scispace - formally typeset
Search or ask a question
Journal ArticleDOI

Particulate matter emitted from poultry and pig houses: Source identification and quantification

01 Apr 2011-Transactions of the ASABE (American Society of Agricultural and Biological Engineers)-Vol. 54, Iss: 2, pp 629-642
TL;DR: In this paper, the chemical and morphological characteristics of fine and coarse airborne particulate matter (PM) from known sources collected from animal houses were compared using two methods: classification rules based on decision trees and multiple linear regression.
Abstract: There is need to identify and quantify the contribution of different sources to airborne particulate matter (PM) emissions from animal houses. To this end, we compared the chemical and morphological characteristics of fine and coarse PM from known sources collected from animal houses with the characteristics of on-farm fine and coarse airborne PM using two methods: classification rules based on decision trees and multiple linear regression. Fourteen different farms corresponding to seven different housing systems for poultry and pigs were sampled during winter. A total of 28 fine and 28 coarse on-farm airborne PM samples were collected, together with a representative sample of each known source per farm (56 known source samples in total). Source contributions were calculated as relative percentage contributions in particle numbers and then estimated in particle mass. Based on particle numbers, results showed that in poultry houses, most on-farm airborne PM originated from feathers (ranging from 4% to 43% in fine PM and from 6% to 35% in coarse PM) and manure (ranging from 9% to 85% in fine PM and from 30% to 94% in coarse PM). For pigs, most on-farm airborne PM originated from manure (ranging from 70% to 98% in fine PM and from 41% to 94% in coarse PM). Based on particle mass, for poultry most on-farm airborne PM still originated from feathers and manure; for pigs, however, most PM originated from skin and manure. Feed had a negligible contribution to on-farm airborne PM compared with other sources. Results presented in this study improve the understanding of sources of PM in different animal housing systems, which may be valuable when choosing optimal PM reduction techniques.

Summary (4 min read)

Introduction

  • Linear regression is used to estimate the relative contribution of each known source as the linear sum of products of source compositions and source contributions, based on predetermined source profiles (Hopke, 1991).
  • The objective of this study was to identify and quantify the contribution of different sources to primary fine (PM2.5) and coarse (PM10‐2.5) PM emissions from animal houses based on chemical and morphological characteristics of particles.
  • It also gives an insight into the environmental hazards of PM and their potential health effects by providing knowledge on sources of PM through properties such as particle morphology and chemical composition.

HOUSING AND ANIMALS

  • Table 1 lists the surveyed animal species, type of housing system, ventilation system, number of animals, and animal age where airborne and source samples were collected.
  • Two different farm locations were sampled for each animal housing system in The Netherlands during winter season.
  • All surveyed animal houses used automatically distributed feeding systems with crumbs or pelleted feed.
  • Broilers and turkey houses used new wood shavings as bedding.
  • Table 1. Description of surveyed animal houses.

ON‐FARM AIRBORNE AND SOURCE SAMPLES

  • Virtual cascade impactors (RespiCon, Helmut Hund GmbH, Wetzlar, Germany) were used on each farm to sample airborne fine and coarse PM onto separate polycarbonate filters (37 mm dia., 5 m pore size).
  • Virtual impactors are similar to conventional impactors, but the impaction surface is replaced with a virtual space of stagnant or slow‐moving air, consequently reducing sampling problems common to conventional impactors, such as overloading and particle bounce losses.
  • Besides 14 background samples, a total of 42 known source samples were sampled, including: concentrate feed (all farms), manure (fresh excreta in poultry and fresh feces in pigs), feathers (in poultry), and wood shavings used as bedding material (present only for broilers and turkeys).
  • Dried and milled samples were stored at room temperature, and then airborne PM was generated in a laboratory dust generator to collect airborne fine and coarse PM samples from each source.
  • The filter cassettes had a static‐dissipative nature and were chosen to protect filters, minimize sample losses, and avoid contamination during storage.

MORPHO‐CHEMICAL ANALYSIS OF AIRBORNE AND SOURCE SAMPLES

  • High‐resolution scanning electron microscopy (SEM) (JSM‐5410, JEOL Ltd., Tokyo, Japan) combined with energy‐dispersive x‐ray analysis (EDX) (Link Tetra analyzer, Oxford Instruments, Abingdon, U.K.) was used to obtain particle‐by‐particle chemical and morphological data.
  • The SEM‐EDX was conducted manually and operated under the same conditions throughout the study: 10 keV accelerating voltage, 15 mm working distance, 3 nA electron probe current, 1000× magnification for coarse PM and 1800× for fine PM, and x‐ray acquisition time of 60 s per particle.
  • At least three fields of view per filter sample were analyzed.
  • Within each field, the minimum projected area diameter for 632 TRANSACTIONS OF THE ASABE the coarse particles was set at 1 m.
  • Based on morphological characteristics, each particle was characterized by 23 variables.

SOURCE APPORTIONMENT METHODS

  • Fine and coarse source samples as well as on‐farm airborne fine and coarse PM samples from each animal house were used in source apportionment using classification rules based on decision trees and multiple linear regression.
  • Single‐particle chemical and morphological characteristics obtained using SEM‐EDX were used as data sources.
  • Apportionment results were calculated as relative percentage contributions in number and then estimated in terms of mass.
  • Results provided by the two methods are compared and discussed.

Classification Rules Based on Decision Trees

  • Decision trees were used to develop a set of rules for each group of sources from each animal house.
  • Single‐particle chemical and morphological characteristics from known sources obtained using SEM‐EDX were joined in a combined database and used in this process.
  • Decision trees were built using See 5 software, using the C5.0 classification algorithm, which is the latest version of the algorithms ID3 and C4.5 developed by Quinlan (1993).
  • This method searched the features that best separated one source from the other by dividing data using mutually exclusive conditions until the newly generated subgroups were homogeneous, i.e., all the elements in a subgroup belonged to the same class or a stopping condition was fulfilled.
  • Overall measure of prediction accuracy for number of particles was obtained by dividing the total correct validations in each source by the total number of classified particles.

Multiple Linear Regression

  • Multiple linear regression was also used to apportion airborne PM sampled on the farms to the known sources.
  • Bulk source chemical characteristics from known sources obtained from the average of single‐particle chemical characteristics using SEM‐EDX, were used in this process.
  • The average PM concentration of elements in fine and coarse airborne on‐farm samples were used as dependent variables, and the average fine and coarse PM concentrations of elements in each source were used as independent variables.

Mass Estimation

  • Results from the classification rules based on decision trees and multiple linear regression were given in particle numbers.
  • Particle number contributions were transformed into mass contributions based on the average mass of particles in each source.
  • The mass for each single particle (m) was calculated from the projected area diameter (Dp) provided by the SEM images, based on a density value and shape factor, following the equation for the mass of a particle (eq. 2) (Ott et al., 2008).

ON FARM AIRBORNE PM MEASUREMENTS

  • Environmental conditions during sampling are shown in table 2, including average PM10 concentrations measured using a light‐scattering system, relative humidity, and temperature measured inside and outside the animal houses.
  • Values in the table represent sampling time averages over 5 to 60 min and standard errors between the two surveyed houses for the same animal category.
  • Sources were identified through individual particle morphologies based on SEM observations.
  • In turkey houses, bent, sharp‐edged particles from wood shavings or feathers, and few spherical particles from excreta were identified (fig. 1d).
  • A mixture of layered manure particles and large, flattened skin particles was identified in collected PM from piglet houses (fig. 1f) and from growing‐finishing pigs (fig. 1g).

Using Classification Rules Based on Decision Trees

  • Results using classification rules based on decision trees shown in table 7 (fine PM) and table 8 (coarse PM) show different relative source contributions from number contributions.
  • The contribution of feathers decreased for broilers but increased or did not vary for laying hens and turkeys when expressed in mass.
  • In mass, the relative percentage contribution of manure was higher for laying hens compared with broilers and turkeys (same as for numbers), but the relative percentage contribution of feathers was higher for laying hens, especially compared with broilers.
  • The mass contribution of skin considerably increased compared with number contributions, in some cases ten‐fold, ranging from 29% to 68% when expressed in mass, and thus decreasing the contribution of manure to below 65% in fine PM and below 41% in coarse PM.
  • Wood shavings showed approximately a two‐fold increase in mass compared with number contributions, whereas the relative percentage contribution of feed and outside did not vary or was generally lower compared with number contributions.

Using Multiple Linear Regression

  • Results using multiple linear regression are shown in table 9 (fine PM) and table 10 (coarse PM).
  • These results are comparable to using classification rules based on decision trees, showing similar trends and differences when compared with number contributions, increasing the relative percentage contribution of feathers for laying hens, of manure for broilers and turkeys, and of skin for pigs.

COMPARISON BETWEEN METHODS

  • Results between classification rules based on decision trees and multiple linear regression in number of particles showed relatively high linear correlations (R2 = 0.75 for fine PM and R2 = 0.61 for coarse PM) (fig. 2).
  • Correlations were higher for fine PM compared with coarse PM, probably influenced by the disagreement in the contribution of skin for piglets in coarse PM between methods.

DISCUSSION

  • The authors results have been presented and analyzed as relative percentage contributions.
  • Fecal particles can morphologically resemble feed particles.
  • Other studies have reported similar source mass contributions.
  • Overall, both methods used in their study to quantify PM source contributions from animal houses presented comparable results.

CONCLUSIONS

  • Results presented in this study improve the understanding of sources of PM in different animal housing systems, not only in numbers but also in mass contributions, which may be valuable when choosing optimal PM reduction techniques.
  • Using two independent methods, source apportionment results were consistent between classification rules based on decision trees and multiple linear regression (R2 = 0.75 for fine PM and R2 = 0.61 for coarse PM), and with detailed and specific chemical and morphological source profiles, both methods presented sufficient levels of accuracy for the aim of this study.
  • Based on particle numbers, in poultry houses, most on‐ farm airborne PM originated from feathers (relative percentage contribution ranging from 4% to 43% in fine and from 6% to 35% in coarse PM) and manure (ranging from 9% to 85% in fine and from 30% to 94% in coarse PM).
  • Based on particle numbers, in pig houses, most on‐farm airborne PM originated from manure (relative percentage contribution ranging from 70% to 98% in fine and from 41% to 94% in coarse PM).
  • Feed had a negligible contribution to on‐farm airborne PM compared with the rest of the sources.

Did you find this useful? Give us your feedback

Figures (11)

Content maybe subject to copyright    Report

Transactions of the ASABE
Vol. 54(2): 629-642 E 2011 American Society of Agricultural and Biological Engineers ISSN 2151-0032 629
P
ARTICULATE MATTER EMITTED FROM POULTRY AND PIG HOUSES:
S
OURCE IDENTIFICATION AND QUANTIFICATION
M. Cambra‐López, T. Hermosilla, H. T. L. Lai, A. J. A. Aarnink, N. W. M. Ogink
ABSTRACT. There is need to identify and quantify the contribution of different sources to airborne particulate matter (PM)
emissions from animal houses. To this end, we compared the chemical and morphological characteristics of fine and coarse
PM from known sources collected from animal houses with the characteristics of on‐farm fine and coarse airborne PM using
two methods: classification rules based on decision trees and multiple linear regression. Fourteen different farms
corresponding to seven different housing systems for poultry and pigs were sampled during winter. A total of 28 fine and
28coarse on‐farm airborne PM samples were collected, together with a representative sample of each known source per farm
(56 known source samples in total). Source contributions were calculated as relative percentage contributions in particle
numbers and then estimated in particle mass. Based on particle numbers, results showed that in poultry houses, most on‐farm
airborne PM originated from feathers (ranging from 4% to 43% in fine PM and from 6% to 35% in coarse PM) and manure
(ranging from 9% to 85% in fine PM and from 30% to 94% in coarse PM). For pigs, most on‐farm airborne PM originated
from manure (ranging from 70% to 98% in fine PM and from 41% to 94% in coarse PM). Based on particle mass, for poultry
most on‐farm airborne PM still originated from feathers and manure; for pigs, however, most PM originated from skin and
manure. Feed had a negligible contribution to on‐farm airborne PM compared with other sources. Results presented in this
study improve the understanding of sources of PM in different animal housing systems, which may be valuable when choosing
optimal PM reduction techniques.
Keywords. Animal housing, Dust, Emissions, Source apportionment.
arge amounts of particulate matter (PM) are emitted
from animal houses, which can compromise animal
and human respiratory health (Radon et al., 2001;
Zuskin et al., 1995) and the environment as well.
The scientific community and stakeholders (farmers and lo‐
cal authorities) are seeking technically feasible and economi‐
cally viable solutions to reduce these emissions to comply
with air quality regulations. Preventing dust release from its
source not only reduces emissions from the animal house but
improves the indoor climate as well. To develop such reduc‐
tion techniques, it is necessary to identify and quantify the
sources that contribute to PM in animal houses.
A complete assessment can be achieved by quantifying
PM contributions from each source according to particle
numbers and mass. Knowledge of the relationship between
Submitted for review in July 2010 as manuscript number SE 8658;
approved for publication by the Structures & Environment Division of
ASABE in February 2011. Presented at the 2010 ASABE International
Symposium on Air Quality and Manure Management for Agriculture as
Paper No. 711P0510cd.
The authors are María Cambra‐López, ASABE Member Engineer,
Research Agricultural Engineer, Institute of Animal Science and
Technology, Universidad Politécnica de Valencia, Valencia, Spain;
Txomin Hermosilla, Researcher, Geo‐Environmental Cartography and
Remote Sensing Research Group, Universidad Politécnica de Valencia,
Valencia, Spain; and Huong T. L. Lai, Researcher, André J. A. Aarnink,
ASABE Member Engineer, Senior Researcher, and Nico W. M. Ogink,
Senior Researcher, Wageningen UR Livestock Research, Lelystad, The
Netherlands. Corresponding author: María Cambra‐López, Institute of
Animal Science and Technology, Universidad Politécnica de Valencia,
Camino de Vera s.n., 46022 Valencia, Spain; phone: +34‐96‐387‐98‐85;
fax: +34‐96‐387‐74‐39; e‐mail: macamlo@upvnet.upv.es.
particle number and mass contributions is essential because
it gives an insight into particle size and morphology related
to different particle types (sources). Moreover, particle size
and morphology are related to a particle's aerodynamic be‐
havior, which is closely related to lung deposition mecha‐
nisms in the human airways: inertial impaction, sedimen-
tation, interception, and diffusion (Zhang, 2004). Although
current European and U.S. regulations set limits to PM con‐
centrations based on mass, a mass‐only approach to reduce
PM would have very little effect on the number concentra‐
tions of smaller particles found in the fine fraction. This frac‐
tion contains fine and ultra‐fine particles that pose greater
risks of adverse health effects because these particles can go
beyond the larynx and penetrate into the unciliated respirato‐
ry system (CEN, 1993). The control of particles larger than
2.5 m m in diameter, however, is also relevant, because these
particles can also cause adverse health effects through depo‐
sition in the upper respiratory airways. Furthermore, particles
larger than 2 mm in diameter found in animal houses have
been shown to contain high amounts of odorants (Cai et al.,
2006) and micro‐organisms (Lee et al., 2006). Consequently,
both PM number and mass concentrations should be mea‐
sured to tackle PM pollution related issues within animal
houses, to develop reduction techniques, and to assess their
effects.
Analytical methods used to characterize PM, such as mi‐
croscopic analysis, can supply useful but limited data on par‐
ticle or source chemical composition and morphological
characteristics. To further identify and quantify source con‐
tributions, source apportionment models can be used. These
models are versatile because they can be used in different sce‐
narios (Watson et al., 2002).
L

630 TRANSACTIONS OF THE ASABE
Source apportionment models based on multivariate lin‐
ear regression permit quantitative source apportionment and
can be used to investigate the relationship between the chem‐
ical and physical properties of the source and the properties
measured at the site. Linear regression is used to estimate the
relative contribution of each known source as the linear sum
of products of source compositions and source contributions,
based on predetermined source profiles (Hopke, 1991). Fur‐
thermore, expert systems based on supervised methods can
be used to analyze data systematically. An expert system is
software that simulates the judgment and behavior of a hu‐
man with expert knowledge and experience in a particular
field (Jensen, 2005). These systems contain a knowledge
base with accumulated experience (data) and a set of rules for
applying the knowledge base to each particular situation that
is described to the program. Expert systems can be applied as
knowledge‐engineering tools in any field to interpret, pre‐
dict, diagnose, design, plan, monitor, and control systems
(Kim and Hopke, 1988). Moreover, expert systems can be
used to develop custom rules in the form of a decision tree
based on examples or training samples with known variables
and then classify data according to these rules. User‐defined
rules based on decision trees have been used to sort and clas‐
sify particles based on large datasets (Hopke, 2008; Hopke
and Song, 1997; Kim and Hopke, 1988; Wienke et al., 1995).
Based on known source profiles, rules can also sort and clas‐
sify measured airborne particles into predetermined and se‐
lected classes or sources.
Attempts to identify and quantify primary sources of PM
in animal houses have been made for pigs and poultry using
different approaches (Feddes et al., 1992; Heber et al., 1988;
Honey and McQuitty, 1979; Qi et al., 1992), but most of these
studies provide limited data from specific production sys‐
tems related to single animal categories (turkeys, growing‐
finishing pigs, and caged layers). Therefore, these studies are
valuable for identifying the most likely sources present in
specific animal production systems, but there is a lack of
comparable source contributions for other production sys‐
tems, between and within animal categories, for different
sized‐particles. To this end, specific methodologies need to
be developed that include statistical methods to calculate
source contributions, and measurement protocols to charac‐
terize the morphology and composition of PM in different
size fractions in animal houses.
Moreover, it is generally accepted that to apply source ap
portionment models in animal houses, it is necessary to ob‐
tain particle chemical characteristics. However, the presence
of similar chemical elements in most sources related to ani‐
mal PM can complicate discrimination among them
(Cambra‐López et al., 2011). Hence, the use of specific and
detailed source profiles based on additional particle charac‐
teristics is necessary. Cambra‐López et al. (2011) reported
that, in addition to chemical data, particle morphological
characteristics could add value in source apportionment in
animal houses because, in some cases, animal‐related PM
can be more heterogeneous in size and morphology than in
chemical composition. Therefore, chemical‐only or com‐
bined chemical and morphological particle characteristics
can be used to apportion single sources to on‐farm airborne
PM and improve the knowledge on the quantitative impor‐
tance of the different PM sources in terms of number and
mass contributions.
The objective of this study was to identify and quantify the
contribution of different sources to primary fine (PM
2.5
) and
coarse (PM
10‐2.5
) PM emissions from animal houses based on
chemical and morphological characteristics of particles. A
comprehensive list of animal categories was surveyed,
including seven different housing systems: broilers on
bedding, laying hens on floor, laying hens in aviary, turkeys
on bedding, piglets, growing‐finishing pigs, and dry and
pregnant sows on slatted floor. The relative contribution from
each source to PM was estimated in terms of number and
mass by comparing the chemical and morphological charac-
teristics of fine and coarse PM from each source with the
characteristics of fine and coarse airborne PM from the
animal houses. Two methods were used to estimate source
contributions: classification rules based on decision trees and
multiple linear regression. This study provides a better
understanding of sources of PM, which is essential to
improve reduction programs applicable to animal houses. It
also gives an insight into the environmental hazards of PM
and their potential health effects by providing knowledge on
sources of PM through properties such as particle mor-
phology and chemical composition.
MATERIAL AND METHODS
To identify and quantify the contribution of different
sources to fine and coarse PM emissions from different
housing systems for poultry and pigs, we sampled airborne
fine and coarse PM on‐farm and collected samples from
known potential PM sources.
HOUSING AND ANIMALS
Table 1 lists the surveyed animal species, type of housing
system, ventilation system, number of animals, and animal
age where airborne and source samples were collected. Two
different farm locations were sampled for each animal
housing system in The Netherlands during winter season. All
surveyed animal houses used automatically distributed
feeding systems with crumbs or pelleted feed. Broilers and
turkey houses used new wood shavings as bedding.
Table 1. Description of surveyed animal houses.
Housing System
Farm
Location
Vent
System
No. of
Animals
Age
(weeks)
Poultry
Broilers, bedding 1 Tunnel 50,400 4
2 Roof 2675 3
Laying hens, floor 1 Tunnel 3850 71
2 Tunnel 16,500 22
Laying hens, aviary 1 Tunnel 24,712 71
2 Tunnel 35,000 50
Turkeys, bedding 1 Ridge 5,000 12
2 Ridge 4,040 10
Pigs
Piglets, slatted floor 1 Roof 125 8
2 Roof 75 9
Growing‐finishing pigs,
partially slatted floor
1 Roof 120 16
2 Roof 60 20
Dry and pregnant sows,
group housing
1 Roof 39 Diverse
2 Roof 46 Diverse

631Vol. 54(2): 629-642
ON‐FARM AIRBORNE AND SOURCE SAMPLES
Virtual cascade impactors (RespiCon, Helmut Hund
GmbH, Wetzlar, Germany) were used on each farm to sample
airborne fine and coarse PM onto separate polycarbonate
filters (37 mm dia., 5 mm pore size). This device is a two‐stage
virtual impactor that follows the convention of the European
Standard (CEN, 1993) with a 50% cutoff at an aerodynamic
diameter of 2.5 mm (for fine PM) and 10 mm (for coarse PM).
According to Li et al. (2000), it exhibits differences less than
17% between measured efficiencies and the curves following
the European Standard (CEN, 1993). Virtual impactors are
similar to conventional impactors, but the impaction surface
is replaced with a virtual space of stagnant or slow‐moving
air, consequently reducing sampling problems common to
conventional impactors, such as overloading and particle
bounce losses. Portable pumps (Genie VSS5, Buck, Inc.,
Orlando, Fla.) were used to draw air through each virtual
cascade impactor at a constant flow of 3.11 L min
‐1
.
Sampling was conducted during morning (from 09:00 to
12:00) at each animal house. Duplicate airborne fine and
coarse PM samples were collected simultaneously near the
exhaust. For all the surveyed animal houses, a total of 28 fine
and 28 coarse on‐farm airborne PM samples were collected
indoors. Sampling time varied from 5 to 60 min, adjusted to
obtain particle loads of 5 to 20 mg particles cm
‐2
filter, to
minimize particle overlap (Willis et al., 2002). One
background (outside) sample was taken from 10 to 15 m
upwind of each farm in the same way as indoor samples.
Sampling time outside varied from 30 to 60 min.
Additionally, a light‐scattering system (DustTrak aerosol
monitor, model 8520, TSI, Inc., Shoreview, Minn.) was used
for on‐line continuous airborne PM
10
concentration
measurement inside and outside on each farm. Sampling time
was 30 to 60 min. One‐minute values were recorded and
stored. Temperature and relative humidity were also
recorded during each sampling, both inside and outside the
animal house, using temperature and relative humidity
sensors (iLog data logger, Escort Data Loggers, Inc.,
Buchanan, Va.).
For source samples, composite samples of potential PM
sources were collected per source and farm by randomly
sampling different locations in the animal house. Besides 14
background samples, a total of 42 known source samples
were sampled, including: concentrate feed (all farms),
manure (fresh excreta in poultry and fresh feces in pigs),
feathers (in poultry), and wood shavings used as bedding
material (present only for broilers and turkeys). We also
collected skin samples in pig houses, but only from sows
because it was impractical to collect such samples from
younger animals (piglets and growing‐finishing pigs) whose
skin was not as loose as a sow's dandruff. However, we used
the skin collected from the sows as a representative example
of “skin source” in the other pig categories. Approximately
200 to 500 g of representative samples of feed, manure, and
clean wood shavings were collected per farm. For feathers
and skin, 10 to 50 g samples were collected. All samples were
stored in clean poly-ethylene bags.
Each source sample per farm was dried for 12 h at 70°C
and then crushed in a ball mill for 1.5 min at 250 rpm. Dried
and milled samples were stored at room temperature, and
then airborne PM was generated in a laboratory dust
generator to collect airborne fine and coarse PM samples
from each source. The dust generator consisted of a stainless
steel cylinder of 20 cm diameter and 30 cm height with an
airtight lid, which had a mechanical agitation system with
rotary blades. A varying quantity, from 0.2 g (feathers) to
40g (feed), of each milled source per farm was introduced in
the dust generator and agitated at 200 rpm. The generated PM
was collected using a virtual cascade impactor (RespiCon,
Helmut Hund GmbH, Wetzlar, Germany) with polycarbonate
filters (37 mm dia., 5 mm pore size), which was placed inside
the generator. A portable pump (Genie VSS5, Buck, Inc.,
Orlando, Fla.) was used to draw air through the impactor from
the dust generator. A detailed description of the dust
generation process and setup can be found in Cambra‐López
et al. (2011).
Sampling time during dust generation varied from 1 min
to 7 h, depending on the amount of particles generated,
aiming at particle loads of 5 to 20 mg particles cm
‐2
filter
(Willis et al., 2002). This generation procedure simulated the
process by which PM can be generated in animal houses.
According to Gill et al. (2006), generating, collecting, and
measuring PM in a controlled laboratory setting is a useful
tool for determining the emission potential per mass of
source, as well as the physical, morphological, and chemical
characteristics of the emission. The laboratory dust
generation procedure used in our study worked by generating
a large cloud of particles and then collecting a small amount
of them. The filter samples generated in the laboratory
(46fine and 46 coarse PM samples) were stored at room
temperature (20°C to 25°C) for several months before
analysis, in sealed filter cassettes (Omega Specialty
Instrument Co., Houston, Tex.). The filter cassettes had a
static‐dissipative nature and were chosen to protect filters,
minimize sample losses, and avoid contamination during
storage. Freezing of filters was avoided to prevent physical
changes, such as particle fragmentation, during freezing.
MORPHO‐CHEMICAL ANALYSIS OF AIRBORNE AND SOURCE
SAMPLES
High‐resolution scanning electron microscopy (SEM)
(JSM‐5410, JEOL Ltd., Tokyo, Japan) combined with
energy‐dispersive x‐ray analysis (EDX) (Link Tetra analyzer,
Oxford Instruments, Abingdon, U.K.) was used to obtain
particle‐by‐particle chemical and morphological data. A
small section (approximately 1 cm
2
) of the as‐collected
polycarbonate filter from fine and coarse fractions was cut
and mounted on a 12 mm carbon stub with double‐sided
carbon adhesive tape. Samples were then coated with carbon
using a vacuum evaporator to provide electrical conductivity
and create a conductive coating for exposure to the SEM
electron beam. Detection of elements with atomic number >6
(carbon) was obtained from element x‐ray spectra.
The SEM‐EDX was conducted manually and operated
under the same conditions throughout the study: 10 keV
accelerating voltage, 15 mm working distance, 3 nA electron
probe current, 1000× magnification for coarse PM and
1800× for fine PM, and x‐ray acquisition time of 60 s per
particle. Secondary electron mode was used for particle
location, measurement, analysis, and image acquisition.
At least three fields of view (spots) per filter sample were
analyzed. On each analyzed field, both an image
(photomicrograph at 1000× or 1800×, saved in tiff format
1024 × 768 resolution) and the single‐particle x‐ray spectra
of every particle found in that field were obtained and stored.
Within each field, the minimum projected area diameter for

632 TRANSACTIONS OF THE ASABE
the coarse particles was set at 1 mm. The minimum projected
area diameter for the fine particles was set at 0.1 mm (Conner
et al., 2001). Since the particles were not flat but included
complex sizes and shapes, the SEM electron beam and beam
energy could be affected by particle morphology. Therefore,
these size limits were set to minimize the amount of data
acquired for non‐particle features (e.g., filter substrate) at the
magnifications used. For each airborne sample, a total of 50
to 75 particles were chemically analyzed in each duplicate
sample. For each source sample, a total of 25 to 50 particles
were chemically analyzed. All x‐ray spectra were processed
with INCA software (Oxford Instruments, Abingdon, U.K.),
confirmed manually to correct for element omission or
confusion, and checked to eliminate the contribution of the
filter material (carbon and oxygen). In fact, in this study, hair
source was not included in the analysis because it showed
very high carbon and oxygen peaks in the SEM‐EDX, which
was confused with the background filter composition.
The stored images (SEM photomicrographs of each field
of view) were analyzed using the object‐based image
analysis (OBIA) approach (Blaschke, 2010) using FETEX
2.0 software (Ruiz et al., 2011). All images were
radiometrically corrected by background values to avoid
spectral differences due to acquisition conditions and to
equalize the background value to compare intensity values
between images. Individual particles were defined by means
of segmentation using thresholding. The OBIA software
extracted spectral, textural, and shape‐based features for
each detected particle (object).
Therefore, based on chemistry, each particle was
characterized by 25 elements: nitrogen (N), sodium (Na),
magnesium (Mg), aluminum (Al), silicon (Si), phosphorus
(P), sulfur (S), chlorine (Cl), potassium (K), calcium (Ca),
iron (Fe), nickel (Ni), copper (Cu), zinc (Zn), silver (Ag),
lead (Pb), tin (Sn), chromium (Cr), cobalt (Co), barium (Ba),
bromide (Br), titanium (Ti), vanadium (V), antimony (Sb),
and gold (Au). Based on morphological characteristics, each
particle was characterized by 23 variables. In total, each
particle was exhaustively characterized by 48 variables.
S
OURCE APPORTIONMENT METHODS
Fine and coarse source samples as well as on‐farm
airborne fine and coarse PM samples from each animal house
were used in source apportionment using classification rules
based on decision trees and multiple linear regression.
Single‐particle chemical and morphological characteristics
obtained using SEM‐EDX were used as data sources.
Apportionment results were calculated as relative percentage
contributions in number and then estimated in terms of mass.
Results provided by the two methods are compared and
discussed.
Classification Rules Based on Decision Trees
Decision trees were used to develop a set of rules for each
group of sources from each animal house. Single‐particle
chemical and morphological characteristics from known
sources obtained using SEM‐EDX were joined in a combined
database and used in this process. Decision trees were built
using See 5 software, using the C5.0 classification algorithm,
which is the latest version of the algorithms ID3 and C4.5
developed by Quinlan (1993). Decision trees were created
following the boosting multiclassifier method (Freund,
1995). This method searched the features that best separated
one source from the other by dividing data using mutually
exclusive conditions until the newly generated subgroups
were homogeneous, i.e., all the elements in a subgroup
belonged to the same class or a stopping condition was
fulfilled. The rules developed using the known sources were
then applied to classify the airborne on‐farm samples into one
of the known sources based on their chemical and
morphological characteristics.
Accuracy of this method was predicted through leave‐
one‐out cross‐validation using a single observation from the
source samples as validation data and the remaining
observations as training data. The cross‐validation statistical
method worked by applying the rules to the source samples
and comparing the source assigned to each particle using the
rules with its reference source per farm. Overall measure of
prediction accuracy for number of particles was obtained by
dividing the total correct validations in each source by the
total number of classified particles.
Multiple Linear Regression
Multiple linear regression was also used to apportion
airborne PM sampled on the farms to the known sources.
Bulk source chemical characteristics from known sources
obtained from the average of single‐particle chemical
characteristics using SEM‐EDX, were used in this process.
The average PM concentration of elements in fine and coarse
airborne on‐farm samples were used as dependent variables,
and the average fine and coarse PM concentrations of
elements in each source were used as independent variables.
All elements were included at once in the model using
Genstat (Genstat Committee, 2008) following equation 1:
()
=
×=
n
k
ikmikmim
FfY
1
(1)
where
Y
im
= relative concentration of the ith element in
collected airborne fine or coarse PM on the mth
farm (average of duplicate samples)
f
ikm
= number contribution of the ith element of the kth
source to airborne fine or coarse PM on the mth
farm (the sum of the fractions was set to 1)
F
ikm
= average relative concentration of the ith element in
the kth source on the mth farm.
Mass Estimation
Results from the classification rules based on decision
trees and multiple linear regression were given in particle
numbers. Particle number contributions were transformed
into mass contributions based on the average mass of
particles in each source. The mass for each single particle (m)
was calculated from the projected area diameter (D
p
)
provided by the SEM images, based on a density value and
shape factor, following the equation for the mass of a particle
(eq. 2) (Ott et al., 2008). From single‐particle masses, the
average particle mass per source was calculated:
63
4
3
3
π×ρ
=
π×ρ=×ρ=
v
p
p
ppp
S
D
rvm
(2)

633Vol. 54(2): 629-642
where
m = particle mass
ρ
p
= particle density
v
p
= particle volume
r = equivalent radius of a spherical particle
D
p
= projected area diameter,
π
×=
Area
D
p
2
S
v
= volume shape factor, a correction factor to convert D
p
to equivalent volume diameter, defined as the
diameter of a sphere having the same volume as the
irregular particle.
We assumed average values for density of 1.2 g cm
‐3
(feathers), 2.6 g cm
‐3
(feed), 1.3 g cm
‐3
(hair), 1.5 g cm
‐3
(manure and wood shavings), 1.4 g cm
‐3
(skin), and 2.1 g
cm
‐3
(outside) (McCrone, 1992). Shape factors used in the
mass calculation were obtained from Zhang (2004),
assigning values of 1.06 (feathers and wood shavings), 1.08
(feed and outside), 1.15 (poultry manure), 1.36 (pig manure),
and 1.88 (skin).
RESULTS
ON FARM AIRBORNE PM MEASUREMENTS
Environmental conditions during sampling are shown in
table 2, including average PM
10
concentrations measured
using a light‐scattering system, relative humidity, and
temperature measured inside and outside the animal houses.
Values in the table represent sampling time averages over 5
to 60 min and standard errors between the two surveyed
houses for the same animal category.
SOURCE IDENTIFICATION
Sources were identified through individual particle
morphologies based on SEM observations. Different types of
particles collected from different animal housing systems
were identified by comparison to known standards
(McCrone, 1992; Cambra‐López et al., 2011). Figure 1 shows
examples of particle types from different animal housing
systems. In broiler houses, a mixture of bent, soft, and loose
particles probably from feathers, and flattened agglomerates
is shown in figure 1a. Bent, sharp‐edged particles from wood
shavings and spherical particles from excreta (sometimes
agglomerated) could also be identified (fig. 1a). For laying
hens, spherical particles from excreta were dominant in
collected PM in floor housing systems (fig. 1b) and also in
aviary systems (fig. 1c). In turkey houses, bent, sharp‐edged
particles from wood shavings or feathers, and few spherical
particles from excreta were identified (fig. 1d). In piglet
houses, deposited round gray, smooth particles were
identified, together with some brighter layered manure
particles (fig. 1e). A mixture of layered manure particles and
large, flattened skin particles was identified in collected PM
from piglet houses (fig. 1f) and from growing‐finishing pigs
(fig. 1g). Large, folded skin particles were identified in
collected PM from dry and pregnant sow houses (fig. 1h).
CONTRIBUTION OF SOURCES TO ON‐FARM AIRBORNE PM
EXPRESSED IN NUMBER
Source apportionment using classification rules based on
decision trees and multiple linear regression resulted in
relative percentage contributions of sources to on‐farm
airborne PM expressed in particle numbers. A total of
912individual particles were apportioned in fine and 1071 in
coarse PM using classification rules based on decision trees.
A total of 1546 individual particles were apportioned in fine
and 1670 in coarse PM using multiple linear regression.
Using Classification Rules Based on Decision Trees
Results using classification rules based on decision trees
are shown in table 3 (fine PM) and table 4 (coarse PM),
together with method accuracies. Results indicated that for
poultry, most of the PM originated from feathers and manure.
The relative percentage contribution of manure was
generally higher in coarse PM (ranging from 30% to 87%)
compared with fine PM (ranging from 9% to 85%). Even
though the number of sources was not equal among poultry
categories due to the presence of wood shavings for broilers
and turkeys and not for laying hens, the relative percentage
contributions of manure were generally higher for laying
hens compared with broilers and turkeys; whereas feather
contribution was higher for broilers and turkeys compared
with laying hens. Where present, wood shavings contributed
less than 20% of particle numbers. For pigs, most of the PM
originated from manure. The relative percentage contribu-
tion of manure was higher in fine PM (ranging from 70% to
89%) compared with coarse PM (ranging from 41% to 84%)
for all pig categories. Skin and feed were the other most
important contributing sources for pigs. The relative
percentage contribution of skin varied from 2% to 33%,
varying between pig categories, being generally higher in
coarse PM compared with fine PM. The relative percentage
contribution of feed was found at or below 16% for all animal
categories. It was slightly higher for pigs compared with
poultry, being the highest for piglets, in both fine and coarse
PM. Outside particles had a relevant contribution for broilers
and turkeys, especially in fine PM.
Table 2. Summary of average PM
10
measurements, temperature (T) and relative humidity (RH) inside and outside the surveyed
animal houses. Standard error (SE) represents variation between both surveyed animal houses for the same animal category.
[a]
Animal Category
PM
10
Inside
(mg m
‐3
)
PM
10
Outside
(mg m
‐3
)
T Inside
(°C)
RH Inside
(%)
T Outside
(°C)
RH Outside
(%)
Avg. SE Avg. SE Avg. SE Avg. SE Avg. SE Avg. SE
Broilers 1.96 0.55 0.08 0.05 23.2 ND 81.6 ND 13.2 ND 50.6 ND
Laying hens, floor 3.94 0.69 0.03 0.00 16.2 1.7 74.8 0.5 10.3 0.4 74.4 18.8
Laying hens, aviary 3.06 1.54 0.03 0.00 15.6 3.2 70.4 3.2 10.3 0.4 74.4 18.8
Turkeys 2.32 0.99 0.08 0.05 19.4 2.5 63.3 7.0 11.3 0.1 54.3 0.2
Piglets 1.44 0.11 0.03 0.01 25.2 0.1 75.8 0.3 11.4 1.8 55.0 20.5
Growing‐finishing pigs 1.27 0.35 0.03 0.01 21.9 0.8 62.3 9.0 11.4 1.8 55.0 20.5
Dry and pregnant sows 0.39 0.01 0.03 0.01 23.9 ND 75.6 ND 13.3 ND 34.5 ND
[a]
ND = no data due to equipment failure in one of the farms.

Citations
More filters
Journal ArticleDOI
TL;DR: In this paper, the authors reviewed the current literature on the relationship between air pollution and food security from the perspective of food system and highlighted that agricultural emissions which substantially contribute to air pollution could happen at every stage along the food supply chain.

66 citations

Journal ArticleDOI
TL;DR: A better understanding is provided of the aerosolization, prevalence, and persistence of S. suis in the air of swine confinement buildings (SCB) and in healthy SCB without reported infections.
Abstract: Streptococcus suis is an important swine pathogen that can cause septicemia, meningitis, and pneumonia. Also recognized as an emerging zoonotic agent, it is responsible for outbreaks of human infections in Asian countries. Serotype 2 is the predominant isolate from diseased animals and humans. The aerosolization of S. suis in the air of swine confinement buildings (SCB) was studied. The presence of S. suis in bioaerosols was monitored in SCB where cases of infection had been reported and in healthy SCB without reported infections. Using a quantitative-PCR (qPCR) method, we determined the total number of bacteria (1 × 10(8) to 2 × 10(8) airborne/m(3)), total number of S. suis bacteria (4 × 10(5) to 10 × 10(5) airborne/m(3)), and number of S. suis serotype 2 and 1/2 bacteria (1 × 10(3) to 30 × 10(3) airborne/m(3)) present in the air. S. suis serotypes 2 and 1/2 were detected in the air of all growing/finishing SCB that had documented cases of S. suis infection and in 50% of healthy SCB. The total number of bacteria and total numbers of S. suis and S. suis serotype 2 and 1/2 bacteria were monitored in one positive SCB during a 5-week period, and it was shown that the aerosolized S. suis serotypes 2 and 1/2 remain airborne for a prolonged period. When the effect of aerosolization on S. suis was observed, the percentage of intact S. suis bacteria (showing cell membrane integrity) in the air might have been up to 13%. Finally S. suis was found in nasal swabs from 14 out of 21 healthy finishing-SCB workers, suggesting significant exposure to the pathogen. This report provides a better understanding of the aerosolization, prevalence, and persistence of S. suis in SCB.

51 citations


Cites background from "Particulate matter emitted from pou..."

  • ...have shown that most airborne particles in pig houses originate from manure and skin (8)....

    [...]

Journal ArticleDOI
TL;DR: PM2.5 from pig house is a significant risk for immune response in alveolar macrophages and the role of endotoxin and ROS played important roles in PM2.

49 citations

Journal ArticleDOI
TL;DR: In this article, the authors investigate indoor concentrations of particulate matter (PM1, PM2.5 and PM10), NH3, N2O, CH4 and CO2 in different conventional pig fattening facilities and one low-emission fattens facility.

36 citations

Journal ArticleDOI
TL;DR: In this paper, the concentration of ammonia and particulate matter (PM) that passes through a size-selective inlet with a 50% cut-off at 10-μm aerodynamic-equivalent diameter (PM10) and emissions into atmosphere in the following three types of laying-hen houses: traditional battery cages with aerated open-manure storage (BSP) and two best available technique (BAT) housing types, namely, an aviary-system housing (ASH) and a vertical tiered cage with manure belts and forced-air drying (V
Abstract: The aim of the present study was to measure the concentration of ammonia and particulate matter (PM) that passes through a size‐selective inlet with a 50% cut‐off at 10-μm aerodynamic-equivalent diameter (PM10) and emissions into atmosphere in the following three types of laying-hen houses: traditional battery cages with aerated open-manure storage (BSP) and two best available technique (BAT) housing types, namely, an aviary-system housing (ASH) and a vertical tiered cage with manure belts and forced-air drying (VTC). Measurements were taken continuously for a period of 1 year in each house. Ammonia concentration was measured continuously in each house using an infrared photoacoustic detector with a 15-min sampling interval. PM10 was measured continuously using a scatter light photometer, corrected by the traditional gravimetric-technique concentration to lower the measurement error. The same instrument was also used to collect PM10 through a traditional gravimetric technique. This procedure was performed to adjust the particulate matter-specific gravity of PM that is typical and specific for every animal house. PM10 and ammonia measurements were carried out together with measurements of inside and outside temperature, inside and outside relative humidity and ventilation rate. For the high PM10 concentrations measured in the ASH house during a preliminary survey, concentrations of total suspended particles (TSP) and fine PM (particles <2.5 microns) were also measured to evaluate the dustiness in the building during the working hours. The ammonia concentration was 5.37 mg/m3 in the traditional BSP house (the reference for cage-housing system), 4.95 mg/m3 in the VTC and 3.85 mg/m3 in the ASH. The ammonia-emission factors were 15.445 mg/h.hen place (0.135 kg/year.hen place) for BSP, 8.258 mg/h.hen place (0.072 kg/year.hen place) for VTC, and 23.704 mg/h.hen place (0.208 kg/year.hen place) for ASH. Ammonia emission-reduction efficiency of VTC v. the BSP was 53%, according to thresholds assessed by Integrated Prevention Pollution Control. The ammonia-reduction efficiency of ASH v. that of the standard Reference Housing system for non-cage housing was 68%. Average yearly PM10 concentration was remarkably higher in the ASH, with 0.215 mg/m3 v. 0.108 mg/m3 for the VTC and 0.094 mg/m3 for BSP. In the ASH, the concentration of total suspended particles (TSP) was 0.444 mg/m3 and that of PM2.5 was 0.032 mg/m3. In this facility, a great variation of PM10 concentration occurred in the morning hours. Recorded values for the PM10 emission were 0.433 mg/h.hen for BSP and 0.081 mg/h.hen for VTC, while the ASH showed the highest PM10 emission (1.230 mg/h.hen), with clear peaks occurring in the morning hours during daily farming operations.

31 citations

References
More filters
Journal ArticleDOI
TL;DR: Receptor models infer contributions from particulate matter (PM) source types using multivariate measurements of particle chemical and physical properties, and complement source models that estimate concentrations from emissions inventories and transport meteorology.

286 citations


"Particulate matter emitted from pou..." refers background in this paper

  • ...These models are versatile because they can be used in different sce‐ narios (Watson et al., 2002)....

    [...]

Book ChapterDOI
30 Oct 1997
TL;DR: In this paper, the authors present an introduction to the Receptor Modeling and its application to solving local air quality problems in the context of ambient PM-10 air quality standard for Particulate Matter.
Abstract: 1. An Introduction to Receptor Modeling (P.K. Hopke). 2. Sampling and Analysis Methods for Ambient PM-10 Aerosol (T.G. Dzubay and R.K. Stevens). 3. Source Sampling for Receptor Modeling (J.E. Houck). 4. Chemical Mass Balance (J.G. Watson, J.C. Chow and T. G. Pace). 5. Multivariate Receptor Models (R.C. Henry). 6. Scanning Electron Microscopy (P.K. Hopke and G.S. Casuccio). 7. Receptor Modeling for Volatile Organic Compounds (P.A. Scheff and R.A. Wadden). 8. Receptor Modeling in the Context of Ambient Air Quality Standard for Particulate Matter (T.G. Pace). 9. Application of Receptor Modeling to Solving Local Air Quality Problems (J.E. Core). Index.

263 citations


"Particulate matter emitted from pou..." refers methods in this paper

  • ...Linear regression is used to estimate the relative contribution of each known source as the linear sum of products of source compositions and source contributions, based on predetermined source profiles (Hopke, 1991)....

    [...]

Journal ArticleDOI
TL;DR: As part of a project to evaluate health hazards for workers in swine confinement buildings, the air in 21 different buildings was sampled with 37 mm cassette filters with and without cyclone preselectors and with cascade impactors, finding the respirable fraction was less in finishing buildings than in farrowing and nursery buildings.
Abstract: As part of a project to evaluate health hazards for workers in swine confinement buildings, the air in 21 different buildings was sampled with 37 mm cassette filters with and without cyclone preselectors and with cascade impactors. Filter results yielded a mean total aerosol of 6.3 mg/m3, a mean respirable aerosol of 0.5 mg/m3; the geometric mean diameter was 2.9 microns. Cascade impactor measurements revealed a mean total aerosol of 7.6 mg/ m3, a respirable aerosol of 2.5 mg/ m3 and a mass median diameter of 9.6 microns. The two major constituents in these aerosols were grain particles and dried fecal matter. The grain particles were larger than fecal particles and proportionately more abundant in finishing buildings where 50 kg-100 kg animals are housed. Therefore the respirable fraction was less in finishing buildings than in farrowing and nursery buildings. Culturing of settled dusts yielded six different mold species, with the highest counts for Verticillium sp. (5 × 102 cfu/mg dry dust) grown at 37°...

182 citations

Journal ArticleDOI
TL;DR: The combined exposure to airborne dust and microorganisms was found to be more severe during harvesting than in animal confinements, and the highest contribution of large particles (3–10 μ m) in total particles was found during grain harvesting, whereas the size distribution was dominated by smaller particles (< 3 μ m), which was partially attributed to the increase in the concentration of the fungal spores.
Abstract: Airborne dust and microorganisms are associated with respiratory diseases and increased mortality and morbidity. Farmers are at high risk of exposure to both of these hazards. Very limited information, however, is available on the combined exposures to both hazards on different types of farms. Moreover, most of the previous studies have measured the mass concentration of particles ignoring the particle size. In this study, farmers' exposure to airborne dust and microorganisms was studied using our newly developed personal sampling system. Particle number concentration and size distribution were measured with an optical particle counter. Simultaneously, particles were collected on a filter and analyzed for microorganisms. The field measurements were conducted in animal confinements (swine, poultry, and dairy) and during grain harvesting (corn and soybean). The results show the following average concentrations on the workers' breathing zone: 1.7 × 10 6 to 2.9 × 10 7 particles/m 3 for total dust, 0.9 × 10 3 ...

172 citations


"Particulate matter emitted from pou..." refers background in this paper

  • ...Furthermore, particles larger than 2 m in diameter found in animal houses have been shown to contain high amounts of odorants (Cai et al., 2006) and micro‐organisms (Lee et al., 2006)....

    [...]

Book
01 Jan 2005
TL;DR: Indoor Air Quality Engineering as mentioned in this paper covers a wide range of indoor air quality engineering principles and applications, providing guidelines for identifying and analyzing indoor Air quality problems as well as designing a system to mitigate these problems.
Abstract: Indoor Air Quality Engineering covers a wide range of indoor air quality engineering principles and applications, providing guidelines for identifying and analyzing indoor air quality problems as well as designing a system to mitigate these problems. Structured into three sections - properties and behavior of airborne pollutants, measurement and sampling efficiency, and air quality enhancement technologies - this book uses real-life examples, design problems, and solutions to illustrate engineering principles. Professionals and students in engineering, environmental sciences, public health, and industrial hygiene concerned with indoor air quality control will find Indoor Air Quality Engineering provides effective methods, technologies, and principles not traditionally covered in other texts.

154 citations


"Particulate matter emitted from pou..." refers background or methods in this paper

  • ...Moreover, particle size and morphology are related to a particle's aerodynamic be‐ havior, which is closely related to lung deposition mecha‐ nisms in the human airways: inertial impaction, sedimentation, interception, and diffusion (Zhang, 2004)....

    [...]

  • ...Shape factors used in the mass calculation were obtained from Zhang (2004), assigning values of 1.06 (feathers and wood shavings), 1.08 (feed and outside), 1.15 (poultry manure), 1.36 (pig manure), and 1.88 (skin)....

    [...]

Related Papers (5)
Frequently Asked Questions (1)
Q1. What are the contributions in this paper?

Results presented in this study improve the understanding of sources of PM in different animal housing systems, which may be valuable when choosing optimal PM reduction techniques.