with an 85% reduction in NH3 emission after field application. Methane emission from the manure storage was also reduced, which decreased the carbon footprint of milk by about 15%. The increased cost of the cover and injection equipment along with increased operating costs for fuel and labor reduced annual farm profit by \$27 per cow.

These simulations illustrate the ability of the farm model to capture dynamic interactions among farm processes to predict whole farm NH3 emissions along with other environmental effects and farm economics. This also demonstrates how the model can be used to evaluate and compare the effects of mitigation strategies and other farm management changes.

## Conclusions

A process model that predicts NH3 emissions from cattle manure was incorporated in a farm simulation model (Integrated Farm System Model), where it is used to estimate daily and annual whole farm NH3 emissions from dairy and beef cattle production systems. Performance of the emission component was evaluated through a comprehensive comparison of simulated emissions to measured and published emission data from differing barn designs, protein feeding levels, manure storage types, field application procedures, and pasture systems.

The expanded farm model provides a tool for evaluating management effects on NH3 emissions and the interacting effects on nitrate leaching, greenhouse gas emissions, nutrient runoff losses, and farm profitability. Simulation of a representative dairy farm in Pennsylvania illustrates the benefits and costs of the use of different barn designs, use of grazing, long-term manure storage, a manure storage cover, and subsurface application of manure.

## Acknowledgments

This research was funded by the U.S. Department of Agriculture and by the National Dairy Board and the Dairy Research Institute.

## Footnotes

##### Ammonia Emission Model for Whole Farm Evaluation of Dairy Production Systems

C. Alan Rotz, Felipe Montes, Sasha D. Hafner, Albert J. Heber and Richard H. Grant
10.2134/jeq2013.04.0121