Professional Documents
Culture Documents
ScienceDirect
Research Paper
article info
Poultry production significantly contributes to atmospheric ammonia (NH3) emissions,
Article history: causing concerns over their environmental impacts and the effects on human and bird
Received 15 December 2019 health inside the poultry houses. A model is needed to estimate NH3 emissions from
Received in revised form poultry facilities in order to develop effective mitigation strategies and science-based
7 May 2020 regulations. The laboratory-scale mechanistic NH3 emission model developed by Tong,
Accepted 11 May 2020 Zhao, Heber, and Ni (2020) requires extensive monitoring and laboratory work for the in-
puts and therefore cannot be conveniently used by producers or regulatory agencies. Based
on this mechanistic model, a farm-scale model was developed for estimating dynamic NH3
Keywords: emission rates from commercial manure-belt layer houses. Sub-models were developed to
Air quality estimate the inputs of the mechanistic model from farm-scale conditions that are
Animal agriculture commonly known during poultry production. Computational fluid dynamics (CFD) models
Computational fluid dynamics and analyses of manure characteristics were used to estimate air conditions, manure
Emission model properties, and NH3 emission areas inside layer houses. The model estimated daily NH3
Poultry housing emission from typical commercial manure-belt layer houses for given dates with input
Regulation data consisting of hen inventory, ambient air temperature, house ventilation mode and
rate, and manure management practices. Model performance was evaluated by compari-
son with NH3 emission data from four commercial manure-belt layer houses with different
ventilation systems. Uncertainties of 26e32% relative to the mechanistic model were ob-
tained based on the uncertainties of individual sub-models along with their corresponding
sensitivities. The model had an acceptable performance during ordinary egg production (p-
value ¼ 0.10e0.61), but it needs further development for special events such as moulting
and introduction of new flocks.
© 2020 IAgrE. Published by Elsevier Ltd. All rights reserved.
* Corresponding author.
E-mail addresses: tong.129@buckeyemail.osu.edu (X. Tong), zhao.119@osu.edu (L. Zhao), heber@purdue.edu (A.J. Heber), jiqin@pur-
due.edu (J.-Q. Ni).
https://doi.org/10.1016/j.biosystemseng.2020.05.008
1537-5110/© 2020 IAgrE. Published by Elsevier Ltd. All rights reserved.
68 b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
involved in NH3 emission. For layer manure, these processes and is calculated using Eq. (5); Cg;∞ is NH3 concentration in the
include 1) partitioning of total ammonia nitrogen (TAN), the free air stream above the manure, which is assumed to be
sum of ammonia (NH3) and ammonium (NHþ þ
4 ), into NH4 in the represented by the NH3 concentration in the layer house, mg
solid and liquid phases and NH3 in the liquid phase, 2) con- m3; and KG is convective mass transfer coefficient (m h1),
version of liquid-phase NH3 to gas-phase NH3 at the manure which is calculated with Eq. (6).
surface, and 3) convective mass transfer of NH3 gas at the
manure surface to the free airstream facilitated by airflow Kd ¼ aKd0 (2)
(Liu, Wang, Beasley, & Shah, 2009; Ni, 1999; Tong, Zhao, Heber,
and Ni, 2020). In a laboratory-scale mechanistic model for a ¼ 10ð10:558621:43143pH0:15670MCþ0:01885pH,MCÞ ; MC 50%
(3)
estimating NH3 emissions from layer manure (Tong et al., a ¼ 10ð2:6480:42013pH0:00072MCÞ ; MC > 50%
2020), the key inputs included air temperature and velocity
at manure surface, and manure properties, such as pH and 2788
0:05Tþ273:15
moisture content (MC). Kd0 ¼ 10 (4)
The major limitation of the mechanistic model for practical
2 3
uses is that its inputs are not readily available for commercial 1825
6:123Tþ273:15
poultry production operations. Computational fluid dynamics 0:2138 4 5
Kh ¼ 10 (5)
T þ 273:15
(CFD) modelling has proven capacities to predict three-
dimensional distributions of airflow and environmental con-
ditions in animal production facilities (Bjerg, Svidt, Zhang, KG ¼ 27:9 þ 2:08T 9:40v þ 0:638Tv 0:0373T2 (6)
Morsing, & Johnsen, 2002; Kim et al., 2008; Blanes-Vidal, Gui-
where a is the dissociation constant ratio; Kd0 is the dissocia-
jarro, Balasch, & Torres, 2008; Tong, Hong, & Zhao, 2019a,
tion constant of the water-NH3 solution; T is the air temper-
2019b, 2019c). Manure properties can be estimated based on
ature (ºC) above the manure, which is assumed to be
long-term measurements in typical commercial layer houses.
represented by the air temperature in the bird zone; and v is
To assist producers and regulatory agencies in their efforts to
the average air velocity (m s1) inside the tunnel or cages. The
conveniently estimate NH3 emissions without extra labora-
key inputs to Eqs. (2)e(6) include air temperature and velocity
tory work or monitoring efforts, the key inputs of the mech-
in the bird zone, and manure pH and MC.
anistic model need to be estimated based on farm-scale
factors that are readily available at farms through statistical
2.2. Manure-belt layer houses for model validation
modelling of manure properties and CFD modelling of envi-
ronmental conditions.
Typical manure-belt layer houses use either cross ventilation,
This study aims to develop a farm-scale NH3 emission
tunnel ventilation, or a combination of the two systems. As
estimation model based on a published mechanistic model
examples of these buildings, two manure-belt layer houses in
(Tong et al., 2020) to estimate NH3 emission from typical
Indiana, USA (denoted as “IN2B00 ) were representative of cross-
commercial manure-belt layer houses for potential regulatory
ventilated houses. Two newly-built manure-belt layer houses
uses and evaluation of mitigation strategies.
with tunnel ventilation in Ohio, USA (denoted as “OH2B-T00 )
were representative of tunnel-ventilated houses. In addition,
two manure-belt layer houses retrofitted from a high-rise
2. Materials and methods
system in Ohio (denoted as “OH2B00 ) were representative of a
combination of the two ventilation types. Among these
2.1. Mechanistic model for estimating ammonia
houses, long-term continuous monitoring data of the NH3
emissions
emission rates were only available for the IN2B and OH2B
layer houses. The measurement data from IN2B and OH2B
According to the mechanistic model developed by Tong et al.
layer houses were therefore used for model validation.
(2020), NH3 emissions from layer manure can be estimated
using Eq. (1), which mathematically models different pro-
2.2.1. Cross ventilation houses at IN2B
cesses involved in NH3 emissions. Multiple empirical equa-
The two IN2B layer houses, B-A and B-B, were 140 m long,
tions (Eqs. (2)e(6)) can be used to calculate the input
19.5 m wide, and 11.5 m high at the ridge and 7 m high at the
parameters required for Eq. (1).
sidewalls (Ni et al., 2017). There were 250,110 ± 10,249 and
( 1
10pH TAN 106 241,487 ± 30,937 laying hens in B-A and B-B, respectively.
EF ¼ 1þ Kh 1000 Laying hens were stocked in seven rows of 10-tier cages, with
Kd 1 MC=100
) (1) five tiers on the second floor and five tiers on the first floor.
100 MC Manure belts underneath the cages were moved 1/3 of the
Cg;∞ KG
MC
total length every day except Sunday to remove and transport
where EF is NH3 emission flux, mg m2 h1; pH is pH value of manure through drying tunnels. The manure was dried for 3 d
layer manure; Kd is dissociation constant of layer manure (Eqs. in the drying tunnels before being conveyed to a manure
(2)e(4)); TAN is wet-based concentration of TAN, which is storage shed. A total of 88 exhaust fans of 1.32 m diameter
assumed 6.943 103 g g1 of layer manure; MC is moisture were installed in each house, including eight fans on the east
content of layer manure, %; Kh is Henry’s Law constant, end wall and 40 fans on each of the two side walls. Fourteen of
defined as (g [NH3eN] m3 [air])/(g [NH3eN] m3 [solution]), the 88 fans were variable-speed fans. Each house had seven
70 b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
adjustable air inlet openings in the ceiling, one above and performance curve with the manufacturing fan performance
along each row of cages. The exhaust fans were controlled in curve (Eq. (7)) measured by the Bioenvironmental and Struc-
12 stages based on indoor air temperature setpoints. tural Systems (BESS) Laboratory of the University of Illinois, a
New birds were introduced at the beginning of the moni- degradation factor (Eq. (8)) of 0.591 was determined to account
toring study, which started on Jan. 01, 2008, in B-A and from for airflow reduction effects of the exhaust fans used in OH2B
Oct. 4 to Nov. 8, 2008 in B-B. Moulting of birds occurred from houses caused by age of the fans, loose belts, and dust accu-
Jan. 10 to Feb. 28, 2009 in B-A and from Dec. 05, 2009 to Jan. 16, mulation on the fan blades. House ventilation rate was
2010 in B-B. More details about the IN2B monitoring were calculated (Eq. (8)) by multiplying the average airflow rate per
provided by Ni et al. (2012; 2017). fan calculated as a function of differential static pressure by
the number of operating fans, which were detected using vi-
2.2.2. Tunnel ventilation houses at OH2B-T bration sensors (Model OSU-06, Ohio State University, Co-
The two OH2B-T layer houses (Tong et al., 2019a) were 137 m lumbus, OH, USA) (Darr, Zhao, Ni, & Gecik, 2007).
long, 19 m wide, and 6.7 m high and equipped with seven rows
of 9-tier cages. Each house was capable of stocking up to Q0 ¼ 0:0004jpj2 0:0542jpj þ 13:059 (7)
196,000 laying hens. Manure was removed from the manure
belts twice a week. A total of 44 single-speed 1.32 m diameter Q ¼ nQ0 D (8)
exhaust fans were installed in the house, 22 fans in each end where Q0 is the airflow rate through one operating fan, m s ; 3 1
wall. There were eleven adjustable inlet baffles on the ceiling, p is the differential static pressure of the house, Pa; Q is the
where seven inlet baffles were located in the middle of the house ventilation rate, m3 s1; n is the number of operating
ceiling and four of them were located near the end walls. The fans; and D is the degradation factor.
ventilation operation was automatically controlled in 11 A similar methodology was used to determine house
stages based on indoor air temperature setpoints. An air ventilation rate at the IN2B layer houses (Chai et al., 2012; Ni
recirculation system was used for manure drying. More de- et al., 2017). In OH2B-T houses, the ventilation rate was
tails about the tunnel-ventilated manure-belt layer houses determined based on multiple measurements of air velocity at
were provided by Tong et al. (2019a). the surface of light traps that cover the fan annex, and a
degradation factor of 0.8 was obtained using the
2.2.3. Mixed ventilation houses at OH2B manufacturing fan curve from the BESS Laboratory (Tong
The two OH2B layer houses, houses 1 and 2, were retrofitted et al., 2019a).
from two-story high-rise deep-pit layer houses (Tong, Hong, & For producers who cannot conduct on-site measurements,
Zhao, 2019b). The houses were 122 m long, 19.5 m wide, 7.7 m the house ventilation rates can be estimated using the venti-
high at the ridge, and 6 m high at sidewalls. There were lation fan curve, which is usually provided by the fan manu-
153,588 ± 9751 and 155,183 ± 7207 laying hens in houses 1 and facturer or available at the BESS Lab website (http://bess.
2, respectively. Eight rows of 8-tier stacked cages had been illinois.edu), and an experiential estimate of degradation fac-
installed to stock laying hens, with three tiers located on the tor based on the conditions of the fans (Eq. (9); Table 1). In
first floor and five tiers located on the second floor. A concrete addition, producers can calculate the ventilation rate in mild
floor separated the two levels with openings underneath each and hot weather using Eq. (10), which is based on an energy
row of cages. Manure collected on the manure belts was balance of the ventilation process (Lu, Hayes, Stinn, Brown-
removed every Monday and Thursday to one end of the layer brandl, & Xin, 2017). In cold weather, poultry houses main-
houses and transported using trucks to a separate manure tain a fixed minimum ventilation rate, which can be estimated
composting facility. A total of 48 single-speed 1.22 m diameter using Eq. (11).
exhaust fans were installed in the house: 12 fans on each of
two end walls on the second floor and 12 fans on each of two D ¼ Dcondition Doutlet Dshutter (9)
side walls on the first floor. A continuous air inlet was located
where Qcondition is the design factor based on conditions of the
at the centre of the ceiling and had two baffles with adjustable
fans (Table 1); Doutlet is the design factor based on the type of
openings. The exhaust fans were controlled in 10 stages based
fan outlet, with a value of 1 for cone outlet and a value of 0.85 if
on indoor air temperature setpoints. The houses had a com-
the outlet is elbow shaped or a partial enclosure; and Dshutter is
bination of tunnel and cross ventilation. Moulting was
the design factor based on type of fan shutter, with a value of 1
induced from July 21 to September 1, 2007 in house 2. A new
for powered shutter and value of 0.95 for non-powered shut-
flock was introduced into house 1 on June 3, 2007. More details
ters that need to be blown open by the wind.
about the OH2B monitoring were provided by Tong et al.
(2019b). 1 h X i
QT ¼ SHP UA ðTin Tout Þ (10)
cp r ðTin Tout Þ
2.3. Determination of house ventilation rate
WVP
QM ¼ (11)
Airflow rates of some representative exhaust fans at the OH2B r ðwin wout Þ 1000
site under various differential static pressure were measured
where QT is the design ventilation rate for temperature con-
using a fan assessment numeration system (FANS) (Gates,
trol, m3 s1; r is ambient air density, kg m3; cp is specific heat
Casey, Xin, Wheeler, & Simmons, 2004) to establish a fan
of air, J kg1 ºC1; Tin is indoor temperature, ºC; Tout is outdoor
performance curve for the exhaust fans (Tong et al.,
temperature, ºC; SHP is sensible heat production from layers,
Unpublished Results). By comparing the actual fan
b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7 71
Table 1 e Experimental estimates of design factor based on conditions of the fans (Dcondition ).
Level Conditions Dcondition
New Brand new > 0.9
Good Tight belt, well maintained, no visible dust accumulation 0.8e0.9
Average Loose belt, visible dust accumulation 0.6e0.8
Poor Loose belt, poorly maintained, much dust accumulation < 0.6
W; A is area of house component, m2; U is thermal conduc- 2.4.2. Division of cages into multiple zones
tance of house component, W m2 ºC1; QM is design ventila- The cages were divided into multiple zones (Fig. 1) based on
tion rate for moisture control, m3 s1; WVP is water vapour their similarities in environmental conditions to account for
production from layers, g s1; win is humidity ratio of indoor the spatial variation of cage temperatures in the whole house.
air, kg water kg1 dry air; and wout is humidity ratio of outdoor The zone splitting of cages in one-quarter of a manure-belt
air, kg water kg1 dry air. layer house (Fig. 1) was the same as that in the other three
quarters of the house with mirrored symmetry assuming that
2.4. Estimation of thermal and airflow conditions on housing structure and ventilation operation were symmetric.
manure surface using CFD simulations Based on simulated air temperatures and velocities in cages
for the three types of layer houses, this zone splitting method
2.4.1. CFD simulations allowed for standard deviation to be smaller than 2.5 C for air
The CFD models developed by Tong et al. (2019a; 2019b; 2019c) temperature and 0.2 m s1 for air velocity within each zone in
estimated the three-dimensional spatial distributions and more than 90% of the zones during the summer. The standard
seasonal variations of airflow, thermal environment, and air deviation within a zone was higher in winter due to air stag-
pollutant concentrations in commercial manure-belt layer nation, which partially contributed to the uncertainty in
houses of typical ventilation systems with acceptable accu- estimating NH3 emissions in winter.
racies (normalised mean square errors of 0.001 0.134). These The cages in IN2B layer houses were divided (Table 3) based
CFD models were used for predicting air conditions (i.e., on their relative locations in the cross-sectional plane (Fig. 1a).
temperature and velocity) in cages under various weather They were not partitioned with respect to part number
conditions and ventilation rates with modifications on the (Fig. 1b) because changes in air temperatures and velocities in
boundary conditions to account for different inventories, cages along the building length were relatively small due to
house dimensions, and applied ventilation rates in different cross ventilation.
houses. Using a different approach, because the tunnel ventilation
Multiple cases (Table 2) were simulated using the CFD was applied on the second floor, the cages in the OH2B layer
models to account for typical weather conditions encountered houses were divided (Table 4) based on both the relative lo-
by commercial manure-belt layer houses in different seasons. cations in the cross-sectional plane and along the building
The boundary conditions of these cases were based on the length. Because different zones contained different numbers
field measurements at the IN2B, OH2B-T, and OH2B layer of cages, the proportion of cages included in each zone was
houses as described in section 2.2. weighted for the determination of total NH3 emission.
Fig. 1 e A general method of zone splitting of cages in one-quarter of a manure-belt layer house.
Fig. 2 e Comparison between measured exhaust temperatures ( ) and estimated cage temperatures ( ) in row 1 and tiers
1e3 in house B-A at IN2B site based on daily means, and the absolute difference between them ( ).
74 b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
temperature in the ith zone, ºC; and vi is average air velocity in addition, the paired t-test was performed in JMP 11.0 Statis-
the ith zone, m s1. tical Analysis Software (SAS Institute Inc., Cary, NC, USA) to
compare the measurements and estimations of cage tem-
0
ERi ¼ 2:4 105 $EFi ,As $b (18) peratures. The Nonparametric Wilcoxon Signed Rank test was
conducted in JMP to compare measured and estimated NH3
!,
X
N
X
N emissions rates. A significance level of 0.05 was used for all
0
ER ¼ ni $ ER0i ni (19)
statistical tests. The data at moulting and introduction of new
n¼1 n¼1
flocks were removed for model validation.
where ER0i is the NH3 emission rate of the whole house
Therefore the heat generated from layers could not be effi- Cage temperatures also significantly differed along the
ciently dissipated by the limited amount of air passing building length, especially in winter (Fig. A3). In winter, the
through them because of low ventilation rate in winter. In all cage temperatures at the first row were higher near the end
seasons, the cage temperatures were the lowest in tiers 1 and walls (part 1) compared with those near the air inlets (part 3).
8 among all tiers because they were located the closest to the This was because only the fans on the second floor operated in
ceiling and floor where air velocity was high because of short- winter, and allowed the formation of tunnel ventilation on the
circuiting of airflow. upper level. While no fan operated on the first floor, some cool
It was also observed that variations in cage temperatures air dropped from the second floor to the first floor and enabled
were quite small in summer (cases 1 and 2), and they a similar air temperature gradient on the first floor. In sum-
increased as ambient air temperature decreased with the mer, however, ventilation was a mixture of tunnel and cross
largest standard deviation in winter (cases 5 and 6). This ventilation, and added complexity to the spatial variation in
pattern was expected because a maximum ventilation rate air temperatures.
was used in hot summers and this allowed for an adequate
amount of air passing through all the cages. The cross venti- 3.2. Validation of CFD simulations and interpolation of
lation also resulted in relatively uniform airflow with similar cage temperatures
air velocities in cages, and thus the thermal environment in-
side cages did not have large fluctuations. In winter, the 3.2.1. IN2B houses
ventilation rate was only 10 14% of the maximum ventila- The estimations of cage temperatures were capable of indi-
tion. The airflow became less uniform because of the limited cating seasonal variations in the IN2B houses, but not during
number of operating exhaust fans and smaller openings of special events, such as moulting and a small period following
partial inlets. This led to air stagnation in the areas far away introduction of new flocks, during which the measured exhaust
from the inlets or exhausts. As a result, cage temperatures had air temperatures were much lower than the estimated tem-
larger fluctuations in winter compared with summer. peratures (Fig. 2). Furthermore, the estimated cage tempera-
tures were slightly higher than the measured exhaust air
3.1.2. Tunnel-ventilated house (OH2B-T) temperatures for most of the periods. This result was expected
Cage temperatures were higher in summer compared with because cage temperatures were normally 1 to 3 C higher than
winter (Table A2, Fig. A2). A small variation of cage tempera- those in the neighbouring aisle (Green, Wesley, Trampel, & Xin,
ture among tiers was observed in summer because all the 2009). A similar pattern was also observed in the CFD simula-
cages were well ventilated with maximum ventilation rate. tion results (Tong et al., 2019a; 2019b). The data for house B-B
However, in winter, cage temperatures were significantly validation is not shown for brevity, but was included for sta-
different among tiers because of air stagnation in some tiers tistical evaluation of model performance.
due to limited ventilation. Based on the paired t-test, the estimated cage tempera-
It was also observed that cage temperatures near the end tures were significantly higher (p < 0.01) than the measured
walls (part 1) were systematically higher than near the air exhaust air temperatures. Despite that, they were highly
inlets (part 3) in summer as was expected under tunnel correlated (r ¼ 0.85). After adjusting the estimations by the
ventilation. The difference between parts 1 and 3 was not as systematic difference, the adjusted estimations were not
obvious in winter because plug flow could not be formed with significantly different (p ¼ 0.94) than the measurement. It was
a lower ventilation rate and partially closed inlet baffles. concluded that CFD simulations and linear interpolations
Instead of creating a large lengthwise gradient of air temper- (Eqs. (12) and (13)) performed adequately in estimating cage
ature, vortex airflows (Tong et al., 2019a) were formed temperatures for normal egg production excluding periods of
resulting in smaller variations of air temperatures along the moulting and flock replacement.
building length in winter.
3.2.2. OH2B houses
3.1.3. Retrofitted house (OH2B) In the OH2B layer houses, the estimated cage temperatures
Variations in cage temperatures among tiers were relatively agreed well with the measurements (Fig. 3). The agreement
stable in summer, whereas in winter, the variations were between the measurements and estimations were better in
quite high in tiers 2 and 3 on the first floor and in tier 7 on the warm seasons than in cold seasons. This was likely due to
second floor of the house (Table A3, Fig. A3). The low air relatively stable airflow in summer when most exhaust fans
temperatures were observed in tiers 4 and 5 located near the operated, whereas fan staging changed often in cold seasons,
floor. This pattern resulted from the house ventilation rates causing large fluctuations in thermal conditions. In addition,
and the housing structure. In summer, the maximum venti- cage temperatures were more sensitive to their locations in
lation rate was applied with exhaust fans operating at both winter because only a small amount of cold air could pass
first and second floors and provided efficient ventilation to the through some cages. The pathways of cold air jets fluctuated
cages. In contrast, only a few exhaust fans operated on the with time due to frequently changed fan operation staging,
second floor in winter with low-velocity air in cages in tiers 2, and therefore the estimation had larger uncertainties in cold
3, and 7, leading to higher cage temperatures. Low air tem- seasons. The estimations were lower than the measurements
peratures in tiers 4 and 5 occurred because they were located at the later moulting period that occurred in house 2 at OH2B
near the floor where high-speed cold air jets passed through a (Fig. 3). This was different from the case in the IN2B houses
larger gap. Another reason was that cold air tends to drop where the model overestimated cage temperatures during
because of its larger density. moulting (Fig. 2).
76 b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
Fig. 3 e The measured ( ) and the estimated ( ) temperatures at UCE in house 2 at OH2B site, and the absolute difference
between them ( ).
Good agreement was observed between measurements 3.3.3. Retrofitted house (OH2B)
and estimations of cage temperatures in both IN2B and Air velocities in cages were the highest in summer and lowest
OH2B houses (Table 5). All of the statistical criteria were in winter in general (Table B3, Fig. B3). In winter, the highest
met, indicating that the CFD simulations and linear inter- air velocities were observed at tier 4 because the high-speed
polation had adequate performance in estimating cage air jets mainly flowed through the gap between the floor and
temperatures. the 4th tier. The peak of air velocities in winter decreased from
the 1st row to the 4th row because air jets had the highest
3.3. Estimation of air velocity in cages using CFD velocity near the side walls and this reduced as the air jets
flowed towards the centre of the house. In contrast, the
3.3.1. Cross-ventilated house (IN2B) highest air velocities in summer were observed in tiers 5e7 of
Air velocities in cages were observed to increase as ambient the 4th row. This was because cages in row 4 were the closest
temperature increased (Table B1, Fig. B1). Among all the to the inlet baffles, and thus high-speed incoming air first
tiers, the cages at the top tier had the highest air velocity flowed into the tiers located on the upper level. Overall, spatial
because they were the closest to the inlets with no blocking variations of cage air velocities in the retrofitted houses were
effect from housing structures. It was also observed that more complex than those in the cross-ventilated and tunnel-
cage air velocities were the highest in row 1 and the lowest ventilated houses.
in row 4. This was because row 1 was directly exposed to
sidewall exhaust fans, whereas row 4 was located in the 3.4. Manure pH
middle of the house and was blocked by in-between rows of
cages. Obvious temporal variations of manure pH were observed for
both of the IN2B and the OH2B houses. The change of manure
3.3.2. Tunnel-ventilated house (OH2B-T) pH with time were not consistent between the two sites
The variations of air velocity in cages among different tiers (Fig. 4).
and rows were relatively small due to tunnel ventilation Since no data on temporal layer manure pH changes was
(Table B2, Fig. B2). On the other hand, air velocities in cages found in the literature, the manure pH in this study was
located near the exhaust fans on the end walls (part 1) were estimated based on measured data. In particular, the rela-
systematically higher than those located near the air inlets tionship of manure pH with month was fitted separately for
(part 3). This agreed with the observation that a large gradient each site using a third order polynomial regression to explain
of air velocity was formed along the length of the building the temporal variation in layer manure pH. The fitted equa-
because of the nature of tunnel ventilation. tions for estimating manure pH in the IN2B houses and the
OH2B houses explained 92.4% and 98.0% of the total variations
in manure pH. Table 6 shows the estimated manure pH in
different months for the IN2B and OH2B sites based on fitted
Table 5 e Model performance in estimating cage
temperatures. equations. Since manure pH is the most sensitive factor
affecting NH3 emission (Tong et al., 2020), it is recommended
Criteria IN2B OH2B
that producers sample manure from belts at least monthly for
NMSE (< 0.25) 0.01 0.04
characterising the pH used to estimate NH3 emission.
|FB| (< 0.3) 0.03 0.10
MG (0.7e1.3) 0.97 1.10
VG (< 4) 0.94 1.22
3.5. Manure moisture content
FAC2 (> 0.5) 1.00 1.00
p-value of paired t test (> 0.05) 0.94 0.52 Manure MC in the IN2B houses ranged from 62.4 to 76.1%,
Note: the criteria in parenthesis indicate an adequate model
which were generally higher than those in the OH2B houses
performance. ranging from 49.0 to 68.6% (Fig. 5). The difference in manure
b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7 77
Fig. 4 e Manure pH in different months for IN2B and OH2B Fig. 5 e Manure MC in different months for IN2B and OH2B
layer houses and the fitted polynomial regression. The layer houses. The error bars indicate standard deviations
error bars indicate standard deviations from multiple from multiple manure pH measurements from each
manure pH measurements from each sampling event. sampling event.
Fig. 6 e The measured ( ) and the estimated ( ) NH3 emissions from house B-B at IN2B site based on (a) hourly averages and
(b) daily averages, and the absolute difference ( ) between daily measurements and estimations.
Fig. 7 e The measured ( ) and the estimated ( ) NH3 emissions from house 1 at OH2B site based on (a) hourly averages and
(b) daily averages, and the absolute difference ( ) between daily measurements and estimations.
Table 8 e NMSE of the model in different seasons. Table 9 e Uncertainty of farm-scale model for IN2B and
OH2B houses.
Seasons IN2B OH2B
IN2B OH2B
Hourly Daily Hourly Daily
Average Change in Average Change in
Summer 0.24 0.18 0.19 0.10
Standard ERa (%) Standard ERa (%)
Spring/Autumn 0.25 0.20 0.35 0.24
Deviation Deviation
Winter 0.80 0.55 1.64 1.07
pH 0.10 11.7 0.08 9.38
Note: the criteria in parenthesis indicate an adequate model
Moisture Content 2.8% 0.93 4.0% 1.32
performance.
Temperature 2.1 C 27.2 1.5 C 19.4
Air Velocity 0.14 m s1 11.6 0.18 m s1 14.9
Uncertainty 31.8% 26.2%
performance did not satisfy hourly estimations because the a
The change in ER corresponds to the change in each input factor
criteria were nearly met for FB or MG, especially for OH2B by its average standard deviation while other input factors are
houses. In contrast, the model was considered adequate for remained at their mean values.
daily estimations because more than half of the criteria were
met for both sites (Hanna & Chang, 2011). The p-values larger
than 0.05 for Wilcoxon test further showed the model ade-
quacy in daily NH3 emission estimations. emission was introduced when using farm-scale factors to es-
Based on the Wilcoxon signed rank test, both the hourly timate the inputs of the mechanistic model (Tong et al., 2020).
and daily estimations of NH3 emission in the IN2B houses
agreed well with measurement. For the OH2B houses, how- 3.6.5. Model application
ever, significant differences were observed between the The model can be used by egg producers and regulatory
hourly estimations and hourly measurements (p ¼ 0.02). The agencies to estimate average NH3 emission from a typical
model was therefore capable of estimating daily averages of manure-belt layer house on any given day. Egg producers can
NH3 emission, but not hourly averages of NH3 emission. The readily use the model to estimate the NH3 emission from their
daily resolution seemed reasonable for estimation because farms as long as they have access to ambient air temperature
the process of NH3 emission from manure involved various and house ventilation rate data in addition to the house and
chemical, biological, and physical processes which were less egg production information. Ambient air temperature can be
likely to yield an acute response on an hourly basis. either monitored onsite or obtained from a nearby weather
station. House ventilation rate is determined based on airflow
3.6.4. Model uncertainty rate of representative fans and the operational status of all the
The uncertainty (Table 9) of the farm-scale model relative to the fans during production. Producers can determine the house
lab-scale mechanistic model (Tong et al., 2020) was 31.8% for ventilation rate using the manufacture fan curve and an
the IN2B houses and 26.2% for the OH2B houses using Eq. (25). estimated degradation factor based on the conditions of fans
This indicated that an additional 26.2 31.8% of error in NH3 (Table 1). In addition, producers can calculate the ventilation
80 b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
rate based on energy balance of the ventilation process using further improved in the future through better estimations of
Eqs. (10) and (11). layer manure pH.
Regarding manure properties, producers are encouraged to Overall, the model exhibited adequate performance in
collect manure samples at their own farms every month and estimating daily average NH3 emissions in representative
send them to a certified lab to test manure pH, moisture manure-belt layer houses during ordinary production. Over-
content, and TAN to obtain a more precise NH3 emission es- estimation of NH3 emissions was observed during special
timate. Alternatively, producers could use the averaged values events such as moulting and introduction of new flocks.
provided in this study, as they are reasonable assumptions for The model can be used as a tool for producers to estimate
the manures in typical manure-belt layer houses. effects of various management practices or mitigation stra-
The model can also be used by regulatory agencies to tegies on NH3 emissions from manure-belt layer houses in the
establish science-based reporting rules. The generalisation of USA for reduced environmental and health impacts.
the model to different types of layer houses, such as cage-free
and enriched colony systems, is possible once CFD models
become available for such layer houses to estimate the air Declaration of Competing Interest
temperature and velocity near manure surfaces.
None declared.
4. Conclusion
Acknowledgement
The developed farm-scale NH3 emission estimation model
can be used for estimating NH3 emission from typical manure-
This study was supported by the USDA-NIFA Grant 2018-
belt layer houses with tunnel, cross-flow, and combined
67019-27803. The authors are thankful for the managers and
ventilation systems. The input factors of the farm-scale NH3
staff of the collaborating poultry farm for their support and
emission model that are accessible to producers and regula-
assistance to the project. The IN2B data were from the Na-
tory agencies consist of hen inventory, ambient temperature,
tional Air Emission Monitoring Study that received adminis-
house ventilation mode and rate, manure production rate,
trative and financial support from the Agricultural Air
and manure removal schedule.
Research Council and the American Egg Board, and collabo-
The farm-scale model had a 26 32% uncertainty relative
rations of the Indiana egg producer. The work of A. J. Heber
to the mechanistic model (Tong et al., 2020) in NH3 emission
and J.-Q. Ni was supported by the USDA-NIFA Hatch project
estimation based on the sensitivity of the multiple input fac-
1011562.
tors of the mechanistic model. Based on analyses of VG, MG,
FAC2, and p-value of Wilcoxon signed rank test, the model is Appendix A. CFD simulation results on cage
adequate for predicting daily NH3 emissions (p ¼ 0.10e0.61), temperatures
but not adequate for accurately predicting hourly NH3 emis-
sions (p ¼ 0.02e0.15). The performance of the model could be
Table A1 e Average cage temperatures ( C) ± standard deviations in each zone in a cross-ventilated layer house (IN2B).
Zone Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
A 31.6 ± 0.3 31.8 ± 0.4 29.6 ± 0.8 28.5 ± 1.6 24.5 ± 2.8 24.2 ± 2.8
B 31.5 ± 0.4 31.8 ± 0.6 29.4 ± 1.0 29.3 ± 2.4 27.3 ± 5.3 26.0 ± 5.0
C 32.2 ± 0.8 32.1 ± 0.5 29.8 ± 1.1 27.7 ± 3.8 26.8 ± 4.4 26.5 ± 4.5
D 32.9 ± 0.6 33.2 ± 0.9 31.4 ± 2.1 30.3 ± 2.4 29.2 ± 4.8 29.5 ± 4.5
E 30.5 ± 0.1 30.5 ± 0.1 27.0 ± 0.3 22.8 ± 1.1 13.8 ± 1.8 13.2 ± 2.1
F 32.0 ± 0.7 32.1 ± 0.7 29.9 ± 1.5 28.5 ± 1.9 24.2 ± 2.9 23.8 ± 2.9
G 32.0 ± 0.9 32.1 ± 1.2 30.0 ± 2.2 28.5 ± 3.5 29.4 ± 5.6 28.5 ± 5.6
H 32.2 ± 0.9 32.6 ± 1.3 31.0 ± 3.0 28.7 ± 3.7 31.9 ± 7.1 32.2 ± 6.5
I 32.7 ± 1.2 33.1 ± 1.0 31.7 ± 3.0 31.2 ± 3.6 26.2 ± 12.9 25.1 ± 11.4
J 30.5 ± 0.2 30.5 ± 0.2 27.0 ± 0.5 22.5 ± 1.5 18.8 ± 7.0 17.3 ± 7.3
K 31.7 ± 0.3 32.1 ± 0.5 30.2 ± 1.1 21.5 ± 2.8 26.5 ± 2.8 25.5 ± 2.8
L 32.0 ± 0.6 32.4 ± 0.9 30.9 ± 2.9 30.3 ± 3.1 32.5 ± 3.5 33.2 ± 3.4
M 33.8 ± 1.1 33.9 ± 1.1 33.5 ± 2.8 31.8 ± 4.0 32.5 ± 6.6 33.2 ± 5.5
N 33.6 ± 1.0 34.0 ± 0.9 33.8 ± 2.9 30.9 ± 4.6 24.8 ± 9.3 24.2 ± 9.9
O 30.4 ± 0.1 30.4 ± 0.1 26.8 ± 0.3 22.5 ± 1.3 17.6 ± 2.2 16.4 ± 2.9
b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7 81
b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
velocity in cages
Appendix B. CFD simulation results on air
Table B1 e Air velocity ± standard deviation (m s¡1) in each zone in a cross-ventilated layer house (IN2B).
Zone Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
A 0.50 ± 0.10 0.40 ± 0.05 0.28 ± 0.09 0.14 ± 0.07 0.12 ± 0.04 0.12 ± 0.04
B 0.74 ± 0.16 0.66 ± 0.14 0.48 ± 0.26 0.20 ± 0.16 0.18 ± 0.15 0.17 ± 0.16
C 0.64 ± 0.26 0.53 ± 0.13 0.36 ± 0.20 0.18 ± 0.11 0.15 ± 0.11 0.15 ± 0.10
D 0.46 ± 0.28 0.27 ± 0.13 0.21 ± 0.11 0.14 ± 0.09 0.13 ± 0.11 0.12 ± 0.12
E 1.64 ± 0.30 1.26 ± 0.18 1.26 ± 0.18 0.38 ± 0.08 0.49 ± 0.09 0.42 ± 0.08
F 0.42 ± 0.13 0.34 ± 0.08 0.23 ± 0.08 0.10 0.10 0.10
G 0.56 ± 0.38 0.42 ± 0.27 0.30 ± 0.26 0.13 ± 0.13 0.10 0.10
H 0.52 ± 0.42 0.34 ± 0.21 0.24 ± 0.22 0.13 ± 0.11 0.10 0.10
I 0.44 ± 0.37 0.26 ± 0.21 0.17 ± 0.14 0.10 0.17 ± 0.22 0.15 ± 0.24
J 1.64 ± 0.35 1.27 ± 0.23 0.76 ± 0.14 0.37 ± 0.11 0.33 ± 0.13 0.32 ± 0.15
K 0.39 ± 0.21 0.21 ± 0.10 0.14 ± 0.05 0.10 0.10 0.10
L 0.47 ± 0.28 0.24 ± 0.16 0.15 ± 0.09 0.10 0.10 0.10
M 0.31 ± 0.21 0.15 ± 0.09 0.10 0.10 0.10 0.10
N 0.42 ± 0.36 0.20 ± 0.16 0.10 0.10 0.15 ± 0.13 0.13 ± 0.10
O 1.57 ± 0.27 1.25 ± 0.16 0.75 ± 0.10 0.31 ± 0.09 0.26 ± 0.10 0.25 ± 0.07
1
Air velocities equal to or lower than 0.10 m s were shown as 0.10 m s1.
83
84 b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7
Bjerg, B., Svidt, K., Zhang, G., Morsing, S., & Johnsen, J. O. (2002).
Modeling of air inlets in CFD prediction of airflow in ventilated
animal houses. Computers and Electronics in Agriculture, 34(1e3),
223e235. https://doi.org/10.1016/S0168-1699(01)00189-2.
Blanes-Vidal, V., Guijarro, E., Balasch, S., & Torres, A. G. (2008).
Application of computational fluid dynamics to the prediction
of airflow in a mechanically ventilated commercial poultry
building. Biosystems Engineering, 100(1), 105e116. https://
doi.org/10.1016/j.biosystemseng.2008.02.004.
Chai, L., Ni, J.-Q., Diehl, C. A., Kilic, I., Heber, A. J., Chen, Y., …
Chen, L. (2012). Ventilation rates in large commercial layer hen
houses with two-year continuous monitoring. British Poultry
Science, 53(1), 19e31. https://doi.org/10.1080/
00071668.2011.643766.
Chang, J. C., & Hanna, S. R. (2004). Air quality model performance
evaluation. Meteorology and Atmospheric Physics, 87(1e3),
167e196. https://doi.org/10.1007/s00703-003-0070-7.
Chepete, J. H., Xin, H., & Li, H. (2011). Ammonia emissions of
laying-hen manure as affected by accumulation time. The
Journal of Poultry Science, 48(2), 133e138. https://doi.org/10.2141/
jpsa.010087.
Darr, M. J., Zhao, L., Ni, J.-Q., & Gecik, C. (2007). A robust sensor for
monitoring the operational status of agricultural ventilation
fans. Transactions of the ASABE, 50(3), 1019e1027. https://
doi.org/10.13031/2013.23142.
Gates, R. S., Casey, K. D., Xin, H., Wheeler, E. F., & Simmons, J. D.
(2004). Fan assessment numeration systems (FANS) design
and calibration specifications. Transactions of the ASAE, 47(5),
1709e1715. https://doi.org/10.13031/2013.17613.
Green, A. R., Wesley, I., Trampel, D. W., & Xin, H. (2009). Air
quality and bird health status in three types of commercial egg
layer houses. Journal of Applied Poultry Research, 18(3), 605e621.
https://doi.org/10.3382/japr.2007-00086.
Hadjab, M., & Medjahed, B. (2012). Comparison and statistical
validation of a model of a photovoltaic module. International
Journal of Energy, 4(6), 133e140.
Hanna, S. R., & Chang, J. (2011). Setting acceptance criteria for air
quality models. In D. Steyn, & S. Trini Castelli (Eds.), Air
Pollution Modeling and its Application XXI. NATO Science for Peace
and Security Series C: Environmental Security (pp. 479e484).
Dordrecht: Springer.
Heber, A. J. (2010). Emissions data from two manure-belt layer
houses in Indiana. Final report for site IN2B of the National Air
Emissions Monitoring Study. Durham, NC: U.S.EPA. https://
archive.epa.gov/airquality/afo2012/web/pdf/
in2bsummaryreport.pdf. (Accessed 10 May 2019).
Heber, A. J., Bogan, W. W., Ni, J.-Q., Lim, T. T., Ramirez-
Dorronsoro, J. C., Cortus, E. L., … Zhang, R. (2008). The national
air emissions monitoring study: Overview of barn sources.
Fig. B3 e Average air velocities in cages with tier number at Livestock Environment VIII - Proceedings of the 8th International
(a) row 1, (b) row 3, and (c) row 4 in different seasons in a Symposium. https://doi.org/10.13031/2013.25499.
JCGM. (2008). Evaluation of measurement data d Guide to the
retrofitted layer house (OH2B). The error bars indicate
expression of uncertainty in measurement. International
standard deviations.
Organization for Standardization. https://doi.org/10.1373/
clinchem.2003.030528. Geneva ISBN, 50 (September).
Kim, K., Yoon, J. Y., Kwon, H. J., Han, J. H., Eek Son, J., Nam, S. W.,
… Lee, I. B. (2008). 3-D CFD analysis of relative humidity
distribution in greenhouse with a fog cooling system and
refrigerative dehumidifiers. Biosystems Engineering, 100(2),
245e255. https://doi.org/10.1016/j.biosystemseng.2008.03.006.
references Liang, Z. S., Westerman, P. W., & Arogo, J. (2002). Modeling
ammonia emission from swine anaerobic lagoons.
Transactions of the ASABE, 45(3), 787e798. https://doi.org/
ASTM. (2002). Standard guide for statistical evaluation of indoor 10.13031/2013.8859.
air quality models (D5157-97). Annual book of American Society Liang, Y., Xin, H., Wheeler, E. F., Gates, R. S., Li, H.,
for Testing Materials Standards. West Conshohocken, PA: Zajaczkowski, J. S., … Zajaczkowski, F. J. (2005). Ammonia
American Society for Testing and Materials. emissions from US laying hen houses in Iowa and
b i o s y s t e m s e n g i n e e r i n g 1 9 6 ( 2 0 2 0 ) 6 7 e8 7 87
Pennsylvania. Transactions of the ASAE, 48, 1927e1941. https:// Patryl, L., & Galeriu, D. (2011). Statistical performances measures -
doi.org/10.13031/2013.20002. models comparison. International Atomic Energy Agency.
Liu, Z., Wang, L., Beasley, D. B., & Shah, S. B. (2009). Modeling https://www-ns.iaea.org/downloads/rw/projects/emras/
ammonia emissions from broiler litter at laboratory scale. emras-two/first-technical-meeting/sixth-working-group-
Transactions of the ASABE, 52(5), 1683e1694. https://doi.org/ meeting/working-group-presentations/workgroup-7-
10.13031/2013.29131. presentations/presentation-6th-wg7-statistical-
Li, Q. F., Wang-Li, L., Bogan, B. W., Wang, K., Chai, L., Ni, J.-Q., performances.pdf. (Accessed 10 May 2019).
et al. (2013). The National Air Emissions Monitoring Study’s Theobald, M. R., Sanz-Cobena, A., Vallejo, A., & Sutton, M. A.
Southeast layer site: Part IV. Effects of Farm Management. (2015). Suitability and uncertainty of two models for the
Transactions of the ASABE, 56, 1199e1209. https://doi.org/ simulation of ammonia dispersion from a pig farm located in
10.13031/trans.56.9674, 2010. an area with frequent calm conditions. Atmospheric
Lu, Y., Hayes, M., Stinn, J. P., Brown-brandl, T., & Xin, H. Environment, 102, 165e175. https://doi.org/10.1016/
(2017). Evaluating ventilation rates based on new heat and j.atmosenv.2014.11.056.
moisture production data for swine production. Transactions Tong, X., Hong, S. W., & Zhao, L. (2019a). CFD modelling of airflow
of the ASABE, 60(1), 237e245. https://doi.org/10.13031/ pattern and thermal environment in a commercial manure-
trans.11888. belt layer house with tunnel ventilation. Biosystems
Maliselo, P. S., & Nkonde, G. K. (2015). Ammonia production in Engineering, 178, 275e293. https://doi.org/10.1016/
poultry houses and its effect on the growth of Gallus Gallus j.biosystemseng.2018.08.008.
Domestica (broiler chickens): A case study of a small scale Tong, X., Hong, S. W., & Zhao, L. (2019b). CFD modelling of airflow,
poultry house in Riverside, Kitwe, Zambia. International Journal thermal environment, and ammonia concentration
of Scientific & Technology Research, 4(4), 141e145. distribution in a commercial manure-belt layer house with
Ni, J.-Q. (1999). Mechanistic models of ammonia release from mixed ventilation systems. Computers and Electronics in
liquid manure: A review. Journal of Agricultural Engineering Agriculture, 162, 281e299. https://doi.org/10.1016/
Research, 72(1), 1e17. https://doi.org/10.1006/jaer.1998.0342. j.compag.2019.03.031.
Ni, J.-Q., Chai, L., Chen, L., Bogan, B. W., Wang, K., Cortus, E. L., … Tong, X., Hong, S. W., & Zhao, L. (2019c). Using CFD simulations to
Diehl, C. A. (2012). Characteristics of ammonia, hydrogen develop an upward airflow displacement ventilation system
sulfide, carbon dioxide, and particulate matter concentrations for manure-belt layer houses to improve the indoor
in high-rise and manure-belt layer hen houses. Atmospheric environment. Biosystems Engineering, 178, 294e308. https://
Environment, 57, 165e174. https://doi.org/10.1016/ doi.org/10.1016/j.biosystemseng.2018.08.006.
j.atmosenv.2012.04.023. Tong, X., Zhao, L., Heber, A. J., & Ni, J.-Q. (2020). Mechanistic
Ni, J.-Q., Cortus, E. L., & Heber, A. J. (2011). Improving ammonia modelling of ammonia emission from laying hen manure at
emission modeling and inventories by data mining and laboratory scale. Biosystems Engineering, 192, 24e41. https://
intelligent interpretation of the national air emission doi.org/10.1016/j.biosystemseng.2020.01.004.
monitoring study database. Atmosphere, 2(2), 110e128. https:// Tong, X., Zhao, L., Manuzon, R. B., Darr, M. J., Heber, A. J., & Ni, J.-
doi.org/10.3390/atmos2020110. Q. Unpublished Results. Ammonia concentrations and
Ni, J.-Q., Diehl, C. A., Chai, L., Chen, Y., Heber, A. J., Lim, T.-T., emissions at two commercial manure-belt layer houses with a
et al. (2017). Factors and characteristics of ammonia, hydrogen mixed used of tunnel and cross ventilation systems.
sulfide, carbon dioxide, and particulate matter emissions from Biosystems Engineering, under review YBENG-D-20-00173
two manure-belt layer hen houses. Atmospheric Environment, Wang, S., Zhao, L., Wang, X., Manuzon, R., Darr, M., Li, H., et al.
156, 113e124. https://doi.org/10.1016/j.atmosenv.2017.02.033. (2014). Estimation of ammonia emission from manure belt
OSHA. (2019). How much ammonia is too much? Part 2. https://www. poultry layer houses using an alternative mass balance
osha.gov/sites/default/files/2019-03/fs4-howmuch2.pdf. method. Transactions of the ASABE, 57(3), 937e947. https://
(Accessed 10 May 2019). doi.org/10.13031/trans.57.10506.