sustainability

Article
The Impact of Heavy Vehicle Traffic Trends on the
Overdesign of Flexible Asphalt Pavements
Paolo Intini *, Nicola Berloco, Pasquale Colonna and Vittorio Ranieri
Department of Civil, Environmental, Land, Building Engineering and Chemistry, Polytechnic University of Bari,
70126 Bari, Italy; nicola.berloco@poliba.it (N.B.); pasquale.colonna@poliba.it (P.C.);
vittorio.ranieri@poliba.it (V.R.)
* Correspondence: paolo.intini@poliba.it; Tel.: +39-080-596-3390




Received: 30 January 2020; Accepted: 26 March 2020; Published: 30 March 2020


Abstract: Given their environmental impact, the careful design of asphalt pavements is crucial.
Previous research has highlighted the influence of several parameters on the outputs of different
pavement design methods. In this study, the focus is on heavy vehicle trends, considering both
the percentage of heavy vehicles in the average traffic flow and its evolution over time, which is
usually included as a growth factor in the design inputs. Since these factors are very often assumed
to be based on old estimates, the first aim of this study was to update them by exploring a recent
series of continuous data collected on the Italian motorway network and showing how to infer
estimates from historical traffic data. Subsequently, the variability of these input factors is introduced
in standard pavement design methods to assess their influence on the design process and to quantify
the risk of overdesign. While the analysis of historical heavy vehicle traffic data may reveal an overall
zero-growth traffic tendency, different scenarios should be considered and assessed in cost-benefit
analyses given the not negligible influence of growth factors on pavement thicknesses. This influence
is shown here in different simulated design conditions, with different initial traffic volumes, share of
heavy vehicles, and resilient moduli.

Keywords: heavy vehicles; traffic growth rate; asphalt pavements; pavement design; pavement thickness

1. Introduction
The construction of asphalt pavement systems has a strong impact on the environment. Different
approaches to improve the sustainability of road pavement systems have been proposed based on
modified asphalts, recycling, life-cycle assessments, and optimized maintenance plans [1–5]. However,
independently of the type of pavement system, there are several uncertainties in the design process
which may result in inappropriately determining the thickness of pavement layers. Whilst considering
traditional asphalt pavement systems, a possible disproportionate thickness resulting from overdesign
is a potential source of environmental damage.
Hence, careful design of road pavement systems may prevent squandering of resources. Different
methods can be used for designing road pavement systems. Among them, semi-empirical or mechanistic
empirical methods such as those based on the AASHTO guides [6,7], also based on multilayer elastic
system analysis (see [8]), are used in practice. Several studies have been conducted with the aim of
assessing the influence of the different inputs on the outcomes of such methods [9–11]. As a result of the
sensitivity analyses, the important role of some factors was highlighted (for example, the determination
of the subgrade resilient modulus, see [10]).
Apart from construction materials properties, traffic-related factors were also investigated.
For example, the variability of the spectrum of heavy vehicles which will travel on the road to be built
was assessed [12]. In fact, pavements are normally designed and analyzed as based on the amount of

Sustainability 2020, 12, 2688; doi:10.3390/su12072688 www.mdpi.com/journal/sustainability

Sustainability 2020, 12, 2688 2 of 13

heavy vehicles on the road, assuming that the influence of light vehicles is negligible. Depending on
the importance of the facility and the desired serviceability, forecasts of heavy vehicle traffic growth
are made for different time ranges. Hence, the analysis of the evolution of trucking is crucial for the
pavement design [13]. However, less attention was paid to this specific issue in previous research that
investigated the influence of input factors on pavement design.
Estimating the heavy vehicle growth tendency is especially important in the long term. The growth
factors which are used in practice usually refer to studies developed decades ago. However, it should
be considered that both short-term and long-term macro-economic tendencies may influence traffic
volume growth or decrease. Related to this, among the other factors generating uncertainty, the impact
of traffic growth variability on pavement design is often neglected. Nevertheless, the use of high
growth factors, where not strictly needed, may lead to unnecessary layer thicknesses (overdesign).
In this study, the impact of heavy vehicle traffic trends on the design of flexible asphalt pavements
is investigated, starting from the analysis of real traffic data. Hence, this study aims to answer the
following research questions:

• How is it possible to infer heavy vehicle traffic growth factors from continuous traffic data?
• How important is the influence of heavy vehicle traffic trends on pavement design?
• Does the overdesign tendency depend on the specific pavement design method adopted?

Historical motorway traffic data have been explored to achieve these objectives. Such data are
useful to assess how to infer traffic growth factors from continuous data. Those factors are then used to
simulate the pavement design process, considering other influential variables. Two pavement design
methods are considered, the semi-empirical and mechanistic empirical, to reveal different possible
overdesign patterns.
The remainder of the paper is structured as follows: the methods used in this study are described
in the “Methods” section. Furthermore, the results obtained are presented and discussed in the
following “Results” and “Discussion” sections, respectively. Finally, some conclusions from the study
are drawn.

2. Methods
Firstly, the dataset on which the traffic analysis is based is described together with the elaborations
conducted to infer the parameters needed. Secondly, the two pavement design methods are briefly
introduced. The use of these methods in the context of the present research is explained as well as the
considered assumptions.

2.1. Traffic Data Analysis

Motorway traffic data were examined with the purpose of identifying heavy vehicle traffic trends.
In particular, the information retrieved in [14] was used. The average vehicular traffic volume on
motorways belonging to the Italian motorway networks (see Figure 1) have been considered for a
total length of 4846 kilometers of roads (belonging to 46 motorway sections with an average length of
105.3 km), with continuously recorded traffic data in the period of 2001–2018. Some motorway sections
were excluded because they were built or modified during the studied time period.
The information reported in [14] includes traffic volumes divided into light vehicles and heavy
vehicles for each motorway section and each year.
Hence, for each year, it was possible to compute the following values:

Average daily heavy vehicle tra f f ic volume per motorway section

%HV = (1)
Total average daily tra f f ic volume per motorway section

Average daily heavy vehicle tra f f ic volume per motorway section

HV per km = (2)
Length o f the motorway section [km]
Sustainability 2020, 12, 2688 3 of 13

Sustainability 2020, 12, x FOR PEER REVIEW 3 of 14

These values were then averaged over all the motorway sections in order to obtain a single value
of both %HV Theseand HVwere
values per km
thenfor each year.
averaged Once
over all the average
the motorway yearly
sections estimates
in order of the
to obtain heavy
a single vehicle
value
both %HV
trafficofvolume per and HV perof
kilometer kmtheformotorway
each year. network
Once the (HVperkm)
average yearly
andestimates of the
the relative heavy vehicle
percentage of heavy
traffic
vehicles overvolume per kilometer
the whole of the (%HV)
traffic volume motorway havenetwork (HVperkm)their
been obtained, and annual
the relative percentage
variation of
is examined.
heavy vehicles
This information over the
is useful forwhole
makingtraffic volume (%HV)
assessments abouthave been
their obtained,
influence ontheir annualdesign,
pavement variation is
which is
examined. This information is useful for making assessments about
one of the research questions of this study, as further shown in the Results section.their influence on pavement
design, which is one of the research questions of this study, as further shown in the Results section.

Figure 1. Sections of the Italian motorway network for which data are available until the end of 2018
Figure 1. Sections of the Italian motorway network for which data are available until the end of 2018
(adapted from [14]).
(adapted from [14]).
2.2. Methods
2.2. Methods for Pavement
for Pavement Design
Design Simulations
Simulations
Two Two methods were used for simulating the pavement design process, while considering the
methods were used for simulating the pavement design process, while considering the
influence of the heavy vehicle traffic variability. Both the semi-empirical AASHTO procedure [6] and
influence of the heavy vehicle traffic variability. Both the semi-empirical AASHTO procedure [6] and
the KENLAYER software (see [8]), which can iteratively solve the elastic multi-layer system problem,
the KENLAYER software (see [8]), which can iteratively solve the elastic multi-layer system problem,
were used. Note that the theoretical comparability between the two methods is outside the scopes of
were this
used. Note that thethe
study. However, theoretical comparability
two methods between
were considered the
for the two methods
simulations is outside
in order to assessthe scopes of
whether
the influence
this study. of the
However, heavy
the twovehicle
methodstraffic
weregrowth, which translates
considered into an increase
for the simulations in the to
in order design traffic,
assess whether
the influence of the heavy vehicle traffic growth, which translates into an increase in the design traffic,
may significantly vary by considering different methods. The two methods are briefly introduced
as follows.
Sustainability 2020, 12, 2688 4 of 13

2.2.1. Semi-Empirical AASHTO Method for Pavement Design

The semi-empirical AASHTO method for pavement design is based on the following
experimentally driven equation:

∆PSI
log ( 4.2−1.5 )
log(ESAL) = ZR SO + 9.36 log(SN + 1) − 0.20 + 1094
+ 2.32 log(MR ) − 8.07 (3)
0.40 +
(SN +1)5.19

where:

ESAL = Number of equivalent single axle loads (18 kips ≈ 80 KN axles);

ZR , S0 = Coefficients depending on the accepted reliability level and the overall design uncertainty;
SN = Structural Number, computed as a function of the layer thicknesses and materials used;
∆PSI = Accepted serviceability loss at the end of the pavement design life;
MR = Resilient modulus of the subgrade.

The number of ESALs travelling on the road pavement in the design life is computed as follows:

(1 + gr)L − 1
ESAL = AADT ∗ %HV ∗ F ∗ Ce ∗ (4)
gr

where:

ESAL, %HV = measures previously defined;

AADT = Annual average daily traffic in the first year of the design life;
F = Factor considering the division of traffic into directions and lanes;
Ce = Overall conversion factor of the total passages of different heavy vehicle axles into ESALs;
gr = heavy vehicle annual growth rate;
L = Design life in years.

The Italian motorway network is considered in this study as a data source for heavy vehicle
traffic volumes. Hence, the parameters in equations (3) and (4) are set according to the importance of
the road (reliability level: 90%, SO = 0.425, PSIi at the beginning of the design life = 4.2, PSI f at the
end of the design life = 2.5, ∆PSI = 1.7, L = 30 years). The typical heavy vehicle spectrum for Italian
motorways [15] was used to compute the Ce factor. The simulations were conducted by taking into
account the variability of the following parameters:

• AADT;
• %HV;
• gr;
• MR .

Traffic volumes were varied up to 90,000 vehicles/day. The latter traffic volume is assumed to be
related to eight lanes motorways (four in each direction), which results in modifying the F factor in
Equation (4) (for motorways with more than four total lanes). The variability of the %HV and gr factors
is assumed to be based on the results from the traffic data analysis described in the previous step.
Two values are assumed for the resilient modulus of the subgrade: 90 and 120 MPa. In the case of such
values, a four-layer asphalt pavement is recommended for motorways, as based on Italian guidelines
for selecting layer thicknesses [15]. For higher resilient moduli, three-layers of hot-mix asphalt may
suffice, while very low moduli are deemed not to be compatible with the importance of the road.
In the presence of such values of the resilient modulus of the subgrade, the following layer
thicknesses are recommended for motorway pavements based on the Italian guidelines [15]: surface
course 6 cm, binder course 7 cm, a variable base course between 14 and 32 cm depending on the
Sustainability 2020, 12, 2688 5 of 13
Sustainability 2020, 12, x FOR PEER REVIEW 5 of 14

is iteratively
heavy vehiclesolved
traffic in different
volume, andscenarios
a subbase (considering
course of 15the variability
cm. Hence, inofthethesimulations,
different factors)
Equationby
varying the base course thickness (and in the case of very high heavy vehicle traffic volumes, the
(3) is iteratively solved in different scenarios (considering the variability of the different factors)
subbase
by varying course).
the base course thickness (and in the case of very high heavy vehicle traffic volumes,
the subbase course).
2.2.2. Method Based on an Elastic Multi-Layer System
2.2.2. Method Based on an Elastic Multi-Layer System
The other method used for simulating the pavement design process is that based on the solution
of anTheelastic
othermulti-layer
method used system. In this study,
for simulating the KENLAYER
the pavement application
design process forbased
is that flexible pavements
on the solution
included in the KENPAVE software is used (see [8]). It computes stresses and strains in the multi-
of an elastic multi-layer system. In this study, the KENLAYER application for flexible pavements
layer
includedpavement system (see
in the KENPAVE Figure is
software 2)used
by conducting
(see [8]). It acomputes
damage stresses
analysisand
andstrains
estimating
in thethe life of a
multi-layer
pavement system. The following main inputs are needed:
pavement system (see Figure 2) by conducting a damage analysis and estimating the life of a pavement
system. The following main inputs are needed:
• The analysis of the heavy vehicle traffic volume by differentiating the heavy vehicle spectrum
• Theinto
analysis of theofheavy
passages vehicle
different axletraffic
loadsvolume by differentiating
(of different the heavy vehicle
weights, configurations, andspectrum
numbersinto of
passages
tires); of different axle loads (of different weights, configurations, and numbers of tires);
• The thickness of each layer of the pavement system;
• •TheThe thickness
elastic of each
parameters of layer of theof
each layer pavement system;
the pavement system.
• The elastic parameters of each layer of the pavement system.

Figure
Figure 2.
2. Multi-layer
Multi-layer elastic
elastic pavement
pavement system.
system.

The
The traffic
traffic inputs
inputs introduced
introduced in in the
the KENLAYER
KENLAYER software
software are
are selected
selected inin order
order toto be
be coherent
coherent
with those considered in the semi-empirical AASHTO method (in terms of heavy vehicle spectra for
with those considered in the semi-empirical AASHTO method (in terms of heavy vehicle spectra for
motorways,
motorways, and and the
the different
different traffic volumes and
traffic volumes and growth
growth rates
rates considered).
considered). The The thickness
thickness of of each
each
layer of the pavement system is iteratively attempted with the same criteria defined in the previous
layer of the pavement system is iteratively attempted with the same criteria defined in the previous
sub-chapter
sub-chapter (variability
(variability of
of the
the base
base and
and subbase
subbase course,
course, considering
considering thethesame
samedesign
designlifelifeof
of30
30years).
years).
The elasticmoduli
The elastic moduliofof thethe pavement
pavement layers
layers werewere obtained
obtained by converting
by converting the minimum
the minimum features
features required
required in [15] (Marshall stability of the surface course: 1100 kg, of the binder course: 1000 kg, ofbase
in [15] (Marshall stability of the surface course: 1100 kg, of the binder course: 1000 kg, of the the
base course:
course: 800 kg,800hypothesizing
kg, hypothesizing high high
heavyheavy vehicle
vehicle volume,
volume, CBR CBR
valuevalue
for theforsubbase
the subbase
course:course:
30%)
30%)
throughthrough the conversion
the conversion tables reported
tables reported in [8].
in [8]. The The modulus
modulus of the subgrade
of the subgrade is assumed is in
assumed
coherence in
coherence with assumptions made in the previous sub-section (90 and 120 MPa).
with assumptions made in the previous sub-section (90 and 120 MPa).
Considering homogeneous periods,
Considering homogeneous periods, the
theequation
equationwhich
whichisisused
usedfor
forestimating
estimatingthe thedesign
design life
life LL is
is reported as follows [8]:
reported as follows [8]:

𝑚𝑚
m −1 X 𝑚𝑚
m −1
𝑛𝑛i𝑖𝑖 −1
n 𝑛𝑛n𝑖𝑖 i −1
X
−1
L𝐿𝐿==(DR )
(𝐷𝐷𝐷𝐷) −1=
min[([(� N )) ;; (�
=min (
Ni,P) ]
) ] (5)
(5)
i=1 𝑁𝑁𝑖𝑖,𝐹𝐹
𝑖𝑖=1
i,F i=1 𝑁𝑁𝑖𝑖,𝑃𝑃
𝑖𝑖=1

Where:
where:
ni = predicted load repetitions for each i-th load;
Sustainability 2020, 12, 2688 6 of 13

ni = predicted load repetitions for each i-th load;

Ni = allowed load repetitions for each i-th load, considering both the fatigue cracking Ni,F and the
permanent deformation Ni,P criteria (defined below [16]);
DR = Damage Ratio.

Ni,F = a εT −b E−c ; Ni,P = d εC − f (6)

where:
εT = tensile strain at the bottom of the asphalt layer;
E = asphalt layer modulus;
εC = compressive strain at the top of the subgrade;
a,b,c,d,f = constants.
Moreover, in the KENLAYER application, it is possible to introduce different material properties
for different periods of the year, thus considering the influence of climate and temperature. This aspect
is not considered in this study (as in the AASHTO method) since it is outside the scope of the research
questions posed.

3. Results
The results obtained in this study are presented in this section and divided according to the stages
considered in the methods section.

3.1. Trends from the Traffic Data Analysis

The motorway traffic data analysis has revealed the heavy vehicle trends reported in the following
figures. Means, standard deviations, and yearly variations are presented in Table 1.

Table 1. Means, standard deviations (S.D.), and yearly variations (YV) of %HV and HVperkm over all
the motorway sections inquired in the observation period.

%HV HVperkm
Year Mean (%) S.D. (%) YV (%) mean S.D. YV (%)
2001 19.5 5.5 - 217.04 187.28 -
2002 19.7 5.6 1.38 222.29 185.44 2.42
2003 19.9 5.5 0.89 230.39 188.29 3.65
2004 20.3 5.4 2.15 241.01 194.44 4.61
2005 20.5 5.3 0.81 243.99 192.49 1.24
2006 20.5 5.2 −0.17 250.31 194.73 2.59
2007 20.4 5.3 −0.38 255.37 198.18 2.02
2008 20.0 5.2 −1.70 249.31 195.88 −2.37
2009 18.6 4.6 −7.09 232.82 185.48 −6.61
2010 18.9 4.8 1.51 236.67 187.49 1.66
2011 19.2 4.9 1.38 237.03 185.86 0.15
2012 19.1 4.9 −0.52 219.63 174.89 −7.34
2013 19.0 4.8 −0.54 215.33 172.07 −1.96
2014 18.9 4.9 −0.11 215.80 170.05 0.22
2015 19.0 4.8 0.15 221.57 171.39 2.68
2016 19.0 4.9 0.05 226.12 174.11 2.05
2017 19.1 5.2 0.83 233.41 177.38 3.23
2018 19.4 5.5 1.29 237.37 179.50 1.70
Average 19.5 5.1 −0.00 232.53 184.16 0.58

Based on the figures reported above, it is possible to note that the percentage of heavy vehicles in
the average daily traffic (%HV) was fairly stable over the years from 2001 to 2018. The average %HV
Sustainability 2020, 12, 2688 7 of 13

was close to 20%. Considering the standard errors reported in Figure 3 for each year, it can be stated
that in almost2020,
Sustainability 70%12,of motorways
x FOR sections, %HV varies between approximately 15% and 25%.7 of 14
PEER REVIEW
Sustainability 2020, 12, x FOR PEER REVIEW 7 of 14
(Percentageofof

traffic)
the

0.30
dailytraffic)
vehiclesininthe

0.30
%HV(Percentage

0.20
0.20
heavyvehicles
averagedaily

0.10 y = -0.0007x + 1.6683

0.10 y = -0.0007x + 1.6683
R² = 0.390
average

R² = 0.390
%HV
heavy

0.00
0.00 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Years
Years

Figure Trendof
3. Trend
Figure 3. ofthe
theheavy
heavyvehicles
vehiclesshare
share
inin
thethe average
average daily
daily traffic
traffic (%HV)
(%HV) in the
in the analyzed
analyzed period.
period
Figure 3. Trend of the heavy vehicles share in the average daily traffic (%HV) in the analyzed period
TheThe average
average number
numberofofheavy heavyvehicles
vehiclesperper kmkm of motorwaytrunk
of motorway trunk(HVperkm)
(HVperkm)isisfairly fairly stable
stable over
over
thethe The average
observation number
period. It of heavy vehicles
fluctuates between peraround
km of motorway
215 and trunk
255 (HVperkm)
HVperkm over is fairly
the stable over
years. In this
observation period. It fluctuates between around 215 and 255 HVperkm over the years. In this case,
theconsidering
observation the period. It fluctuates between around 2154and 255 HVperkm over the years. In this case,
considering the standard errors reported in Figure 4 for each year, in most motorways, HVperkm
case, standard errors reported in Figure for each year, in most motorways, HVperkm
considering the standard errors and reported The in Figure 4 for each year, in most motorways, HVperkm
varies
varies between
between approximately50
approximately 50 and400. 400. The overall
overall average
average trend
trendeveneven decreased
decreased over
over thethe
years,
years,
varies between approximately 50 and 400. The overall average trend even decreased over the years,
while
while thethe computed
computed annualgrowth
annual growthrate rateaveraged
averaged over the the years
yearsisisequal
equaltoto0.58%,
0.58%,which which practically
practically
while the computed annual growth rate averaged over the years is equal to 0.58%, which practically
means
means there
there waswas almostzero
almost zerogrowth
growthover overthe
the observation
observation period.
period.The Thegrowth
growthrate rateisisestimated
estimated based
based
means there was almost zero growth over the observation period. The growth rate is estimated based
on on
thethe evolution
evolution ofofthethevalue HVperkm over
valueofofHVperkm over time,
time, since
since ininthis
thisway,
way,the theraterateisiscomputed
computed over
overa real
a real
on the evolution of the value of HVperkm over time, since in this way, the rate is computed over a real
amount
amount of of heavy
heavy vehicles(rather
vehicles (ratherthanthanaapercentage)
percentage) considering
considering the theactual
actualaverage
averagedistribution
distribution of of
amount of heavy vehicles (rather than a percentage) considering the actual average distribution of
the vehicles per km of the whole motorway section. Moreover, as
the vehicles per km of the whole motorway section. Moreover, as can be noticed from the estimates can be noticed from the estimates
the vehicles per km of the whole motorway section. Moreover, as can be noticed from the estimates
shown
shown in in Table
Table 1, 1, this
this strategyleads
strategy leadsto toconservative
conservative estimates
estimates of ofgrowth
growthrates rateswith withrespect
respect toto
those
those
shown in Table 1, this strategy leads to conservative estimates of growth rates with respect to those
estimated
estimated based onpercentages
percentages(since (sincethetheyearly
yearly variations
variations based on HV%
HV% are very small, given the
estimated based on percentages (since the yearly variations based on HV% are very small, given thethe
based on based on are very small, given
very
very low low increase
increase in
inin percentages,
percentages, if any).The influence of setting the growth rate is one of the research
very low increase percentages,ififany). The influence
any).The influence of ofsetting
settingthethegrowth
growthrate rateisis oneoneofofthethe research
research
questions
questions of of this
this study.
study. Hence,
Hence, in in addition
addition to to average
average tendencies
tendencies which
which leadleadto to almost
almost zero
zero growth,it is
growth,
questions of this study. Hence, in addition to average tendencies which lead to almost zero growth,
it is worth breaking down the analysis period into smaller periods to reveal some local tendencies. In
it is worth
worth breaking breaking
down down the analysis
the analysis period period into smaller
into smaller periods
periods to reveal
to reveal some somelocallocal tendencies.
tendencies. InIn
fact,
fact, while in recent decades almost zero growth was noted, in relation to the different effects of a
fact,inwhile
while recent indecades
recent decadesalmost almost
zero growthzero growth was noted,
was noted, in relation
in relation to the different
to the different effects effects of a
of a financial
financial crisis [17], it is quite unusual and risky to consider zero traffic growth for a future time span
financial
crisis [17], itcrisis [17],unusual
is quite it is quiteandunusual
risky and risky to consider
to consider zero
zero traffic traffic growth
growth for a futurefor a time
future timeequal
span span to
equal to the design life of the road (in this case, 30 years). Local growth tendencies are shown in the
theequal
design to life
the ofdesign
the roadlife of(inthe road
this case,(in30
this case, Local
years). 30 years).
growthLocal growth tendencies
tendencies are shownare inshown
the next infigure.
the
next figure.
next figure.
500 y = -0.4822x + 1201.5
kmofof

500 y = -0.4822x + 1201.5

(Numberofof

R² = 0.0427
400 R² = 0.0427
trunk)
perkm

400
HVperkm(Number

motorwaytrunk)

300
vehiclesper

300
200
heavyvehicles
motorway

200
HVperkm

100
100
0
heavy

02001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Years
Years
Figure 4. Trend of the average heavy vehicles per km of motorway trunk in the analyzed period.
Figure 4. Trend of the average heavy vehicles per km of motorway trunk in the analyzed period.
Figure 4. Trend of the average heavy vehicles per km of motorway trunk in the analyzed period.
Based on Figure 5, some local growth tendencies can be found in three different periods, which
Based on Figure 5, some local growth tendencies can be found in three different periods, which
cover more
Based onthan half5,ofsome
Figure the observation
local growthperiod. The average
tendencies can becomputed
found inyearly growth tendencies
three different are:
periods, which
cover more than half of the observation period. The average computed yearly growth tendencies are:
2.8%
cover in the
more firsthalf
than period (Figure
of the 5a), 0.9%period.
observation in the second period (Figure
The average 5b),yearly
computed and 2.0% in thetendencies
growth third periodare:
2.8% in the first period (Figure 5a), 0.9% in the second period (Figure 5b), and 2.0% in the third period
(Figure
2.8% in the5c). This
first means
period that even
(Figure considering
5a), 0.9% local tendencies,
in the second the yearly
period (Figure HV2.0%
5b), and growth ratethird
in the doesperiod
not
(Figure 5c). This means that even considering local tendencies, the yearly HV growth rate does not
exceed 3%.
exceed 3%.
Sustainability 2020, 12, 2688 8 of 13

(Figure 5c). This means that even considering local tendencies, the yearly HV growth rate does not
Sustainability
exceed 3%. 2020, 12, x FOR PEER REVIEW 8 of 14

260 250 240

250 y = 2.1063x - 3998.2

HVperkm
R² = 0.813 230
HVperkm

HVperkm
240
240
230
220
220
y = 6.5937x - 12977 y = 4.789x - 9427.2
R² = 0.983 R² = 0.967
210 230 210
2001 2003 2005 2007 2009 2011 2013 2015 2017
Years Years Years
(a) (b) (c)
Figure
Figure 5. Local
5. Local HVperkm
HVperkm increasingtrends
increasing trendsininthe
theperiods:
periods: a)
(a)2001–2007,
2001–2007,b)(b)
2009–2011, c) 2013–2018.
2009–2011, (c) 2013–2018.

These
These estimatesare
estimates arebased
basedon oncountry-wide
country-wide averages.
averages.However,
However, the
themotorway
motorway network
network covers
covers
the whole country and Italy is made up of different areas with specific socio-economic
the whole country and Italy is made up of different areas with specific socio-economic features [17]. features [17].
Moreover,
Moreover, as as is evident
is evident fromthe
from thestandard
standard deviation
deviation values,
values,the
theheavy
heavy vehicle
vehicletraffic volume
traffic volumeis highly
is highly
variable across the motorway network. Hence, the above reported growth rates were computed again
variable across the motorway network. Hence, the above reported growth rates were computed again
by differentiating the whole sample of motorways into groups based on macro-regions (using the
by differentiating the whole sample of motorways into groups based on macro-regions (using the
NUTS-1 classification [18]) and traffic volume ranges. Traffic volume ranges were defined by
NUTS-1 classification [18]) and traffic volume ranges. Traffic volume ranges were defined by dividing
dividing motorways into homogeneously numerous clusters. The following classes were thus
motorways
considered: intolow homogeneously
volume (motorway numerous
sectionsclusters. The following
with average classes
daily volume were vehicles/day),
<50,000 thus considered:
lowmedium
volumevolume
(motorway with average daily volume <50,000
(50,000–100,000 vehicles/day), high volume (100,000–150,000 vehicles/day), volume
sections vehicles/day), medium and
very high volume (>150,000 vehicles/day).
(50,000–100,000 vehicles/day), high volume (100,000–150,000 vehicles/day), and very high volume
Asvehicles/day).
(>150,000 it can be noted from Table 2, the previously reported average estimates do not notably vary
when
As itdifferent
can be notedregionsfrom
and traffic
Table clusters are considered.
2, the previously In particular,
reported average the averagedo
estimates growth rates arevary
not notably
always
when between
different 0% and
regions and1% (except
traffic for low-volume
clusters motorways,
are considered. where itthe
In particular, is 1.3%).
averageMost maximum
growth rates are
growth rates are always between 2% and 3%. However, there is one noticeable
always between 0% and 1% (except for low-volume motorways, where it is 1.3%). Most maximum value belonging to
low-volume
growth rates are motorways (4.6%). 2% and 3%. However, there is one noticeable value belonging to
always between
low-volume motorways (4.6%).
Table 2. Growth rates of motorways divided into region and traffic groups.

Group Type
Table 2. Growth rates ofRegion Average
motorways divided into region and Heavy Vehicle Volume
traffic groups.
North- North- Very
Group
Group Type Centre
Region South * High Heavy
Average Medium Low
Vehicle Volume
West East High
Group North-West North-East Centre South * Very High High Medium Low
Number of
21 13 11 13 11 10 14 11
Motorways
Number of**
21 13 11 13 11 10 14 11
Motorways **
Average Growth
0.62 0.79 0.99 0.44 0.47 0.43 0.66 1.26
Rate (%)
Average Growth
0.62 0.79 0.99 0.44 0.47 0.43 0.66 1.26
Rate (%)
Maximum Growth
3.03 2.86 2.44 2.51 2.31 2.91 3.34 4.60
Rate (%)
Maximum Growth
3.03 2.86 2.44 2.51 2.31 2.91 3.34 4.60
Rate (%)
Period (Maximum
Period ‘01–‘07 ‘13–‘18 ‘01–‘07 ‘01–‘07 ‘13–‘18 ‘01–‘07 ‘02–‘07 ‘01–‘07
Growth Rate) ‘01–‘07 ‘13–‘18 ‘01–‘07 ‘01–‘07 ‘13–‘18 ‘01–‘07 ‘02–‘07 ‘01–‘07
(Maximum Growth Rate)
* Includes the only motorway for the Islands region. Two years of data for the A56 motorway were
* Includes the only motorway for the Islands region. Two years of data for the A56 motorway were extrapolated.
extrapolated.
** Some ** Some
motorways motorways
are considered forare considered
more than one for moresince
region, thanthey
onesignificantly
region, sincecross
theythe
significantly
borders of the
cross themacro-regions.
considered borders of the considered macro-regions.

The variability of the heavy vehicle traffic trends is considered in the pavement design
simulations performed.
The variability According
of the heavy to the
vehicle results
traffic presented
trends in this sub-section,
is considered the following
in the pavement values
design simulations
are assumed for the previously defined variables:
performed. According to the results presented in this sub-section, the following values are assumed
for the previously defined variables:
• %𝐻𝐻𝐻𝐻―10%, 20%, 30%;
Sustainability 2020, 12, 2688 9 of 13

• %HV—10%, 20%, 30%;

• gr—0.00, 0.01, 0.02, 0.03.

Note that gr = 0.03 is considered only for some traffic volumes, because in other cases it may result
in the long term in
Sustainability volumes
2020, incompatible
12, x FOR PEER REVIEW with the road depending on the initial volume. 9Note of 14 that only
in the case of low-volume motorways, a higher growth rate could have been considered (up to 4.6%).
• gr―0.00, 0.01, 0.02, 0.03.
However, given the scarcity of these cases, and the average growth rates always variable between 0%
and 1%. TheNote maximumthat gr = growth
0.03 is considered
rate wasonly for3%
set to some
in traffic volumes,
accordance because
with in other
average cases it may values.
country-wide
result in the long term in volumes incompatible with the road depending on the initial volume. Note
that only
3.2. Results frominthethePavement
case of low-volume motorways, a higher growth rate could have been considered (up
Design Simulations
to 4.6%). However, given the scarcity of these cases, and the average growth rates always variable
Thebetween
results0% and 1%. from
obtained The maximum growth design
the pavement rate wassimulations
set to 3% in accordance
using bothwith
theaverage
AASHTO country-
method and
wide values.
the KENLAYER software are shown in the next figures. In these figures, the minimum thickness of
the base course is shown for different combinations of the resilient modulus (MR ), the share of heavy
3.2. Results from the Pavement Design Simulations
vehicles (%HV), the heavy vehicle growth rate (gr), and the initial annual average daily traffic (AADT).
The results obtained from the pavement design simulations using both the AASHTO method
The results for low-traffic motorways (with AADT < 50,000 vehicles/day, likely four-lanes motorways)
and the KENLAYER software are shown in the next figures. In these figures, the minimum thickness
are shown in base
of the Figure 6. The
course results
is shown forfor other combinations
different motorways (with AADT ≥modulus
of the resilient 50,000 (M vehicles/day,
R), the share of
likely to be
heavy vehicles
six or eight-lanes (%HV), the heavy
motorways) vehiclein
are shown growth
Figure rate
7.(gr), and the initial annual average daily traffic
(AADT).
Apart fromThe results
trivial for low-traffic
remarks on themotorways
variability (with
of theAADT < 50,000
layer vehicles/day,
thickness with the likely four-lanes
initial traffic volume,
motorways) are shown in Figure 6. The results for other motorways (with
the share of heavy vehicles (resulting in thickness increasing) and the resilient modulus of the subgrade AADT ≥ 50,000
vehicles/day, likely to be six or eight-lanes motorways) are shown in Figure 7.
(resulting in thickness decreasing), there is an evident influence of the heavy vehicle growth rate.
Apart from trivial remarks on the variability of the layer thickness with the initial traffic volume,
In fact, depending
the share of on the vehicles
heavy considered combination
(resulting in thickness of variables,
increasing) itand
may theresult in amodulus
resilient thickness variation of
of the
1 to 4 cmsubgrade
between (resulting in thickness
zero growth ratedecreasing), there is an evident
and the maximum influence
considered rateof(3%).
the heavy
Thisvehicle
is lessgrowth
evident when
rate. In
considering fact, depending
results on the considered
from the KENLAYER combination
simulations, butof minimum
variables, itrequired
may result in a thickness
thicknesses are smaller
variation of 1 to 4 cm between zero growth rate and the maximum considered rate
in that case and so their variability is smaller as well. The significant influence of the traffic growth rate (3%). This is less
evident when considering results from the KENLAYER simulations, but minimum required
can be noted for different resilient moduli, different traffic volumes, and percentages of heavy vehicles.
thicknesses are smaller in that case and so their variability is smaller as well. The significant influence
In several cases,
of the even
traffic a 1%rate
growth gr increase is related
can be noted to thickness
for different resilient increasing, resulting
moduli, different trafficin a significant
volumes, and overall
increasepercentages
in the total of heavy
heavy vehicles.
vehicleInpassages.
several cases,
In some 1% gr increase
even a limited cases,is once
relatedthe
to thickness
maximum increasing,
thickness of the
resulting
base course in a significant
is reached (32 cm,overall increase
according to in the total
[15]), heavy vehicle
the subgrade passages.
course In some
thickness is limited cases,
also increased under
once the
the different maximum
growth ratesthickness of the base
with respect to thecourse is reached
standard (32 cm,thickness
constant according ofto [15]),
15 cmthe[15],
subgrade
such as in the
course thickness is also increased under the different growth rates with respect to the standard
case of: MR = 90 MPa, %HV = 20–30% (Figure 7a).
constant thickness of 15 cm [15], such as in the case of: MR = 90 MPa, %HV = 20–30% (Figure 7a).

(a)

Figure 6. Cont.
 Sustainability
Sustainability 268812, x FOR PEER REVIEW
2020, 12,2020, 10 of 14 10 of 13

Sustainability 2020, 12, x FOR PEER REVIEW 10 of 14

(b)
6. Minimum thickness required for the base course considering the variation of MR, %HV,
Figure 6.Figure
Minimum thickness required for the base course considering the variation of MR , %HV, AADT,
AADT,gr, for low-traffic motorways as an output of: (a) the AASHTO method, (b) the KENLAYER
gr, for low-traffic motorways as an output of: (a) the AASHTO method, (b) the KENLAYER software.
software.

The KENLAYER software relies on the application of a mechanistic model. Through additional
simulations performed, it was verified that results from the KENLAYER software are comparable
with those obtained from the AASHTO method, assuming a reliability percentage of 50% (ZR SO = 0
in Equation (3)), far lower than that assumed in the simulations (results plotted in Figures 6a and
7a). Hence, the variability of the heavy vehicle traffic growth on the outputs from the considered
design methods could be possibly interpreted as an effect of a different assumed design reliability. In
other words, a significant effect of the heavy vehicle growth factor can be noted independently of
the assumed reliability level. Clearly, the possible overdesign effect is more evident for high-volume
motorways and for lower values of the resilient moduli. For example, in several cases of low-volume
motorways (MR = 120 MPa, %HV = 10–20%), a minimum pavement thickness (with a base course
of 14 cm) seems adequate, independently of the(b) growth factor. However, a thickness varying (from
Figure
1 cm to 3 cm) with6.the
Minimum
growththickness required
factor for the (from
increasing base course
0% considering
to 3%) canthealways of M
variationbe R, %HV,
seen for high-volume
AADT,gr, for low-traffic motorways as an output of: (a) the AASHTO method, (b) the KENLAYER
motorways, in the case of MR = 90 MPa.
software.

(a)

(a)

Figure 7. Cont.
Sustainability 2020, 12, 2688 11 of 13
Sustainability 2020, 12, x FOR PEER REVIEW 11 of 14

(b)

Figure 7.Figure 7. Minimum

Minimum thickness
thickness required
required for for
thethe base
base course
course considering the
considering variationofofMM
the variation R, %HV,
R , %HV, AADT,
AADT, gr, for higher traffic motorways as an output of: (a) the AASHTO method, (b) the KENLAYER
gr, for higher traffic motorways as an output of: (a) the AASHTO method, (b) the KENLAYER software.
software.

4. Discussion
The KENLAYER software relies on the application of a mechanistic model. Through additional
Thesimulations
traffic data performed,
analysis ithaswasrevealed
verified that results
recent fromfor
trends theheavy
KENLAYERvehicle software
trafficare comparable
volumes. In particular,
with those obtained from the AASHTO method, assuming a reliability percentage of 50% (𝑍𝑍𝑅𝑅 𝑆𝑆𝑂𝑂 = 0
it highlighted that, from 2001 to 2018, the share of heavy vehicles in the traffic volume on motorways
in equation (3)), far lower than that assumed in the simulations (results plotted in Figures 6a and 7a).
in Italy did notthe
Hence, increase, and
variability norheavy
of the did the average
vehicle trafficnumber
growth onofthe heavy vehicles
outputs from theper kilometer
considered of motorway
design
methods could be possibly
section (significantly). Increasing interpreted as an effect
tendencies of a different
in local assumed design
traffic growth reliability.for
were limited In other
some years to
about 3% words, a significant
at most, except foreffect of the
some heavy vehicle
limited growth factor
sub-categories can bevolume
of lower noted independently
motorways.of the
assumed reliability level. Clearly, the possible overdesign effect is more evident for high-volume
This result from a recent continuous set of traffic data suggests the importance of considering
motorways and for lower values of the resilient moduli. For example, in several cases of low-volume
recent trends for estimating the traffic growth factors, which may be much more limited than values
motorways (MR = 120 MPa, %HV = 10–20%), a minimum pavement thickness (with a base course of
considered in some
14 cm) less recent
seems adequate, sources (see
independently [8]).growth
of the The same
factor. advantage is valid varying
However, a thickness for the (from
estimates
1 of the
percentagecm toof3 heavy
cm) withvehicles
the growth in factor increasing
the traffic flow. (from 0% to 3%)
Clearly, forcan always
both be seen for
estimates, high-volume
local factors related to
motorways,
infrastructure in the case ofsocio-economic
development, MR = 90 MPa. factors, transport policies, and changes in the market [13]
of specific countries may be even more determinant than the availability of recent data. For example,
4. Discussion
we have shown how the growth factor may vary between regions and traffic ranges even within the
The traffic data analysis has revealed recent trends for heavy vehicle traffic volumes. In
sample of Italian motorways.
particular, it highlighted that, from 2001 to 2018, the share of heavy vehicles in the traffic volume on
Themotorways
first research
in Italyquestion concerned
did not increase, and northedidpossibility of estimating
the average number long-term
of heavy vehicles trends useful for
per kilometer
pavement of design
motorway fromsection (significantly).
historical Increasing
continuous traffic tendencies
data. Weinhave local shown
traffic growth were limited
that estimates for on recent
based
some years to about 3% at most, except for some limited sub-categories of lower
data may even suggest a zero-growth traffic trend (such as in this case), or in any case fluctuating over volume motorways.
This result from a recent continuous set of traffic data suggests the importance of considering
the years. Hence, it may be arduous to infer a single estimate from historic traffic trends, especially in
recent trends for estimating the traffic growth factors, which may be much more limited than values
the presence of fluctuating
considered in some lesseconomic conditions,
recent sources (see [8]). Thewhich may influence
same advantage is validtraffic volumes.of In
for the estimates thethis study,
differentpercentage
scenariosofwere heavyhypothesized
vehicles in the traffic flow. Clearly,
to account for this forissue,
both estimates,
including local
thefactors related to
“best-case” scenario of
infrastructure
zero-growth traffic anddevelopment, socio-economic
the “worst-case” factors,
scenario transport
based policies,
on the highestand changes in the market
country-wide yearly growth
rate found[13]in
ofthe
specific countries may
observation periodbe even
for atmore
leastdeterminant
two yearsthan in athe availability of recent data. For
row.
example, we have shown how the growth factor may vary between regions and traffic ranges even
Thewithin
otherthe research questions concerned the influence of the heavy vehicle variability on pavement
sample of Italian motorways.
design, considering different possible methods. The possible overdesign effect due to high hypothesized
traffic growth rates was verified by means of two commonly used pavement design methods
(the AASHTO semi-empirical [6] and mechanistic model, through the KENLAYER software [8]).
A significant influence of the heavy vehicle variability was highlighted for both the considered methods.
Concerning the study aims, it is possible to note that the variability of the pavement thickness due to
the growth rate is more evident in the results obtained from the AASHTO method. However, this may
be intrinsically related to the chosen design reliability level, and thus, the greater thicknesses obtained.
It is evident that other factors may have a greater impact on thickness, such as the resilient modulus
Sustainability 2020, 12, 2688 12 of 13

of the subgrade [10] which is clear from simulation results. Even the growth rate factor, which is
one of the most uncertain parameters to be estimated, has a significant influence on the pavement
layer thicknesses because it obviously relates to an increase in the total heavy vehicle traffic. In some
cases, variations in the gr rate from 0% to 3 % resulted in an increased thickness of up to 3–4 cm, for a
design life of 30 years. To provide some reference numbers, a base course thickness increase of 4 cm
on a 10 km long eight-lane motorway section may result in an order of magnitude of €2 million cost
increase and an additional 1.5 million cubic meters of asphalt mixture. These values clearly indicate
a possible economic and environmental issue. Higher traffic growth rates, which may be used in
common practice, may then result in even greater variations.
The proposed strategy for estimating increasing trends is to consider different scenarios based
on recent estimates. Considering this approach in accordance with the results from simulations,
some remarks are evident, which may be useful in practice. In fact, different traffic growth scenarios
should be included among the important factors to be implemented in detailed life-cycle assessments,
or at least in the general framework of a cost-benefit analysis related to pavement management
systems [19]. Otherwise, from a practical perspective, in some cases designers may still rely uniquely
on old estimates, which may not be appropriate in the particular local context and which may result in
the overdesign of pavement layer thicknesses.

5. Conclusions
This study was conceived in order to answer some research questions which may be of interest
for both practitioners and for the further development of research. It was aimed at assessing the
relationships between heavy vehicle historical continuous traffic volumes and the related estimates
useful for pavement design. Moreover, it was aimed at assessing the influence of these estimates on
pavement design by means of different methods.
The difficult connections between historical heavy vehicle traffic data and parameter estimates to
be used for future periods have been highlighted. In particular, an almost zero-growth tendency was
found in the recent period of almost 20 years, as based on data related to the Italian motorway network.
This was confirmed by disaggregating the sample of motorways into sub-groups according to different
regions and traffic ranges. Hence, with the aim of considering non null growth factors, other local
values were estimated based on consecutive years during which an increasing traffic tendency was
noted, leading to a maximum country-wide estimated growth factor of 3% (which may have local
variations).
Pavement design is consistently influenced by the growth factor on the considered road type due
to the high traffic volumes reached in the overall design life. This effect can be noted from both the
methods used for simulations (which were found to be comparable in given conditions), particularly
for the semi-empirical method. An increase in the base course thickness up to 4 cm was noted for traffic
growth factors increasing from 0% to 3%, and in limited cases, in the subbase course thickness as well,
indicating a possible overdesign effect. Hence, the results from this study encourage consideration of
different possible scenarios for the traffic growth factor by including them in the cost-benefit analysis
and/or life-cycle assessment.
These remarks are based on traditional asphalt pavements as an example, but they could also be
applicable for pavements made with different materials. Moreover, further research may apply the
same procedure with comparison within countries for different road types or within a larger procedure
of cost-benefit analysis with specific focus on heavy vehicle traffic trends. Further sensitivity analysis
considering several influential variables may also provide new insights into possible overdesign effects
specifically due to given pavement design methods.

Author Contributions: Conceptualization, P.I., N.B., P.C. and V.R.; methodology, P.I., N.B. and V.R.; software,
P.I.; investigation, P.I.; data curation, P.I.; writing—original draft preparation, P.I.; writing—review and editing,
P.I., N.B., P.C. and V.R.; supervision, P.C. and V.R. All authors have read and agreed to the published version of
the manuscript.
Sustainability 2020, 12, 2688 13 of 13

Funding: This research received no external funding.

Acknowledgments: Cristian Lazzari is acknowledged for the preparatory analysis performed by using two
pavement design methods during his master’s thesis.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Tutu, K.A.; Tuffour, Y.A. Warm-mix asphalt and pavement sustainability: A review. Open J. Civ. Eng. 2016,
6, 84. [CrossRef]
2. Giustozzi, F.; Crispino, M.; Flintsch, G. Multi-attribute life cycle assessment of preventive maintenance
treatments on road pavements for achieving environmental sustainability. Int. J. Life Cycle Assess. 2012,
17, 409–419. [CrossRef]
3. Praticò, F.G.; Giunta, M.; Mistretta, M.; Gulotta, T.M. Energy and Environmental Life Cycle Assessment of
Sustainable Pavement Materials and Technologies for Urban Roads. Sustainability 2020, 12, 704. [CrossRef]
4. Santero, N.J.; Masanet, E.; Horvath, A. Life-cycle assessment of pavements. Part I: Critical review.
Resour. Conserv. Recycl. 2011, 55, 801–809. [CrossRef]
5. Bressi, S.; Santos, J.; Giunta, M.; Pistonesi, L.; Lo Presti, D. A comparative life-cycle assessment of asphalt
mixtures for railway sub-ballast containing alternative materials. Resour. Conserv. Recycl. 2018, 137, 76–88.
[CrossRef]
6. American Association of Highway and Transportation Officials (AASHTO). Guide for Design of Pavement
Structures; AASHTO: Washington, DC, USA, 1993.
7. American Association of Highway and Transportation Officials (AASHTO). Mechanistic-Empirical Pavement
Design Guide: A Manual of Practice; AASHTO: Washington, DC, USA, 2008.
8. Huang, Y.H. Pavement Analysis and Design, 2nd ed.; Prentice Hall Inc.: Upper Saddle River, NJ, USA, 2004.
9. Stroup-Gardiner, M.; Turochy, R.E.; Carter, A. Influence of various material and traffic inputs on flexible
pavement design methods for Alabama roadways. J. Transp. Eng. 2007, 133, 455–461. [CrossRef]
10. Baus, R.L.; Fogg, J.A. AASHTO flexible pavement design equation study. J. Transp. Eng. 1989, 115, 559–564.
[CrossRef]
11. Schwartz, C.W.; Li, R.; Ceylan, H.; Kim, S.; Gopalakrishnan, K. Global Sensitivity Analysis of
Mechanistic–Empirical Performance Predictions for Flexible Pavements. Transp. Res. Rec. 2013, 2368, 12–23.
[CrossRef]
12. Haider, S.W.; Harichandran, R.S. Effect of axle load spectrum characteristics on flexible pavement performance.
Transp. Res. Rec. 2009, 2095, 101–114. [CrossRef]
13. Hajek, J.J.; Billing, J.R. Trucking trends and changes that affect pavements. Transp. Res. Rec. 2002, 1816, 96–103.
[CrossRef]
14. Associazione Italiana Società Concessionarie Autostrade e Trafori (AISCAT). Informazioni (Notiziario
Trimestrale). (In English: Information). 2001–2018. Available online: http://www.aiscat.it/pubblicazioni.htm
(accessed on 23 January 2020).
15. Consiglio Nazionale delle Ricerche (CNR), (Italian National Research Council). Catalogo Delle Pavimentazioni
Stradali (In English: Road Pavements Catalogue); CNR: Roma, Italy, 1995.
16. Asphalt Institute. Thickness Design—Asphalt Pavements for Highways and Streets (No. 1); Asphalt Institute:
Lexington, KY, USA, 1981.
17. Stuckler, D.; Basu, S.; Suhrcke, M.; McKee, M. The health implications of financial crisis: A review of the
evidence. Ulst. Med. J. 2009, 78, 142.
18. Eurostat. Regions in the European Union. Nomenclature of Territorial Units for Statistics—NUTS 2016/EU-28;
Publication Office of the European Union: Luxembourg, 2018.
19. Santos, J.; Ferreira, A. Pavement design optimization considering costs and preventive interventions. J. Transp.
Eng. 2012, 138, 911–923. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).