Investigation of the Scalability of Air Transportation ...web.mit.edu/airlines/industry_outreach/board_meeting_presentation_files/meeting-oct...• Key constraints of the current air

M I T I n t e r n a t i o n a l C e n t e r f o r A i r T r a n s p o r t a t i o nM I T I n t e r n a t i o n a l C e n t e r f o r A i r T r a n s p o r t a t i o n

MIT MIT ICAT ICAT

Investigation of the Scalability of Investigation of the Scalability of Air Transportation Networks Air Transportation Networks

Philippe [email protected]

R. John [email protected]

Global Airline Industry Program/Industry Advisory Board Meeting

October 26th 2006

MIT MIT ICAT ICAT

2

Motivation & Approach

Scalability: the ability of a system, a network or a process to change its scale in order to meet growing volumes of demand

Relevance to the air transportation system• Growing demand for air transportation

FAA forecast growth rate: 2005-2017 forecast (enplanements air carrier: +3.1% per year, regional carriers: +4.3%, general aviation turbojet operations: +6.0%)

• Key constraints of the current air transportation systemInfrastructure (i.e. airport & airspace capacity)

• Challenges and implications of not meeting demand Generation and propagation of delays throughout the systemEconomic impacts (time loss for travelers, operational inefficiencies for airlines, environmental cost through excess fuel burn)

Need to investigate ways to augment the scale of the air transportation system in order to meet future demand

Approach• Analysis of air transportation network topology and evolution

Data: Enhanced Traffic Management System (ETMS) and TAF traffic datafrom October 1st 2004 to September 30th 2005 (20.5 million flights analyzed) & traffic from 1976 to 2005

• Application of scalable (scale free) network theory• Case study approach (23 case studies of regional airport systems)• Development of network evolutionary dynamic models

MIT MIT ICAT ICAT

3

Introduction:Network and System Dynamics Representation of the Air Transportation System

Infrastructure(airport nodes)

Transport Network

Demand(Latent demand)

High Level System Dynamics Model

Air carrier & Reg. Av. layer

Airport utilization

Buss. & Gen. aviation layer

Demand

Capacity Inadequacy

# airports

Airport system capacityCapacity of airports

PerformanceLOS / Delays

MIT MIT ICAT ICAT

4

Evolution of Demand for Air Transportation

Historical and projected growth of demand (enplanements) for air transportation

Greater number of operations are expected in the NAS in the upcoming years

Factors amplifying the problem• Decreasing size of aircraft: Influence of Regional Jets• Entry of small aircraft in the NAS in the upcoming years: VLJs, UAVs

Total Enplanements in the U.S.

1976

1979

1990

2000 2004

0

100

200

300

400

500

600

700

800

1975 1980 1985 1990 1995 2000 2005

Mill

ions

Enpl

anem

ents

* Graph represents realized demand



Airport utilization


Demand

Capacity Inadequacy

# airports



MIT MIT ICAT ICAT

5

Capacity the National Airport System

Capacity of the National Airport System• Airports in the United States in 2006

Total:19,847 airports5,261 public airports

• Capacity also exist in high density metropolitan areas(number of runways in major regional airport systems)

05

10152025

303540

4550

Dallas

Chicag

oLo

s Ang

eles

Detroit

New York

Boston

Housto

n

San Fr

ansis

coPhil

adelp

hiaAtla

ntaWash

ington

Miami

Phoen

ixCinc

innati

Saint L

ouis

Mineap

olis

Num

ber o

f run

way

s (lo

nger

than

5,0

00 ft

)

Surrounding airportsCore airports

Number of runways

longer than 5,000ft



Airport utilization


Demand

Capacity Inadequacy

# airports



MIT MIT ICAT ICAT

6

Airport Utilization in the United States

Concentration of traffic at key airports in the system

• 80% of the air carrier operations are handled at top 50 airports (4% of usable airports)

• 80% of the total itinerant operations are handled at820 airports (8% of usable airports)

0%

20%

40%

60%

80%

100%

120%

0 500 1000 1500 2000 2500 3000 3500

Airports (sorted by decreasing traffic)

Cum

ulat

ive

traffi

c sh

are

(ope

ratio

ns)

Air Carrier Operations

Total Itinerant Operations



Airport utilization


Demand

Capacity Inadequacy

# airports



MIT MIT ICAT ICAT

7

Inadequacy of Capacity at Key Points in the System

Demand-Supply Mismatch at key points in the system

• e.g. La Guardia (LGA) in 2000demand exceeded capacity by a factor of 2

• e.g. Chicago O’Hare (ORD) in 2003

Demand growth is adding pressure on key airports



Airport utilization


Demand

Capacity Inadequacy

# airports



La Guardia airport (operations from 07:00 to 21:59)

0

10

20

30

40

50

60

70

80

jan feb mar apr may jun jul aug sep oct nov dec

Thou

sand

sO

pera

tions

Total demandCapacityOperations

in 2000

MIT MIT ICAT ICAT

8

Delays as an Indicator of Airport Capacity Inadequacy

Implications of capacity inadequacy• Generation of delays at key points in the system

Peak of delays in 2000 due to the general growth of demand and the La Guardia problem that propagated throughout the systemDelays are back to 2000 levels and are stabilized

• Propagation of delays throughout the network

National delays from 1995 to 2006

0

0.5

1

1.5

2

2.5

3

3.5

Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06

Milli

ons

Del

ays

(in m

in.)

National Delays12 per. Mov. Avg. (National Delays)



Airport utilization


Demand

Capacity Inadequacy

# airports



MIT MIT ICAT ICAT

9

Scaling “Up” the Network:Adjustment of Capacity of Airports

Evolution of throughput of airport (nodes) in the air transportation network between 1976 and 2004


Airport utilization


Demand

Capacity Inadequacy

# airports



Capacity adjustmentScaling “Up”


19761980

1990

2000

2004

Airport t

hroughput:

Scaling “u

p”

MIT MIT ICAT ICAT

10

Scale Free (i.e. Scalable) Networks

Definition & Properties

Scale free networks exhibit power law degree distributions

Notations and basic network characterization concepts:

k: degree of a node = number of connections to other nodes

e.g. kin = 2

kout = 2

k = 4

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0 100 200 300 400 500 600 700

Degree

Freq

uenc

y

Degree distribution

i.e. A network with a power law degree distribution is represented by an affine function on a log–log scale graph

Scale free networks have the ability to change scale in order to meet any level of demand

1

10

100

1000

1 10 100 1000

Degree

Nod

e R

ank

log-log scale

Scaling “Up”Mode

time

MIT MIT ICAT ICAT

11

Flight Weighted Degree Distribution of the Air Transportation Network

The air transportation network exhibits a partial power law distribution (scalable network)The cut-off is explained by nodal capacity constraints that limit the ability of the network to scale up

0.001

0.01

0.1

11 10 100 1000

Weighted degree

Cum

ulat

ive

Freq

uenc

y p(

W >

w )

w cut-off w max= airport operations

*

MIT MIT ICAT ICAT

12

Delay/Impossibility of Adjusting the Capacity of Key Nodes in the Network


Airport utilization


Demand

Capacity Inadequacy

# airports




High Level System Dynamics ModelAirport Airport Percentage of OEP new runway project

code name operations (date completion/delayed capacity benefit)

EWR Newark 8.8%ATL Atlanta 6.8% 2006 / + 33%LGA LaGuardia 6.7%ORD Chicago 5.8% ?PHL Philadelphia 5.0% 2008 / ?JFK Kennedy 4.0%BOS Boston 2.8% 2006 / ?SFO San Francisco 2.6%PHX Phoenix 2.4%IAH Houston 2.0%IAD Dulles 1.9% 2008 / +12%LAS Las Vegas 1.5%CLT Charlotte 0.9%DTW Detroit 0.8%MSP Minn./St. Paul 0.7%DCA Reagan National 0.6%DFW Dallas/Ft.Worth 0.6%CVG Cincinnati 0.6%MIA Miami 0.4%SAN San Diego 0.4%BWI Balt.-Wash. Intl 0.4%MEM Memphis 0.3%SEA Seattle 0.3% 2008 / + 46%DEN Denver 0.3%LAX Los Angeles 0.3% 2008 / Not Avail.MCO Orlando 0.3%SLC Salt Lake City 0.2%TPA Tampa 0.2%STL St. Louis 0.1% 2006 / + 48%PIT Pittsburgh 0.1%

Factors limiting airport capacity adjustment:

• Constrained airport layout (limited ability to expand the footprint of the airport)

• Layout of existing runways (legacy system)• Environmental constraints

Data source: [Delay data: FAA Operational Network, OPSNET], [Capacity improvement: FAA Operational Evolution Plan OEP] in 2005.

MIT MIT ICAT ICAT

13

Scaling “Out” to new nodes:Construction of new airports


Airport utilization


Demand

Capacity Inadequacy

# airports




Construction of new airports Scaling “Out” new nodes

High Level System Dynamics ModelLimited ability to add new airports:• Last major airport opening: DEN 1995

Evolution of the number of public airports in the United States from 1980 to 2005Average loss of airport from 1985 to 2004: 30 airports per year

Factors influencing the inability to add new airport and close existing airports:

• Land availability (in areas of high demand for air transportation & high density of population)

• Environmental constraints• Pressure from real estate development

0

1,000

2,000

3,000

4,000

5,000

6,000

1980 1985 1990 1995 2000 2005

Num

ber o

f airp

orts

Public airport

Certificated (Part 139) airport

MIT MIT ICAT ICAT

14

Scaling “Out” to Existing Nodes:Utilization of Existing Nodes in the Network


Airport utilization


Change the utilization of nodes in the network:

Scaling “Out” existing nodes

Demand

Capacity Inadequacy

# airports





High Level System Dynamics ModelScaling “out” to existing underutilized airportsEmergence of secondary airportsAverage age of existing secondary airports (from opening): 73 yearsFuture of secondary airports

• Use of secondary airports has been one of the key mechanisms by which demand was met in congested metropolitan areas

• Strengthening role in the future• Key to the national plans for meeting future demand

(e.g. NGATS Plan)

SFO

LAX

MSP

DAL

HOU

DTWORD

STL

CVG

ATL

DCA

PHL

LGA / JFK / EWR

BOS

MIA

PHXCore airport (Original)

Secondary airport

BUR

OAK/SJC

ONT

SNA

FLL

BWI ISP

MHT

MDWPVD

LGB

Core airport (Emerged)

DFW

IAH

IAD

PHX

LAS

SLC DEN

MEM CLT

MCOTPA

PIT

SEA

SAN

SRQ

PIE MLB

SFB Secondary airport (re-emerged from original core)

SFO

LAX

MSP

DAL

HOU

DTWORD

STL

CVG

ATL

DCA

PHL

LGA / JFK / EWR

BOS

MIA

PHXCore airport (Original)

Secondary airport

BUR

OAK/SJC

ONT

SNA

FLL

BWI ISP

MHT

MDWPVD

LGB

Core airport (Emerged)

DFW

IAH

IAD

PHX

LAS

SLC DEN

MEM CLT

MCOTPA

PIT

SEA

SAN

SRQ

PIE MLB

SFB Secondary airport (re-emerged from original core)

Core and secondary airports in the U.S.

MIT MIT ICAT ICAT

15

Scaling “Out” through Mode Shift:Utilization of small aircraft & small airports

Utilization of existing small airports(airports with 3000ft+ runways)Emergence of new services (i.e. on-demand air taxi) enabled by a “technology push” and a system “performance pull”


Airport utilization


Change the utilization of nodes in the network:

Scaling “Out” existing nodes

Demand

Capacity Inadequacy

# airports






Scaling “Out” through Mode Shift

Air carrier & Reg. Av. Business & Gen. Av.

Very Light Jets

MIT MIT ICAT ICAT

16

Implications for Air Traffic ControlImplications for Air Traffic Control

La GuardiaLa Guardia LGALGAEssexEssex CDWCDW

NewarkNewark EWREWR

KennedyKennedy JFKJFK

TeterboroTeterboro TEBTEB

WestchesterWestchester HPNHPN

LindenLinden LDJLDJFarmingdaleFarmingdale FRGFRG

MorristownMorristown MMUMMU

GreenwoodGreenwood 4N14N1

IslipIslip ISPISP

Old BridgeOld Bridge 3N63N6

Central JerseyCentral Jersey 47N47N

MonmouthMonmouth BLMBLM

SMQSMQ

DanburyDanbury DXRDXR

BridgeportBridgeport BDRBDR

SolsbergSolsberg N51N51

SussexSussex FWMFWM

PrincetonPrinceton 39N39N

CoreCore

SecondarySecondary

GA (>3000ft GA (>3000ft rwyrwy))

Business Av. /Business Av. /High density GAHigh density GA

Legend: airports

GA (<3000ft GA (<3000ft rwyrwy))

La GuardiaLa Guardia LGALGAEssexEssex CDWCDW

NewarkNewark EWREWR

KennedyKennedy JFKJFK

TeterboroTeterboro TEBTEB

WestchesterWestchester HPNHPN

LindenLinden LDJLDJFarmingdaleFarmingdale FRGFRG

MorristownMorristown MMUMMU

GreenwoodGreenwood 4N14N1

IslipIslip ISPISP

Old BridgeOld Bridge 3N63N6

Central JerseyCentral Jersey 47N47N

MonmouthMonmouth BLMBLM

SMQSMQ

DanburyDanbury DXRDXR

BridgeportBridgeport BDRBDR

SolsbergSolsberg N51N51

SussexSussex FWMFWM

PrincetonPrinceton 39N39N

CoreCore

SecondarySecondary



Legend: airports


CoreCore

SecondarySecondary



Legend: airports


Scaling “out” to new or existing nodes and trough mode shift impact terminal areas

• Concentration of traffic of light jet traffic as an indication of future concentration of traffic by VLJs

• ETMS data analysis: 64% of the flights performed by Light Jets had either their departure or arrival in one of the top 23 regional airport systems

Implications for ATC: • Larger number of airports with significant

volume of operations in the regional airport system

• Emergence of interactions between airports(e.g. New York airport system: interactions

New York regional airport systemIllustration of future complex multi airport systems

between arrival and departure streams between LGA, TEB, EWR, JFK)

MIT MIT ICAT ICAT

17

Analysis• 10 multi-airport regional airport

systems• 30 airports • 445 airport-airport routes

(out of a maximum of 870 feasible airport-airport routes)

Data• Bureau of Transportation Statistics

DB1 database segment data (March 2005)

Implications for Airlines:Emergence of Parallel Networks/Markets

OD Market

Parallel airport-airport route(sec. to sec. airport route)

Semi-parallel airport-airport route(sec. to sec. airport route)

Base Network (Core to core

airports)

Semi Parallel Network

(Core to Sec. airports)

Parallel Network (Sec. to Sec.)

0.0

1.0

2.0

3.0

4.0

5.0

6.0

Mill

ions

Pass

enge

rs

55%

45%Base airport-airport route(core to core airport route)

LegendCore airport

Secondary airport

MIT MIT ICAT ICAT

18

Reachable Destinations from Core and Secondary Airports

0

20

40

60

80

100

120

140

LAX BUR ONT SNA LGB

Num

ber o

f des

tinat

ions

Destinations from thecore airport

Destination accessibleonly from thesecondary airport

Shared destinationswith the core airport

On average (for 10 regional airport systems), 38% of destinations reachable from the core airport are also reachable from secondary airports.

Illustration with three regional airport systems:

Implications of parallel networks for airlines:• Competition between carriers operating airport-to-airport routes within the same OD market• Cost implications:

Infrastructure cost (generally higher at core airports than secondary airports)Higher reliability of operations at secondary airports due to lower average delays than at core airportsDilution of operations for air carriers when operating at core and secondary airports simultaneously.

0

20

40

60

80

100

120

BOS PVD MHT

Num

ber o

f des

tinat

ions


Destinationaccessible only fromthe secondary airport


Boston Los Angeles

0

20

40

60

80

100

120

140

160

180

ORD MDW

Num

ber o

f des

tinat

ions


Destination accessibleonly from the secondaryairport


Chicago

MIT MIT ICAT ICAT

19

Conclusions

Scaling modes of air transportation networks:• Scaling “up” an existing network by adding resources• Scaling “out” to new nodes: construction of new airports• Scaling “out” to existing nodes: emergence of secondary airports• Scaling “out” through mode shift: emergence of air transportation services utilizing small

airports and small aircraft

Limited potential of adding capacity at major airports and building new airports

Increase the attractiveness of existing underutilized airports• Existing secondary airports will gain more traffic• New secondary airports are going to emerge• General aviation & business aviation reliever airports will also become key to

accommodating future demand growth

Implications • Air traffic control: larger number of airports with significant volume of traffic and coupled

operations• Airlines: emergence of parallel airport-airport routes within OD markets

MIT MIT ICAT ICAT

20

Questions&

Comments

Investigation of the Scalability of Air Transportation ...web.mit.edu/airlines/industry_outreach/board_meeting_presentation_files/meeting-oct...• Key constraints of the current air

Documents