Air quality control in a Southern Romanian [631718]

Air quality control in a Southern Romanian
metropolitan area during the first half of 2019
Clementina Moldovan1, Liana -Simona Sbîrnă2, and Sebastian Sbîrnă3
1University of Petro șani, Faculty of Mines, Department of Environmental Engineering and Geology ,
Petro șani, Romania
2University of Craiova, Faculty of Sciences, Department of Chemistry , Craiova, Romania
3Technical University of Denmark, Institute of Applied Mathematics and Computer Science, DTU
Compute, Digital Media Engineering Master Programme , Copenhagen, Denmark

Abstract. This paper aims to present the concentration profiles of the main
air pollutants – i.e., nitrogen oxides (NO x), sulfur dioxide (SO 2), carbon
monoxide (CO) and suspended particulate matter (PM 10) – results recorded
during the first half of 2019 by two air quality monitoring stations in
Craiova, which is an important metropolitan area in Southern Romania.
Another goal of the paper is finding the best numerical diffusion model to
fit the recorded values for PM 10, as this pollutant seems to be the major
problem, because its daily average is often higher than the European Union
threshold, meaning that imperative measures have to be taken for reducing
particulate m atter concentration in Craiova (like in some other major
Romanian metropolitan areas ), in order for Romania to get the exoneration
regarding air pollution from the European Union and, of course, for its
citizens to imp rove the quality of their lives .
1 Introduction
A large number of epidemiological studies conducted around the world lead to associations
between a ir pollutants and excesses in daily mortality and morbidity [1 -4].
As air pollution represents such a stringent problem of the modern society, the current
work consists in gathering and mathematically modeling a series of results which were
obtained in Craiova, during the first half of 2019 , picturing the evolution of atmospheric
pollutants’ concentration (the most represe ntative air pollutants were produced within the
urban agglomeration of Craiova by linear or point pollution sources – mainly city transports
and coal -fired power plants, respectively).
To achieve an increasing air quality in the urban agglomeration of Cra iova (which is
obviously a fundamental aspect in increasing life quality), several measures have to be taken for
reducing the concentration of particles found in suspension in the atmosphere and also the
concentrations of the gaseous pollutants that are pr esent in the inspired air, mainly nitrogen
oxides (NO x), sulfur dioxide (SO 2), carbon monoxide (CO) and suspended particulate matter
(PM 10).
Air pollution is a serious problem in the urban agglomerations, because of its detrimental
impact on public health and living standards.

2 Preliminaries
Values recorded for the four major atmospheric pollutants previously mentioned at two modern
automatic air quality monitoring stations shall be presented in this paper, in order to compare
them with the limit -values.
2.1 Limit -value s for the main pollutants
According to the Law 104 /June 15th, 2011 [5], the hourly limit -value for NO 2 is 200 μg/m3, in order
to preserve health protection.
Similarly, the same law states that 350 μg/m3 is the hourly limit -value for SO 2 and that 10
mg/m3 is the daily limit -value of the eight hours averages for CO, whereas the daily average value
of PM 10 concentration should not exceed 50 µg/m3.
2.2 Automatic air quality monitoring stations
The two selected stations are representative for Craiova urban agglomeration: RO0078A (DJ 1),
situated on the main street (Calea Bucure ști) and RO0079A (DJ 2), which is located near City Hall :
data points were plotted in four charts – one for each major air pollutant previously mentioned ;
these charts will be shown in what follows (there are obviously too many data to be given as a table).
3 Values recorded for the main pollutants
The concentrations of the air pollutants are influenced by the air pollutants’ levels, meteorological
conditions and topography.
3.1 Values recorded for nitrogen oxides
The nitrogen oxides are generally denoted as NO x. The first chart represents the hourly values
recorded for NO 2 (the main part of NO x) at both stations, during the first half of 2019.

Fig. 1. Temporal variability of NO 2 (main part of NO x) concentration values , recorded at both DJ 1 and DJ 2
stations .

With a single exception, t he graph shows no alarming temporal variability of NO 2 (and therefore
for NO x).
3.2 Values recorded for sulfur dioxide
The second chart represents the hourly values recorded for sulfur dioxide , SO 2, at both DJ 1 and
DJ 2 stations, during the above mentioned period.

Fig. 2. Temporal variability of SO 2 concentration values , recorded at both DJ 1 and DJ 2 stations.
Once again, the graph shows no realy worrying temporal variability, although the imposed
limit of 350 µg/m3 was exceeded sometimes .
3.3 Values recorded for carbon mon oxide
The third chart represents the eight -hour average values recorded for carbon monoxide , CO, at
both DJ 1 and DJ 2.

Fig. 3. Temporal variability of CO concentration values , recorded at both DJ 1 and DJ 2 stations.

The graph shows no upsetting temporal variability of CO, the values being far away from the
eight-hour limit .
3.4 Values recorded for suspended particulate matter
The fourth chart represents the daily values recorded for PM 10 at DJ 1 only, during the investigation
periods (this one being closer to the meteorological station which recorded the wind speed).
Indeed, these data were tha n correlated to the wind speed, showing that, as it was expected, a
low value for the wind velocity usually leads to high values of the particulate matter in the
atmosphere (according to Fick’s law).

Fig. 4. Temporal variability of PM 10 concentration values , recorded at DJ 1 station only
Most of the time, no disquieting values were recorded, but the number of times when the limit
value of 50 µg/m3 was exceeded requires more concern about the human health. The correlation
between the PM 10 concentration and the wind velocity were also plotted into a chart.

Fig. 5. Correlation between the PM 10 concentration values and the wind speed

4 Statistical analysis
Statistical analysis and modeling has important benefits, such as: application of appropriate
statistical analysis techniques; development of appropriate conclusions and key learning from
the data; ensuring results address experimental objectives; maximizing information gained
from the data; maximizing chances of the interpretation bein g successful.
4.1 Log-normal distribution
In the probability theory, a log -normal distribution is a probability distribution of a random
variable whose logarithm is normally distributed. Parameters of the log -normal distribution,
namely the geometric mean µg and the standard geometric deviation, σ g, are defined by (1) and
(2), respectively.
4.2 Probability density function
The probability density function of the log -normal distribution, the cumulative distribution
function, which is defined as the probability of the variable x to be smaller than a critical
value x 0 and the complementary distribution function, which is defined as the probability of
the variable x exceeding the critical value x 0, are given by ( 3), (4) and ( 5), respectively.
The variable m that is involved in (4) is calculated by ( 6). In ( 1)-(6), variable x represents
the concentration of a pollutant.
μg= (x1∙x2∙… ∙ xN)1
N (1)
σg=exp {1
N[∑ (ln xi−lnμg)2N
i=1 ]}1
2 (2)
p(xi)=1
√2π× xi×lnσgexp [− (lnxi−lnμg)2
2 ×ln2σg] (3)
F(x)=Pr[xi< x0]= 1
√2π ∫ e−t2
2dtm
−∞ (4)
F′(x)=Pr[xi> x0]=1−F(x) (5)
m=(lnx0−lnμg)
lnσg (6)
4.3 Statistical indexes
The suitability of a distribution can be examined via statistical indexes, to perform the
computational analysis. The mean bias error (MBE), the mean absolute error (MAE), the
root mean square error (RMSE), its systematic (RMSE s) and unsystematic (RMSE u)
components and the index of agreement (d) are those that will be calculated within this study.
A drawback of RMSE is that few large errors in the sum may produce a significant
increase in its value, whereas a drawback of MBE means that an overestimation in one
observation may be canceled by an underestimation in the other. P and O represent the
predicted and the observed values respectively, whereas the overbar indicates mean values.
MBE = P̅− O̅ (7)
MAE =1
N∑ |Pi−N
i=1 Oi| (8)
RMSE =[1
N∑ (Pi−Oi)2 N
i=1 ]1
2 (9)

RMSE s= [1
N∑ (Pi∗−Oi)2 N
i=1 ]1
2 (10)
RMSE u= [1
N∑ (Pi∗−Pi)2 N
i=1 ]1
2 (11)
d=1− ∑ (Pi−Oi)2 N
i=1
∑ (|N
i=1Pi−O̅|+|Oi−O̅|)2 (12)
t=[(N−1)MBE2
RMSE2− MBE2]1
2 (13)

In (10) and (11), P i* is calculated as a + b ·Oi, where a and b are the intercept and the slope
of the least squares line between the predicted and the observed values, whereas N is the
number of concentration classes in which the data are divided. The reduction R in current
emissions to meet air quality standards is often calculated by “the rollback equation ” (14),
where E{C} is the current annual mean value of the pollutant’s concentration, E {C} S is the
annual mean value corresponding to air quality standards and C b is the background
concentration.
R=(E{C}−E{C}s)
(E{C}−Cb) (14)
The mean concentration for log -normal distribution can be therefore calculated .
lnE{C}=lnμg+ 1
2ln2σg (15)
When the parent probability distribution of air pollutants is properly chosen, this specific
distribution can be used to estimate the mean concentration, the number of times when the
air quality standards is exceed ed and the emission sources reduction, to meet the air quality
standards . Table 1 presents this accordance in figures , whereas Fig. 6 shows the log-normal
distribution that fits very well to the distribu tion of the PM 10 concentration values
4.4 Verifying goodness -of-fit
The goodness -of-fit is verified by calculating the statistical indicators previously presented. The
mean of the observed values is identical to the mean of the predicted values (O = P).
Table 1. Accordance between parent and observed distribution
Class PM 10 (µg/m3) Real
percent Calcd.
percent Diff.*
1 0-9.99 5.0 5.1 +0.1
2 10-19.99 20.2 20.4 +0.2
3 20-29.99 25.0 24.8 -0.2
4 30-39.99 23.2 23.1 -0.1
5 40-49.99 14.1 14.4 +0. 3
6 50-59.99 8.2 8.1 -0.1
7 60-69.99 4.2 4.4 +0.2
8 70-79.99 2.1 2.4 +0.3
9 80-89.99 1.3 1.0 -0.3
10 90-99.99 1.1 1.0 -0.1
11 100-109.99 0.8 0.8 0
12 110-119.99 0.7 0.7 0
* Difference between calculated and real percentage
Verifying goodness -of-fit shows that the log -normal distribution is capable to simulate
the experimental data the most adequately .

Fig. 6. Log-normal distribution of the real data
Conclusion
In this study, the log -normal distribution is applied in order to describe the observed
distribution of PM 10 levels in Craiova city – capital of Dolj County (south -western
Romania).
The major conclusion is that t he PM 10 levels (slightly higher compared to the air quality
standards imposed by the European Union) are very good fitted by a log -normal
distribution, which also is valuable as a prediction.
Taking into account the results of this study , the following measures are recommended :
closing the historical center for road traffic and revitalizing it, which would be advantageous
in both short and long term; developing green areas ; implementing an efficient traffic routing
and traffic lights system in order to provide functional links between the central roads and
the other roads ; implementing roundabouts in order to eliminate increasing concentrations of
air pollutants and reducing road congestion by default ; improving public transport , using
biofuels in the bus fleet ; using electric vehicles , thus eliminating the generation of additional
emissions ; carrying out desulphurizatio n installation s by the local energy complex ; wetting
the ash and slag deposits in the dry season, in order to eliminate flying ash involvement .
It is estimated that, by implementing these measures, air pollution would decrease by 5%
by closing the road and revitalizing its historical center , also with 2 % by developing and
introducing roundabouts and proper traffic lights routing system , with another 4 % after
completing the subway passage and putting it into use and other 4% by using electric vehicles
and alternative fuels.
References
1. S.K. Chaulya, R. Trivedi, A Kumar, R.K. Tiwary, R.S. Singh, P.K. Pandey, R. Kumar,
Atmos Pollut Res 10 (3), 675 -688 (2019 )
2. E. Canepa, C.F. Ratto, Environ Model Softw 18, 365-372 (2003 )
3. D.M. Moreira, M.T. Vilhena, D. Buske, T. Tirabassi, Atmos Environ 40, 3186 -3194
(2006 )
4. S. Baroutian, A. Mohebbi, A. Soltani Goharrizi, J. Hazard Mater 136, 468 -474 (2006 )
5. Available on http://www.calitateaer.ro/public/assessment -page/pollutants -page/