Results of measurements of particulate matter concentrations inside a pig fattening facility

(1) Institute for Agricultural and Fisheries Research (ILVO). Animal Sciences Unit Pig Husbandry. Scheldeweg, 68. BE-9090 Melle (Belgium). (2) Institute for Agricultural and Fisheries Research (ILVO). Technology and Food Science Unit Agricultural Engineering. Burgemeester Van Gansberghelaan, 115-1. BE-Merelbeke (Belgium). E-mail: Peter.Demeyer@ilvo.vlaanderen.be (3) Ghent University. Faculty of Bioscience Engineering. Department of Applied Analytical and Physical Chemistry. Coupure links, 653. BE-9000 Ghent (Belgium). (4) Ghent University. Faculty of Bioscience Engineering. Department of Sustainable Organic Chemistry and Technology. Coupure links, 653. BE-9000 Ghent (Belgium).


INTRODUCTION
Inside livestock barns, a wide variety of aerial pollutants, including particulate matter (PM), can affect indoor air quality (NRC, 2003).Unlike the chemically well-defined gaseous pollutants, PM is a mixture of many types of particles that differ in size, shape, chemical composition, and density (Pedersen et al., 2000).To cope with the heterogeneous nature of PM and the associated highly irregular shape and variety in density of the particles, the behavior of the different particles is commonly described by the aerodynamic diameter (AED).The AED of an irregularly shaped particle is defined as the diameter of a sphere with a standard density (1 000 kg .m -³) that would have the same settling velocity in air as the particle (Zhang, 2004).A wide range of particles with different AED can be found inside livestock houses (Harry, 1978).Particle size distribution (PSD) analysis may help to describe this heterogeneity in AED and is perhaps the most important physical parameter determining particle behavior.
Therefore, the aims of the current study were to investigate the correlations between the different size fractions of indoor PM in a commercial fattening pig barn and to perform PSD analysis in order to get an overview of the dominant size ranges in fattening pig facilities.

Experimental design
The measurements were performed at one commercial fattening pig facility in Diksmuide, Belgium during two fattening periods.However, measurements were performed in two types of housing systems (conventional and low-ammonia-emission) and two cleaning protocols (dry and wet) were performed in order to increase the variation in PM concentrations.A detailed description of the housing systems and the cleaning protocols can be found in Ulens et al. (2014).
Both the Grimm spectrometers and the GPCs make use of a laser diode light source.Scattered light is collected and focused onto a photo diode that converts the bursts of light into electrical pulses.The amplitude of the pulses is used as the measure of the particle size.
More information about the measuring setup and characteristics of the different instruments can be found in Ulens et al. (2014).

Data analysis
Correlations.Comparison of the data from the different housing systems and cleaning protocols revealed no significant differences in PM concentrations (Ulens et al., 2014).Therefore, all correlations were calculated based on the full dataset.The full dataset contained approximately 18 000 hourly means of PM concentrations.The Grimm spectrometer and GPC datasets contained both approximately 9 000 hourly mean PM concentrations.
Correlations between the different PM fractions were calculated using SPSS Statistics 21.0 (SPSS Inc., Chicago, IL, USA) for the entire dataset and for the two subsets separately.Using the Kolmogorov-Smirnov test and based on visual inspection of QQ-plots, it was shown that the data were not normally distributed (p < 0.05).Therefore, Spearman's rank correlation coefficients were calculated.All statistical tests were performed at 0.05 significance level.
To represent the PSD, the count median diameter (CMD) and the mass median diameter (MMD), together with their respective geometric standard deviation (GSD) were calculated, using equations adapted from Zhang (2004).Both diameters were calculated based upon the number of particles for the 30 size ranges of the Grimm spectrometers.For purposes of calculation we assumed that all particles were spherical and had the same density.
The CMD (in µm) is defined as the geometric mean diameter of the number-weighted PSD.For a lognormal distribution, the geometric mean equals the median.The CMD was calculated using Equation 1: where F i : number of particles per m³ in size range i, ∑F i : total number of particles per m³, d i : mean diameter of the lower and upper limit of size range i, in µm.
The MMD (in µm) is defined as the diameter for which half the total mass of particles is larger and half is smaller than this size.The MMD was calculated using Equation 2: The CGSD and MGSD are dimensionless quantities with a value greater than 1.0 and are a measure for the width of the number-weighted or mass-weighted aerodynamic particle size distribution.The CGSD for the number-weighted PSD was calculated using Equation 3 and the MGSD was calculated using Equation 4: CMD, MMD, and their respective GSD were calculated on hourly data from the two consecutive fattening periods in the four compartments.These calculations were automated in R3.0.1 (R Core Team, 2013).

Correlations between different particulate matter size fractions
Very high correlations (R > 0.95) were found between PM 10 and PM 2.5 when analyzing data from both the Grimm spectrometers and GPCs (Table 1).The observed high correlations between PM 10 and PM 2.5 indoor concentrations were also found by Van Ransbeeck et al. (2013) inside livestock buildings and by Marcazzan et al. (2001) in ambient air.This can partially be explained by the fact that PM 2.5 is a substantial part of PM 10 .Nevertheless, in the current study the mean ratio PM 10 :PM 2.5 was about 10:1, while Marcazzan et al. (2001) found a ratio of 3:2 in ambient air.Most of the PM inside livestock buildings is primary in origin and can mainly be found in the coarse (PM 10 -PM 2.5 ) fraction.This is especially the case for PM originating from feed, animal hair, and skin as well as manure (Cambra-López et al., 2011).Particles in the fine (PM 2.5 ) fraction are mostly formed through chemical reactions between gases and particles.These secondary processes occur to a lesser extent inside livestock buildings and part of the mechanically generated particles can fall into the PM 2.5 size range (Cambra-López et al., 2010).
Lower correlations (R < 0.4) were found between PM 10 and PM 1 and between PM 2.5 and PM 1 when analyzing the data from both measuring instruments together and from the GPCs (Table 1).However, higher correlations (R > 0.5) between PM 10 and PM 1 and between PM 2.5 and PM 1 were found when analyzing the data from the Grimm spectrometers (Table 1).Using the same Grimm spectrometers, Van Ransbeeck et al. (2013) found a high correlation between PM 2.5 and PM 1 indoor concentrations (R = 0.77) and a lower correlation between PM 10 and PM 1 indoor concentrations (R = 0.46).However, when using the GPCs, correlations were much lower in our study.This indicates that the observed correlations with PM 1 are dependent upon the measuring instrument used.However, both instruments claim a counting efficiency of 50% at AED of 0.3 µm and of 100% for all particles larger than 0.45 µm (manufacturer's website; Schmoll et al., 2010).The relative humidity inside the stable can also play an important role.At high relative humidity, water molecules risk of being recognized as particles by the optical instrumentation which can falsify the measurements.To overcome this problem, the Grimm spectrometers are equipped with an air mixing device which can add particle-free dry air to the sample airflow.This system is activated when the relative humidity exceeds 85% (manual Grimm spectrometer).The GPCs however are not equipped with such a device and therefore do not correct for high relative humidity.

Particle size distribution
Figure 1 shows a typical example of the differential number-weighted and differential mass-weighted particle size distribution in the pig barn.It can be seen that the smallest particle sizes (< 1 µm) are highest in number, whereas the highest mass-weighted fractions occur near 10 µm.Few differences were found between the mean CMD values for the different compartments and fattening periods with values ranging from 0.43 to 0.49 µm.These values are similar to the value (0.40 µm) found by Lai et al. (2012) who analyzed the PSD in different pig buildings using identical Grimm spectrometers.The mean MMD values, ranging from 10.73 to 12.18 µm, found in the current study correspond well with values found by Maghirang et al. (1997) in a pig nursery (ranging from 10 to 19 µm) using a cascade impactor.The GSD values found in the current study for the number-and mass-weighted distribution were all larger than 1.22, indicating that the aerosols in all compartments were polydisperse.Despite the different housing systems and cleaning protocols observed in this study, very similar PSDs were found.However, as reported previously (Ulens et al., 2014), indoor mass concentrations of PM 10 , PM 2.5 , and PM 1 changed throughout the fattening periods.Furthermore, the lack of a clear pattern over a day (data not shown) or over a fattening period (data not shown) is in contrast with the observed diurnal pattern and day to day pattern during a fattening period found for PM 10 , PM 2.5 , and PM 1 concentrations (Van Ransbeeck et al., 2012).This indicates that, although the total mass of particles (PM concentrations) changed significantly (during a day and during a fattening period) inside the barn, the CMD and MMD values remained about the same.

CONCLUSIONS
The results from the present study showed high correlations between the indoor concentrations of PM 10 and PM 2.5 .No differences in PSD could be found between different housing systems or cleaning protocols.) 0 1 000 2 000 3 000 4 000 5 000 6 000

Table 1 .
Spearman's rank correlation coefficients between the different particule matters (PM) fractions based upon PM data from the Grimm spectrometers (GS), the Graywolf particle counters (GPC) and data from both the GS and GPC -Coefficients de corrélations de Spearman entre différentes classes de taille de particules (PM) déterminées à partir de spectromètres de Grimm (GS), de compteurs de particules Graywolf (GPC) et des données cumulées des GS et GPC.