OrganiccharacterizationofPM2.5 in the Emilia-Romagna region(I) · OrganiccharacterizationofPM2.5 in...
Transcript of OrganiccharacterizationofPM2.5 in the Emilia-Romagna region(I) · OrganiccharacterizationofPM2.5 in...
Organic characterization of PM2.5 in the
Emilia-Romagna region (I) S. Ferrari1, F. Scotto1, D. Bacco1, I. Ricciardelli1, A. Trentini1 , M.C. Pietrogrande2, M. Visentin2, P. Ugolini1, T. D'Alessandro1 and
V. Poluzzi1
1Emilia-Romagna Regional Agency for Prevention and Environment-Bologna, Emilia-Romagna, I-40138, Italy1Emilia-Romagna Regional Agency for Prevention and Environment-Bologna, Emilia-Romagna, I-40138, Italy2 Department of Chemical and Pharmaceutical Sciences, University of Ferrara, I-44121, Italy
Presenting author e-mail: [email protected]
INTRODUCTION
In the Project “Supersito” (www.superisto-er.it) eight observation campaigns were carried out in two sites (Fig. 1)
of Emilia-Romagna region (Italy): Bologna-Urban background (MS), residential centre in the Po Valley near
Apennines, and San Pietro Capofiume-Rural site (SPC), 30 Km north-east far from Bologna.Apennines, and San Pietro Capofiume-Rural site (SPC), 30 Km north-east far from Bologna.
The campaigns were performed in different periods of the year in order to obtain an overview of the diverse
weather conditions, multiple emission sources and/or chemical transformation.
The sampling periods are shown in the table 1.
The determined daily analytics were alkanes, polycyclic aromatic hydrocarbons, the 2-nitro+3-nitrofluorantene,
carboxylic acids and sugars.
The aim of this study was increase the information about the chemical organic characterization of the PM2.5 and
evaluate the relationship meteorological conditions on the chemical compounds.
It is currently being evaluated the performance of the Positive Matrix Factorization. It was been achievable an
Table 1 - Campaigns and sampling periods of intensive observations.
N. Campaign Date
1a
- fall campaign 14/11/2011 - 06/12/2011
2a
- spring/summ. campaign 13/06/2012 - 11/07/2012
3a
- fall campaign 23/10/2012 - 11/11/201It is currently being evaluated the performance of the Positive Matrix Factorization. It was been achievable an
exploratory investigation by means of Principal Component Analysis (PCA). The PCA was carried out on the
meteorological data, in order to obtain a characterization of the different periods of measurement during the
three years. Similarity a second PCA was applied to the chemical dataset of 42 species to summarize the obtained
results and highlight the differences and similarities in compounds profile in inter- and intra-situ.
METEOROLOGICAL CHARACTERISATION OF CAMPAIGNS
Two different PCAs were performed on meteorological data in both sites, considering the following variables: wind speed,
3a
- fall campaign 23/10/2012 - 11/11/201
4a
- winter campaign 30/01/2013 - 27/05/2013
5a- spring campaign 07/05/2013 - 27/05/2013
6a- fall campaign 27/09/2013 - 25/10/2013
7a- winter campaign 28/01/2014 - 01/03/2014
8a
- spring campaign 13/05/2014 - 11/06/2014
Two different PCAs were performed on meteorological data in both sites, considering the following variables: wind speed,
relative humidity, absolute humidity, mean temperature, sun radiation, atmospheric pressure and precipitation. The derived
results (in Fig. 2) are similar in both the sites and show that the first Principal Component (PC) is mainly characterized by
high values of mean temperature and sun radiation and low values of relative humidity, therefore it points out the seasonal
pattern. Summer days have high score values and winter days have low score values. The second PC is a contrast between
atmospheric pressure and abundant precipitations. As a consequence the observations with low scores in the second PC are
rainy days. The third PC is dominated by the wind speed.
According to this analysis, it is possible to sum up the meteorological features of the two sites through the first three PCs,
that accounts for about the 85% of the total variability. The eight campaigns are seasonally disposed, as expected, but athat accounts for about the 85% of the total variability. The eight campaigns are seasonally disposed, as expected, but a
particular behaviour of the first campaign could be observed. In effect, even if it was an autumnal campaign it is placed
among the winter campaigns and the entire period was characterized by high pressure. These conditions - winter high
pressures that bring to the typical thermal inversions - brought to particularly elevated concentration of pollutants in the
atmosphere, as it is evidenced in the following analysis.
AIR QUALITY INDEX FOR ORGANIC COUMPOUNDS
As second step, a PCA was performed on the chemical organic
compounds. Also in this case two different analysis were carried outA B compounds. Also in this case two different analysis were carried out
for the two sites (Figs. 3 A-B). The scores related to the first two PCs
were plotted and it could be noted that the patterns of the two
locations are quite similar. For the both analysis the first PC was a
weighted sum of the pollutants concentrations, while the second PC
was a contrast between alkanes and the other pollutants. Therefore
the analysis clearly points out that the first campaign registered
higher levels of pollutants than the others campaigns of the same
season. Then, as expected, the winter and fall campaigns had higher
A B
RELATION BETWEEN METEOROLOGICAL PRINCIPAL COMPONENTS AND ORGANIC SPECIES
Since the first PC of the pollutants was a weighted sum of all them and it explained about the 50%
of the overall variability, it was decided to employ it as a global indicator of the organic compounds
level. Then it was used as a response variable (PCpoll) and a relationship between this one and the
first two PCs of the meteorological variables (PCmet1 and PCmet2) was investigated. After a
A B
season. Then, as expected, the winter and fall campaigns had higher
levels of pollution than the spring or summer campaigns. In the MS,
a particular feature of the sixth campaign could be noted: in those
days increasing high values of alkanes were measured with respect
to the other pollutants.
Fig. 3 – Score plot of PC1 and PC2 components of PCA model for organic compounds dataset in urban
background (A) and in rural site (B).
first two PCs of the meteorological variables (PCmet1 and PCmet2) was investigated. After a
logarithmic transformation of the response and starting from a polynomial linear model, the
following two models were chosen as optimal:
The first result, highlighted in Figs. 4 A-C, is the linear relationship between the log of the response
and the first meteorological PC, with a high degree of negative correlation (β1 estimated -0.76 in MS
and -0,60 in SPC). This means that summer days have lower pollution levels than winter days.
As far as the second PC is concerned a quadratic term is required in the SPC data. However both MS C D
SPC model:
log(PCpoll) = β0+β1 ∙PCmet1+β2 ∙PCmet2+β3 ∙PCmet22
R2 = 0.75
MS model:
log(PCpoll) = β0+β1 ∙PCmet1+β2 ∙PCmet2
R2 = 0.76
As far as the second PC is concerned a quadratic term is required in the SPC data. However both MS
and SPC show that high pressure (i.e. high values of PCmet2)is related with more elevated pollution
level (Figs. 4 B-D). In this way a satisfying model that is able to explain the correlation between
pollution and the weather conditions was obtained. However it is worth to remark that about a half
of the variability of the considered pollutants is modelled through the first PC, the remaining 50% is
again attributed to the meteorological conditions, emissions and transformation processes in
atmosphere.
REMARKS
– All organic compounds determined in 8 measurement campaigns had the same behaviour both in
urban site and rural;
This research was conducted as part of the Supersito Project, which was supported and financed by Emilia-Romagna
Region and Regional Agency for Prevention and Environment under Deliberation Regional Government n. 1971/13.
The authors are thankful to Aldo Gardini, Statistical Sciences Department - Bologna University, for the statistic
processing.
urban site and rural;
– About the half of the variability of the considered pollutants had a strong correlation with the first
PCs of the meteorological data.
Fig. 4 – Plots of the log of the response vs the covariates with the lines fitted by the models in
rural site (A-B) and in urban background (C-D).