Conference PaperPDF Available


The air transportation network is essential to ensure mobility and connectivity between important metropolises and regions in Europe. Within the overall system, full-service network carriers operate sub-systems consisting of hub and spoke airports. These airlines enable various connections from different destinations by a low number of routes via the hub. Due to increasing traffic volumes in recent years, the system has become more congested, and several hub airports operate at their capacity limits. This has downstream effects on the hub airlines’ operation performance. Capacity expansions are needed to handle the additional traffic efficiently. To assess the impact of these measures, local performance developments and propagation effects into the whole air transportation system should be considered. In this study, a simulation of relevant European airports is used to analyze network behavior under capacity enhancements for selected hub airports and airlines. Airports are rated based on flight delays and queue times at runways. The rate of reached flight connections is used to evaluate the performance of full-service network carriers. The results underline the dependency of airlines on the corresponding hub airport operations. A 10% capacity increase at London Heathrow decreases in- and outbound delay by 42% and 80% and improves the rate of successful flight connections from British Airways by approx. 10%. Significant propagation effects for carriers using London Heathrow as a spoke airport cannot be observed due to the complex large airline networks and the resulting low importance of one spoke airport for the overall network service quality.
Modeling the European Air Transportation Network
considering inter-airport coordination
Daniel Lubig1, Michael Schultz1, Jan Evler1, Hartmut Fricke1, Floris Herrema2, Roc´
ıo Barrag´
an Montes2, Bruno Desart2
1Institute of Logistics and Aviation, University of Technology Dresden, Dresden, Germany
2EUROCONTROL, Brussels, Belgium
Abstract—The air transportation network is essential to en-
sure mobility and connectivity between important metropolises
and regions in Europe. Within the overall system, full-service
network carriers operate sub-systems consisting of hub and
spoke airports. These airlines enable various connections from
different destinations by a low number of routes via the hub.
Due to increasing traffic volumes in recent years, the system
has become more congested, and several hub airports operate
at their capacity limits. This has downstream effects on the
hub airlines’ operation performance. Capacity expansions are
needed to handle the additional traffic efficiently. To assess the
impact of these measures, local performance developments and
propagation effects into the whole air transportation system
should be considered. In this study, a simulation of relevant
European airports is used to analyze network behavior under
capacity enhancements for selected hub airports and airlines.
Airports are rated based on flight delays and queue times at
runways. The rate of reached flight connections is used to
evaluate the performance of full-service network carriers. The
results underline the dependency of airlines on the corresponding
hub airport operations. A 10% capacity increase at London
Heathrow decreases in- and outbound delay by 42% and 80%
and improves the rate of successful flight connections from British
Airways by approx. 10%. Significant propagation effects for
carriers using London Heathrow as a spoke airport cannot be
observed due to the complex large airline networks and the
resulting low importance of one spoke airport for the overall
network service quality.
Keywords—Air traffic network, Capacity management, Airport
performance, Airline performance, Benefit propagation
Europe is one of the major regions in terms of aviation
traffic volume besides North America and Asia [1]. In 2019,
the last complete unaffected year from the significant impact
of the COVID-19 pandemic, over nine million departures
were operated, and more than one billion passengers were
carried. The recent years are characterized by steady growth
in Instrument Flight Rules (IFR) movements and handled
passengers [2]. In parallel, EUROCONTROL, as responsible
central management unit of the European aviation network,
monitors an unsatisfying increase in delay and lower service
qualities partly resulting from a saturation of available ca-
pacities at airports and airspaces. The capacity of an aviation
system is determined by several input factors [3]. This includes
for airports e.g., the available runway configuration, expected
weather influences, and the available Air Traffic Control (ATC)
equipment. Airspace capacities depend e.g., on the traffic
complexity, the airspace design, and separation standards.
An airport consists of many sub-systems, each limited by a
maximum capacity. The determination of these capacities is a
complex procedure that requires the use of different methods,
including mathematical models and observation results during
daily business. The most constraining sub-system limits the
overall airport capacity (bottleneck) [4]. If the demand exceeds
the available capacity, the system becomes congested, which
results in disadvantageous delays [5]. Increasing delays leads
to higher costs e.g., for missed connections from transfer
passengers, and could impact subsequent operations of an
airline [6]–[8]. The management of the scarce resources in
Europe is the responsibility of the Network Manager (NM)
EUROCONTROL predicts a traffic increase of 53% more
flight events until 2040 compared to the traffic values in
2017 [10]. However, the forecast did not consider the impact
of COVID-19, which will probably lead to overestimated
development figures. Other crises and demand decreases (e.g.,
during the financial crisis in 2007) have shown the fast
recovery potential of the air transportation system [11]. Fur-
ther, EUROCONTROL and the European Commission defined
challenging service quality and performance level for the
European aviation system [12,13]. It will not be possible to
achieve these performance targets with the currently avail-
able capacities while maintaining high safety standards [14].
Challenging aspects are the variety of connections and the
coupled air/ground processes within the air transportation
network [15]–[17]. These can lead to downstream effects of
capacity lacks on non-direct participated network systems and
could impact the entire operating day and cause a massive
impact on air traffic performance.
A. Status quo
To increase the capacity of an airport, a wide range of
measures is available, which can be divided into two major
categories. The first one includes investment options like the
physical extension at an airport [18]. This covers, e.g., the
construction of new runways/terminals, improved taxiways
(Rapid Exit Taxiways (RETs) and End-Around Taxiways
(EATs)), or upgraded ATC equipment. As proved in several
types of research, new infrastructure has a hugely beneficial
effect on the airport capacity [19]–[21]. However, regional
political constraints and the structural landscape around the
airport can complicate or avoid physical airport expansions
due to the significant consequences on third parties [18]. Since
infrastructure enhancing measures have a major impact on
the whole airport system and the vicinity environment, the
planning, assessment, and implementation is a time-consuming
process and is based on future traffic forecasts [22]. Structural
extensions can be over- or undersized when the forecasts
are inaccurate. Hence, investment options are connected with
unavoidable risk. The second group of measures includes non-
investment opportunities, which contains improved procedures
and processes to handle the air traffic [18]. This includes e.g.,
improved usage of slots for runway occupying activities, the
reduction of safety distances due to wake turbulence, or a
traffic shift to nearby less congested airports to relax the traffic
Local capacity enhancements or disturbances are directly
linked to propagation effects within a network, which are
analyzed from two perspectives in previous researches. The
first examines the impact of a single aircraft on airline per-
formance. Therefore, the well-known Delay Multiplier (DM)
is a commonly used evaluation tool, which describes the ratio
between an initial delay and the resulting downstream delay
within the network [23]–[25]. This occurs due to the usage of
aircraft and flight crews for further flight cycles and transfer
passengers who need to reach their follow-up flight. For
evaluation of DMs, delay trees are an appropriate instrument
since these also illustrate the complex airline linkages. Flights
associated with high DMs are network critical. Airlines should
be aware of these flights and intervene if the operation is
adversely affected. The magnitude of the DM depends on
multiple factors. Flights departing in the morning hours tend
to have higher DMs than flights in the later hours due to
a larger number of connected flights [23,24,26]. Most air
traffic activities are stopped in the late evening because of
night flight bans, which allows the air transportation system
to recover from disruptions. The flight schedule structure of
an airline depends on the used business model. This leads
to different initial conditions for the propagation of delays.
Full-Service Network Carrier (FSNC) operate a hub-and-spoke
or multi-hub-and-spoke system where the hub is used as a
central transfer airport [27,28], which allows the connection
of multiple destinations with a low amount of flight routes.
In the last decade, worldwide alliances and joint ventures of
FSNCs have been established to get access to other regions
of the world, reduce the competition on the market, and
overcome regional administration barriers [29,30]. Low Cost
Carrier (LCC) are using a decentralized point-to-point network
to link two destinations without any transfer procedures [31].
However, LCC operate base airports to enable centralized
maintenance and customer service activities. Flights from a
hub-and-spoke network tend to lower DMs compared to point-
to-point network flights under the same initial delay conditions
[25,32]. Further, more flights are adversely affected in a point-
to-point system, and the recovery process from the initial
delay needs more flight cycles compared to a hub-and-spoke
network. FSNC have various instruments at their hub airport
to absorb delays (e.g., spare flight aircraft or crews). Robust
scheduling of airline resources can prevent an airline from the
adverse effects from delay propagation [33].
B. Focus and structure of the document
In our contribution, we want to analyze the effect of local
capacity improvements on the performance of FSNC and the
corresponding hub airport. In a first step, major European
network carriers and alliances are described, and a method
to identify possible designed flight connections is presented.
Further, a simulation considering important European airports
is implemented to determine the network performance during
a busy air traffic volume period.
Based on two data sets provided from EUROCONTROL, a
simulation of relevant European network airports is executed.
The data includes post-operation information for more than
3.2 million reported flight events during the summer period
between July and September 2019. Table Ishows the relevant
information used for the model development. Based on the
data, relevant airports are extracted, and the basic simulation
logic is implemented. Detailed information about the setup
are explained by the authors in a previous study [34]. In the
following, a summary of the central simulation elements and
extensions resp. adjustments are explained.
Information Description Example
ADEP Departure airport EDDF
ADES Destination airport EGLL
SOBT/AOBT Scheduled/Actual time for Off-
Block at the departure airport
Taxi-out duration - 00:12:36
STA/ATA Scheduled/Actual time of arrival 12:40:00
Taxi-in duration - 00:08:00
Aircraft type ICAO designator for used aircraft A320
Airline designator 3 digit ICAO code for an airline DLH
A. Basic simulation logic
The model considers 212 relevant European airports located
in 41 countries, where the NM is responsible for Air Traffic
Flow and Capacity Management (ATFCM) activities (NM
area). Each of them is represented by the available number
of runways used for take-off or landing procedures. To ensure
safe operations, separation distances between two consecutive
starting or landing aircraft are defined based on the wake
turbulence category of the used aircraft type. Further, only
one aircraft operating on a runway is allowed. This processing
time equals the Arrival Runway Occupancy Time (AROT) or
Departure Runway Occupancy Time (DROT). These safety
time buffers are limiting the runway capacity. The airports
are connected via flights. A flight schedule from a busy
weekday during the summer period is used as simulation input,
which includes 35,098 flights and 58,478 movements operated
at the 212 network airports. Flights to other airports are
handled by a dummy airport, which has infinite capacity. This
prevents the network from external disruptions triggered by
non-network airports. Airlines are using aircraft for multiple
flight legs during an operating day resulting in flight cycles.
A general example is shown in Fig. 1. The cycle is described
by four timestamps for the off-block, departure, arrival, and
in-block event. Between these timestamps, processes need to
be executed (taxi-out, flight, taxi-in, turnaround).
The process times are determined stochastically based on
the reported times within the data set. Since the runway
capacity is limited at airports, additional waiting times can
occur at the origin airport (additional taxi-out time) or at
the destination airport (additional Arrival Sequencing and
Metering Area (ASMA) time) if the runway is not approved for
use. To investigate the effect of enhancing runway capacity on
the airport and FSNC performance, the time-based separation
distances are decreased in 5% steps up to 30%. This reduction
leads to a comparable increase in capacity if the runway
is under pressure. The baseline scenario does not consider
any capacity enhancements. Every scenario is executed 10
times, and each run takes about 1 hour of processing time
using an Intel(R) Core(TM) i7-1065G7 CPU processor. The
simulation is implemented using a Python programming plat-
form. Propagation effects occur only if flights from an airline
are connected. Therefore, a data-driven spatial method for
determining designed flight connections based on the airline
flight schedule is developed.
Figure 1. Principle of the general flight cycle.
B. Definition of designed flight connections from FSNC
The business model from FSNC airlines equals a hub-and-
spoke network to connect a variety of airports via a central
transfer airport. This allows passengers to access the entire
network of an FSNC by only one transfer. Selected relevant
FSNC with a hub airport located in Europe are shown in
Table II.
FSNC Hub airport location Traffic share at business
days (median) [%]
Lufthansa Frankfurt a.M. 62
Munich 59
British Airways London 52
KLM Amsterdam 57
Air France Paris 44
Iberia Madrid 59
The number of hub airports differs depending on the
FSNC. E.g., Lufthansa uses Frankfurt a.M. and Munich airport
as a hub, whereas British Airways maintains only London
Heathrow as a transfer option. The particular hub airline
predominates the traffic volume at the corresponding hub
airport. In Frankfurt a.M. 62% of flights are operated by
Lufthansa airlines at the median weekday within the data
set period. This results in a high dependence between the
FSNC and the corresponding hub airport performance. Every
FSNC can only generate the benefits of the hub-and-spoke
network when the connectivity between flights is ensured.
If the hub airport operation is staggering or unreliable, the
FSNC suffers from unstable conditions. To mitigate this effect,
hub airlines are directly involved in airport decision-making
and future development resolutions. E.g., Lufthansa operates
a dedicated area at Frankfurt a.M. Airport (EDDF) and is a
shareholder in the airport operator company (Fraport AG) [35].
European FSNC are characterized by a dense network of short-
and medium-haul flights within Europe combined with several
long-haul flights to major metropolises located in the rest of
the world. Co-operations, joint ventures, and alliances between
airlines are arranged to enhance the offered network for the
customers. Lufthansa maintains a strong collaboration with
United Airlines, which is one of the important FSNC in the
USA. This enables advantages for both parties, as transferring
to flights of the respective airline partner is feasible, and access
to the network is provided. Table III gives an overview of three
major airline co-operation groups that have emerged in recent
Airline group Included passage airlines Joint-venture airlines
Lufthansa group
Lufthansa, Lufthansa
CityLine, Air Dolomiti,
Swiss, Austrian Airlines,
Brussels Airlines,
United Airlines, Air
Canada, All Nippon
Airlines, Singapore
Airlines, Air China
IAG group
British Airways, British
Airways CityFlyer,
Sun Air, Iberia, Iberia
Express, Vueling Airlines,
LEVEL, Aer Lingus
American Airlines,
Finnair, Japan
Airlines, Qatar
Airways, China
Southern Airlines
Air France-KLM
Air France, Air France
Hop, Transavia France,
KLM, KLM Cityhopper,
Delta Air Lines,
Alitalia, Virgin
Atlantic, China
Southern Airlines,
Xiamen Airlines
Since it is challenging to obtain holistic information on
transfer passengers because this involves sensitive airline data
subject to confidentiality (International Air Transport Associ-
ation (IATA) and several Global Distribution Systems (GDSs)
provide selected transfer data at some cost), a three-step
approach is used to identify the intended flight connections
as input to the simulation [39]. The general logic of a transfer
flight between an Origin and Destination (O&D) looks as
Origin (O) Flight 1 (F1)
Hub Flight 2 (F2)
Destination (D)
Two flights are needed to connect the O&D. F1 carries the
passenger from the origin to the hub airport. After the transfer
process at the hub, F2 is used to finish the journey and
reach the final destination. The applied principle is explained
using an example flight operated by Lufthansa from EDDF
to Stockholm Arlanda Airport (ESSA) as F2 and a general
scenario for a long-haul O&D via EDDF.
1) Determining the transfer time horizon: A designed flight
connection from a FSNC is characterized by a transfer at the
hub airport. This process includes the entrance into the airport
terminal, an optional id card control as well as a security
check, and the boarding pass control at the departure gate [40].
In addition, there could be a long distance to cover between the
arrival gate of F1 and the departure gate of F2. It may even be
necessary to switch between two terminals at the hub airport.
These activities are time-consuming and form a required time
buffer for a passenger and his luggage to reach a connecting
flight, which is represented by the Minimum Connecting Time
(MCT) [41]. The MCT varies between different airports and
is mainly depending on the size. In addition to this minimum
transfer time, the airlines aim at planning connections as time-
effective as possible to keep travel times short. To represent
this criterion, a maximum transfer time of 3 hours is used as an
assumption. Longer stopovers at hub airports are possible, but
the corresponding flights are not handled as a designed flight
connection within the simulation logic. Finally, the period for
planned connections can be determined using the Scheduled
Off-Block Time (SOBT) of F2 and Scheduled Time of Arrival
(STA) of F1 for all flights of an FSNC. The transfer time
between F1 and F2 (TTF1 F2) equals the difference from the
SOBT from the second flight leg and the STA from the first
flight leg (1).
A F1 has a designed connection with F2 if the TTF1F2 is
within the transfer time limits (2).
connection =(Yes if MCTHub TTF1F2 3 hours
No else (2)
This algorithm is applied for every flight of an airline
group and the corresponding hub airports shown in Table II
and III. 103 connections fulfill the time-based criteria for the
example flight from EDDF to ESSA. The corresponding 80
airports are shown in Fig. 2(left).
2) Definition of the excluded area - part 1: The time-
based criteria for the classification of flight connections don’t
consider the geographical position from the O&D airports.
This allows the definition of flights from two closed located
airports as designed flight connections. In the worst-case, the
O&D airports are located in the same city, which results in a
highly uncommon journey route. Usually, these short-haul trips
are traveled by other transportation modes (e.g., high-speed
trains or long-distance buses). To exclude unusual routing
with closely located airports, a distance- and an angle-based
criterion is applied.
In a first step a Lower Distance Boundary (LDB) and a
Upper Distance Boundary (UDB) is determined to limit the
excluded area around the destination airport. The basis is the
orthodromic distance between the hub and destination airport
(dHubD). The location of both airports is characterized by
a longitudinal (λ) and a latitudinal (ϕ) coordinate in degrees.
Equation (3) is used to calculate dHubDin kilometres. LDB
and UDB are determined by subtraction resp. addition of
600 km as a threshold value from dHubD.
dHubD=arccos(sin(ϕHub)·sin(ϕD) + cos(ϕHub )
·cos(ϕD)·cos(λHub λD)) ·6.378 km (3)
The distance borders for the example flight are shown in
Fig. 2(left). EDDF and ESSA are located about 1,220 km
apart from each other, which results in a LDB of 620 km
and an UDB of 1,820 km. To further restrict the excluded
area around the destination airport, an angle-based limitation is
used. The basis is the location of the destination in reference to
the hub airport and the north direction (N). These three points
form the angle θN, Hub, D shown in Fig. 2(left). The angle
border in the negative direction of rotation is called Negative
Angle Boundary (NAB). The Positive Angle Boundary (PAB)
equals the border in the positive direction of rotation. Both
are determined similarly to the distance limits by subtracting
or adding the threshold value of 30°. Equation (4) shows the
determination of θN, Hub, D using the scalar product.
θN, Hub, D =arccos
PHub PN·
LDB, UDB, NAB, and PAB create a sector around the
destination airport which corresponds to the excluded area. If
an origin airport is located in this area, the resulting O&D is
defined as uncommon and is not considered in the simulation.
Fig. 2(left) shows the excluded area in red for the example
flight. Flights from seven departure airports (ESSA included)
are neglected due to the close distance to ESSA. In this
special case, this also affects seven flights, which leads to 96
remaining valid leg 1 flights for the example case.
3) Definition of the excluded area - part 2: The explained
method for determining the excluded area works well for the
European area but has issues with increasing distance from
the hub to origin and destination airports. Fig. 2(right) shows
a scenario for F2 from EDDF to New York John F. Kennedy
Airport (KJFK). The resulting excluded area around KJFK
becomes significantly larger than for ESSA. This results from
the exponential expansion of ellipses with increasing distances
between hub and destination airport. The curious shape of the
excluded area arises due to the equirectangular projection of
the map and the resulting deformation. The sector includes
several major airports located at the upper US east coast
Figure 2. Evaluation method to identify designed flight connections by FSNC airlines.
(e.g. Washington Dulles Airport (KIAD) or Boston Logan
Airport (KBOS)), but airports located in the southeast area
(e.g., Atlanta Airport (KATL) or Miami Airport (KMIA)) are
handled as valid connection to KJFK via EDDF. In practice,
however, it is not a viable route. To prevent this unintentional
declaration, a second distance-based criterion is applied for
long-haul O&Ds if both airports are more than 3,000 km apart
from the hub airport. In this case a Detour Factor (DF) is deter-
mined to precise the valid O&D definition. For this purpose
the distance between origin and destination (dOD), origin
and hub (dOHub) as well as between hub and destination
airport (dHubD) is calculated using the corresponding airport
coordinates and the adjusted equation for the determination of
the orthodromic distance (3). The definition of the DF for an
O&D (DFO&D) is given in (5).
DFO&D =dOHub +dHubD
The identification of valid long-haul O&Ds is based on a
threshold value for the DF. If the distance sum from both
flight legs exceeds the distance between origin and destination
airport by more than 45% (DF = 1.45), the corresponding
O&D is excluded (6).
O&D valid =(Yes if DFO&D 1.45
No else (6)
For the shown case in Fig. 2(right) with KMIA as origin,
EDDF as hub, and KJFK as destination the DF equals 7.94 and
therefore the O&D is declined. This method prevents long-haul
O&Ds from and back to the same region of the world. The DF
threshold value is a result from a set of example O&Ds with
EDDF as hub airport shown in Table IV. Green highlighted
cells show possible O&Ds using the defined threshold DF
value of 1.45.
Table Vshows the results for the evaluation of designed
flight connections from major FSNC airlines. Lufthansa op-
erates the highest amount of outbound hub-flights (500) and
overall potential connections (31,110). The highest determined
value of valid feeder flights equals 102 for two Lufthansa
DF Destination (D)
Origin (O)
WSSS - 3.71 2.59 4.02 3.94 1.39 1.19 1.26 2.03
RJTT 3.71 - 1.79 8.2 2.64 2.12 1.67 1.02 1.27
OMDB 2.59 1.79 - 2.16 5.02 1.06 1 1.21 1.86
ZBAA 4.02 8.2 2.16 - 3.65 1.7 1.39 1 1.33
VIDP 3.94 2.64 5.02 3.65 - 1.2 1.07 1.12 1.67
KLAX 1.39 2.12 1.06 1.7 1.2 - 7.55 1.86 1.16
MMMX 1.19 1.67 1 1.39 1.07 7.55 - 2.49 1.38
SBGL 1.26 1.02 1.21 1 1.12 1.86 2.49 - 3.11
FACT 2.03 1.27 1.86 1.33 1.67 1.16 1.38 3.11 -
flights departing at 8 PM. However, it should be stated that
a designed flight connection describes only a possible link
between two flights. If the O&D is not booked from any
passenger, the successful connection of those flights has no
impact on the transfer performance from the airline. The
number of connecting flights is comparatively low for early
morning flights because only a limited number of arriving
flights have arrived at the hub airport before them.
FSNC Hub airport
flights (F2)
(group airlines
Feeder flights (F1)
min max avg sum
Lufthansa Frankfurt a.M. 500 8 102 62 31,110
Munich 391 4 74 46 17,799
British Airways London 368 1 44 27 9,837
KLM Amsterdam 389 8 89 53 20,842
Air France Paris 363 1 45 22 7,899
Iberia Madrid 369 1 52 21 7,570
C. Unimpeded process times
The data set from EUROCONTROL includes taxi-out and
flight times, which are used as simulation input for the
stochastic determination of the process times. Fig. 3shows
the mean and standard deviation for taxi-out times for EDDF
(left) and EGLL (right), depending on the number of aircraft
taxiing on the apron. The number of taxiing aircraft is counted
Figure 3. Mean taxi-out times (grey) and standard deviations (black) based on apron traffic for EDDF and London Heathrow Airport (EGLL).
in 5-minute steps, representing the current traffic volume on
the apron. The red lines show the share of periods for all
observed traffic volume conditions. Standard deviations are
not calculated for high traffic volumes on the apron because
this condition occurs only rarely. High percentages for low
traffic volumes result from the night hour time periods. The
average taxi-out time considers all flights starting their taxi-
out process under the same initial apron traffic volume. For
EDDF the taxi-out times are stable with increasing apron
traffic, which indicates that the apron design of the airport can
absorb the resulting congestion. The taxi-out times at EGLL
are steadily increasing due to congestion effects if a threshold
of simultaneously taxiing aircraft is exceeded. Stable roll time
parameters can be monitored with lower traffic volumes on the
apron. The transition point indicates the amount of manageable
aircraft by the apron infrastructure. Every additional aircraft
leads to a loss in quality and an increase in taxi-out time due to
congestion effects. The threshold value depends on the apron
size as well as the design and differs between airports.
Taxi-out times for lower apron traffic volumes vary due to
different taxi-out distances depending on the gate position and
active runway configuration. For the simulation, the use of
an unimpeded taxi-out time without queue times is necessary.
The simulation creates additional waiting times for departures
and arrivals on its own due to the exclusive use of a server
(runway) by only one flight. An unimpeded taxi-out time
doesn’t consider any additional times due to congestion effects
on the apron. Fig. 3indicates the increase of mean taxi-out
time for EGLL if more than seven aircraft are simultaneously
moving on the apron. Analysis of other network airports shows
that even when more than five flights are taxiing, interactions
occur amongst them. We assume that an unaffected taxi
operation at all network airports is guaranteed when the apron
traffic volume does not exceed this threshold. The used taxi-
out parameters considered in the simulation are calculated
based on flights that started their taxi procedure meeting this
apron traffic volume criterion.
A similar approach is used to determine the flight times
between an O&D for a specific aircraft type. Fig. 4exhibits
the distribution of reported flight times from EGLL to EDDF
flown by A320neo. The red line shows the corresponding
normal distribution. Due to the consideration of outliers,
the curve runs flat. Comparably low flight times are results
from e.g., advantageous wind conditions. Holdings around
the departure airport due to congestion effects, necessary re-
routings, or severe weather can be reasons for high travel
times. Flight times outside the 2σare located in the outlier
area and are excluded, which results in a better shaping curve
(green). Valid flight times for the simulation are within the
3σinterval calculated using the parameters from the green
curve to avoid extremely short or long times due to the natural
behavior of the normal distribution.
Figure 4. Distribution of reported flight times from EGLL to EDDF for flights
using A320neo aircraft.
The reported values of taxi-in times for an airport are
constant, but they differ from airport to airport. A study
using open-source data for London Gatwick Airport (EGKK)
indicates a lower dependence of taxi-in times on apron traffic
volumes [42]. Hence, these fixed values are used for the
simulation. The turnaround time is calculated using an as-
sumption based on the wake turbulence category from the
used aircraft type. Therefore reference aircraft for medium
(A320neo), heavy (A350-900), and super (A380-800) cat-
egories are specified, and the indicated full servicing turn
round time (A320neo: 44 min [43]; A350-900: 61 min [44])
resp. the typical turn-round time - standard servicing via
main deck and upper deck (A380-800: 90 min [45]) is used
as expected value (µ). The corresponding assumed standard
deviation equals 10% of µ. Based on the resulting normal
distribution, the turnaround times are generated within the
simulation. An expected turnaround time of 20 minutes and a
standard deviation of 2 minutes is assumed for light aircraft.
The simulation is executed considering capacity enhance-
ments for two major European airports (EDDF and EGLL),
which serves as hub for a FSNC. Fig. 5shows the results for
the following four performance indicators for both airports:
Inbound delay: Difference between Actual Time of
Arrival (ATA) and STA of arrival flights
Outbound delay: Difference between Actual Off-Block
Time (AOBT) and SOBT of departure flights
additional taxi-out time: queue time at the runway from
departing flights
ASMA additional time: queue time at the runway from
arriving flights
The results exhibit low delay figures for EDDF and manage-
able additional waiting times at the runways for the baseline
case. Capacity enhancement measures don’t show a significant
improvement for the performance of EDDF since the delay
is not a result of traffic congestion at this airport. On the
other hand, EGLL is characterized by high delays and queue
times at the runway. In contrast to EDDF, a capacity increase
leads to a significant improvement in airport performance. 10%
more capacity reduces the inbound delay by 10,832 minutes
(-42%), the outbound delay by 2,884 minutes (-80%), the
additional taxi-out time by 7,815 minutes (-33%), and the
ASMA additional time by 8,931 minutes (-56%). However,
a saturation effect can be observed for high rates of capacity
enhancements. This indicates structural delay from a too-tight
flight schedule or created at other network airports.
Figure 5. Development of the airport performance indicators from EDDF (top)
and EGLL (bottom) under increasing capacity.
To investigate the improvements for FSNC the connection
rate is analyzed, which equals the ratio between fulfilled flight
connections and the number of defined flight connections from
a FSNC outbound flights at the hub airport. A feeder flight
Figure 6. Median connection rate of major FSNC under increased capacity at
EDDF (top) and EGLL (bottom).
is connected successfully with the outbound flight if the in-
block time from the feeder occurs before the off-block time
from the outbound flight minus the MCT of the hub airport.
Fig. 6shows the range of median connection rates for all
10 simulation runs for increased capacity at EDDF (top) and
EGLL (bottom). Between 87% and 90% of feeder flights
arrive in time to catch the outbound flight of the median
Lufthansa flight. This ratio does not improve by increasing
capacity at EDDF. The missed connections occur majorly due
to disturbances at the spoke airports. The median connection
ratio of other FSNC is also not improving significantly when
the capacity of EDDF is enhanced. For British Airways, a
connection rate of 55% up to 60% can be observed. The
comparable low ratio reasons from the high congestion at
EGLL. Suppose the capacity at EGLL is enhanced, the median
connection rate increases significantly. Similar to airport per-
formance, a saturation effect and lower growth occur at higher
rates of capacity increase. A 10% capacity increase at EGLL
leads to a similar increase in fulfilled flight connections for
the median flight from British Airways. This underlines the
importance of the hub airport performance for the resident
FSNC. However, no measurable effect can be observed for
other FSNCs. The median connection rates are stable and
independent of the improved performance at EGLL. The
airport is only one spoke in the network for the other FSNC.
Lufthansa serves 136 further spoke airports from EDDF on the
simulated operation day. The capacity enhancement at EGLL
improves primary the direct flights to this airport. Benefits to
downstream flight events using the same resources for a flight
are possible if the flight is delayed in the baseline scenario and
the planned ground buffer is insufficient and exceeded. Nev-
ertheless, Lufthansa operates 24 flights between EDDF and
EGLL in the used simulation flight schedule. In comparison
to the overall number of 900 Lufthansa flights via EDDF the
improvement potential is limited. An enhanced spoke airport
has not a quantifiable impact on the Lufthansa connection rates
since the operated networks are too comprehensive.
FSNC are essential providers of inner-European and inter-
continental connections via a hub airport. To offer a wide
range of O&D linkages, a seamless transfer between the
first and second flight leg is essential. A simulation of one
operation day from relevant airports of the European air
transportation network is used to investigate the development
of flight connection rates from FSNC under increased local
airport capacity. The results exhibit the importance of the
hub airport performance for FSNC. A 10% increased capacity
at EGLL leads to significantly less delay and waiting time
at the airport resources compared to the baseline scenario.
Simultaneously the ratio of successfully reached designed
flight connections from British Airways increased by the
same percentage. The growth rates slightly decrease with
steady capacity enhancements, leading to a natural saturation
in airport delay figures and connection rates. To overcome
these barriers, other measures like flight schedule adjustments
are necessary. Other FSNC connection rates (which serves
EGLL as spoke airport) do not improve significantly from
the capacity increase. The direct connections to EGLL are
positively affected by the airport improvement, but propagation
effects on downstream airline flights are not observed. The
large expansion of FSNC networks and the high number of
operated flights during an operation day leads to a negligible
effect of improved connections to only one spoke airport.
This publication was supported and funded by EUROCON-
TROL via contract NO.19-220468-C. Its contents are solely
the authors’ responsibility and do not necessarily represent the
official views of EUROCONTROL.
[1] ICAO, “Annual report 2019: Presentation of 2019 air transport statistical
results,” 2019.
[2] EUROCONTROL, “Performance Review Report 2019,” 2020.
[3] ICAO, “Doc. 9971 - manual on collaborative air traffic flow management
(ATFM),” 2018.
[4] EUROCONTROL, “Airport capacity assessment methodology (ACAM)
manual,” 2016.
[5] ——, “CODA digest - annual report for 2019,” 2020.
[6] A. Cook and G. Tanner, “The cost of passenger delay to airlines in
Europe-consultation document,” 2015.
[7] J. Rosenow, P. Michling, M. Schultz, and J. Sch¨
onberger, “Evaluation
of strategies to reduce the cost impacts of flight delays on total network
costs,” Aerospace, vol. 7, no. 11, 2020.
[8] J. Evler, M. Schultz, H. Fricke, and A. Cook, “Development of Stochas-
tic Delay Cost Functions,” in 10th SID, 2020.
[9] EUROCONTROL, “ATFCM users manual,” 2018.
[10] EUROCONTROL, “European aviation in 2040: Challenges of growth -
Annex 1 flight forecast to 2040,” 2018.
[11] Industry High Level Group (IHLG), “Aviation benefits report,” 2019.
[12] EUROPEAN COMMISSION, “Flightpath 2050 Europe’s vision for
aviation: Report of the high level group on aviation research,” 2011.
traffic management master plan,” 2009.
[14] F. Knabe and M. Schultz, “A new way to indicate airport airside
performance from an economic perspective,Transp. Res. Procedia,
vol. 14, 2016.
[15] EUROCONTROL, “European aviation in 2040: Challenges of growth -
Annex 4 network congestion,” 2018.
[16] J. Rosenow and M. Schultz, “Coupling of turnaround and trajectory
optimization based on delay cost,” in WSC, 2018.
[17] J. Evler, M. Schultz, and H. Fricke, “Flight prioritization and turnaround
recovery,” in 14th ATM Research and Development Seminar, 2021.
[18] P. Berster, M. C. Gelhausen, and D. Wilken, “Options to mitigate
negative effects of airport capacity constraints,” in ATRS, 2019.
[19] M. Hansen, “Post-deployment analysis of capacity and delay impacts of
an airport enhancement,” ATCQ, vol. 12, no. 4, 2004.
[20] M. M. Mota, P. Scala, and G. Boosten, “Simulation-based capacity
analysis for a future airport,” in APCASE, 2014.
[21] E. Aydo˘
gan and C. C¸ etek, “Runway capacity enhancement analysis of
end-around taxiway at istanbul new airport,D¨
uzce ¨
Universitesi Bilim
ve Teknoloji Dergisi, vol. 7, no. 1, 2019.
[22] L. Dray, “An empirical analysis of airport capacity expansion,” JATM,
vol. 87, 2020.
[23] R. Beatty, R. Hsu, L. Berry, and J. Rome, “Preliminary evaluation of
flight delay propagation through an airline schedule,” ATCQ, vol. 7,
no. 4, 1999.
[24] A. Kondo, “Delay propagation and multiplier,” in 51st Annual Transp.
Res. Forum, 2010.
[25] N. Kafle and B. Zou, “Modeling flight delay propagation,” Transp. Res.
Part B: Methodological, vol. 93, 2016.
[26] J. Evler, M. Lindner, M. Schultz, and H. Fricke, “Integration of
turnaround and aircraft recovery to mitigate delay propagation in airline
networks,” Computers & Operations Research, vol. 138, 2022.
[27] O. Lordan, “Study of the full-service and low-cost carriers network
configuration,” J Ind Eng Manag, vol. 7, no. 5, 2014.
[28] O. Lordan, J. M. Sallan, N. Escorihuela, and D. Gonzalez-Prieto,
“Robustness of airline route networks,Physica A, vol. 445, 2016.
[29] T. C. Usta¨
omer, V. Durmaz, and Z. Lei, “The effect of joint ventures
on airline competition,” Procedia Soc Behav Sci, vol. 210, 2015.
[30] I. Douglas and D. Tan, “Global airline alliances and profitability,
Transp. Res. Part A: Policy and Practice, vol. 103, 2017.
[31] X. Fu, H. Jin, S. Liu, T. H. Oum, and J. Yan, “Exploring network effects
of point-to-point networks,” Transport Policy, vol. 76, 2019.
[32] S. AhmadBeygi, A. Cohn, Y. Guan, and P. Belobaba, “Analysis of the
potential for delay propagation in passenger airline networks,” JATM,
vol. 14, no. 5, 2008.
[33] W. Wu and C.-L. Wu, “Enhanced delay propagation tree model with
bayesian network for modelling flight delay propagation,” Transp. Plan-
ning and Technology, vol. 41, no. 3, 2018.
[34] D. Lubig, M. Schultz, H. Fricke, F. Herrema, R. Barrag´
an Montes,
and B. Desart, “Propagation of airport capacity improvements to the
air transport network,” in AIAA/IEEE DASC, 2021.
[35] Fraport AG, “Annual report 2020,” 2021. [Online]. Available:
[36] Lufthansa, “Network airlines; alliances and partner airlines,” 2021.
[Online]. Available:
[37] IAG, “Our brands; alliances and joint businesses,” 2021. [Online].
[38] AirFranceKLM Group, “Brands; group,” 2021. [Online]. Available:
[39] J. Evler, “To be published: Resource-constrained airline ground oper-
ations: Optimising airline schedule recovery under uncertainty,” Ph.D.
dissertation, 2022.
[40] A. G. de Barros, A. Somasundaraswaran, and S. Wirasinghe, “Evaluation
of level of service for transfer passengers at airports,JATM, vol. 13,
no. 5, 2007.
[41] IATA, “Minimum connecting time - user guide,” 2020.
[42] M. Schultz, J. Rosenow, and X. Olive, “A-CDM Lite,” in 9th SID, 2019.
[43] AIRBUS, “A320: Aircraft characteristics, airport and maintenance plan-
ning,” 2017.
[44] ——, “A350: Aircraft characteristics, airport and maintenance planning,
[45] ——, “A380: Aircraft characteristics, airport and maintenance planning,
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Competitive price pressure and economic cost pressure constantly force airlines to improve their optimization strategies. Besides predictable operational costs, delay costs are a significant cost driver for airlines. Especially reactionary delay costs can endanger the profitability of such a company. These time-dependent costs depend on the number of sensitive transfer passengers. This cost component is represented by the number of missed flights and the connectivity of onward flights, i.e., the offer of alternative flight connections. The airline has several options to compensate for reactionary delays, for example, by increasing cruising speeds, shortening turnaround times, rebookings and cancellations. The effects of these options on the cost balance of airline total operating costs have been examined in detail, considering a flight-specific number of transfer passengers. The results have been applied to a 24-h rotation schedule of a large German hub airport. We found, that the fast turnaround and increasing cruise speed are the most effective strategies to compensate for passenger-specific delay costs. The results could be used in a multi-criteria trajectory optimization to find a balance between environmentally-driven and cost-index-driven detours and speed adjustments
Full-text available
Theoretical analyses of the impact of airport capacity expansion must model or make assumptions about the effect of capacity on demand, airline competition, aircraft types, fares and other characteristics of a given airport. In this paper, we use empirical data on historical schedules, fares, delays and demand for the busiest 150 airports in 2015 to examine the typical impact of historical capacity expansions. We find significant diversity in outcomes, with over half the expanded airports either using less than their pre-expansion capacity or remaining constrained even at post-expansion capacity by 2016. Many of the expected impacts, such as reductions in typical aircraft size, either do not materialise or are dominated by other effects (for example, recessions; airlines beginning or ending operations at an airport; changes in regulation). Behaviour on expansion is affected by slot control regulations and whether the airport is initially capacity-constrained. In particular, slot-controlled airports typically add new destinations and carriers on expansion rather than making significant changes to existing schedules.
Conference Paper
Full-text available
This study successfully implements flight specific delay costs in an air traffic simulation with a multi-criteria trajectory optimization and exemplifies a coupling of turnaround and trajectory optimization of historical real flights. Therein, delay costs and detour costs for the reduction of contrail formation are individually calculated for each flight and considered in a flight specific multi-criteria trajectory optimization with the air traffic simulation environment TOMATO. Detours in the optimized trajectories are mainly caused by the intent of avoiding contrail formation. With this case study, the historical flight plan could be stretched and departures and arrivals could be more homogeneously distributed during the analyzed three hours while at the same time ecological costs could be saved by 15 per cent. Therewith the promising potential of System Wide Information Management between airports, airlines, air traffic control and customers could once again be shown.
Full-text available
Flight delay presents a widespread phenomenon in the air transportation system, costing billions of dollars every year. Some delay originating from an upstream flight spreads to downstream flights. This phenomenon is defined as delay propagation. To understand the delay propagation patterns and associated mitigation measures, this study proposes a novel analytical-econometric approach. Considering that airlines deliberately insert buffer into flight schedules and ground turnaround operations, an analytical model is developed to quantify propagated and newly formed delays that occur to each sequence of flights that an aircraft flies in a day, from three perspectives on the ways that delays are absorbed by the buffer. With delays computed from the analytical model, we further develop a joint discrete-continuous econometric model and use the Heckman's two-step procedure to reveal the effects of various influencing factors on the initiation and progression of propagated delays. Results from the econometric analysis provide estimates on how much propagated delay will be generated out of each minute of newly formed delay, for the US domestic aviation system as well as for individual major airports and airlines. The impacts of various factors on the initiation and progression of propagated delay are quantified. These results may help aviation system planners gain additional insights into flight delay propagation patterns and consequently prioritize resource allocation while improving system overall performance. Airlines can also be better informed to assign buffer to their flight schedules to mitigate delay propagation.
Full-text available
Airport capacity is a key performance indicator for the ability of the airport to handle an expected traffic demand. The classical way to quote airport capacities and flows on the airside is to count the number of potential airframe movements per unit of time (e.g. per hour or per year). This approach seems to be historically reasonable from a perspective of planning and organizing predicted aircraft movements and available airside resources. However from an economic point of view a more sophisticated indicator system could be beneficial to represent and compare future airport performance. The method presented here uses additional parameters like aircraft size, number of passengers, cargo load or value of passengers (e.g. premium vs. economy) to indicate the economic performance of the airport system (social welfare optimization). The results can be used to address possible improvements in ATM with respect to economic relevance for the operational air traffic stakeholders (airport, airline, air navigation service provider). Since a single flight movement is now differently rated regarding to the calculated airport performance contribution, it is expected that the accompanied prioritization procedures could significantly influence airline and airport operations. Further on, the mutual dependencies between airlines (concurrent demand for sufficient airport resources) and airline-airport (provide efficient air transport) will be emphasized in a transparent and reliable way. To exemplarily show the consequences and the benefits of the introduced metric exemplarily, we will use different airports (airport categories, runway configurations and operational scenarios), different parameter settings (short/medium/long haul flights, heavy/medium/light aircraft) and relevant combination of the identified input parameters. Special emphasis will be given to the discussion of the weighting of parameters depending on the needs of the stakeholders in air traffic management and in air transportation. It will be sketched how the results of the new method can be used in various time horizons for planning airside movements at an airport, i.e. from pre-tactical flow management in a performance based airport management environment to the evaluation of long-term strategic expansion options of an airport. A comparison of a set of airports having similar performance using the conventional capacity scale will demonstrate the significance of the new indicator set and the possibilities of its application.
This article presents a novel approach to incorporate the aircraft turnaround, which has recently been identified as one of the major contributors to airline delay, into existing concepts for integrated aircraft, crew, and passenger recovery. We aim to fill the research gap on how to holistically model network delay propagation as tactical decision support for airline schedule recovery. Our model introduces a heterogeneous vehicle routing problem with time windows for the assignment of aircraft to flight routes and integrates it with an extended version of the resource-constrained project schedule problem for the allocation of scarce resources to turnarounds at the central hub airport, such that we can proactively estimate delay propagation in an airline network. Passenger and crew itineraries are modelled as links between flights, such that needed transfer times influence the stand allocation and resource assignment. These links may only be broken if reserve capacities are available and the related rebooking and compensation costs are more efficient than accepting departure delays to maintain transfers. With this approach, we are able to calculate flight-specific delay cost functions and find substantial dependencies about the time of the day, the number of succeeding flight legs and particular downstream destinations. The integrated recovery model is implemented into a rolling horizon algorithm and applied to a case study setting to analyse its performance in comparison to the individual turnaround and aircraft recovery models. Within different delay scenarios, we find that the incorporation of turnaround recovery options significantly improves the resilience of the airline network. Especially in low and moderate delay situations, we achieve a full recovery of the flight schedule simply by rebooking passengers, reallocating aircraft among stands and accelerating ground operations. Thus, often considered recovery options, such as aircraft swaps and flight cancellations, are not required for delays around 30 minutes in our case study. This reduces total costs in comparison to the conventional aircraft recovery model by 49%. Despite the lower efficiency of turnaround recovery in medium and high delay scenarios, the combination of flexible aircraft assignments and ground operations still generates additional cost savings of at least 21% and helps to reduce the necessary amount of optimal recovery options.
An enhanced Delay Propagation Tree model with Bayesian Network (DPT-BN) is developed to model multi-flight delay propagation and delay interdependencies. Using a set of real airline data, results show that flights have non-homogeneous delay propagation effects. The DPT-BN model is used to infer posterior delay profiles with different delay and scheduling scenarios. The major contribution of the DPT-BN model is to demonstrate how the modelling of non-independent and identically distributed delay profiles is more realistic for the observed delay propagation mechanism, and how robust airline scheduling methodologies can benefit from this probability-based delay model.
Alliances and partnerships between airlines are entered into to create competitive advantage, reduce costs, and expand network reach. Three global alliance clusters emerged, with founding partners located in the major geographic regions, and often already involved in bilateral partnerships with other founders. This study examines whether the formation of global airline alliances, with its related expansion of network reach, resulted in an increase in profitability for the founding members. Employing difference-in-difference regressions, this study has found no evidence that the formation of global alliances improved the profitability of founding member airlines, or conferred an economic advantage over airlines that were not founding members. This result is robust across geographic regions, individual global alliances, and alternative event dates. The results of this research suggest that regulators should be less anxious that by sanctioning these closer relationships they are providing major carriers with opportunities for excessive profits.