You are on page 1of 13

ARTICLE IN PRESS

Atmospheric Environment 41 (2007) 6432–6444


www.elsevier.com/locate/atmosenv

Application of a Lagrangian particle model to assess the


impact of harbour, industrial and urban activities on air
quality in the Taranto area, Italy
Claudio Gariazzoa,, Vincenzo Papaleoa, Armando Pelliccionia,
Giuseppe Calorib, Paola Radiceb, Gianni Tinarellib
a
Italian Institute for Occupational Safety and Health (ISPESL), Research Centre, Via Fontana Candida 1,
00040 Monteporzio Catone, RM, Italy
b
Arianet, Via Gilino 9, 20128 Milano, Italy
Received 9 August 2006; received in revised form 5 June 2007; accepted 8 June 2007

Abstract

This paper evaluates the relative impact on air quality of harbour emissions, with respect to other emission sources
located in the same area. The impact assessment study was conducted in the city of Taranto, Italy. This area was
considered as representative of a typical Mediterranean harbour region, where shipping, industries and urban activities co-
exist at a short distance, producing an ideal case to study the interaction among these different sources. Chemical and
meteorological field campaigns were carried out to provide data to this study. An emission inventory has been developed
taking into account industrial sources, traffic, domestic heating, fugitive and harbour emissions. A 3D Lagrangian particle
dispersion model (SPRAY) has then been applied to the study area using reconstructed meteorological fields calculated by
the diagnostic meteorological model MINERVE. 3D short term hourly concentrations have been computed for both all
and specific sources. Industrial activities are found to be the main contributor to SO2. Industry and traffic emissions are
mainly responsible for NOx simulated concentrations. CO concentrations are found to be mainly related to traffic
emissions, while primary PM10 simulated concentrations tend to be linked to industrial and fugitive emissions.
Contributions of harbour activities to the seasonal average concentrations of SO2 and NOx are predicted to be up to 5 and
30 mg m3, respectively to be compared to a overall peak values of 60 mg m3 for SO2 and 70 mg m3 for NOx. At selected
urban monitoring stations, SO2 and NOx average source contributions are predicted to be both of about 9% from harbour
activities, while 87% and 41% respectively of total concentrations are predicted to be of industrial origin.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Harbour; Industrial; Emission; Model; SODAR

1. Introduction

Corresponding author. Tel.: +39 06 94181525; The growing number of shipping movements and
fax: +39 06 94181527. the related release of air pollutants have drawn
E-mail address: claudio.gariazzo@ispesl.it (C. Gariazzo). attention onto this kind of emission source.

1352-2310/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.atmosenv.2007.06.005
ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6433

Furthermore, ship emissions have not been as 2. Methodology


tightly controlled as many land-based emission
sources, and the international context of the 2.1. Case study
shipping sector has posed difficulties to achieve
progress in improving environmental performance. The city of Taranto is one of the most popula-
Ship emissions can make a significant contribution ted–industrialized areas in Italy. Here, the typical
to air pollution problems in the European Commu- urban emissions are superimposed with the indus-
nity. In particular, these are a major source trial ones located in proximity of the city bound-
of SO2 and NOx, which leads to acidification aries. A map of the studied area is depicted in Fig. 1,
and eutrophication as well as formation of ground which shows both a large and a zoomed view.
level ozone. A recent European Commission (EC) Among the industrial activities located here, the
study (Entec, 2002) quantifies as 2.6 and biggest one in terms of both emissions quantity and
3.6 Kte year1 the SO2 and NOx emitted by ships extension of working areas, is a large steel plant.
in EU countries in the year 2000, with a growing Carbon-coke transformation processes are con-
rate of 3.3 and 4.6 Kte year1 up to 2010. A recent ducted here. A power plant with tall emission stacks
report produced by the EC (Entec, 2005) indi- is also located inside this area. A significant
cates to range between 50% and 1%, depending industrial source is also an oil refinery with
on the pollutant and the EU country, the ratio of hydrocarbons transformation processes and storage
ship to total emissions, as calculated by RAINS of crude and final products. It is the third most
model used for the Clean Air for Europe important refinery in Italy in terms of production.
programme. A big cement facility is also part of the emission
According to the EC study, respectively 4.5% and sources. Other smaller emission sources can be
6.2% of the total SO2 and NOx emitted by ships are found in this territory, related to both the main
due to in-port activities as maneuvering, loading/ industrial activities and the local economy. The
unloading and hotelling. A high number of major above industrial activities use the harbour of
ports are located close to or in proximity of Taranto to download primary goods and to delivery
main cities. One third of top 100 highest emission final products. These tasks are often related to
EU ports in 2000 were located in the Mediterranean emissions of particulate matter. Combustion pro-
area, with 18 of them sited in Italy relatively cesses during wharf operations also produce pollu-
close to high-urbanized areas. Ports located in tants emissions.
proximity to urbanized area, produce a combined Although emissions related to harbour activities
environmental effect due to the superposition of are not considered as the major contributing sources
ship emissions with those related to urban sources. for the local air quality, their overall assessment is
For technical and economical reasons, ports can often unknown as well as their relative contribution
also be close to industrial areas. Consequently the to the local air quality.
emission of air pollutants, mainly particulate
matter, is expected as a result of unloading of 2.2. Modelling system
primary goods and loading of the final products
activities. To conduct this assessment study, a modelling
According to the above facts any environmental system based on three numerical codes was used: the
plan, conducted to improve the air quality in such MINERVE meteorological model, the SURFPRO
areas, needs the assessment of the relative impor- turbulence pre-processor and the SPRAY Lagran-
tance of harbour activities emissions with respect to gian particles dispersion model. Three-dimensional
other emission sources, to optimize the cost/effec- fields of wind and temperature have been calculated
tiveness of planned interventions. using the diagnostic code MINERVE (Aria Tech-
On these premises, a case study has been nologies, 2001) applying a mass-consistent objective
conducted in one of the most important Italian analysis scheme. The SPRAY 3.0 model (Tinarelli
ports, where harbour, industrial and urban activities et al., 1994; Gariazzo et al., 2004) was used to
are close together, producing an ideal case to study reconstruct concentration fields produced by pollu-
the interaction among these different sources, their tants emitted by the considered sources. SPRAY is a
effects on the environment and their contributions 3D model that simulates air pollution dispersion
to the air quality. in the atmosphere in non-homogenous and
ARTICLE IN PRESS
6434 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444

Fig. 1. Map of the studied area with a large and zoomed view showing location (UTM 33 [Km]) of the monitoring stations and the main
industrial, harbour and urban areas. Legend: Meteo (r); Mobile Labs (+); Meteo and Chemical (J); Chemical ().

non-stationary conditions. SPRAY uses a certain were used, located both downtown and in the city
number of computational (fictitious) particles and neighbourhoods. Pollutant concentration measure-
each particle is ‘‘moved’’ at each time step by ments were also carried out in the above stations. In
pseudo-velocities, solutions of stochastic Langevin addition, three mobile laboratories were located in
equations with Gaussian random forcing (Thom- the area. The Italian Institute for Occupational
son, 1987), taking into account the three basic Safety and Health (ISPESL) mobile meteorological
dispersion components: the transport due to the laboratory (MML) was placed in the harbour of
mean fluid velocity; the random turbulent fluctua- Taranto. Besides the main standard meteorological
tion of wind components; and the molecular data, the MML calculates average values of
diffusion. Turbulent fluctuation of wind compo- turbulence parameters starting from data collected
nents (su, sv, sw) is used by SPRAY to determine by sonic anemometers. Two Mobile Chemical
the random motion causing the dispersion. The Meteorological Laboratories (MCML) were located
latter are calculated by means of parameterization in the rural area of Palagiano, about 30 km west the
codes (Hanna, 1982) based on scaling variables city of Taranto, and in the urban area of Statte,
derived by the SURFPRO code on the basis of the nearly 15 km north of Taranto, where the impact of
Monin–Obukhov similarity theory and surface pollutants emitted from the industrial areas is
energy budget evaluations (Van Ulden and Holt- expected to be detected. To get information on
slag, 1985), starting from two-dimensional arrays of meteorological parameters in the eastern and north-
surface parameters (albedo, bowen ratio, roughness ern parts of the territory, two other meteorological
length z0) and ground meteorological parameters stations were placed in the S. Giorgio and Monte
given as input. Mesola villages.
Upper air measurements were carried out by
2.3. Field campaigns description means of the ISPESL SODAR/RASS system,
located close to the MML station in front of sea
Two field campaigns were conducted in the winter and at nearly 2 km far from the industrial area.
and summer seasons to feed the model with real Wind, turbulence and temperature profiles up to
data and to validate modelling results. 400 m a.g.l. were measured. Such data were com-
To provide meteorological data to the modelling pleted by wind, temperature and pressure profiles
system, various stations were available (Fig. 1). up to 24 km obtained from radio soundings,
Among the local monitoring network, five stations collected every 6 h at the city of Brindisi, 80 km
(Garibaldi, Peripato, Orsini, Paolo VI and Dante) East of Taranto.
ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6435

Meteorological data collected during the field shown in order to underline the contribution of
campaigns, as well as available historical data various activities to the total emissions of specific
routinely collected by the local monitoring network, pollutants; the road transport has the greatest
were used to detect the main circulation patterns relevance, representing more than 50% of CO and
occurring in the area. Results show that in summer, NMVOC emissions, and less than 20% for NOx
high-pressure systems are active and the high ones. Industrial sources are by far the major source
insulation produces convective atmospheric condi- of SO2, NOx and PM10.
tions with a well-mixed boundary layer. At daytime, Data concerning stack emissions from industrial
the wind mostly blows from south-southwest sources were collected directly from plants owners.
(S-SW) with speed up to 5 m s1, as a consequence Such point sources represent installations that are
of the sea breeze triggered by the land/sea tempera- relevant at the scale of the analysis and that are
ture gradients. It then turns to north-northwest known individually. Up to 380 different sources
(N-NW) during nighttime due to a land breeze were considered in this study. Particulate matter
effect. In winter, the prevailing flows are from fugitive emissions related to industrial activities
N-NW and south-southeast (S-SE). Cold or warm were also taken into account. Storage areas and
fronts often pass through the area with synoptic quarries were considered as possible emission
wind conditions characterized by speeds up to sources of re-suspended matter. Their emission
8 m s1. A few episodes with persistent high pressure estimation is based on the methodology described
systems, associated with weak pressure gradients in Parret (1992). To calculate the emissions it needs
and low wind conditions were detected in this some inputs data such as: type of goods, total
season, determining conditions favouring accumu- amount of goods per year, extension of the
lation of pollutants in the lower layers. area used for storage of goods and silt content of
goods. In addition, in order to estimate the yearly
2.4. Emission inventory amount of re-suspended dust due to meteorological
conditions, the methodology also needs the
The emission inventory developed for this study number of rainy days in the considered year and
contains yearly emissions of the main chemical the relative amount (in percent) of wind speed above
species linked to the most significant anthropic a threshold of 5.4 m s1 during the year, considered
activity sectors: industrial sources, road traffic, as the level above which wind-driven dust re-
residential heating, harbour activities. This is an suspension takes place. According to above require-
upgraded version of an emission inventory devel- ments, data related to the year 2002 were then
oped for the city of Taranto in a previous study collected and used for the fugitive emissions
(ATI et al., 2003). estimation.
In Fig. 2, sources emissions, considered as a Road traffic emissions are represented in two
whole and grouped into SNAP categories, are ways: as line and as areal sources. The road network

100%
Maritime
activities
80% Road
transport
60% Production
processes
Combustion in
40% industry
Domestic
20% heating
Combustion
Energy
0%
CO

NMVOC

SO2
NOX

PM10

Fig. 2. Sources contribution to the total emissions of some considered pollutants.


ARTICLE IN PRESS
6436 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444

and the related vehicular flows were available for 2.5. Models setup
the urban area of Taranto, for highways, and for
provincial and state roads. In all these cases A 35  35 km2 model domain was considered to
emissions are determined by means of TRaffic cover all possible impacts on the surrounding areas
Emission Factor Improved Calculation (TREFIC) and relevant towns. The domain has been horizon-
software (Nanni et al., 2005), based upon the tally divided into 71  71 grid cells with 500 m
COPERT III methodology integrated with more resolution and vertically splitted from the ground
detailed emission factors for particulate from brakes level to the top, set to 1500 m. Surface parameters
and tire wear (IIASA, 2001). Traffic emissions for (albedo, roughness length and Bowen ratio) were
all other populated places located in the modelled estimated on the basis of the Corine land cover map
domain are treated as area sources and surrogated (ETC/LC, 1997) with a 500 m resolution. A lookup
from the previous ones, using the number of table was used to assign to each cover class a value
registered vehicles as a proxy variable. of albedo, roughness length and Bowen ratio.
The emission estimation from Taranto residential A total of 33 days were selected to simulate air
heating is based on yearly fuel consumption, while pollution dispersion around the studied area. Four-
for all other surrounding towns, emissions are teen of them were related to winter season and 19 to
surrogated from those of Taranto, using their summer. Their selection was based on both typical
number of inhabitants. local atmospheric circulations/synoptical conditions
Harbour emissions from loading/unloading activ- in the considered season and on the occurrence of
ities and combustion processes during wharf opera- pollutants concentration peaks clearly linked with
tions are estimated from mercantile fleet data, industrial emissions as revealed by observations at
average in-port idling time and type of transported the monitoring stations.
goods. Emissions linked to the quays were derived The 3D wind and temperature fields were
from detailed information on goods handling. Ship obtained by means of the MINERVE model, using
emissions take into account three different stages: meteorological data collected during the field
maneuvering, hotelling and cruising. Different data campaigns. As the SODAR/RASS system provided
are employed for the emission estimation: shipping wind and temperature profile data up to 300 m a.g.l.
movements in the studied area, vessel data on average, the remaining part of the model domain
(e.g. engine power, fuels) and emission factors was not covered by data. To overcome this problem
defined from the existing literature (Entec, 2002; the upper part of the atmospheric boundary layer
Lloyd’s Register, 1995, 1999). In the year consid- (from 700 m on) above the Taranto area was
ered, a total of 1999 movements were registered by considered to be equivalent to that observed at the
the Italian Statistical Institute (ISTAT) in the Brindisi station, 80 km away from the model
harbour of Taranto. Hotelling time in the port is domain. According to this assumption, the upper
evaluated to be 2.5 days, cruising time (from the part of the model domain was covered by the
entrance to the berth) less than an hour, while the Brindisi radiosounding data. A 2D Cressman
maneuvering time is evaluated to be half an hour. It interpolation scheme was used to spatially inter-
is assumed that all ships are characterized by slow polate both surface and upper air data at each
speed diesel engine; while inside the port, an average model vertical level, to obtain a first guess
speed service of 6 knots is assumed. Based on these meteorological field. Surface wind was then cor-
data, Lloyd’s emission factors (Lloyd’s Register, rected according to both surface roughness (z0) and
1995) and elaboration of fuel consumption data at topographic altitude, using a scheme based on
full power, ship emissions in different phases are similarity theory. The wind field was then adjusted
evaluated according to formulas reported by Trozzi to fulfil mass-consistent objective analysis con-
and Vaccaro (1998). straints. Minimum values of sw data were set
To feed the dispersion model, yearly emissions according to data collected at both MML by sonic
are temporally and spatially disaggregated using anemometers and SODAR/RASS systems. The 3D
specific information related to the area. Areal wind and temperature fields generated by the
emissions from the considered activities are allo- MINERVE meteorological model, the atmospheric
cated according to the spatial distributions of the turbulence produced by SURFPRO and the emission
related classes of land use, while time modulation is data, were then provided as input to the SPRAY
based on activity-specific proxies. dispersion model. 3D hourly concentrations of NOx,
ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6437

SO2, CO and primary PM10 were calculated as total Table 1


overall concentration and in terms of contributions of Statistical evaluation of SPRAY model performance over all
specific sources. available stations

C̄ p (mg m3) C̄ o (mg m3) NBIAS NMSE


3. Results
NOx 26.3 44.6 0.42 1.59
3.1. Comparison with monitoring data SO2 13.3 6.0 1.58 10.74
CO 82.5 966.0 0.91 12.67
PM10 17.5 50.0 0.63 4.14
Dispersion models are modern tools to obtain
spatially resolved concentrations, but their applic- Key: C̄ p and C̄ o are the averages over the data set for the model
ability and reliability are uncertain in many cases. predictions and the observations. NBIAS is the normalized bias,
NMSE is the normalised mean square error.
Stringent validation and comparison with measure-
ments are essential before results are used to assess
the impact of emissions on the territory. To evaluate for this evaluation: one located downtown (Dante)
the model performances, observed NOx, SO2, CO and the other (Orsini) in an urban district in
and primary PM10 hourly concentrations were proximity to the industrial area. Fig. 3 shows a
compared with the corresponding modelled values. comparison of observed and SPRAY modelled NOx
As far as the background level of pollutants is concentrations at Dante and Orsini monitoring
concerned, it should be considered that Lagrangian stations, for winter and summer seasons, respec-
particle model cannot take into account boundary tively. While at the Dante station the model tends to
conditions, like other kind of models (e.g. Eulerian) overestimate observed values, the opposite happens
can do. Nevertheless, according to observations at Orsini, likely due to an incorrect evaluation of the
collected during the seasonal field campaigns, the actual emissions. In general, the NOx model results
background levels for both NOx and SO2 are found exhibit a good reproduction of observed values both
to be at least one order of magnitude lower than in time and concentration value.
peak levels. For these pollutants, the lack of The SO2 model results (not shown) in general
background concentrations does not introduce exhibit peaks higher than observed, sometimes with
significant model underestimations, but only a slight modelled peaks not revealed by observations.
negative bias. Conversely, according to observa- Possibly, this is due to the above-mentioned
tions, the lack of background concentrations for CO problems on measurements reliability.
and PM10 could significantly introduce underesti- For both CO and primary PM10, model results
mations in model predictions. indicate an overall underestimation compared to
Table 1 summarizes a statistical evaluation of measured concentrations. For both pollutants this
model performances over all the available stations result might be mainly due to the regional back-
and all simulated days. In general, model predic- ground contribution, which is not taken into
tions underestimate the observed average concen- account by Lagrangian models, as mentioned
trations of all pollutants, but SO2, which is before. As far as PM10 is concerned, the lack of
overestimated by the model. Best results are other emission sources or an incorrect evaluation
obtained for NOx, while poor performances are of the emission factors, as well as the effect of
obtained for both CO and primary PM10. To secondary pollutants, could also contribute to the
partially explain these statistical results, it should model underestimation. Consequently, results of
be considered that, especially for SO2 and PM10 primary PM10 obtained in this study can only be
pollutants, data collected by the local monitoring considered as the consequences produced by the
network were not found as reliable as expected. combined effect of both local emissions and
Therefore, some disagreements in the values of the dispersion conditions. However, these results can
statistical indexes are probably due to a mismatch still be used for an assessment study and definition
between model results and observations. of control strategies, as they allow to handle the key
A further evaluation of model results can be factors (e.g. emissions), for which the local autho-
obtained by comparing the time series of observa- rities should have prior responsibility on control of
tions with model results. The better performance air quality. A more detailed study on PM would
obtained for NOx results can also be observed in the require the use of a comprehensive chemical
time series comparison. Two stations were selected transport model.
ARTICLE IN PRESS
6438 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444

Winter
100

Modeled
Observed
80

60
NOx (ppb)

40

20

0
04

04

04

04

04
04

04

04

04

04

04

04

04

04
20

20

20

20

20
20

20

20

20

20

20

20

20

20
2/

3/

3/

3/

3/
2/

3/

3/

3/

3/

3/

3/

3/

3/
/0

/0

/0

/0

/0
/0

/0

/0

/0

/0

/0

/0

/0

/0
21

04

09

12

15
20

02

03

05

06

10

13

14

16
Summer
120
Observed
Modeled
100

80
NOx (ppb)

60

40

20

0
15 4

16 4

19 4

20 4

4
/0 04

/0 04
/0 04

/0 04

/0 04

/0 04

14 4
/0 04

/0 04

/0 04

/0 04
/0 04

/0 04

/0 04

7/ 00

7/ 00

7/ 00

7/ 00
00
7/ 00
25 /20

11 /20
20 /20

26 20

05 /20

06 20

12 /20

13 /20
08 /20

09 /20
19 /20

24 /20

07 /20

/2

/2

/2

/2

/2
2
6/

7/

7/
6

7
6

7
6

7
/0
18

Fig. 3. Comparison of observed and SPRAY modelled NOx concentrations at Dante (winter) and Orsini (summer) stations.

3.2. Seasonal-averaged concentration maps During wintertime, NOx average values up to


50 mg m3 are foreseen, while at summertime higher
Although the comparison of model results with peak concentrations (70 mg m3) are predicted due
observations is sometimes misleading, it is interest- to a stronger mixing of the local atmosphere, which
ing to investigate the concentration results in terms drops down pollutants emitted by elevated stacks.
of the impacts on the local territory produced by the The highest winter NOx peak is located in the urban
industrial, harbour and urban emissions, as well as area of Taranto and a secondary peak is also placed
the seasonal effect caused by the different meteor- in the Tamburi district (Orsini station), close to the
ological conditions. The 3D hourly concentrations industrial area. Other hot spot NOx peaks can also
results have been used to calculate average winter be observed in the nearby towns. At summertime, a
and summer concentration maps. Fig. 4a shows large NOx peak (up to 50 mg m3) is located around
winter- and summer-averaged maps of NOx, SO2, the industrial area, whilst the predicted concentra-
and primary PM10 concentrations predicted by the tion in the city of Taranto is lower (30 mg m3). Hot
SPRAY model. spots on the surrounding towns are clearly visible as
ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6439

Fig. 4. (a) Average concentration maps of NOx, SO2 and primary PM10, predicted by the SPRAY model in winter and summer seasons.
(b) Seasonal NOx concentration maps ascribed to harbour activities as predicted by the SPRAY model.
ARTICLE IN PRESS
6440 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444

Fig. 4. (Continued)

for winter results. The effect of land/sea breeze on urban area of Taranto. During summertime, the
concentrations is also clearly visible in the summer average concentration map shows a maximum
results. A greater part of the inland territory is located, as for winter results, in the industrial area,
influenced by the NOx concentrations, which with values of up to 60 mg m3. Concentrations
extends up to 15 km from the coastline. The spatial between 10 and 20 mg m3 are predicted over the
dispersion of NOx concentrations over the sea is urban area of Taranto. As for NOx results, the
also greater in summer than in winter. spatial distribution of summer and winter SO2
According to the SPRAY calculated harbour concentrations is rather different due to different
contributions shown in Fig. 4b, during wintertime meteorological conditions. According to the
average values of up to 20 mg m3 are produced by SPRAY model estimations, both winter and sum-
the harbour activities, mainly caused by the mer harbour contributions are not more than
‘‘Polisettoriale’’ quay, 6 km west from the city of 5 mg m3, mainly due to the ‘‘Polisettoriale’’ quay
Taranto, and its effect spreads over the sea up to emissions.
5 km from the coastline and covers part of it. This The winter and summer CO concentration maps
quay is the most important one in the harbour, (not shown) show a traffic related origin, with peaks
being used by a large number of container ships. located in the urban areas and close to the
The harbour contributions from around this quay connecting roads. Summer concentration values
are dominant with respect to the other contributor are foreseen to be higher than winter (200 vs.
sources. At summertime, a peak value of 30 mg m3 100 mg m3) due to the frequent summer time low
can be ascribed to the harbour activities. Its wind speed conditions. Concentrations are confined
contribution can be clearly split in two main around the emission areas as a consequence of their
components: one ascribed to the ‘‘Polisettoriale’’ emission at ground level.
quay and the other to the harbour closer to the city Primary PM10 seasonal concentration maps
of Taranto, with the former being stronger than the (Fig. 4a) exhibit an impact area restricted to the
latter one, due to the higher number of ships industrial district and its surrounding with an
involved. The spread due to breeze effects is even extension of 4  4 km2. The city of Taranto is only
larger both over the sea (about 10 km) and inland partially influenced by this kind of emission.
(about 5 km). According to the model prediction, the summer
The SO2 winter average concentration map concentration is higher (120 mg m3) than in winter
(Fig. 4a) shows three peaks with average value of (70 mg m3), due to the low dispersion conditions
25 mg m3, located close to industrial area and in the mentioned above.
ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6441

In order to get information on air quality 50 mg m3. Its proximity to the industrial area is
predicted by SPRAY at particular locations and the main reason for this result. The station of
its compliance with the European legislation, daily Peripato seems also to be affected by high daily
and seasonally averaged pollutant concentrations, PM10 concentrations (50 and 63 mg m3 maximum
as well as their maximum hourly values, were daily concentrations at winter and summer, respec-
calculated for selected stations. Results are shown tively). The annual legal limit value of 40 mg m3 is
in Table 2. The EU mother directive 96/62/CE and probably overcame at the Orsini station, according
the two daughter directives 99/30/CE and 2000/69/ to the seasonal concentrations derived in this study.
CE, establish the air quality standards for the NOx, The CO concentrations are below the legal limits,
SO2, CO and particulate pollutants. As far as the although contributions from background concen-
model results for SO2 are concerned, the daily air trations have not been accounted for in this study
quality standard (150 mg m3) is fulfilled. The hourly due to model limitations.
one is close to the limit value of 350 mg m3 at Orsini
station (263 mg m3 at wintertime), due to its 3.3. Evaluation of sources contribution
position close to the industrial areas. Critical values
are predicted for hourly NOx concentrations at In a Lagrangian particle dispersion model, the
Orsini, Dante and Peripato stations, with values airborne pollutant is simulated by means of a
either close or above the limit of hourly concentra- certain number of ‘virtual’ or ‘pseudo’ particles,
tion (200 mg m3) given for human health protec- each of them representing a certain pollutant mass
tion. Hourly concentrations close to the legal limit emitted by a specific source. For any particle, the
are predicted at summertime in the small town of model knows both the source where it was emitted
Statte (210 mg m3), mainly due to pollutants from and the mass of the pollutant it accounts for.
transported by sea breeze. Although annual con- So the concentration of a pollutant produced in a
centrations cannot be obtained from the results of model cell by a specific source or group of them, can
this study, the calculated SO2 and NOx seasonal be easily calculated by counting the number of
concentrations, indicate a possibility that annual particles which can be ascribed to the selected
concentrations might be close to or higher than the source, summing up their mass contribution and
legal limits at Orsini, Dante and Peripato stations. divide it by the cell volume. In practice, the model
The station of Orsini and its district of Tamburi was set up to calculate a 3D pollutants concentra-
may also be affected by daily PM10 concentrations tion field for groups of different sources such
higher than the corresponding legal limit of as industrial, traffic, domestic heatings, harbour

Table 2
Seasonal, daily and maximum hourly ground concentrations of pollutants predicted by the SPRAY model at selected monitoring stations

Station NOx SO2 CO PM10

Seas. Daily Max 1 h Seas. Daily Max 1 h Seas. Daily Max 1 h Seas. Daily Max 1 h

Min Max Min Max Min Max Min Max

Winter
Orsini (Urb/Ind) 37 12 56 304 25 0 48 263 59 25 106 320 51 6 99 193
Dante (Urban) 41 22 64 180 16 0 36 126 132 66 193 632 14 2 26 86
Peripato (Urban Park) 45 21 68 212 23 0 42 208 94 41 158 406 23 2 50 148
Palagiano (Rural) 6 1 8 50 1 0 4 37 40 14 75 358 3 1 5 32
Paolo VI (Urban) 4 0 15 81 2 0 10 84 15 6 31 106 2 0 9 97
Statte (Urban) 19 7 42 130 5 0 28 88 84 40 146 502 3 0 18 79
Summer
Orsini (Urb/Ind) 40 12 70 236 26 3 51 202 75 20 143 700 77 9 123 603
Dante (Urban) 38 6 64 191 14 0 25 134 138 40 242 677 18 1 44 210
Peripato (Urban Park) 43 10 68 201 19 0 39 175 108 33 181 582 29 2 63 408
Palagiano (Rural) 3 0 7 54 1 0 6 65 40 12 77 461 2 1 4 33
Paolo VI (Urban) 12 0 36 131 9 0 26 100 27 8 59 229 7 0 25 104
Statte (Urban) 21 6 50 210 6 0 29 168 104 34 216 810 4 0 17 88
ARTICLE IN PRESS
6442 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444

Table 3
Seasonal SPRAY modelled source contributions at some selected monitoring stations (WI as winter; SU as summer)

Source Dante Orsini Palagiano Paolo VI Peripato Statte

WI SU WI SU WI SU WI SU WI SU WI SU

SO2 (%)
Industry 92 86 97 89 70 71 91 92 95 89 94 91
Traffic 2 3 1 1 7 14 2 1 1 2 2 2
Domestic heating 0 N.A 0 N.A 0 N.A 0 N.A 0 N.A 0 N.A
Harbour activities 6 11 3 10 23 15 7 7 4 9 4 7
Total concen. (mg m3) 15 14 25 26 1 1 2 9 22 19 4 6

NOx (%)
Industry 38 30 65 53 15 16 47 62 64 54 23 22
Traffic 43 59 24 30 66 74 28 24 24 35 65 73
Domestic heating 13 N.A 6 N.A 7 N.A 15 N.A 7 N.A 9 N.A
Harbour activities 6 11 5 17 12 10 10 14 5 11 3 5
Total concen. (mg m3) 40 38 36 41 6 3 4 12 44 43 18 21
CO (%)
Industry 7 3 7 9 1 1 7 14 12 5 3 2
Traffic 78 97 80 90 97 99 79 85 77 94 93 98
Domestic heating 15 N.A 13 N.A 2 N.A 14 N.A 11 N.A 4 N.A
Harbour activities 0 0 0 1 0 0 0 1 0 1 0 0
Total concen. (mg m3) 131 139 59 76 39 41 15 27 94 108 84 105
PM10 (%)
Industry 72 73 62 62 78 65 58 74 82 80 65 69
Traffic 14 15 2 2 18 33 11 7 6 6 14 18
Domestic heating 4 N.A 0 N.A 1 N.A 2 N.A 1 N.A 2 N.A
Harbour activities 0 0 0 0 0 0 0 0 0 0 0 0
Fugitive 10 12 36 36 3 2 29 19 11 14 19 13
Total concen. (mg m3) 14 18 51 78 3 2 2 7 22 29 3 4

activities as well as a 3D concentration field for all tribution with 3–7% and 7–11% in winter and
the considered sources. Contributions for each summer, respectively, not considering the result at
group of sources were then calculated for each Palagiano station (23–15%), due to the low total
simulated pollutant. SO2 concentrations (1 mg m3). Low or negligible
In order to evaluate the contribution of specific contributions are instead foreseen from both traffic
groups of sources at particular locations of interest, and domestic heating.
six monitoring stations were selected: three of them The NOx contributions show some differences
are located in the urban area of Taranto (Orsini, among both sources and stations. Industry and
Dante and Peripato), two in the surrounding cities traffic emissions are the two main contributors to
(Palagiano and Statte) and one in the new urban the estimated total ground concentrations with an
district of Paolo VI. Results are shown in Table 3 for average 41% and 45%, respectively. Harbour
the considered pollutants, in terms of percentage of activities contribute by 3–10% in winter and
source contribution to the total estimated ground 5–17% in summer, while domestic heating by about
concentration (expressed in mg m3) at each selected 9% on average. Results at the selected monitoring
monitoring station. Contributions from industry as a stations show some difference among source con-
whole, traffic, domestic heating, fugitive and harbour tributions. The urban stations of Dante, Orsini and
activities were considered for contribution analysis. Peripato indicate a different impact from industry
It can be shown that industrial activities are the and traffic. Whilst the Dante station, located
main contributor to the total SO2 concentrations downtown (see Fig. 1), shows a dominant traffic
(up to 97%), regardless of both station and season. contribution (43–59%), with an industry influence
Harbour activities exhibit the second larger con- up to 30–38%, the Orsini station, located between
ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6443

the industrial area and the port, exhibits a prevailing 4. Conclusions


industry impact (53–65%). The same effect can also
be observed at the Peripato urban station, located in An air pollution assessment study has been
a urban park close to Mar Piccolo, with an open conducted for the most relevant emission sources
area facing the industry facilities (e.g., maps in located in the city of Taranto. 3D SO2, NOx, CO and
Fig. 1). The urban district of Paolo VI, located primary PM10 concentration fields have been calcu-
north to Mar Piccolo, seems to be mainly affected lated by means of the SPRAY model for the studied
by industrial sources (47–62%), while traffic con- area in selected periods covering both winter and
tribution at this station can be accounted for summer seasons. The model system has been supplied
24–28% of the total NOx concentration. The town with both ground and upper air meteorological data
of Statte, located north to the industrial area (e.g., and with an emission inventory taking into account
Fig. 1), reveals an industry contribution of up to harbour activities and other local emission sources.
23%, whereas traffic is estimated to be main Results show that the impact of emitted pollu-
contributor source in this area. The rural area of tants extends to a large area of the local territory,
Palagiano (west of Taranto) also shows a dominant depending on the pollutant itself. Typical industrial
traffic influence (66–74%) but the total estimated contaminant, such as NOx and SO2, locate their
NOx concentration is very low (3–6 mg m3). The concentration peaks around the industrial areas.
harbour NOx contributions exhibit higher values at However, significant concentrations are found in
summertime due to the influence of sea breeze. In the surrounding urban and rural areas. A seasonal
particular, the Orsini station, closest to the harbour, effect on the ground concentration fields is observed
exhibits the largest values (17%), followed by Paolo to be mainly caused by the different meteorological
VI (14%). The urban stations of Dante and conditions. In particular the summer breeze seems
Peripato both show a harbour contribution at to produce a larger spread of the concentration
summertime of about 11%. The NOx summer fields for SO2, NOx and PM10 pollutants.
harbour contribution at Statte is instead about 5%. According to the SPRAY calculated NOx source
CO source contributions are dominated by traffic contributions, harbour activities are not found to be
emission (89% on average) regardless of season and the main contributor source on average. Up to 20
station. The remaining contributions can be as- and 30 mg m3 can be ascribed to this source at,
cribed to the domestic heating and industries respectively, wintertime and summertime, mainly
emissions for about the same amount. caused by the ‘‘Polisettoriale’’ quay and to a lesser
Primary PM10 concentrations can mainly be extent by the harbour closer to the city of Taranto.
attributed to industry emissions (70%). Traffic and Up to 5 mg m3 of SO2 can instead be ascribed to
fugitive emissions being the two other contributing this source. Concentrations produced by ship
sources. Among the results obtained in the con- emissions are foreseen up to 5 km inland and
sidered monitoring stations, it is worth noticing the 10 km from the coastline, depending on the season.
fugitive and traffic contributions at the Orsini Although harbour contributions are considered not
station. Here fugitive contributions of up to 36% to be relevant in proximity of the urban area, where
are predicted, while the traffic one is nearly vehicles and industrial emissions are dominant, their
negligible (2%). This result can be explained with relative importance increases in the surrounding of
its proximity to the industrial area. The urban the ‘‘Polisettoriale’’ quay where it is found to be the
stations of Dante show a traffic influence of about main emission source.
14–15% and fugitive contribution of 10–12%. Source contributions for the modelled pollutants
A lower traffic PM10 contribution (6%) is estimated have been calculated at selected monitoring sta-
at the Peripato station, in agreement with the lower tions. Results indicate a prevailing contribution of
NOx contributions predicted at the same station for industrial activities on the estimated ground con-
this source. The PM10 contribution results at centrations of SO2 and PM10. NOx contribution
Palagiano, Paolo VI and Statte stations are nearly results, point out a more diffuse responsibility, with
of the same order of magnitude or lower than other industry and traffic sources mainly involved, fol-
stations, however, the total estimated concentra- lowed by domestic heating and harbour activities
tions are significantly low (2–7 mg m3), due to with a much lower contribution. The harbour
distance from the industrial area considered to be activities have been evaluated to account for NOx
the main emission source for this pollutant. and SO2 ground concentrations with an average
ARTICLE IN PRESS
6444 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444

3–17% and 3–11%, respectively, while negligible Entec UK Limited, 2005. European Commission Directorate
contributions to CO and PM10 are foreseen. General Environment—Service contract on ship emissions:
Seasonal differences in spatial distribution of assignment, abatement and market-based instruments. Task
1—preliminary assignment of ship emission to European
harbour contribution to NOx and SO2 concentra- countries. Final Report.
tions in the studied area have also been evidenced. ETC/LC, 1997. European Topic Centre of Land Cover (ETC/
The results of this study show that in the presence LC)—European Environmental Agency (EEA). Environment
of industrial/urban emission sources, as for the Satellite Data Centre, P.O. Box 806, SE-981 28, Kiruna,
Taranto area, the harbour contributions are not Sweden.
Gariazzo, C., Pelliccioni, A., Bogliolo, M.P., Scalisi, G., 2004.
found to be the main pollutant source, but its effect Evaluation of a Lagrangian particle model (SPRAY) to assess
cannot be considered as negligible, and conse- environmental impact of an industrial facility in complex
quently, they have to be properly accounted for. terrain. Water, Air and Soil Pollution 155, 137–158.
In other specific situations, where the other main Hanna, S.R., 1982. Applications in air pollution modeling. In:
emission sources are missing, the harbour contribu- Nieuwstadt, F.T.M., Van Dop, H. (Eds.), Atmospheric
Turbulence and Air Pollutin Modelling. Reidel, Dordrecht
tions might become relevant for the local air quality, (Chapter 7).
and its emissions have to be accurately assessed. IIASA, 2001. RAINS-Europe /http://www.iiasa.ac.at/rains/
home.htmlS.
Acknowledgements Lloyd’s Register, 1995. Marine Exhaust Emissions Research
Programme. Loyd’s Register Engineering Services.
Lloyd’s Register, 1999. Marine exhaust emission quantification
The authors thank the Taranto Port Authority study—Mediterranean Sea. Prepared for European Commis-
for hosting our instruments. Special thanks to sion. DG XI Environment, Nuclear Safety and Civil
GECOM company for providing local support. protection. Report no. 99/EE/7044.
Finally we acknowledge Dr. G. De Gennaro of Nanni, A., Radice, P., Piersanti, A., 2005. TRaffic Emission
Department of Chemistry—University of Bari for Factor Improved Calculation (TREFIC). User manual—
Version 4.0, ARIANET R2005.02, Milan, Italy.
providing the emissions data. The research was Parret, F.W., 1992. Dust Emissions. Computational Mechanics
partially funded by the Italian Ministry of Health Publications, Ashurst Lodge, Ashurst, Southampton, USA.
(PMS/022/2002). Tinarelli, G., Anfossi, D., Brusasca, G., Ferrero, E., Giostra, U.,
Morselli, M.G., Moussafir, J., Tampieri, F., Trombetti, F.,
1994. Lagrangian particle simulation of tracer dispersion in
References the Lee of a schematic two-dimensional hill. Journal of
Applied Meteorology 33, 744–756.
Aria Technologies, 2001. Minerve Wind Field Models version Thomson, D.J., 1987. Criteria for the selection of stochastic
7.0, General Design Manual. ARIA Tech. Report, May 2001. models of particle trajectories in turbulent flows. Journal of
ATI, GECOM, CESI, START, 2003. Servizio di valutazione Fluid Mechanics 180, 529–556.
degli effetti dell’inquinamento atmosferico nell’area urbana di Trozzi, C., Vaccaro, R., 1998. Methodologies for estimating air
Taranto. Municipality of Taranto. Tech. Report, vol. I. pollutant emissions from ships. TECHNE Report MEET
Entec UK Limited, 2002. European Commission—Quantifica- RF98.
tion of emissions from ships associated with ship movements Van Ulden, A.P., Holtslag, A.A.M., 1985. Estimation of atmo-
between ports in the European Community. Tech. Report, spheric boundary layer parameters for diffusion application.
July 2002. Journal of Climate and Applied Meteorology 24, 1196–1207.