Transportation Research Part D 8 (2003) 415432

www.elsevier.com/locate/trd

Modelling operational, economic and

environmental performance of an air transport network
Milan Janic
OTB Research Institute, Delft University of Technology, Thijsseweg 11, 2629 JA Delft, The Netherlands
Accepted 17 March 2003

Abstract

The paper models the operational, economic and environmental performance of an air transport net-
work consisting of airports and air routes connecting them. The operational capacity represents the op-
erational performance. Thresholds on the networks environmental burdens reect the environmental
performance. The economic performance comprises the networks prots. Modelling the network perfor-
mance includes using integer programming techniques to maximise total network prots for given oper-
ational capacity and environmental constraints under conditions where environmental externalities are
internalised.

2003 Elsevier Ltd. All rights reserved.

Keywords: Air transport network; Performance; Capacity; Environment; Prots; Externalities; Optimization

1. Introduction

Civil aviation has been one of the worlds fastest growing industries (International Civil Avi-
ation Organisation, 1994; Lyle, 1999). Such growth has had positive and negative eects. The
positive eects have included the employment within and around the sector, and the stimulus to
both the local and global economy. The negative eects have consisted of the externalities such as
the aircraft noise at airports, air pollution on a local and global scale, the incidental contami-
nation of soil and water, change of land use, air accidents, and congestion (Air Transport Action
Group, 1996a,b; European Commission, 1996; Janic, 1999). Both types of eect related to each

E-mail address: janic@otb.tudelft.nl (M. Janic).

1361-9209/$ - see front matter
2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S1361-9209(03)00041-5
416 M. Janic / Transportation Research Part D 8 (2003) 415432

other raising questions how to dene an appropriate balance under given circumstances. 1
(EUROCONTROL/ECAC, 2001a; Environmental Protection Agency, 1999).
This paper models the operational, economic and environmental performance of an air
transport network. The number of ights carried out in the network is determined to maximise
network prots and meet given operational and environmental constraints designed to internalise
environmental externalities.

2. Performance of an air transport network

An air transport network consists of airports and air routes that connecting them. The network
may have dierent spatial congurations, depending on its physical size, the number and spatial
location of airports included, the pattern of the passenger and cargo demand, and the schedules
and aircraft types the airlines use to serve demand.
The network capacity reects the maximum number of ights, which can be carried out over
the networks componentsairports and air routesunder given conditions. The conditions
specify the ight pattern and the constraints implemented to enable safe operations, and to keep
the particular environmental burdens under control (Janic, 1986).
Network prots comprise the prots of the operators such as airlines, airports and air trac
control, determined as the dierences between their operational revenues and costs, including the
internalised costs of particular burdens (Doganis, 1992).
Operational safety constraints involve the application of the air trac control separation rules
to provide safe ights. If the minimum separation rules are applied under conditions of the
constant ight demand, the number of ights served at the airports and on the air routes will be
maximised. This represents the operational capacity and it is usually specied for a given period of
time (Horonje and McKelvey, 1994; Janic, 1986).
Environmental constraints limit the burdens of the air transport networks harmful impacts on
peoples health and environment. In many cases, these constraints have the institutional or po-
litical character. They are implemented as the thresholds for the particular types of burdens
(Intergovernmental Panel on Climate Change, 1999). In order to keep the actual burdens below
the thresholds, their main generatorsthe number of ights serving the passenger and cargo
demandneeds to be constrained or controlled. In such case, the maximum number of ights that
generates the burden meeting the threshold can be considered as the environmental capacity. 2
This capacity can be determined for a given network component, for example an airport, or with
respect to dierent burdens, for example, noise or air pollution.

1
Many busy European airports have control noise, and Zurich and Stockholm Arlanda also control air pollution. In
the US, noise has been controlled at about 400 airports.
2
At European airports, one of the earliest constraints on operations to protect local people from noise was at
London Heathrow airport in the early 1970s. The annual number of operations was limited to 270 000. However, under
strong pressure from airlines this was lifted in 1991. A recent case has involved Amsterdam Schiphol airport where
legislation limited annual number of ights to 380,000 in 1998 with the possibility of an annual increase of 20000 until
2003. At Zurich and Washington International airport, quotas on the number of ights have been introduced to control
air pollution and noise, respectively (Oerman and Bakker, 1998).
 M. Janic / Transportation Research Part D 8 (2003) 415432 417

3. Modelling performance of an air transport network

3.1. Problem formulation

Modelling of the air transport network performance aims to maximise the network pro-
ts under given operational and environmental constraints. Usually, an objective function is
dened to represent prots. The constraints are represented by the capacity of the particular
network components in terms of the operational safety, noise and air pollution thresholds,
respectively. The objective function and constraints are assumed to be linear functions
of the decision variablesthe maximum number of ights allowed to be carried out
in the network. These variables are the non-negative integers, that together with the linearity
of the objective function and constraints enable the use Integer Programming (IP) techniques
to obtain optimal solutions. This approach represents a continuation of the work of Ferrar
(1974) that dealt with allocation of an airports departure capacity under dierent types of
constraints.
Let t be the time period in which modelling of the network performance is carried out, and
where:

N , K, L are the number of the airports, airlines and aircraft types in the network, respectively;
i, j are indices of the ight origin and destination airport, respectively i 6 j; i; j 2 N ;
k, l are indices of airline and aircraft types, k 2 K; l 2 L;
Rijkl , Cijkl are the revenue and cost per ight, carried out by the aircraft type l of airline k
between airports i and j;
Xijkl is the maximum number of the ights allowed to be carried out by aircraft type l of air-
line k between airports i and j;
ba=jikl , bd=ijkl are the amounts of the air pollutants emitted by aircraft type l of airline k at the
airport i, while coming from and leaving for the airport j;
cjikl , cijkl are the amounts of the air pollutants emitted by aircraft type l of airline k while
cruising along the route between airports j and i;
aa=jikl , ad=ijkl are the noise levels generated by aircraft type l of airline k at airport i, while
coming from and leaving for airport j;
Kai , Kdi are the operational capacities of airport i for the arrivals and departures;
paji , qdij are the portions of the operational capacity of airport i allocated to the ights on the
incoming route and outgoing route ji and ij;
Bi is the air pollution quota for airport i;
Bij , Bji are the air pollution quotas for the en-route airspace between the airports i and
j;
Aai , Adi are the noise quotas at airport i for the arriving and departing ights;
rikl , sikl are the portions of the available noise quota assigned to the aircraft type l of airline
k during its arrival at and departure from airport i;
Ki is the optimal capacity of airport i i 2 N ;
Kij is the optimal capacity of a route between the airports i and j, i; j 2 N ; and
K is the capacity of the air transport network that maximises prots under given operational
and environmental constraints.
418 M. Janic / Transportation Research Part D 8 (2003) 415432

The IP formulation of the problem can be stated as (Winston, 1994):

Maximise the objective function, i.e., the prots of given air transport network:
X
max Rijkl  Cijkl Xijkl 1
ijkl

subject to :
Constraints on the operational capacity of airports
Allocation of the airports operational capacity to the arrival and departure routes:

Incoming tracarrivals
X
Xjikl 6 paji  Kai i; j 2 N 2a
kl=j6i

Outgoing tracdepartures
X
Xijkl 6 qdij  Kdi i; j 2 N 2b
kl=i6j

P P
where j paji j qdij 1.

Balancing the number of the incoming and outgoing ights:

X X
Xjikl  Xijkl P 0 i 2 N 2c
jkl=j6i jkl=i6j

Balancing the airline and aircraft types on the incoming and outgoing routes:
X X
Xjikl  Xijkl P 0 i; j 2 N 2d
kl=j6i kl=i6j

Constraints on the environmental thresholds at airports

Aircraft noise:
Policy 1: The noise quota is allocated in proportion to the noise from the expected ights on the
incoming and outgoing routes.

Incoming ightsarrivals
X
aa=jikl  Xjikl 6 paji  Aai i; j 2 N 3a
kl=j6i

Outgoing ightsdepartures
X
ad=ijkl  Xijkl 6 qdij  Adi i; j 2 N 3b
kl=i6j
 M. Janic / Transportation Research Part D 8 (2003) 415432 419

Policy 2: The noise quota is allocated in proportion to ight and aircraft types.

Incoming ightsarrivals
X
aa=jikl  Xjikl 6 rikl  Aai k 2 K; l 2 L; i 2 N 3c
j=j6i

Outgoing ightsdepartures
X
ad=ijkl  Xijkl 6 sikl  Adi k 2 K; l 2 L; i 2 N 3d
j=i6j
P P
where kl rikl kl sikl 1.

Air pollution:

Airports-local scope
X
ba=jikl  Xjikl bd=ijkl  Xijkl 6 Bi i2N 4a
jkl=i6j
En-route airspace-global scope
X
cjikl  Xjikl=t cijkl  Xijkl 6 Bji Bij i; j 2 N; i 6 j 4b
kl=i6j

Constraints on the non-negativity of decision variables

Xijkl P 0 integers for i; j 2 N ; i 6 j; k 2 K; l 2 L 5
The optimal solution equation (1) is the maximum number of ights carried out at the par-
ticular airports and along the air routes under the constraints (2)(5).
Sub-constraints (2a) and (2b) limit the number of ights to the operational capacity of the
airports incoming and outgoing routes. The sub-constraints (2c) and (2d) provide a balance in the
total number of arrivals and departures, the number of incoming and outgoing ights carried out
by the dierent aircraft types operated by dierent airlines, and the number of aircraft on each
airport arrival and departure route.
Constraints (3a)(3d) limit the maximum noise from the ights on the airport incoming and
outgoing routes to the noise quotas. The dierent, mutually exclusive, policies can be applied to
allocate these quotas to the particular routes and ights. Similarly, the constraints (4a) and (4b)
keep the maximum air pollution from ights at the airports and air routes bellow or at the allowed
air pollution quotas. Finally, the constraints (5) ensure the decision variablesthe number of
ightsare the non-negative integers.
If D Xijkl


is the decision matrix, which maximises the objective equation (1) subject to
constraints (2)(5), the optimal capacity of the air transport network and its components can be
synthesised as follows i; j 2 N ; i 6 j; k 2 K; l 2 L; t 2 T :

For an airport i
X
Ki 
Xjikl 
Xijkl 6a
jkl
420 M. Janic / Transportation Research Part D 8 (2003) 415432

For a route between the airports i and j

X
Dij 
Xjikl 
Xijkl 6b
kl

For the air transport network

X X
K Ki Kij 6c
i ij

3.2. Principles of determining the particular parameters of the model

3.2.1. Coecients of the objective function

Revenues
The revenues gained by an air transport network consist of the direct revenues of the network
operators such as the airlines, airports and air trac control. The benets of air travel for the
users and cargo consignors themselves and the indirect benets for local communities are not
taken into account (Button and Stough, 1998).
The revenues of the network operators are expressed per ight. In such a context, the airline
revenue per ight depends on the airfares and the number of passengers on board. It generally
increases as these factors increase (Doganis, 2002). Assuming that the volume of passenger and
freight demand is always at a level sucient to make the maximum allowed number of ights in
the network at least zero-protable, the airlines may have an interest in carrying out as many as
possible greater, fuller and more expensive ights in order to rise the revenues.
The airport revenue per ight consists of the aeronautical and non-aeronautical part. In this
case, the aeronautical part mostly consists of the aircraft and passenger fees and the non-aero-
nautical part consists of the revenues obtained from other not directly aviation businesses. The
revenue from both parts generally increases in line with increasing of the number of ights and
passengers. Consequently, the airports may seek to raise their revenues by accommodating the
larger number of larger aircraft and greater number of passengers.
Air trac control mainly collects revenues by charging the navigation fees to the airline ights.
The particular fee per ight increases with increasing of ying distance and aircraft weight.
Therefore, in order to recover its operating costs and gain prots more easily, air trac control in
a given area may seek to collect fees from as many longer ights as possible.

Costs
The costs of an air transport network include the direct cost of operators and the internalised
externalities. Similarly as in case of revenues, these costs are expressed per ight. The airlines
direct operational cost per ight embraces the costs of the resources (aircraft, labour, and
energy) spent to carry it out. This cost generally increases with increasing of the aircraft
size, number of passengers on board, and ying time (Doganis, 2002). An airports direct op-
erational cost per ight contains the costs of resources spent accommodating the passengers and
aircraft. This cost generally increases with increasing of the aircraft size and number of pas-
sengers.
The operational cost per ight of the air trac control embraces the cost of the resources
(facilities, equipment, energy, and labour) spent on managing and controlling the ight within a
 M. Janic / Transportation Research Part D 8 (2003) 415432 421

given airspace. This cost is mainly proportional to the quality of controlling dependent on so-
phistication of the control facilities and equipment (EUROCONTROL, 1999).
Externalities may include air pollution, noise, and air accidents (Air Transport Action Group,
1996b; Levison et al., 1996). The cost of air pollution, noise and air accident per ight depends on
the pre-emptive or direct cost of damage caused by the related burdens (impacts). In an absolute
sense, each of the above externalities generally increases with increasing of the aircraft size, the
number of passengers on board, and duration of ight. In addition, congestion cost is considered
as an additional externality. In this case, this cost exceptionally includes only the private passenger
and airline cost of the ight delay and not the internalised cost of marginal delays, which this
ight imposes on the other ights during the congestion (Daniel, 1995). This cost generally in-
creases with increasing of the value of passenger time, number of passengers on board, aircraft
operating cost and delay.

3.2.2. Technological coecients and the right-hand side of constraints

Constraints related to the operational capacity
For the sub-constraints (2a) and (2b), the technological coecients have the value one for both
types of the decision variablesthe incoming and outgoing ights at airports, Xjikl and Xijk .
The right-hand sides, Kai and Kdi , as the airports operational arrival and departure capacity,
respectively, based on the minimum air trac control separation rules, can be allocated to the
routes and airlines according to the dierent slot-allocation policies. For example, such policies
can reserve an adequate capacity in terms of slots in proportion to the expected number of ights
on particular routes, exclusively for the ights of particular airlines, or for the specic type of
ights. 3 The parameters paji and qdij reect the use of the rst among the above policies.

Constraints related to the allowed noise quotas

 Technological coecients: In the sub-constraints (3a) and (3b), the technological coecients
aa=jikl and ad=ijkl represent the noise in terms of the sound energy 10Laji=kl =10 and 10Ldij=kl =10 generated
by an arrival and departure ight, respectively. The parameters Laji=kl and Ldij=kl represent the
aircraft noise in dB(A) (decibels), which is usually measured at the noise measurement locations in
the vicinity of airports (International Civil Aviation Organisation, 1993b).
 The right-hand side: The noise quotas at airports, Aai and Adi as the right-hand side of the sub-
constraints (3a) and (3b) comprise the total noise energy in terms of t  10Lai=eq =10 and
t  10Ldi=eq =10 allowed to be generated by the incoming and outgoing ights during time t where
the L:i=eq s are the corresponding average energy sound levels, which accumulate all the sound
energy for multiple ights. 4 (Horonje and McKelvey, 1994).
Dierent policies can be applied to allocate the noise quotas Aai and Adi to the arriving and
departing ights, airlines and aircraft types. For example, Policy 1 can exclusively allocate these
quotas in proportion to the noise from the ights on the particular incoming and outgoing routes.

3
At the capacity-constrained and busy airports such policy may save slots for the short hauls, new entrants, etc.
(Coleman, 1999).
4
The noise quota is zero during the night ying ban. During the day, it can be, for example 57 dB(A) at London
Heathrow, 85 dB(A) at Birmingham, and 73 dB(A) at Frankfurt airport (Department of the Environment, Transport
and Regions, 2000).
422 M. Janic / Transportation Research Part D 8 (2003) 415432

In such case, the parameters paji and qdij in the sub-constraints (3a) and (3b) will be identical to
those in (2a) and (2b). Policy 2 can exclusively allocate the above noise quotas with respect to the
noise performance of the particular aircraft types. The universal criterion to be applied should
encourage the access of quieter and discourage the access of noisy aircraft. The parameters rikl and
sikl in the sub-constraints (3c) and (3d) reect the possible application of this policy.

Constraints related to the air pollution quotas

 Technological coecients: In the sub constraints (4a), the technological coecients ba=jikl and
bd=ijkl as the amount of air pollution emitted by the corresponding ight at given airport (kilo-
grams/ight) depends on the time, i.e., duration of the landing/take-o (LTO) cycle, specic fuel
consumption, and the emission index. The specic fuel consumption depends on the type of
aircraft engines. The emission index is the function of the type of air pollutant, aircraft engine,
specic fuel consumption, and the LTO cycle mode. The particular modes of the LTO cycle are
standardised. In this paper, the rst standardised modeapproach is analogous to an arrival, the
next two modestake-o and climbto a departure, and the last taxi/ground idle standardised
mode to the sum of the ight taxiing-in and taxiing-out time. The air pollution from the airport
passenger and aircraft ground-service vehicles, as well as the air pollution from the airport surface
access systems can also be appropriately added to the air pollution of a given ight (International
Civil Aviation Organisation, 1993a).
In the sub-constraints (4b), the technological coecients cjikl and cijkl represent the amount of
the air pollutants emitted during the cruising phase of the corresponding ight. These amounts are
proportional to the fuel consumption and emission index (both being dependent on the type
of aircraft engines), ight regime (speed/altitude), ying time, and trac and the weather con-
ditions.
 The right-hand side: The right-hand side of the sub-constraints (4a), Bi is dened as the
amount of air pollutants or quota (in kilograms) allowed to be emitted at given airport during
given period of time. This quota can be determining by using either local or global criteria, or
both (Environmental Protection Agency, 1999). In this paper, the quota is calculated as the
product of the allowable concentration of given air pollutants and the volume of the related area
around the airport. If the allowed concentration is gi (kilograms of pollutants per cubic kilo-
metre) and the volume of the relevant airspace is Vi (cubic kilometres), their product will be
Bi gi Vi . Alternatively, the air pollution quota Bi could be determined by trading-o with the air
pollution quotas between airports in the network (Intergovernmental Panel on Climate Change,
1999).
The right-hand side of the sub-constraints (4b), the air pollution quotas Bji and Bij can be
expressed by the amount of air pollution allowed in the en-route airspace between the corre-
sponding airports. The quantifying procedure is similar to that at airports, but it should take into
account specics of the airspace, type and nature of the air pollution. 5

5
At ground level the pollutants of relevance at airports area HCs (hydrocarbons), CO (carbon monoxide) and NOx
(nitrogen oxides) (International Civil Aviation Organisation, 1993a). In the en-route airspace where the cruising phase
of most sub-sonic ights takes place, the troposphere and the low stratosphere, the air pollutants are NOx , CO2 (carbon
dioxide), and H2 O (water vapour).
 M. Janic / Transportation Research Part D 8 (2003) 415432 423

4. An example of modelling the network performance

4.1. Input data

4.1.1. Network congurationthe case around London Heathrow airport

The modelling of the air transport network performance is demonstrated on the simple network
consisted of one airport and the set of the incoming and outgoing routes. The case relates to the
air transport network around London Heathrow airport. At present, about a hundred routes and
hundred airlines operating 40 dierent aircraft types connect the airport to the other worlds
airports. Such high diversity of the routes and airlines implies the existence of the large network
optimization problem of about 2  100  40 decision variables and 10 constraints, which makes
the use of the IP technique complex. Therefore, clustering of the air routes and aircraft types
combined with disregarding the dierences between particular airlines is carried out to reduce the
size of the original problem. Consequently, the clusters of routes between Heathrow airport and
domestic UK, European and US market areas are formed. These clusters comprise about 90% of
the Heathrow trac (Airline Business, 1999). However, due to the still very high diversity in
length of the particular clustered routes, the clusters are additionally grouped, this time into seven
classes, and thus the original problem converted into its manageable surrogate consisting of 28
decision variables (seven routes and four aircraft categories) and 10 constraints (Groenewege,
1999). The classes of routes as well as other relevant inputs such as the aircraft types and cate-
gories, their seat capacity and the average route load factor are given in Table 1.

4.1.2. Constraints
Operational constraints: At present, the system of two parallel runways, one for arrivals and
another for departures, is used at Heathrow airport during the period of t 19:5 h/day. The
change of two parallel runways at 3 p.m. every day and the night ying ban between 11.30 p.m.
and 4.00 a.m. are already in place to relieve the noise burden. Under such circumstances, the
airport declares an operational capacity of 78 operations/h (39 arrivals and 39 departures) (De-
partment of the Environment, Transport and Regions, 2000). The above-mentioned European
and European/US ights use about 90% of this capacity. Consequently, the airports operational
capacity allocated to this trac is determined as: K0 78 (ights/h)  19.5 (h/day)  0.90 1370
(ights/day). Since the number of arrivals and departures is approximately balanced during the
day, the corresponding operational arrival and departure capacity is determined as:
Ka0 Kd0 1370=2 685 (aircraft/day) (the index 0 instead of i is used to simplify the nota-
tion).
Each of the above capacities is allocated to seven route classes in proportion to the expected
number of ights (i.e., according to the present slot-allocation policy). Since Heathrow airport is
mostly origin/destination airport for ights, the portions of the operational capacity (slots) al-
located to each incoming and outgoing route are adopted to be symmetrical and equal to:
pa10 pa20 pa30t pa40 qd01 qd02 qd03 qd04 0:68 (European routes); pa50 qd05 0:15
(European/US East Coast routes); pa60 qd06 0:10 (European/US Central routes);
pa70 qd07 0:07 (European/US West Coast routes) (Airline Business, 1999). The operational
capacity of the en-route airspace between Heathrow and other airports is assumed to be unlim-
ited.
424 M. Janic / Transportation Research Part D 8 (2003) 415432

Table 1
Characteristics of the classes of air routes, aircraft types and trac in example
Route Explanation Average Airlinea Typical Aircraft Aircraft LFc
(index) (j) route aircraft category capacity
length typeb (index) (seats)
(km) (d) (l)
1 UK domestic and 402 European B737/ 1 140 0.70
Closest EU (North) A320
2 Scandinavia and 816 European B737/ 1 140 0.70
Central EU and A320
Central Europe
3 South EU and 1461 European B737/ 1 140 0.70
Mediterranean A320
4 Eastern Europe 2380 European B737/ 1 140 0.70
A320
5 North America 6012 European/US: B767/ 2 250 0.70
(East Coast) Virgin/BA/American/ A330
United
6 North America 7440 European/US: B777/ 3 350 0.70
(Central) Virgin/BA/American/ A340
United
7 North America 8630 European/US: B747 4 370 0.70
(West Coast) Virgin/BA/American/
United
a
Due to the institutional restrictions (Bermuda II), only four airlines y between Heathrow and US airports.
b
These airlines operate more than 90% of this eet news@airwise.com.
c
LFRoute load factor represents to the averages for the industry (Airline Business, 2000).

 Environmental constraints: Noise. In this example, all aircraft of the same category, inde-
pendently on the route, are assumed to generate approximately the same level of arrival and
departure noise. Table 2 contains the noise per individual aircraft category measured at points
relevant for the aircraft noise certication (Smith, 1989).
Table 2 is used to compute the coecients aa=j0=l and ad=0j=l of the noise sub-constraints (3a)
and (3b). The noise quotas for the arrival and departure ights Aa0 and Ad0 are assumed to be
identical and determined for the various values of the parameters Lai=eq and Ldi=eq . These values are
assumed as they are measured at the aircraft noise certication points around the airport. The
noise quotas are allocated according to Policy 1, i.e., similarly as the airports operational ca-
pacity, and according to Policy 2, in proportion to the share of aircraft categories (types). In the
latter case, the allocation looks as follows: r0==1 s0==1 0:68 (Category 1); r0==2 s0==2 0:11
(Category 2); r0==3 s0==3 0:11 (Category 3); and r0==4 s0==4 0:10 (Category 4).
Air pollution. The air pollutions per the LTO cycle, as the coecients ba=j0=l and bd=0j=l of the
sub-constraints (4a) are determined by using data on the intensity of the emission of pollutants
HCs, CO, and NOx per aircraft engine (kg/s), the number of engines, the aircraft maximum take-
o weight, and the duration of particular modes of the LTO cycle (j 17, l 14) (Defence
Evaluation and Research Agency, 1999). Regression technique is used to establish the relation-
ships between the intensity of emissions per mode of the LTO cycle, EIm (dependent variable) and
the aircraft maximum take-o weight W (independent variable). Half of the taxi-idle time of the
 M. Janic / Transportation Research Part D 8 (2003) 415432 425

Table 2
The aircraft noise characteristics in given example
Aircraft category (l) Noise dB(A)a
Arrival, La==l Departure, Ld==l
1 80 74
2 90 75
3 90 78
4 94 93
a
Adapted from Smith (1989).

Table 3
Dependence of the intensity of emission per aircraft on the aircraft take-o weight during LTO cycle
LTO cycle (mode) Code of LTO Duration of LTO Intensity of emissions
mode (m) mode (min)a per aircraft (g/s)b
Take-o 1 0.7 EI1 0:832W 1:026 ,
R2 0:873; N 33
Climb 2 2.2 EI2 0:755W 0:959 ,
R2 0:907; N 33
Taxi-idle 3 26.0 EI3 0:225W 0:754 ,
R2 0:827; N 33
Approach/landing 4 4.0 EI4 0:367W 0:759 ,
R2 0:856; N 33
a
International Civil Aviation Organisation (1993a).
b
Air pollutantsHCs, CO, NOx ; 50 6 W 6 400 tons (W typical aircraft take-o weight).

LTO cycle is allocated to the taxiing-in and another half to the taxiing-out phase, which reects
the most common situation at Heathrow airport. The results are given in Table 3.
By using the regression equations in Table 3 and the aircraft typical take-o weight in Table 4,
the amount of air pollution for the arrival and departure part of the LTO cycle for the particular
aircraft categories is calculated and given in Table 4.
In calculating B0 in the sub-constraints (4a), the relevant volume of airspace around the airport
is assumed to be V0 12 (km3 ) (the airspace is cylindrical with a base of 12 (km2 ) and height of
1 (km)). The total tolerable concentration of the relevant air pollutants HCs, CO, and NOx for the

Table 4
Amount of pollutants emitted by particular aircraft categories during LTO cycle in given example
Aircraft category(l) Typical aircraft take-o The amount of the emitted pollutants(kg/aircraft)
weight (tons)a Arrivals, ba=j0=l W b Departures, bd=0j=l W c
1 75 6.844 13.744
2 206 14.773 34.517
3 306 19.922 49.663
4 375 23.231 59.919
a
Flight International (1999a,b).
b
bad=0j=l W 88:08W 0:759 175:5W 0:754 .
c
bd=0j=l W 175:5W 0:754 34:94W 1:026 99:66W 0:959 .
426 M. Janic / Transportation Research Part D 8 (2003) 415432

period of t 1 day is estimated to be g0 0:0246 (g/m3 ) 24,600 (kg/km3 ), which gives the total
allowable amount of the air pollutants of B0 24; 600 (tons/km3 )  12 (km3 )  (19.5/
24) 239,850 (tons/day) (Department of the Environment, Transport and Regions, 1999).
The air pollution quotas Bj0 and B0j in the sub-constraints (4b) for the air-routes connecting
Heathrow to other airports are unlimited.

4.1.3. Coecients of the objective function

The revenue and cost per ight as the coecients of the objective function 1 are determined for
each route class and each ight category. The revenue is calculated as the sum of the airport and
airline revenue per ight. The cost per ight is calculated as the sum of the airport and airline
operational costs and the internalised externalities per ight. The revenues and costs per ight of
the air trac control are assumed as already contained in the airline revenues. These calculations
implicitly take into account the institutional restrictions (Bermuda II) of the access to Heathrow
airport for European non-EU and US airlines. Table 5 contains the calculated revenues and costs
per route class and ight category (j 17; l 14). As can be seen, if given externalities are
taken into account (internalised), some short haul ights become unprotable. 6

4.2. Analysis of the results

Using inputs from Tables 15, two sets of experiments are carried using the software package
LINDO. The purpose is to investigate how dierent types of environmental constraints aect
network, economic and the operational performances. In the rst set of experiments, the airport
noise quotas Aa0 and Ad0 are varied. In the second set the airport air pollution quota B0 is varied.
In both cases, the operational capacities Ka0 and Kd0 are constant. The results are shown in Figs.
13.

4.2.1. Inuence of the noise constraints

Fig. 1a shows utilisation of the airports operational capacity as aected by noise quotas when
they are allocated to incoming and outgoing routes in proportion to expected trac. The available
noise quotas Aa0 65 and Ad0 65 based on the continuous noise of La0eq Ld0eq 65 dB(A), are
allocated using Policy 1. The air pollution quota is fully usedB0 239:9 tons/day.
Maximum prots are realised by full utilisation of the operational capacity allocated to Routes
3 and 4 (mediumlong European ights), and the partial utilisation of the capacity allocated to
Routes 5, 6 and 7 (European/US ights). Routes 1 and 2 are excluded from the optimal solution
because these handle short range unprotable European ights. Consequently, the optimal ca-
pacity K0 reaches only 420 ights per day, i.e., about 30% of the maximum, while prots are about
70% lower when the noise and air pollution constraints are not included. Under these circum-
stances, Policy 2 for allocation of the same noise quotas has produced a similar eect.
Fig. 1b shows the situation when unprotable ights and corresponding routes are excluded.
Operational capacities are allocated to protable ights. Other conditions remain unchanged.

6
This is similar to the results of some previous studies that support the idea that as externalities are internalised,
some ights might become unprotable and are replaceable by surface transport alternatives (Levison et al., 1996).
 M. Janic / Transportation Research Part D 8 (2003) 415432 427

Table 5
The revenues and costs per ight as coecients of the objective function in given example
Route/ Average cost ($/ight), C0=j=l  Cj=0==l Average revenue Prot
aircraft ($/ight), ($/ight)
category R0=j=l  Rj=0==l
0/j/l Airlinea Airportb Internalised externalities Airlinea Airportb
(Noise/air Congestion
pollution/air (average delay)d
accidents)c
Passen- Airline6
gers5
1/1 7970 2842 158 163 453 8027 3528 )34
2/1 14,978 2842 318 163 535 15,200 3528 )108
3/1 23,786 2842 570 163 537 24,419 3528 49
4/1 32,659 2842 928 163 484 34,089 3528 541
5/2 76,863 5075 4187 583 484 96,637 6300 15,614
5/3 107,525 7105 5891 817 677 131,091 8820 17,896
5/4 113,669 7511 6227 863 715 138,538 9324 18,877
6/2 95,046 5075 5182 583 488 115,878 6300 15,804
6/3 133,065 7105 7254 817 683 162,229 8820 22,125
6/4 140,669 7511 7669 863 722 171,449 9324 23,359
7/2 110,249 5075 6011 583 490 134,412 6300 18,324
7/3 154,348 7105 8415 817 686 188,177 8820 26,226
7/4 163,168 7511 8896 863 725 198,930 9324 27,091
a
Airline operational cost: European routes (1,2,3,4): Cd 0:218e0:000186d ($/pkm); R2 0:644; N 16/European/
US routes (5,6,7): Cd 0:073 ($/pkm) (Airline Business, 2000); Airline revenue: European routes (1,2,3,4):
Rd 0:218e0:000168d ($/pkm); R2 0:525; N 16/European/US routes (5,6,7): Rd 0:089 ($/pkm) (Airline Busi-
ness, 2000) where pkm is passenger-kilometre; d is route lengthkm.
b
The average airport cost and revenue per passenger at Heathrow airport are estimated to be 29 and 36 $/pass,
respectively (Airline Business, 1999).
c
Adapted from Levison et al. (1996): Cost of noise (UK case)0.002 $/pkm; Cost of air pollution independently on
type and scope0.00098 $/pkm; Cost of air accidents0.001$/pkm.
d
Based on the average delay of 5 min per an arriving and departing ighttypical for Heathrow airport (EURO-
CONTROL/ECAC, 2001b); The average ying time on a route is: td 0:62 0:0021d; The value of passenger time is
adopted to be 20 $/pass/h for the European and 40 $/pass/h for the European/US routes (Federal Aviation Admin-
istration, 1995; Levison et al., 1996).

This change increases utilisation of the airports operational capacity to K0 1072 ights per
dayabout 80% of its maximum. However this increases prots by 5% compared to the case in
Fig. 1a.

4.2.2. Inuence of the air pollution constraints

Fig. 2a shows the inuence of the air pollution quota B0 on the airports prots and the util-
isation of its operational capacity. The quota is set at 15.0 tons, i.e., at only 6% of the maximum
quota. The arrival and departure noise quotas are determined for the maximum continuous sound
level of 85 dB(A), i.e., as Aa0 85 and Ad0 85, which means removal of these noise constraints.
Nevertheless, these quotas are allocated to the routes and ights according to Policy 1.
428 M. Janic / Transportation Research Part D 8 (2003) 415432

(a)

(b)

Fig. 1. Inuence of noise constraints on airport performance when (a) operational capacity is allocated to all ights; (b)
operational capacity is allocated only to protable ights.

The results are quite dierent than those in Fig. 1. The air pollution constraints also aect the
airport performance. Prots are maximised by favouring more protable, but more polluting,
long-distance intercontinental ights on Routes 57 and by excluding both the unprotable and
protable European ights on Routes 14. Excluding the European ights and routes signicantly
reduces utilisation of the airports operational capacityto about 70%and yields lower prots
8% compared to the situation with no air pollution quotas.
Fig. 2b shows that allocating the operational capacity exclusively to the protable ights has
not improved the overall airports operational and economic performance in comparison to those
in Fig. 2a.

4.2.3. The inuence of the gradual relaxation of the particular constraints

Fig. 3 shows the relationships between the airport prots and the utilisation of the oper-
ational capacity as inuenced by the relaxation of the noise and air pollution constraints. The
 M. Janic / Transportation Research Part D 8 (2003) 415432 429

(a)

(b)

Fig. 2. Inuence of air pollution constraints on airport performance when (a) operational capacity is allocated to all
ights; (b) operational capacity is allocated only to protable ights.

150 tons 239.9 tons

15.0 tons
85 dB(A)

10.0 tons

5.0 tons 75 dB(A)

65 dB(A)
55 dB (A)

Fig. 3. Dependence of airport performance on the relaxation of environmental constrainst.

430 M. Janic / Transportation Research Part D 8 (2003) 415432

operational capacity is allocated only to the protable ights. The noise capacity is allo-
cated in proportion to the noise of the expected ights on particular routes, i.e., according to
Policy 1.
The lower curve shows how the utilisation of the airports operational capacity and corre-
sponding prots increase after gradual relaxation of the noise constraints Aa0  and Ad0 . The
continuous sound levels are varied in the range 55, 65, 75 and 85 dB(A). The air pollution con-
straint is essentially ineective. Gradual relaxation of the noise constraint increases utilisation of
the airports operational capacity. For example, the relaxation of the noise constraint from 55 to
65 dB(A) increases utilisation of operational capacity to about 70% and prots to 32% of their
maximums without constraints. This implies that noise from the numerous, less noisy and less
protable European ights lls this gap. The noise relaxation gaps from 65 to 75 dB(A) and from
75 to 85 dB(A) enable an additional increase in the utilisation of the airports operational capacity
(30%) and prots (70%). The noise gaps are lled by noisier more protable long-distance Eu-
ropean/US ights.
The upper curve in Fig. 3 shows how the utilisation of the airports operational capacity and
prots increase by relaxing the air pollution constraint. The noise constraints are set at 85
dB(A), i.e., practically ineective. The rst relaxation gap to 15.0 tons/day signicantly increases
prots and only modestly the utilisation of the airports operational capacity to about 90% and
32% of the corresponding maxima. This implies that the more air polluting and more protable
European/US ights use this gap. The next relaxation gaps have raise the maximum air pol-
lution level from 15.0 to 150.0, and to 239.9 tons/day. The less air polluting and less protable
medium and long-distance European ights use these gaps. Consequently, the substantial in-
crease in the utilisation of the airports operational capacity by about 70% increases prots
modestly by 8%.

5. Conclusions

The paper has modelled the operational, economic and environmental performance of an air
transport network consisting of airports and air routes connecting them. The network perfor-
mance comprises the network prots and constraints. The former have been maximised by
maximising the number of ights allowed in the network under given constraints on operational
and environmental performance designed to internalise environmental externalities. The prots
and constraints are linear functions of the decision variables that allowed use of Integer Pro-
gramming techniques.
The modelling has been of a simplied network around London Heathrow Airport. In the
numerical experiments, only noise and air pollution quotas have been used. The results show that,
after internalising externalities, some categories of ights have become unprotable, and have
thus been excluded from the optimal solution. Noise and air pollution quotas have signicantly
aected the airports performance in terms of the utilisation of its operational capacity and its
prots. Dierent policies for allocating noise quotas to ights and routes have not produced
signicantly dierent results. In this context, the inuence of noise and air pollution quotas has
diered during their gradual relaxation, but in general, noise constraints seemed to have a
stronger relative impact. Both however, have lost their inuence when set above a threshold.
 M. Janic / Transportation Research Part D 8 (2003) 415432 431

References

Airline Business, 1999. Airports-top 100 rankings. Airline Business (June), 5870.
Airline Business, 2000. The airline ranking. Airline Business (July), 8694.
Air Transport Action Group, 1996a. The Economic Benets of Air Transport. Air Transport Action Group, Geneva.
Air Transport Action Group, 1996b. Air Transport and the Environment. Air Transport Action Group, Geneva.
Button, J.K., Stough, R., 1998. The Benets of Being a Hub Airport City; Convenient Travel and High-Tech Job
Growth, Report, Aviation Policy Program, George Mason University, Fairfax, Virginia.
Coleman, R.J., 1999. Environmentally sustainable capacity. In: Proceedings of the ECAC/EU Dialogue with the
European Air Transport Industry: Airport CapacityChallenges for the Future, Salzburg.
Daniel, J.I., 1995. Congestion pricing and capacity of large-hub airports: a bottleneck model with stochastic queues.
Econometrica 63, 327370.
Defence Evaluation and Research Agency, 1999. Aircraft Engine Exhaust Emission Databank. Defence Evaluation and
Research Agency, London.
Department of the Environment, Transport and Regions, 2000. Noise from Arriving Aircraft: Final Report of the
ANMAC Technical Working Group, Department of the Environment, Transport and Regions, London.
Doganis, R., 2002. Flying-O Course: The Economics of International Aviation. Routledge, London.
Doganis, R., 1992. The Airport Business. Routledge, London.
European Commission, 1996. Towards Fair and Ecient Pricing in Transport: Policy Options for Internalising the
External Cost of Transport in European Union, Green Paper, Bulletin of the European Union, European
Commission, Supplement 2/96, Brussels.
Environmental Protection Agency, 1999. Evaluation of Air Pollution Emissions from Subsonic Commercial Jet
Aircraft, EPA-420R-99-013, ICF Consulting Group, Environmental Protection Agency, Washington, DC.
EUROCONTROL, 1999. Cost of the En-Route Air Navigational Services in Europe, EEC Note 8/99, EUROCON-
TROL Agency, Bretigny-sur-Orge.
EUROCONTROL/ECAC, 2001a. Study on Constraints to Grow, Report, EUROCONTROL Agency, Brussels.
EUROCONTROL/ECAC, 2001b. ATFM Delays to Air Transport in Europe, Annual Report 2000, CODA Brussels,
EUROCONTROL Agency, Brussels.
Federal Aviation Administration, 1995. Total Cost for Air Carrier Delay for the Years 19871994. Federal Aviation
Administration, ReportAPO-130, Washington, DC.
Ferrar, T.A., 1974. The allocation of airport capacity with emphasis on environmental quality. Transportation
Research 8, 163169.
Flight International, 1999a. Commercial airline directory. Flight International (September), 6267.
Flight International, 1999b. Commercial aircraft/engine directory. Flight International (September), 6274/(December),
5261.
Groenewege, A.D., 1999. Air distances between major world airports. In: Compendium of International Civil Aviation
Organisation, Appendix 5. International Aviation Development Corporation, Montreal.
Horonje, R., McKelvey, F.X., 1994. Planning & Design of Airports, fourth ed McGraw Hill Inc., New York.
International Civil Aviation Organisation, 1993a. Aircraft Engine Emissions, Environmental Protection, Annex 16, vol.
2. International Civil Aviation Organisation, Montreal.
International Civil Aviation Organisation, 1993b. Aircraft Noise, Environmental Protection, Annex 16, vol. 1.
International Civil aviation Organisation, Montreal.
Intergovernmental Panel on Climate Change, 1999. Aviation and the Global Atmosphere, Intergovernmental Panel on
Climate Change, Special ReportSummary for Policy Makers, WMOVNEP, Geneva.
Janic, M., 1986. A practical capacity model of an air route network. Transportation Planning and Technology 11, 87
103.
Janic, M., 1999. Aviation and externalities: the accomplishments and problems. Transportation Research D 4, 159
180.
Levison, D., Gillen, D, Kanafani, A., Mathieu, J.-M., 1996. The Full Cost of Intercity TransportationA Comparison
of High Speed Rail, Air and Highway Transportation in California, UCB-ITS-RR-96-3, Institute of Transportation
Studies, University of California, Berkeley.
432 M. Janic / Transportation Research Part D 8 (2003) 415432

Lyle, C., 1999. Maturing industry must continue to cope with challenges of growth in 21st century. ICAO Journal 54,
2024.
Oerman, H., Bakker, M., 1998. Growing pains of major European airports. In: 2nd USA/Europe Air Trac
Management R&D Seminar, Orlando.
Smith, M.J.T., 1989. Aircraft Noise. Cambridge University Press, Cambridge.
Winston, W.L., 1994. Operations Research: Applications and Algorithms, Third ed Duxbury Press, Belmont.