AtmosphericPollutionResearch5(2014)648Ͳ655 

Atm spheric Pollution Research

www.atmospolres.com 

Compilation of a road transport emission inventory for the Province of

Turin: Advantages and key factors of a bottom–up approach
LucaPallavidino1,RossellaPrandi1,AlessandroBertello2,ElisaBracco2,FrancescoPavone2
1
SimulariaSrl,viaPrincipeTommaso39,Torino,Italy
2
ProvinciadiTorino,ServizioQualitàdell’AriaeRisorseEnergetiche,C.soInghilterra7,Torino,Italy

ABSTRACT 
Traffic is known to be a major source of air pollution, especially in urban areas and it is widely accepted that
effectivenessofmeasuresputintoforcetoimproveairqualitywasinthepastlimitedbyapartialknowledgeofroad
transportemissions.Inthisworkwehaveperformedacomparisonbetweentheroadtransportpollutantemissionsof
Province of Turin calculated with a bottom–up approach and the corresponding emissions of the official Piedmont
Inventory,basedmainlyonatop–downapproach.Thebottom–upinventorywasobtainedusingtheoutputofatraffic
modelreferredonlytoprivatevehiclesandintegratingitwithtrafficsurveydataandmobilityreportstudies.Wewere
abletohighlightthekeyfactorsliableforinfluencingtheresults.Thetrafficsurveyscanchangethecontributionofthe
vehicularcategorieswhile statistically–basedannual mileagesarecrucialfordeterminingtheapportionmentamong
Copert categories and the proxy variable used to estimate urban diffuse emissions (traffic models do not fully CorrespondingAuthor:
reproducetheamountoftrafficflows)exertsagreatinfluenceonthetotalamountofemissionsandonthespatial
distribution. In bottom–up inventory we have obtained a different apportionment of emissions among vehicular
Luca Pallavidino
:+39Ͳ011Ͳ19717302
categories: in example, for urban roads the CO2 of passenger cars has risen from 52.7% in top–down inventory to
:+39Ͳ011Ͳ19717317
79.5%inbottom–up.Totalemissionsofroadtransporthavesignificantlyreducedinbottom–upinventorycompared
:l.pallavidino@simularia.it
to top–down (NOX–16%, CO–41%, NMVOC–66%) and the emissions of Turin have reduced more than in the other
municipalities(CO2–24%,NOX–41%,CO–53%,NMVOC–71%).Wehavefoundthat,ifnotunfeasibleforlackofdata,the 
bottom–upmethodologyshouldbepreferredsinceitallowsamorestraightforwardandtransparentchoiceofinput ArticleHistory:
parameters. Received:02December2013
 Revised:14May2014
Keywords:Emissioninventory,roadtransport,top–down,airquality Accepted:16May2014

doi:10.5094/APR.2014.074

1.Introduction
 ployingafullbottom–upmethodologyandacomparisonbetween
Emission inventories are the basic information needed to
guarantee a good understanding of ambient air pollution, to plan
airqualitymeasuresandestimatingtheirefficiency.Inventoriesare
usually prepared using two different approaches, top–down
(Progiou and Ziomas, 2011) and bottom–up (Wang et al., 2008;
Borge et al., 2012), that sometimes can complement each other
(Beevers et al., 2012). The top–down methodology is the most
effectiveforthecompilationofnationaldatabasesofemissionsas
suggestedbyEMEP/EEAInventoryGuidebook(EEA,2009),sinceit
allows to obtain a geographically complete and methodologically
homogeneous set of information. Starting from the national
inventory, it is then possible to estimate emissions with a higher
spatial resolution using proxy variables like, in the case of road
transport (Borrego et al., 2000; EEA, 2009), the resident
population, the fuel sold, the lengths of roads, etc. The choice of
proxies proves often to be demanding, since it can easily causes
the under/overestimation of emissions in different areas and the
resultinginaccuracyinairqualitymanagement.
 (Pallavidino et al, 2011), in the framework of AERA project we
When it comes to air pollution assessment in local action
plans,infact,thespatialandtemporalresolutionofroadtransport
emissionsisveryimportantandthebottom–upapproachbecomes
preferable (Baldasano et al., 2010). This methodology requires
moredetailedinputactivitydataandlongertimesareexpectedfor
thecompilation,butitisrecommendedinparticularforurbanair
quality assessment, where road transport is the most relevant
source of pollution and traffic flows show high differences from
areatoarea.
The lack or inaccuracy of input data can pose a risk in emͲ
bottom–up and top–down is always recommendable (Borrego et
al.,2000;EEA,2009).

In Section 3 we present the comparison between the official
emissioninventoryofroadtransportcompiledbyPiedmontRegion
fortheterritoryoftheProvinceofTurinusingmainlyatop–down
approach (Regione Piemonte, 2008) and the road transport
emissions that we have calculated for the same territory with a
bottom–up approach, starting from traffic flow data, on site
surveysandmobilityreportstudies.

The Province of Turin territory, approximately 6800km2,
offers the opportunity to highlight strengths and weaknesses of
thetwoapproachesinamixtureofruralandindustrialareas,with
a spread out urban agglomerate (Turin metropolitan area) in the
centre.

Expanding a previous work on the sole metropolitan area
applied the bottom–up methodology to the full road network of
the Province of Turin with the aim to obtain a more accurate
estimate of the local road transport emissions relying, when
available, on detailed and updated information. As a side
consequence, the comparison between the two methodologies
(top–down vs. bottom–up) could also be analyzed from the point
ofviewofthesensitivityofinputparameters.



©Author(s)2014.ThisworkisdistributedundertheCreativeCommonsAttribution3.0License.
 Pallavidino et al. – Atmospheric Pollution Research (APR)

2.MethodologyandData
 vehicles(LDV)andheavydutyvehicles(HDV)onthebasisoftraffic
2.1.Bottom–upapproach
 motorways(seetheSM,TableS2).
The calculation of emissions in the bottom–up approach has
been performed by means of Trefic (Nanni et al., 2009; Denby,
2011),asoftwarethatimplementsCopertIVemissionfactors(EEA,
2009) for most of pollutants and emission factors provided by
IIASA (Klimont et al., 2002) for PM10 and PM2.5. Since the IIASA
emission factors include exhaust and wear emissions but do not
take into account the resuspension of PM10, this contribution has
been calculated by using the EPA formula published in 2011 (U.S.
EPA, 2011) that takes into account the average weight of
circulating fleet, the surface silt loading of the street and the
numbers of hours of precipitation (see the Supporting Material,
SM, S1). In Copert IV methodology the emission factors vary
dependingontheslopeoftheroad,themeantravelspeedandthe
vehiclecategory,identifiedbythevehicletype(passengercar,light
duty), the vehicle technology (Euro 1, Euro 2), the engine disͲ
placement,thevehicleweight,andfuel.Givenaroadnetworkand
the other requested input data, for each link Trefic can calculate
emissionfactorsforallCopertclasses.
 themasweightingfactors.
Fortransportanalysis,ProvinceofTurinmakesuseofatraffic
model describing private traffic flows on a detailed road network
consisting of 13690 links (C.S.S.T., 2001), which was recently
updated (Torriani et al., 2012) considering the latest available
mobility studies for the assessment of Origin–Destination matrix
(Levinsonetal.,2009),andnewtrafficmeasurementsforabetter
calibration(seetheSM,FigureS1).
 
Intheroadnetwork,motorwaysaretheminorityoflinks(with
atotalof621km),whileurbanroadssumupto6331kmandrural
roads to 2232 km. Inside the metropolitan area, the motorway
linksarepartofaRingRoadthatenclosesTurinonthewest.The
Origin–Destination matrix takes into account the mobility of
peopleamong636zonesand1493963one–waymovementsina
meanworkingday,describingahighfractionofthewholemobility
since2200000peopleresideintheProvinceofTurin.TheOrigin–
Destinationmatrixrefersonlytopeopleusingaprivatevehicle,so
that traffic flows calculated by the traffic model can be assigned
onlytoprivatecarsandtwo–wheelers.
 
In order to evaluate the yearly traffic from the working day
traffic and to split the private traffic between cars and two–
wheelers it was necessary to estimate the factors to apply to the
modeldata,derivingthemfromtheanalysisoflocaltrafficcounts
(fortwo–wheelersfactorsseetheSM,TableS1).
 
TheemissionsofpublictransportbuseshavenotbeenconsidͲ
ered as included in the traffic flow data and have been treated
separately, since the local public transport company made
availableaccuratedataaboutthefuelconsumption,Copertclasses
andtheaveragetravellingspeed(GTT,2008).
 of the kth municipality; POPk is the population of the kth
The traffic model doesn’t give any information about comerͲ
cialvehicles.Fortheassessmentofcommercialtrafficflowsmore
than 390 traffic measurements were used. On the motorways,
where traffic data are accurate and spatially dense, a rough
conservation of traffic flows of commercial vehicles could be
considered,whileforurbanroadswefoundoutanaverage9.2%of
AverageDailyTraffic(ADT)forcommercialvehicleswithrespectto
private vehicles. Rural roads exhibit a variety of behaviors,
depending on many factors (proximity to a motorway or an
industrialarea,accesstoaborderingtunnel,etc.).Onthebasisof
traffic counts, it was assumed a fixed percentage of commercial
vehicles flow with respect to private vehicles ADT with three
options: 10% inside Turin Metropolitan Area and 15% or 5%
outside it. The map reporting the percentages of commercial
vehicles circulating on the streets of the Province of Turin is
representedinFigureS2(seetheSM).
649
Thecommercialvehicleflowswerefurthersplitintolightduty
surveys for rural and urban roads and of payment toll data for

After the estimation of traffic flows of two–wheelers,
passenger cars, light duty vehicles and heavy duty vehicles, each
value had to be divided into Copert classes identified by fuel,
environmental Euro category, engine displacement, and vehicle
weightinordertoapplytheproperemissionfactors.Thissplitting
procedure must be carefully performed because it is known to
have a great influence on calculated emissions. In principle, the
composition of the circulating fleet could be simply derived from
thevehiclePublicRegisterdatawithoutanymanipulation,butfor
amoreaccurateestimateofemissionsitisconvenienttotakeinto
accountthatsomeCopertclasses,usuallytheoldest,travelmuch
less kilometers than others and the relative contribution on the
vehicle category emissions is much lower (Andre et al., 1999;
Caserini et al, 2007a; LAT/AUTh, 2008; Wang et al., 2008). One
common way to consider the “age factor” is to define a set of
yearlymileagesdrivenonaveragebyeachCopertvehicleonurban
roads and rural/motorways, based on statistical data and using

In this study, we applied annual mileages estimated by
Caserinietal.(2011,2013)tothevehicleregisterdataofPiedmont
Region for year 2008. The mileage, divided between urban and
rural/motorways,givestwodifferentfleetcompositions,shownin
the Supporting Material in terms of environmental “Euro”
categories(seetheSM,FiguresS3andS4).
Since the road network does not account for all roads, CO2
emissions calculated from traffic flow data don’t match CO2
emissions estimated by official Piedmont Region Inventory in the
ProvinceofTurinonthebasisofthefuelsold.ThemissingCO2was
chosentobeallocatedtourbandiffuseemissionsandsplitamong
allmunicipalitiesusingthepopulationasproxyvariable(Carlettiet
al., 2013). The amount of pollutants associated to urban diffuse
emissions was calculated on the basis of specific ratios derived
from the amounts of CO2 and other pollutants already calculated
for all the urban roads of network. The procedure is simply
describedbythreeequations:
CO2diff=CO2tot–CO2lin (1)

CO2diff,k=CO2diffxPOPk (2)

POLLi,k=CO2diff,kxPOLLlin–urb,i/CO2lin–urb (3)
where,CO2diffisthequantityofdiffuseCO2,notdirectlyassociated
to the road network; CO2 tot is the quantity of CO2 associated in
PiedmontInventory(IREA)toroadtransport;CO2linisthequantity
ofCO2calculatedwiththebottom–upapproachforthewholeroad
network;CO2diff,kisthequantityofCO2associatedtodiffusetraffic
th
municipality;POLLi,kisthequantityofpollutanti (NOX,SO2,PM10,
th
etc.) associated to the k  municipality; POLLlin–urb, i is the quantity
of pollutant ith emitted from all urban links in the whole road
network; CO2lin–urb is the quantity of CO2 emitted from all urban
linksinthewholeroadnetwork.

2.2.Piedmontregioninventory

The emission inventory about road transport that we obtain
using the bottom–up approach was compared with the official
inventory compiled by the local authority, the Piedmont Region
(RegionePiemonte,2008).

The Piedmont Region Inventory was compiled using Inemar
database (Arpa Lombardia, 2008), which combines both bottom–
up and top–down approaches mainly referring to Copert
 Pallavidino et al. – Atmospheric Pollution Research (APR) 650

IV/Corinair emission factors. In the case of road transport, a and 2 summarize emissions estimated using the two methodolͲ
bottom–up perspective is adopted to assess the emissions of ogies.
motorwaysandmainruralroadswhilethetop–downapproachis 
usedtoestimateurbanemissions. The road network emissions of CO2 estimated by means of
 Treficsoftware(seeTable1)accountfor62.9%ofCO2emissionsof
Motorways and rural roads are described by a road network road transport calculated from fuel sold. Residual CO2 emissions
consistingofabout1700linksintheterritoryofProvinceofTurin were assigned to urban traffic and distributed among municiͲ
(see the SM, Figure S7), with traffic flows separated into two palitiesusingtheresidentpopulationasaproxyvariable.
classes: light vehicles (including two–wheelers and passengers 
cars) and heavy vehicles (including light duty vehicles and heavy When we compare total emissions (last rows in Tables 1 and
dutyvehicles). 2), we see that larger discrepancies arise for all pollutants: given
 that the total CO2 is equal by construction as stated before, the
For the emissions evaluation, Inemar database requires a new inventory accounts only for 83.8% of old NOX emissions,
circulatingfleetasinputdata.Thisonewascompiledbyweighting 59.0% of CO, 34.1% of NMVOC, 95.3% of PM10 (including resusͲ
thefleet,asresultsfromthe2007PublicRegister,withtheannual pension).
mileagestypicallydrivenbyeachCopertvehicleonurban,ruralor 
motorway roads, as results from an older study (Caserini et al., The SO2 values are not readily comparable because in Trefic
2007b) than the one used in bottom–up inventory (see the SM, the sulfur content is 40ppm for both gasoline and diesel, in
FiguresS5andS6).Aftertheestimationoflinearemissionsandof agreement with Copert methodology, while in Piedmont Region
thecorrespondingfuelconsumption,thefuelassignedtoresidual Inventorythesulfurcontentis130ppmforgasolineand300ppm
urban emissions was calculated from the balance of consumption fordieselaccordinglywitholderstandardoffuels(EEA,2009).
from the fuel sold in Piedmont Region (Ministry of Economic 
Development, 2007). Each fuel type was then allocated to all If we compare emissions calculated with a full bottom–up
municipalities using the fleet data weighted as described above approachinthetwoinventoriesforthemotorway/ruralnetworks
andtheconsumptionfactor. (first row) we see that the new inventory reconstructs 137% of
 linearCO2emissionsofPiedmontRegionInventory,becauseofthe
Inemar database can then calculate the emissions of each moredetailednetworkemployedinthisstudy.Acloserinspection
pollutant with the detail of Copert class and municipality. The of specific pollutant ratios (see Table 4, first row) shows a good
inventory of Province of Turin was obtained by the sum of agreement between the two inventories in the road network
emissionsofallmunicipalitiesbelongingtoit. emissions estimate, except for the exhaust, abrasion, wear PM10
 contributionduetothechoiceofdifferentemissionfactors(IIASA
Since 2007 in the inventory of Piedmont Region the vs.Copert).
particulate emissions due to resuspension have been evaluated 
applying a procedure outside the Inemar database and based on Greaterdifferencesappearintheestimateofurbanemissions
theformulapublishedbyU.S.EPA(U.S.EPA,2006),themostup– (see Table4, second row): apart from ammonia, the top–down
to–date when the inventory was compiled. Almost all parameters inventory displays higher pollutant ratios than the bottom–up
of this formula are the same of more recent one used in the inventory,especiallyforNMVOC(almostthreetimesasmuch)and
bottom–up inventory, but the different values of coefficients and PM10(twiceasmuch).
exponents produce different calculated amounts of resuspended 
particles(seetheSM,S2). Differentratiosforpollutants,giventhesameemissionfactors
 usedforeachCopertclass(ing/km),arecausedbydifferencesin
3.ResultsandDiscussion the input data: road networks (flows and/or shares of light and
 heavyvehicles),annualmileagefactorsandregisteredfleets.Inthe
Road transport emissions obtained in this study using a following sections we investigate the role of input parameters in
bottom–up approach and emissions reported in Piedmont InvenͲ order to test the sensitivity of the emission estimate and to
toryshowsomeconsiderabledifferencesforallpollutants.Table1 increasetheaccuracyoffutureprojections,extremelyrelevantfor
theassessmentofairqualityscenarios.

Table1.Pollutantemissionscalculatedusingthebottom–upapproach.Fleetregisteredin2008andnewannualmileagedata
 CO2(kt/y) CO(t/y) NMVOC(t/y) NH3(t/y) NOX(t/y) PM10(t/y) SO2(t/y)
Roadnetworkemissions/Motorways
1971 11293 921 229.7 7635 521 48.1
andruralroadsonly
Roadnetworkemissions/Urbanroads 469 4994 593 64.3 1374 109 11.5
GTTpublicbusesroadnetwork
65 176 89  626 23 0.1
emissions
Urbandiffuseemissions 1373 16042 1734 210 4089 328 33.6
Resuspension      2218 
Totalroadtransportemissions
3878 32505 3337 504 13724 3199 93.3
Bottom–upapproachusedinthisstudy

Table2.PollutantemissionsreportedinPiedmontRegionInventory.Fleetregisteredin2007andoldannualmileagedata
 CO2(kt/y) CO(t/y) NMVOC(t/y) NH3(t/y) NOX(t/y) PM10(t/y) SO2(t/y)
Roadnetworkemissions/Motorways
1435 9848 711 217 5637 478 248
andruralroadsonly
Urbandiffuseemissions 2443 45189 8987 199 10733 1167 402
Resuspension      1712 
Totalroadtransportemissions
3878 55037 9769 416 16370 3357 650
PiedmontRegionInventory
 Pallavidino et al. – Atmospheric Pollution Research (APR)


A detailed comparison of emission factors expressed in
quantity of pollutant emitted (grams) per quantity of fuel used
(kilograms)isreportedintheSM(TableS3)separatingeachSNAP
class(SelectedNomenclatureforsourcesofAirPollution).TableS3
presents the emission factors obtained in this work with the
bottom–upapproachandthoseofPiedmontRegionInventory,and
we also add those of national inventory relative to Province of
Turin to validate the estimate (ISPRA, 2010). The analysis was
focusedonlytopollutantscalculatedinalltheinventorieswiththe
CopertIVemissionfactors(CO,NOX,NH3,NMVOC)anditstrengthͲ
ens what isexplained in the following sections: the differences in
the circulating fleet, mainly caused by the considered annual
mileage, produce different emission factors in particular for
passengercarsinurbanroads.
 approach we used, the uncertainty of the annual mileages data
3.1.Circulatingfleet
 categories with moderate influence in the emission estimations.
As stated above, circulating fleet was defined by applying
weightingfactors(averageannualmileages)onregisteredvehicles
data. Piedmont Inventory and the bottom–up inventory use
differentcirculatingfleetsasinputdata,bothintermsofregistered
vehiclesdata(2007vs.2008)andannualmileages(Caserinietal.,
2007bvs.Caserinietal.,2011).Whileasingleyearvariationinthe
registered fleet doesn’t introduce a substantial effect in the
circulating fleet and consequently on its environmental
performance, substantial changes in the annual mileage factors
tend to increase or decrease distances driven by the oldest
vehicles, often the most polluting vehicles in the registered fleet,
possiblytriggeringconsiderabledifferencesinthetotalamountof
emissionsrelatedtoroadtransport.
 rences in the apportionment of CO2 emissions, and then of fuel
In order to quantify the influence of average mileage on
emissions,wehaveperformedanadditionalbottom–upestimation
ofemissionsusingthesameinputdataasregardstrafficflowsand
registered vehicles data (2008) as done before in the bottom–up
approach, but applying the old annual mileages of Copert classes
as employed in Piedmont Inventory to calculate the circulating
fleet.
 Moving on the CO2 splitting among vehicular types, motorͲ
Table3 reports the downgraded bottom–up emissions: the
comparison with emissions of Table1 clearly shows that the
circulating fleet definition (in terms of Copert classes) is a crucial
factorintheestimationofCOandNMVOC(whoseemissionfactors
stronglydependonthevehicleage),withdifferencesoftheorder
of 50%. NH3 and NOX emissions are less affected (order of 15%),
while other pollutants are only slightly influenced. Similar concͲ
lusions can be drawn by looking at pollutant ratios (see Table4),
with values now much closer to the Piedmont Inventory (top–
down),especiallyinthecaseofurbandiffuseemissions.

3.2.Urbantraffic
 inventory.Sincethelatteriscoherentwithavailabletrafficsurveys
Another important discrepancy between the bottom–up
approach and the top–down methodology used in Piedmont
Inventoryistheestimationofdiffuseurbanemissions.Infactthis
feature influences apportionment of fuel consumption among
vehicularcategoriesandCopertclasses,whicharecharacterizedby
highlydifferentemissionfactors.

Table3.Pollutantemissionscalculatedusingthebottom–upapproach.Fleetregisteredin2008andoldannualmileagedata
 CO2(kt/y) CO(t/y)
Roadnetworkemissions/Motorways
1979 13965
andruralroadsonly
Roadnetworkemissions/Urbanroads 470 8691
GTTpublicbusesroadnetwork
65 176
emissions
Urbandiffuseemissions 1364 25754
Resuspension  
Totalroadtransportemissions 3878 48586
651
In the bottom–up approach we used, urban road emissions
are calculated on the basis of flow data and the splitting into
vehicle types is a transparent procedure, derived from measureͲ
ments and local evaluations. The annual mileage information was
onlyusedtosplitthesinglevehiclecategoriesintoCopertclasses.
On the contrary, in the top–down methodology urban roads
emissions were calculated apportioning fuel consumption among
Copert classes using the registered fleet weighted with urban
annual mileage: vehicular categories (two–wheelers, cars, LDV,
HDV) compete in the apportioning of fuel consumption and the
choiceofaverageannualmileagesofCopertclasses,reasonablein
principle when looking at a single category, can result in an
undesired increase of distances yearly covered by a certain
category in a urban context. In other words, in the bottom–up
onlyaffectstheallocationofCopertclassesintothesinglevehicle
Onthecontrary,inthetop–downapproach,theuncertaintyofthe
annual mileages directly influences flow data and their splitting
into vehicle categories and Copert classes and may cause
overestimationofpollutantemissions.

In order to evidence the differences between two methodolͲ
ogies we analyzed the different apportionment of CO2 (that is a
reliable indicator of fuel consumption) among urban roads, rural
roads and motorways and the apportionment among the vehicle
categoriesforeachroadtype,keepinginmindthattotalemissions
ofCO2arethesameintwoinventoriesbyconstruction.

The charts and the data in Figure1 show remarkable diffeͲ
consumption, in the two inventories. The bottom–up approach
appliedinthispapergivesalargercontributionofmotorwaysand
ruralroads(respectively778and1193ktinbottom–upversus686
and 749kt in Piedmont Inventory) due to the use of a more
detailedroadnetworkandlikelydifferencesintrafficflowsofthe
tworoadnetworksemployed(seetheSM,FiguresS1andS7).

ways exhibit a higher contribution of commercial vehicles, in
particularHDV,forthebottom–upinventory,whilethereisamuch
smaller fraction of emissions of LDV for rural roads as compared
withthePiedmontInventory.

The most important difference resides in urban roads (see
Figure1) that account for 63% of road transport CO2 emission in
thePiedmontInventoryand49%inthebottom–upinventory(34%
urban diffuse): the share associated to commercial vehicles is
muchhigherinthePiedmontInventory,bothforLDVandforHDV,
summingtoatotalof41%ascomparedto18%inthebottom–up
data (returning less than 10% of commercial vehicles flowing on
urbanroads),wesuggestthistobeaside–effectofthetop–down
methodology,wherefuelconsumptionisdistributedbytakinginto
accountboththenumberandtheaveragemileageofeachvehicle
type.
NMVOC(t/y) NH3(t/y) NOX(t/y) PM10(t/y) SO2(t/y)
1171 275.1 8073 534 48
1012 71.9 1760 127 11.3
89  626 23 0.1
2824 232 5361 375 32.6
   2218 
5096 579 15820 3277 92.0
 Pallavidino et al. – Atmospheric Pollution Research (APR) 652

Table4.Specificemissionfactors(forktofCO2)derivedinthetwoinventorieswithdifferentmileagesforbottom–upapproach
forroadnetworkandurbantraffic
CO/CO2 NMVOC/CO2 NH3/CO2 NOX/CO2 PM10/CO2
  (t/kt) (t/kt) (t/kt) (t/kt) (t/kt)
Bottom–up 5.73 0.47 0.12 3.87 0.26
Roadnetworkemissions/Motorways Top–down 6.86 0.50 0.15 3.93 0.33
andruralroadsonly Bottom–up
7.06 0.59 0.14 4.08 0.27
(oldmileage)
Bottom–up 11.28 1.27 0.15 3.28 0.24

Urbandiffuseemissions Top–down 18.50 3.68 0.08 4.39 0.48

Bottom–up
18.15 2.04 0.16 4.19 0.28
(oldmileage)



Figure1.ApportionmentofCO2yearlyemissionsamongvehiclecategoriesforeachroadtypeinPiedmontInventoryandin
thebottom–upinventory.Theemissionslinkedtothedifferentroadtypesarereportedunderthepiecharts.

Ahighershareofcommercialandtwo–wheeledvehiclesleads For Turin, the reduction in bottom–up inventory of other
to higher emissions, because for most pollutants passenger cars pollutantsisthenmuchhigherthanothermunicipalities:–41%for
emission factors are lower, thus producing the underestimation NOX,–53%forCOand–71%forNMVOC.
observed in the bottom–up inventory (see Tables1 and 2). Since 
administrative action plan for air quality improvement are mainly Once again this is a side–effect of the choice of the proxy
addressed to urban road transport, usually the largest source of variablesusedtodistributeurbandiffuseemissions.InFigure3the
pollution in cities, inaccuracy in the apportionment of transport numberofvehiclespercapitahasbeenplotforthemunicipalities
emissionsamongvehicletypesmayresultinineffectivepoliciesin intheProvinceofTurin:ontheleftforlightvehicles(carandtwo–
theabatementofurbanpollutantemissions. wheelers),ontherightforcommercialvehicles.Inbothcasesthe
 cityofTurin,comparedtoneighboringareas,showsahighnumber
3.3.Spatialmapping of vehicles per capita. This could be due to the fact that many
 vehicles,notprivatelyowned,areregisteredinTurinwheremore
The different apportionment of CO2 among road types, and businesses have a registered office addresses and not necessarily
the different criterion chosen to assign urban diffuse emissions operationalsites.
amongmunicipalities(populationinthebottom–upinventoryand 
vehicle fleet in the Piedmont Inventory) returns a very different 3.4.Fuelbalance
spatialdistributionforCO2(or correspondentlyfuelconsumption) 
andalsoforotherpollutants. As a further check of the emission estimates, we have
 performed the comparison of fuel consumption (split into fuel
Figure2shows,intheareaconsidered,thespatialdistribution type)inthebottom–upinventoryandthePiedmontInventorywith
of variations in estimated CO2, expressed in terms of ratios the fuel sold in the territory of Province of Turin, all of them
between bottom–up and Piedmont Inventory emissions. Even if expressed in terms of CO2. In particular, Table5 shows a
the overall CO2 emissions in the province are the same in two remarkableagreementintheshareofdifferentfuels.Whileforthe
inventories by construction, the map in Figure2 highlights some Piedmont Inventory this is expected by construction, in the
evident features: in the bottom–up inventory emissions increase bottom–up inventory just a constraint on the total amount of
(red shades) in those municipalities traversed by highways and CO2 was set: the correct fuel share we found strengthens the
highflowsruralroads,while,outsidethemetropolitanarea,there choices made on traffic flows repartition, fleet composition and
is a decrease (green shades) where the resident population is average annual mileage which are at the basis of this emission
small.WhenlookingattheTurinmetropolitanarea,weseeasharp assessment.
reduction(–24%)inthecityofTurin,onlyslightlycompensatedby 
theriseintheborderingmunicipalities. 
 Pallavidino et al. – Atmospheric Pollution Research (APR) 653



Figure2.MapofratioofCO2roadtransportemissionsbetweenthebottom–upinventoryandPiedmontInventorycalculatedfor
eachmunicipality.

Table5.ComparisonofapportionmentofCO2emissionsassociatedtoeachfuelinthisstudy(bottom–upinventory),inPiedmont
Inventory,andinfuelbulletin
Bottom–upInventory PiedmontInventory
FuelBulletin(ProvinceofTurin)
RoadTransport(ProvinceofTurin) RoadTransport(ProvinceofTurin)
 CO2(kt/y)
CO2(kt/y) CO2(kt/y)
Diesel 2324 2375 2261
Gasoline 1456 1274 1411
LPGandCNG 98 130CNG+100LPG 24CNG+86LPG
Total 3878 3878 3782

4.Conclusions Inventory,bottom–upemissionsofCOincreaseof49%,NMVOCof
 53%andNOXof15%.
In this work we have calculated road transport pollutant 
emissionswithabottom–upapproachusingtheoutputofatraffic Bottom–up and Piedmont Inventory also shows evident
modelandintegratingabigquantityofdataabouttrafficflowsand dissimilaritiesintheapportionmentofemissionsamongvehicular
mobility of people made available by the Province of Turin. For categories, especially for urbanroad transport.The choice of two
PM10 we have calculated all the primary components: exhaust different proxies, population and vehicle fleet, to allocate diffuse
pipe,brakewear,tyrewearandresuspension. urbanemissionsproduceshighvariabilityinthespatialdistribution
 ofemissions.
Despite the same Copert IV emission factors and the same 
amount of total CO2 have been used, we found remarkable Since the bottom–up approach relies on a more transparent
differenceswiththeemissionsreportedinthePiedmontInventory choice of parameters we believe it guarantees a sounder estiͲ
mainly based on a top–down methodology. The total bottom–up mation of road transport emissions and of the apportionment
emissions are much lower for all pollutants except for NH3 and amongvehicletypes.Whenwefocusonurbanareas,wheretraffic
CO2,thelatterbeingforcedtobeequalinthetwocases.Theratio isreckonedtobeamajorsourceforlocalairpollution,actionplan
between bottom–up and Piedmont Inventory is 83.8% for NOX, measures maybe exposed to the risk of pooreffectiveness by an
59.0% for CO, 34.1% for NMVOC, 95.3% for PM10 and 14.3% for inadequateknowledgeofemissiveburdens.
SO2. 
 Evenifthefuelbalanceconfirmsthereliabilityofthedatawe
An extensive analysis on the set of input parameters has haveused(trafficmodeloutput,trafficsurveys,annualmileagesof
shown that discrepancy is mainly due to different vehicle fleets vehiclecategoriestodefineacirculatingfleet),greatimprovement
used in the two inventories. We found that statistically–based canbeexpectedinthenextfuturewhenalargeamountsofdata
weighting factors (annual mileages) have great influence on the (like the ones collected by traffic control CCTV devices and in
emission estimates: using the same mileages of Piedmont general by intelligent traffic systems) is likely to be available for
statisticalpurposes.
 Pallavidino et al. – Atmospheric Pollution Research (APR) 654



a)

b)


Figure3.Mapsofnumberofvehiclespercapita[(a)carandtwo–wheelers,(b) commercial]calculatedforeachmunicipality.

Acknowledgments Piemonte, Regione Piemonte and Provincia di Torino for having
 kindlyprovidedallthedatarequired.
This study is part of AERA Project (Air Environment Regions 
Alcotra),fundedbyEUundertheAlcotra(FESR)programme,which SupportingMaterialAvailable
involved local administration and environmental agencies of 
regions located on the border between Italy and France with the Two wheelers percentage of private traffic (TableS1), Light
task to provide all partners with the information and the instruͲ Duty Vehicles percentage compared with all commercial vehicles
ments to adopt the right strategies to protect air quality in the (TableS2),Emissionfactorsexpresseding/kgoffuelforbottom–
transboundary area. Helpful discussions with A. Nanni and M. P. up inventory, top–down inventory of Piedmont Region and top–
Costa (Arianet Srl), D. Rampone and T. De Carli (Csi Piemonte), F. down inventory of Italian national environmental agency (ISPRA)
Matera and G. Truffo (Regione Piemonte) are warmly acknowlͲ (TableS3),Roadnetworkusedinthebottom–upapproach(Figure
edged. The authors would like to thank 5T s.r.l., Ativa, CSI S1), Map of percentage of commercial vehicles compared with
 Pallavidino et al. – Atmospheric Pollution Research (APR)

private vehicles average daily traffic (ADT) (FigureS2), ApporͲ
tionmentofvehicular categoriesamongenvironmentalcategories
“Euro” for the fleet circulating on urban roads (FigureS3),
Apportionment of vehicular categories among environmental
categories “Euro” for the fleet circulating on rural roads and
motorways (FigureS4), Apportionment of vehicular categories
amongenvironmentalcategories“Euro”forthefleetcirculatingon
urban roads (FigureS5), Apportionment of vehicular categories
amongenvironmentalcategories“Euro”forthefleetcirculatingon
rural roads and motorways (Figure S6), Road network used to
compile Piermont Region Inventory (IREA) (FigureS7). This inforͲ
mation is available free of charge via theInternet at http://www.
atmopolres.com.
 disaggregazione–dellinventario–nazionale–2010/disaggregazione–
References
 February2013.
Andre, M., Hammarstrom, U., Reynaud, I., 1999. Driving Statistics for the
AssessmentofPollutantEmissionsfromRoadTransport,INRETSReport
LTE9906,Bron,France,194pages.
ArpaLombardia,2008.http://www.inemar.eu/xwiki/bin/view/InemarWiki/
Moduli,accessedinOctober2012.
Baldasano, J.M., Goncalves, M., Soret, A., Jimenez–Guerrero, P., 2010. Air
pollutionimpactsofspeedlimitationmeasuresinlargecities:Theneed
for improving traffic data in a metropolitan area. Atmospheric
Environment44,2997–3006.
Beevers, S.D., Westmoreland, E., de Jong, M.C., Williams, M.L., Carslaw,
D.C.,2012.TrendsinNOXandNO2emissionsfromroadtrafficinGreat
Britain.AtmosphericEnvironment54,107–116.
Borge,R.,deMiguel,I.,delaPaz,D.,Lumbreras,J.,Perez,J.,Rodriguez,E.,
2012. Comparison of road traffic emission models in Madrid (Spain).
AtmosphericEnvironment62,461–471.
Borrego, C., Tchepel, O., Barros, N., Miranda, A.I., 2000. Impact of road
traffic emissions on air quality of the Lisbon region. Atmospheric
Environment34,4683–4690.
Carletti, S., Di Nicola, G., Passerini, G., 2013. Precise estimates at
municipality level of airborne pollutant emissions due to road traffic.
th
Proceedings of 19  International Conference on Urban Transport and
theEnvironment,May29–31,2013,Kos,Greece,pp461–472.
Caserini,S.,Pastorello,C.,Gaifami,P.,Ntziachristos,L.,2013.Impactofthe
dropping activity with vehicle age on air pollutant emissions.
AtmosphericPollutionResearch4,282–289.
Caserini, S., Gaifami, P., Moretti, M., 2011. http://groupware.sinanet.
isprambiente.it/expert_panel/library/ept16/caserini_percorrenze,
accessedinOctober2012.
Caserini, S., Giugliano, M., Pastorello, C., 2007a. Emissions scenarios of
particulate and its precursors from vehicular traffic in Lombardy.
IngegneriaAmbientale36,22–28.
Caserini, S., Pastorello, C., Tugnoli, S, 2007b. http://groupware.sinanet.
isprambiente.it/expert_panel/library/ept13/tugnoli–percorrenzepdf,
accessedinOctober2012.
C.S.S.T. (Centro Studi Sistemi Trasporto), 2001. MT Model – Integrated
System of Mathematical Models for the Support to the Decisions of
PlanningoftheTrafficandtheTransport,CSSTReport,Torino,Italy,60
pages.

655
Denby,B.,2011.GuideonModellingNitrogenDioxide(NO2)forAirQuality
Assessment and Planning Relevant to the European Air Quality
Directive,ETC/ACMTechnicalPaper2011/15,Netherlands,89pages.
EEA (European Environment Agency), 2009. EMEP/EEA Air Pollutant
Emission Inventory Guidebook 2009 – Technical Guidance to Prepare
NationalEmissionInventories,ExhaustEmissionsfromRoadTransport
(update May 2012), EEA Technical Report 9/2009, Copenhagen,
Denmark,136pages.
GTT (Gruppo Torinese Trasporti), 2008. http://www.comune.torino.it/gtt/
gruppo/bil_sost/06_ambientale.pdf,accessedinSeptember2012.
ISPRA(IstitutoSuperioreperlaProtezioneelaRicercaAmbientale),2010.
http://www.sinanet.isprambiente.it/it/sia–ispra/inventaria/
dellinventario–nazionale–2010–versione–4.0/view, accessed in
Klimont, Z., Cofala, J., Bertok, I., Amann, M., Heyes, C., Gyarfas, F., 2002.
Modelling Particulate Emissions in Europe: A Framework to Estimate
Reduction Potential and Control Costs, Interim Report IR–02–076,
Luxemburg,169pages.
LAT/AUTh, 2008. European Database of Vehicle Stock for the Calculation
and Forecast of Pollutant and Greenhouse Gases Emissions with
TREMOVE and COPERT, LAT Report No. 08.RE.0009.V2, Thessaloniki,
Greece,260pages.
Levinson, D., Liu, H., Garrison, W., Danczyk, A., Corbett, M., 2009.
http://upload.wikimedia.org/wikipedia/commons/7/79/Fundamentals
_of_Transportation.pdf,accessedinFebruary2014.
MinistryofEconomicDevelopment,2007.FuelBulletin2006,http://dgerm.
sviluppoeconomico.gov.it/dgerm/bollettino/indice.asp?anno=2006&tri
mestre=4,accessedinMarch2012.
Nanni, A., Radice, P., Smith, P., 2009. TREFIC (Traffic Emission Factors
Improved Calculation) User's guide, Report ARIANET R2009.19, Milan,
Italy,50pages.
Pallavidino, L., Prandi, R., Costa, M., Nanni, A., Bertello, A., Bracco, E.,
Pavone, F., 2011. Top–down vs. bottom–up approach in delineating
th
traffic role in air quality scenarios. Proceedings of HARMO14 – 14 
International Conference on Harmonisation within Atmospheric
DispersionModellingforRegulatory Purpose, October 2–6, 2011,Kos,
Greece,pp.307–311.
Progiou, A.G., Ziomas, I.C., 2011. Road traffic emissions impact on air
quality of the Greater Athens Area based on a 20 year emissions
inventory.ScienceoftheTotalEnvironment410,1–7.
Regione Piemonte, 2008. http://www.sistemapiemonte.it/ambiente/irea/,
accessedinMarch2012.
Torriani,L.,Debernardi,A.,Ferrara,E.,2012.DevelopmentofaSimulation
ModelfortheTrafficFlowsintheProvinceofTurin(ModuleforPrivate
Transport),InterimReport,Milan,Italy,60Pages.
U.S. EPA (U.S. Environmental Protection Agency), 2011. AP 42 Section
13.2.1 Paved Roads, http://www.epa.gov/ttn/chief/ap42/ch13/final/
c13s0201.pdf,accessedinMarch2012.
U.S. EPA (U.S. Environmental Protection Agency), 2006. AP 42 Section
13.2.1 Paved Roads, http://www.epa.gov/ttn/chief/ap42/ch13/
bgdocs/b13s0201.pdf,accessedinMarch2012.
Wang,H.K.,Chen,C.H.,Huang,C.,Fu,L.X.,2008.On–roadvehicleemission
inventory and its uncertainty analysis for Shanghai, China. Science of
theTotalEnvironment398,60–67.