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
Furthermore, ship emissions have not been as
tightly controlled as many land-based emission
sources, and the international context of the
shipping sector has posed difficulties to achieve
progress in improving environmental performance.
Ship emissions can make a significant contribution
to air pollution problems in the European Commu-
nity. In particular, these are a major source
of SO2 and NOx, which leads to acidification
and eutrophication as well as formation of ground
level ozone. A recent European Commission (EC)
study (Entec, 2002) quantifies as 2.6 and
3.6 Kte year1 the SO2 and NOx emitted by ships
in EU countries in the year 2000, with a growing
rate of 3.3 and 4.6 Kte year1 up to 2010. A recent
report produced by the EC (Entec, 2005) indi-
cates to range between 50% and 1%, depending
on the pollutant and the EU country, the ratio of
ship to total emissions, as calculated by RAINS
model used for the Clean Air for Europe
programme.
According to the EC study, respectively 4.5% and
6.2% of the total SO2 and NOx emitted by ships are
due to in-port activities as maneuvering, loading/
unloading and hotelling. A high number of major
ports are located close to or in proximity of
main cities. One third of top 100 highest emission
EU ports in 2000 were located in the Mediterranean
area, with 18 of them sited in Italy relatively
close to high-urbanized areas. Ports located in
proximity to urbanized area, produce a combined
environmental effect due to the superposition of
ship emissions with those related to urban sources.
For technical and economical reasons, ports can
also be close to industrial areas. Consequently the
emission of air pollutants, mainly particulate
matter, is expected as a result of unloading of
primary goods and loading of the final products
activities.
According to the above facts any environmental
plan, conducted to improve the air quality in such
areas, needs the assessment of the relative impor-
tance of harbour activities emissions with respect to
other emission sources, to optimize the cost/effec-
tiveness of planned interventions.
On these premises, a case study has been
conducted in one of the most important Italian
ports, where harbour, industrial and urban activities
are close together, producing an ideal case to study
the interaction among these different sources, their
effects on the environment and their contributions
to the air quality.
6433
2. Methodology
2.1. Case study
The city of Taranto is one of the most popula-
ted–industrialized areas in Italy. Here, the typical
urban emissions are superimposed with the indus-
trial ones located in proximity of the city bound-
aries. A map of the studied area is depicted in Fig. 1,
which shows both a large and a zoomed view.
Among the industrial activities located here, the
biggest one in terms of both emissions quantity and
extension of working areas, is a large steel plant.
Carbon-coke transformation processes are con-
ducted here. A power plant with tall emission stacks
is also located inside this area. A significant
industrial source is also an oil refinery with
hydrocarbons transformation processes and storage
of crude and final products. It is the third most
important refinery in Italy in terms of production.
A big cement facility is also part of the emission
sources. Other smaller emission sources can be
found in this territory, related to both the main
industrial activities and the local economy. The
above industrial activities use the harbour of
Taranto to download primary goods and to delivery
final products. These tasks are often related to
emissions of particulate matter. Combustion pro-
cesses during wharf operations also produce pollu-
tants emissions.
Although emissions related to harbour activities
are not considered as the major contributing sources
for the local air quality, their overall assessment is
often unknown as well as their relative contribution
to the local air quality.
2.2. Modelling system
To conduct this assessment study, a modelling
system based on three numerical codes was used: the
MINERVE meteorological model, the SURFPRO
turbulence pre-processor and the SPRAY Lagran-
gian particles dispersion model. Three-dimensional
fields of wind and temperature have been calculated
using the diagnostic code MINERVE (Aria Tech-
nologies, 2001) applying a mass-consistent objective
analysis scheme. The SPRAY 3.0 model (Tinarelli
et al., 1994; Gariazzo et al., 2004) was used to
reconstruct concentration fields produced by pollu-
tants emitted by the considered sources. SPRAY is a
3D model that simulates air pollution dispersion
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
number of computational (fictitious) particles and
each particle is ‘‘moved’’ at each time step by
pseudo-velocities, solutions of stochastic Langevin
equations with Gaussian random forcing (Thom-
son, 1987), taking into account the three basic
dispersion components: the transport due to the
mean fluid velocity; the random turbulent fluctua-
tion of wind components; and the molecular
diffusion. Turbulent fluctuation of wind compo-
nents (su, sv, sw) is used by SPRAY to determine
the random motion causing the dispersion. The
latter are calculated by means of parameterization
codes (Hanna, 1982) based on scaling variables
derived by the SURFPRO code on the basis of the
Monin–Obukhov similarity theory and surface
energy budget evaluations (Van Ulden and Holt-
slag, 1985), starting from two-dimensional arrays of
surface parameters (albedo, bowen ratio, roughness
length z0) and ground meteorological parameters
given as input.
2.3. Field campaigns description
Two field campaigns were conducted in the winter
and summer seasons to feed the model with real
data and to validate modelling results.
To provide meteorological data to the modelling
system, various stations were available (Fig. 1).
Among the local monitoring network, five stations
(Garibaldi, Peripato, Orsini, Paolo VI and Dante)
were used, located both downtown and in the city
neighbourhoods. Pollutant concentration measure-
ments were also carried out in the above stations. In
addition, three mobile laboratories were located in
the area. The Italian Institute for Occupational
Safety and Health (ISPESL) mobile meteorological
laboratory (MML) was placed in the harbour of
Taranto. Besides the main standard meteorological
data, the MML calculates average values of
turbulence parameters starting from data collected
by sonic anemometers. Two Mobile Chemical
Meteorological Laboratories (MCML) were located
in the rural area of Palagiano, about 30 km west the
city of Taranto, and in the urban area of Statte,
nearly 15 km north of Taranto, where the impact of
pollutants emitted from the industrial areas is
expected to be detected. To get information on
meteorological parameters in the eastern and north-
ern parts of the territory, two other meteorological
stations were placed in the S. Giorgio and Monte
Mesola villages.
Upper air measurements were carried out by
means of the ISPESL SODAR/RASS system,
located close to the MML station in front of sea
and at nearly 2 km far from the industrial area.
Wind, turbulence and temperature profiles up to
400 m a.g.l. were measured. Such data were com-
pleted by wind, temperature and pressure profiles
up to 24 km obtained from radio soundings,
collected every 6 h at the city of Brindisi, 80 km
East of Taranto.
 ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444
Meteorological data collected during the field
campaigns, as well as available historical data
routinely collected by the local monitoring network,
were used to detect the main circulation patterns
occurring in the area. Results show that in summer,
high-pressure systems are active and the high
insulation produces convective atmospheric condi-
tions with a well-mixed boundary layer. At daytime,
the wind mostly blows from south-southwest
(S-SW) with speed up to 5 m s1, as a consequence
of the sea breeze triggered by the land/sea tempera-
ture gradients. It then turns to north-northwest
(N-NW) during nighttime due to a land breeze
effect. In winter, the prevailing flows are from
N-NW and south-southeast (S-SE). Cold or warm
fronts often pass through the area with synoptic
wind conditions characterized by speeds up to
8 m s1. A few episodes with persistent high pressure
systems, associated with weak pressure gradients
and low wind conditions were detected in this
season, determining conditions favouring accumu-
lation of pollutants in the lower layers.
2.4. Emission inventory
The emission inventory developed for this study
contains yearly emissions of the main chemical
species linked to the most significant anthropic
activity sectors: industrial sources, road traffic,
residential heating, harbour activities. This is an
upgraded version of an emission inventory devel-
oped for the city of Taranto in a previous study
(ATI et al., 2003).
In Fig. 2, sources emissions, considered as a
whole and grouped into SNAP categories, are
100%
80%
60%
40%
20%
0%
CO
NMVOC
NOX
PM10
Fig. 2. Sources contribution to the total emissions of some considered pollutants.
6435
shown in order to underline the contribution of
various activities to the total emissions of specific
pollutants; the road transport has the greatest
relevance, representing more than 50% of CO and
NMVOC emissions, and less than 20% for NOx
ones. Industrial sources are by far the major source
of SO2, NOx and PM10.
Data concerning stack emissions from industrial
sources were collected directly from plants owners.
Such point sources represent installations that are
relevant at the scale of the analysis and that are
known individually. Up to 380 different sources
were considered in this study. Particulate matter
fugitive emissions related to industrial activities
were also taken into account. Storage areas and
quarries were considered as possible emission
sources of re-suspended matter. Their emission
estimation is based on the methodology described
in Parret (1992). To calculate the emissions it needs
some inputs data such as: type of goods, total
amount of goods per year, extension of the
area used for storage of goods and silt content of
goods. In addition, in order to estimate the yearly
amount of re-suspended dust due to meteorological
conditions, the methodology also needs the
number of rainy days in the considered year and
the relative amount (in percent) of wind speed above
a threshold of 5.4 m s1 during the year, considered
as the level above which wind-driven dust re-
suspension takes place. According to above require-
ments, data related to the year 2002 were then
collected and used for the fugitive emissions
estimation.
Road traffic emissions are represented in two
ways: as line and as areal sources. The road network
Maritime
activities
Road
transport
Production
processes
Combustion in
industry
Domestic
heating
Combustion
Energy
SO2
 ARTICLE IN PRESS
6436 C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444
and the related vehicular flows were available for
the urban area of Taranto, for highways, and for
provincial and state roads. In all these cases
emissions are determined by means of TRaffic
Emission Factor Improved Calculation (TREFIC)
software (Nanni et al., 2005), based upon the
COPERT III methodology integrated with more
detailed emission factors for particulate from brakes
and tire wear (IIASA, 2001). Traffic emissions for
all other populated places located in the modelled
domain are treated as area sources and surrogated
from the previous ones, using the number of
registered vehicles as a proxy variable.
The emission estimation from Taranto residential
heating is based on yearly fuel consumption, while
for all other surrounding towns, emissions are
surrogated from those of Taranto, using their
number of inhabitants.
Harbour emissions from loading/unloading activ-
ities and combustion processes during wharf opera-
tions are estimated from mercantile fleet data,
average in-port idling time and type of transported
goods. Emissions linked to the quays were derived
from detailed information on goods handling. Ship
emissions take into account three different stages:
maneuvering, hotelling and cruising. Different data
are employed for the emission estimation: shipping
movements in the studied area, vessel data
(e.g. engine power, fuels) and emission factors
defined from the existing literature (Entec, 2002;
Lloyd’s Register, 1995, 1999). In the year consid-
ered, a total of 1999 movements were registered by
the Italian Statistical Institute (ISTAT) in the
harbour of Taranto. Hotelling time in the port is
evaluated to be 2.5 days, cruising time (from the
entrance to the berth) less than an hour, while the
maneuvering time is evaluated to be half an hour. It
is assumed that all ships are characterized by slow
speed diesel engine; while inside the port, an average
speed service of 6 knots is assumed. Based on these
data, Lloyd’s emission factors (Lloyd’s Register,
1995) and elaboration of fuel consumption data at
full power, ship emissions in different phases are
evaluated according to formulas reported by Trozzi
and Vaccaro (1998).
To feed the dispersion model, yearly emissions
are temporally and spatially disaggregated using
specific information related to the area. Areal
emissions from the considered activities are allo-
cated according to the spatial distributions of the
related classes of land use, while time modulation is
based on activity-specific proxies.
2.5. Models setup
A 35  35 km2 model domain was considered to
cover all possible impacts on the surrounding areas
and relevant towns. The domain has been horizon-
tally divided into 71  71 grid cells with 500 m
resolution and vertically splitted from the ground
level to the top, set to 1500 m. Surface parameters
(albedo, roughness length and Bowen ratio) were
estimated on the basis of the Corine land cover map
(ETC/LC, 1997) with a 500 m resolution. A lookup
table was used to assign to each cover class a value
of albedo, roughness length and Bowen ratio.
A total of 33 days were selected to simulate air
pollution dispersion around the studied area. Four-
teen of them were related to winter season and 19 to
summer. Their selection was based on both typical
local atmospheric circulations/synoptical conditions
in the considered season and on the occurrence of
pollutants concentration peaks clearly linked with
industrial emissions as revealed by observations at
the monitoring stations.
The 3D wind and temperature fields were
obtained by means of the MINERVE model, using
meteorological data collected during the field
campaigns. As the SODAR/RASS system provided
wind and temperature profile data up to 300 m a.g.l.
on average, the remaining part of the model domain
was not covered by data. To overcome this problem
the upper part of the atmospheric boundary layer
(from 700 m on) above the Taranto area was
considered to be equivalent to that observed at the
Brindisi station, 80 km away from the model
domain. According to this assumption, the upper
part of the model domain was covered by the
Brindisi radiosounding data. A 2D Cressman
interpolation scheme was used to spatially inter-
polate both surface and upper air data at each
model vertical level, to obtain a first guess
meteorological field. Surface wind was then cor-
rected according to both surface roughness (z0) and
topographic altitude, using a scheme based on
similarity theory. The wind field was then adjusted
to fulfil mass-consistent objective analysis con-
straints. Minimum values of sw data were set
according to data collected at both MML by sonic
anemometers and SODAR/RASS systems. The 3D
wind and temperature fields generated by the
MINERVE meteorological model, the atmospheric
turbulence produced by SURFPRO and the emission
data, were then provided as input to the SPRAY
dispersion model. 3D hourly concentrations of NOx,
 ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444
SO2, CO and primary PM10 were calculated as total
overall concentration and in terms of contributions of
specific sources.
3. Results
3.1. Comparison with monitoring data
Dispersion models are modern tools to obtain
spatially resolved concentrations, but their applic-
ability and reliability are uncertain in many cases.
Stringent validation and comparison with measure-
ments are essential before results are used to assess
the impact of emissions on the territory. To evaluate
the model performances, observed NOx, SO2, CO
and primary PM10 hourly concentrations were
compared with the corresponding modelled values.
As far as the background level of pollutants is
concerned, it should be considered that Lagrangian
particle model cannot take into account boundary
conditions, like other kind of models (e.g. Eulerian)
can do. Nevertheless, according to observations
collected during the seasonal field campaigns, the
background levels for both NOx and SO2 are found
to be at least one order of magnitude lower than
peak levels. For these pollutants, the lack of
background concentrations does not introduce
significant model underestimations, but only a slight
negative bias. Conversely, according to observa-
tions, the lack of background concentrations for CO
and PM10 could significantly introduce underesti-
mations in model predictions.
Table 1 summarizes a statistical evaluation of
model performances over all the available stations
and all simulated days. In general, model predic-
tions underestimate the observed average concen-
trations of all pollutants, but SO2, which is
overestimated by the model. Best results are
obtained for NOx, while poor performances are
obtained for both CO and primary PM10. To
partially explain these statistical results, it should
be considered that, especially for SO2 and PM10
pollutants, data collected by the local monitoring
network were not found as reliable as expected.
Therefore, some disagreements in the values of the
statistical indexes are probably due to a mismatch
between model results and observations.
A further evaluation of model results can be
obtained by comparing the time series of observa-
tions with model results. The better performance
obtained for NOx results can also be observed in the
time series comparison. Two stations were selected
6437
Table 1
Statistical evaluation of SPRAY model performance over all
available stations
C̄ p (mg m3) C̄ o (mg m3) NBIAS NMSE
NOx 26.3 44.6 0.42 1.59
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
Key: C̄ p and C̄ o are the averages over the data set for the model
predictions and the observations. NBIAS is the normalized bias,
NMSE is the normalised mean square error.
for this evaluation: one located downtown (Dante)
and the other (Orsini) in an urban district in
proximity to the industrial area. Fig. 3 shows a
comparison of observed and SPRAY modelled NOx
concentrations at Dante and Orsini monitoring
stations, for winter and summer seasons, respec-
tively. While at the Dante station the model tends to
overestimate observed values, the opposite happens
at Orsini, likely due to an incorrect evaluation of the
actual emissions. In general, the NOx model results
exhibit a good reproduction of observed values both
in time and concentration value.
The SO2 model results (not shown) in general
exhibit peaks higher than observed, sometimes with
modelled peaks not revealed by observations.
Possibly, this is due to the above-mentioned
problems on measurements reliability.
For both CO and primary PM10, model results
indicate an overall underestimation compared to
measured concentrations. For both pollutants this
result might be mainly due to the regional back-
ground contribution, which is not taken into
account by Lagrangian models, as mentioned
before. As far as PM10 is concerned, the lack of
other emission sources or an incorrect evaluation
of the emission factors, as well as the effect of
secondary pollutants, could also contribute to the
model underestimation. Consequently, results of
primary PM10 obtained in this study can only be
considered as the consequences produced by the
combined effect of both local emissions and
dispersion conditions. However, these results can
still be used for an assessment study and definition
of control strategies, as they allow to handle the key
factors (e.g. emissions), for which the local autho-
rities should have prior responsibility on control of
air quality. A more detailed study on PM would
require the use of a comprehensive chemical
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
concentrations is also clearly visible in the summer
results. A greater part of the inland territory is
influenced by the NOx concentrations, which
extends up to 15 km from the coastline. The spatial
dispersion of NOx concentrations over the sea is
also greater in summer than in winter.
According to the SPRAY calculated harbour
contributions shown in Fig. 4b, during wintertime
average values of up to 20 mg m3 are produced by
the harbour activities, mainly caused by the
‘‘Polisettoriale’’ quay, 6 km west from the city of
Taranto, and its effect spreads over the sea up to
5 km from the coastline and covers part of it. This
quay is the most important one in the harbour,
being used by a large number of container ships.
The harbour contributions from around this quay
are dominant with respect to the other contributor
sources. At summertime, a peak value of 30 mg m3
can be ascribed to the harbour activities. Its
contribution can be clearly split in two main
components: one ascribed to the ‘‘Polisettoriale’’
quay and the other to the harbour closer to the city
of Taranto, with the former being stronger than the
latter one, due to the higher number of ships
involved. The spread due to breeze effects is even
larger both over the sea (about 10 km) and inland
(about 5 km).
The SO2 winter average concentration map
(Fig. 4a) shows three peaks with average value of
25 mg m3, located close to industrial area and in the
urban area of Taranto. During summertime, the
average concentration map shows a maximum
located, as for winter results, in the industrial area,
with values of up to 60 mg m3. Concentrations
between 10 and 20 mg m3 are predicted over the
urban area of Taranto. As for NOx results, the
spatial distribution of summer and winter SO2
concentrations is rather different due to different
meteorological conditions. According to the
SPRAY model estimations, both winter and sum-
mer harbour contributions are not more than
5 mg m3, mainly due to the ‘‘Polisettoriale’’ quay
emissions.
The winter and summer CO concentration maps
(not shown) show a traffic related origin, with peaks
located in the urban areas and close to the
connecting roads. Summer concentration values
are foreseen to be higher than winter (200 vs.
100 mg m3) due to the frequent summer time low
wind speed conditions. Concentrations are confined
around the emission areas as a consequence of their
emission at ground level.
Primary PM10 seasonal concentration maps
(Fig. 4a) exhibit an impact area restricted to the
industrial district and its surrounding with an
extension of 4  4 km2. The city of Taranto is only
partially influenced by this kind of emission.
According to the model prediction, the summer
concentration is higher (120 mg m3) than in winter
(70 mg m3), due to the low dispersion conditions
mentioned above.
 ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444 6441

In order to get information on air quality
predicted by SPRAY at particular locations and
its compliance with the European legislation, daily
and seasonally averaged pollutant concentrations,
as well as their maximum hourly values, were
calculated for selected stations. Results are shown
in Table 2. The EU mother directive 96/62/CE and
the two daughter directives 99/30/CE and 2000/69/
CE, establish the air quality standards for the NOx,
SO2, CO and particulate pollutants. As far as the
model results for SO2 are concerned, the daily air
quality standard (150 mg m3) is fulfilled. The hourly
one is close to the limit value of 350 mg m3 at Orsini
station (263 mg m3 at wintertime), due to its
position close to the industrial areas. Critical values
are predicted for hourly NOx concentrations at
Orsini, Dante and Peripato stations, with values
either close or above the limit of hourly concentra-
tion (200 mg m3) given for human health protec-
tion. Hourly concentrations close to the legal limit
are predicted at summertime in the small town of
Statte (210 mg m3), mainly due to pollutants
transported by sea breeze. Although annual con-
centrations cannot be obtained from the results of
this study, the calculated SO2 and NOx seasonal
concentrations, indicate a possibility that annual
concentrations might be close to or higher than the
legal limits at Orsini, Dante and Peripato stations.
The station of Orsini and its district of Tamburi
may also be affected by daily PM10 concentrations
higher than the corresponding legal limit of
50 mg m3. Its proximity to the industrial area is
the main reason for this result. The station of
Peripato seems also to be affected by high daily
PM10 concentrations (50 and 63 mg m3 maximum
daily concentrations at winter and summer, respec-
tively). The annual legal limit value of 40 mg m3 is
probably overcame at the Orsini station, according
to the seasonal concentrations derived in this study.
The CO concentrations are below the legal limits,
although contributions from background concen-
trations have not been accounted for in this study
due to model limitations.
3.3. Evaluation of sources contribution
In a Lagrangian particle dispersion model, the
airborne pollutant is simulated by means of a
certain number of ‘virtual’ or ‘pseudo’ particles,
each of them representing a certain pollutant mass
emitted by a specific source. For any particle, the
model knows both the source where it was emitted
from and the mass of the pollutant it accounts for.
So the concentration of a pollutant produced in a
model cell by a specific source or group of them, can
be easily calculated by counting the number of
particles which can be ascribed to the selected
source, summing up their mass contribution and
divide it by the cell volume. In practice, the model
was set up to calculate a 3D pollutants concentra-
tion field for groups of different sources such
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
the considered sources. Contributions for each
group of sources were then calculated for each
simulated pollutant.
In order to evaluate the contribution of specific
groups of sources at particular locations of interest,
six monitoring stations were selected: three of them
are located in the urban area of Taranto (Orsini,
Dante and Peripato), two in the surrounding cities
(Palagiano and Statte) and one in the new urban
district of Paolo VI. Results are shown in Table 3 for
the considered pollutants, in terms of percentage of
source contribution to the total estimated ground
concentration (expressed in mg m3) at each selected
monitoring station. Contributions from industry as a
whole, traffic, domestic heating, fugitive and harbour
activities were considered for contribution analysis.
It can be shown that industrial activities are the
main contributor to the total SO2 concentrations
(up to 97%), regardless of both station and season.
Harbour activities exhibit the second larger con-
tribution with 3–7% and 7–11% in winter and
summer, respectively, not considering the result at
Palagiano station (23–15%), due to the low total
SO2 concentrations (1 mg m3). Low or negligible
contributions are instead foreseen from both traffic
and domestic heating.
The NOx contributions show some differences
among both sources and stations. Industry and
traffic emissions are the two main contributors to
the estimated total ground concentrations with an
average 41% and 45%, respectively. Harbour
activities contribute by 3–10% in winter and
5–17% in summer, while domestic heating by about
9% on average. Results at the selected monitoring
stations show some difference among source con-
tributions. The urban stations of Dante, Orsini and
Peripato indicate a different impact from industry
and traffic. Whilst the Dante station, located
downtown (see Fig. 1), shows a dominant traffic
contribution (43–59%), with an industry influence
up to 30–38%, the Orsini station, located between
 ARTICLE IN PRESS
C. Gariazzo et al. / Atmospheric Environment 41 (2007) 6432–6444
the industrial area and the port, exhibits a prevailing
industry impact (53–65%). The same effect can also
be observed at the Peripato urban station, located in
a urban park close to Mar Piccolo, with an open
area facing the industry facilities (e.g., maps in
Fig. 1). The urban district of Paolo VI, located
north to Mar Piccolo, seems to be mainly affected
by industrial sources (47–62%), while traffic con-
tribution at this station can be accounted for
24–28% of the total NOx concentration. The town
of Statte, located north to the industrial area (e.g.,
Fig. 1), reveals an industry contribution of up to
23%, whereas traffic is estimated to be main
contributor source in this area. The rural area of
Palagiano (west of Taranto) also shows a dominant
traffic influence (66–74%) but the total estimated
NOx concentration is very low (3–6 mg m3). The
harbour NOx contributions exhibit higher values at
summertime due to the influence of sea breeze. In
particular, the Orsini station, closest to the harbour,
exhibits the largest values (17%), followed by Paolo
VI (14%). The urban stations of Dante and
Peripato both show a harbour contribution at
summertime of about 11%. The NOx summer
harbour contribution at Statte is instead about 5%.
CO source contributions are dominated by traffic
emission (89% on average) regardless of season and
station. The remaining contributions can be as-
cribed to the domestic heating and industries
emissions for about the same amount.
Primary PM10 concentrations can mainly be
attributed to industry emissions (70%). Traffic and
fugitive emissions being the two other contributing
sources. Among the results obtained in the con-
sidered monitoring stations, it is worth noticing the
fugitive and traffic contributions at the Orsini
station. Here fugitive contributions of up to 36%
are predicted, while the traffic one is nearly
negligible (2%). This result can be explained with
its proximity to the industrial area. The urban
stations of Dante show a traffic influence of about
14–15% and fugitive contribution of 10–12%.
A lower traffic PM10 contribution (6%) is estimated
at the Peripato station, in agreement with the lower
NOx contributions predicted at the same station for
this source. The PM10 contribution results at
Palagiano, Paolo VI and Statte stations are nearly
of the same order of magnitude or lower than other
stations, however, the total estimated concentra-
tions are significantly low (2–7 mg m3), due to
distance from the industrial area considered to be
the main emission source for this pollutant.
6443
4. Conclusions
An air pollution assessment study has been
conducted for the most relevant emission sources
located in the city of Taranto. 3D SO2, NOx, CO and
primary PM10 concentration fields have been calcu-
lated by means of the SPRAY model for the studied
area in selected periods covering both winter and
summer seasons. The model system has been supplied
with both ground and upper air meteorological data
and with an emission inventory taking into account
harbour activities and other local emission sources.
Results show that the impact of emitted pollu-
tants extends to a large area of the local territory,
depending on the pollutant itself. Typical industrial
contaminant, such as NOx and SO2, locate their
concentration peaks around the industrial areas.
However, significant concentrations are found in
the surrounding urban and rural areas. A seasonal
effect on the ground concentration fields is observed
to be mainly caused by the different meteorological
conditions. In particular the summer breeze seems
to produce a larger spread of the concentration
fields for SO2, NOx and PM10 pollutants.
According to the SPRAY calculated NOx source
contributions, harbour activities are not found to be
the main contributor source on average. Up to 20
and 30 mg m3 can be ascribed to this source at,
respectively, wintertime and summertime, mainly
caused by the ‘‘Polisettoriale’’ quay and to a lesser
extent by the harbour closer to the city of Taranto.
Up to 5 mg m3 of SO2 can instead be ascribed to
this source. Concentrations produced by ship
emissions are foreseen up to 5 km inland and
10 km from the coastline, depending on the season.
Although harbour contributions are considered not
to be relevant in proximity of the urban area, where
vehicles and industrial emissions are dominant, their
relative importance increases in the surrounding of
the ‘‘Polisettoriale’’ quay where it is found to be the
main emission source.
Source contributions for the modelled pollutants
have been calculated at selected monitoring sta-
tions. Results indicate a prevailing contribution of
industrial activities on the estimated ground con-
centrations of SO2 and PM10. NOx contribution
results, point out a more diffuse responsibility, with
industry and traffic sources mainly involved, fol-
lowed by domestic heating and harbour activities
with a much lower contribution. The harbour
activities have been evaluated to account for NOx
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
contributions to CO and PM10 are foreseen.
Seasonal differences in spatial distribution of
harbour contribution to NOx and SO2 concentra-
tions in the studied area have also been evidenced.
The results of this study show that in the presence
of industrial/urban emission sources, as for the
Taranto area, the harbour contributions are not
found to be the main pollutant source, but its effect
cannot be considered as negligible, and conse-
quently, they have to be properly accounted for.
In other specific situations, where the other main
emission sources are missing, the harbour contribu-
tions might become relevant for the local air quality,
and its emissions have to be accurately assessed.
Acknowledgements
The authors thank the Taranto Port Authority
for hosting our instruments. Special thanks to
GECOM company for providing local support.
Finally we acknowledge Dr. G. De Gennaro of
Department of Chemistry—University of Bari for
providing the emissions data. The research was
partially funded by the Italian Ministry of Health
(PMS/022/2002).
References
Aria Technologies, 2001. Minerve Wind Field Models version
7.0, General Design Manual. ARIA Tech. Report, May 2001.
ATI, GECOM, CESI, START, 2003. Servizio di valutazione
degli effetti dell’inquinamento atmosferico nell’area urbana di
Taranto. Municipality of Taranto. Tech. Report, vol. I.
Entec UK Limited, 2002. European Commission—Quantifica-
tion of emissions from ships associated with ship movements
between ports in the European Community. Tech. Report,
July 2002.
Entec UK Limited, 2005. European Commission Directorate
General Environment—Service contract on ship emissions:
assignment, abatement and market-based instruments. Task
1—preliminary assignment of ship emission to European
countries. Final Report.
ETC/LC, 1997. European Topic Centre of Land Cover (ETC/
LC)—European Environmental Agency (EEA). Environment
Satellite Data Centre, P.O. Box 806, SE-981 28, Kiruna,
Sweden.
Gariazzo, C., Pelliccioni, A., Bogliolo, M.P., Scalisi, G., 2004.
Evaluation of a Lagrangian particle model (SPRAY) to assess
environmental impact of an industrial facility in complex
terrain. Water, Air and Soil Pollution 155, 137–158.
Hanna, S.R., 1982. Applications in air pollution modeling. In:
Nieuwstadt, F.T.M., Van Dop, H. (Eds.), Atmospheric
Turbulence and Air Pollutin Modelling. Reidel, Dordrecht
(Chapter 7).
IIASA, 2001. RAINS-Europe /http://www.iiasa.ac.at/rains/
home.htmlS.
Lloyd’s Register, 1995. Marine Exhaust Emissions Research
Programme. Loyd’s Register Engineering Services.
Lloyd’s Register, 1999. Marine exhaust emission quantification
study—Mediterranean Sea. Prepared for European Commis-
sion. DG XI Environment, Nuclear Safety and Civil
protection. Report no. 99/EE/7044.
Nanni, A., Radice, P., Piersanti, A., 2005. TRaffic Emission
Factor Improved Calculation (TREFIC). User manual—
Version 4.0, ARIANET R2005.02, Milan, Italy.
Parret, F.W., 1992. Dust Emissions. Computational Mechanics
Publications, Ashurst Lodge, Ashurst, Southampton, USA.
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
the Lee of a schematic two-dimensional hill. Journal of
Applied Meteorology 33, 744–756.
Thomson, D.J., 1987. Criteria for the selection of stochastic
models of particle trajectories in turbulent flows. Journal of
Fluid Mechanics 180, 529–556.
Trozzi, C., Vaccaro, R., 1998. Methodologies for estimating air
pollutant emissions from ships. TECHNE Report MEET
RF98.
Van Ulden, A.P., Holtslag, A.A.M., 1985. Estimation of atmo-
spheric boundary layer parameters for diffusion application.
Journal of Climate and Applied Meteorology 24, 1196–1207.