NEXTOR
Airport Capacity, Airport Delay,
and Airline Service Supply:
The Case of DFW
Faculty and Staff: D. Gillen, M. Hansen, A. Kanafani, J. Tsao
Visiting Scholar: G. Nero and
Students: S. A. Huang and W. Wei
Institute of Transportation Studies
University of California at Berkeley
1
NEXTOR
Presentation Outline
(
(
(
(
Background Study/Motivation
Delay Changes
Airline Adaptations to New Capacity
Conclusions & Future Work
2
1
NEXTOR
Background
( 10/96 Changes at DFW
• New runway (October 1)
• Airspace redesign (October 10)
( Expected Capacity Impacts
• VFR: from 102 to 146 operations/hr
• IFR: from 66 to 108 operations/hr
( Expected Benefit
• “75% increase in regional capacity; 15-25% nationwide”
( Research Questions
• How did delay change?
• How did airlines adapt?
3
NEXTOR
Airport Average Arrival Delay
(against schedule, per flight)
Average Flight Delay
25
ATL
DFW
ORD
15
10
5
n-
Fe 96
b9
M 6
ar
-9
Ap 6
r
M -96
ay
Ju 96
n9
Ju 6
l-9
Au 6
gSe 96
pO 96
ct
-9
N 6
ov
D 96
ec
Ja 96
nFe 97
b9
M 7
ar
-9
Ap 7
r
M -97
ay
Ju 97
n9
Ju 7
l-9
Au 7
gSe 97
pO 97
ct
-9
N 7
ov
D 97
ec
Ja 97
nFe 98
b98
0
Ja
Average Delay per Flight
20
-5
Month
4
2
NEXTOR
Airline System Arrival Delay
(against schedule, per flight)
18
AA
AS
DL
16
UA
WN
14
Minutes
12
10
8
6
4
2
0
95-1Q
95-2Q
95-3Q
95-4Q
96-1Q
96-2Q
96-3Q
Time
96-4Q
97-1Q
97-2Q
97-3Q
97-4Q
98-1Q
5
NEXTOR
Schedule Padding
( Compare AA scheduled flight
(gate to gate) time between DFW
and 20 largest airports in January
96 and January 98
( Find AA scheduled flight times
increased by 4 minutes
• On - time performance initiative
• Accounts for DFW and AA delay
changes
6
3
NEXTOR
Effective Flight Time
( Effective Flight Time (EFT) is the
time between scheduled departure
and actual arrival of a flight.
( EFT is unaffected by changes in
scheduled flight time.
( EFT can be decomposed into
departure delay, taxi time, and
airborne time.
7
NEXTOR
Aggregation to Daily Level
( Daily Average Flight Time (DAFT) is
the mean daily EFT of all arrivals
from a particular origin.
( Daily Flight Time Index (DFTI) is a
weighted average of the DAFT for a
fixed “market basket” of origins.
( Like EFT, DAFT and DFTI can be
decomposed into departure delay,
taxi time, and airborne time.
8
4
NEXTOR
DFTI Before and After
Capacity Expansion
250.00
DFTI
200.00
150.00
100.00
Before
After
Date
1/1/98
10/1/97
4/1/97
7/1/97
1/1/97
10/1/96
4/1/96
7/1/96
1/1/96
10/1/95
4/1/95
1/1/95
0.00
7/1/95
50.00
( No obvious
impact for good
days.
( Some indication
of decrease in bad
days.
( Some very bad
days both before
and after.
9
NEXTOR
Daily Level DFTI Model
( Statistical model estimated on
daily observations of DFTI at DFW
and explanatory variables.
( Explanatory variables include:
• Demand
• Weather
• Delay at Origin Airports
• Expansion
10
5
NEXTOR
Metrics for Explanatory Variables
Explanator
Demand
Metric
Hypothetical deterministic delay based on
scheduled arrival times and nominal
capacity.
Four weather factors resulting from
performing factor analysis on 15 weather
variables. Factors correspond to
temperature, wind, precipitation, and
visibility.
Average departure delay for flights not
going to DFW region from origin airports
included in “market basket.”
“Dummy” variable =0 prior to 10/1/96 and
1 after 10/31/96.
Weather
Origin airport
delay
Capacity
Expansion
11
NEXTOR
Predicted vs Actual DFTI
Actual DFTI
240.00
220.00
200.00
180.00
160.00
140.00
140.00
160.00
180.00
Predicted DFTI
200.00
220.00
( Model accounts for
most variation in
DFTI.
( Pronounced
heteroskedasticity
poses estimation
problems.
12
6
NEXTOR
Effect of Capacity Expansion
on DFTI
( Strong interaction with visibility
factor.
( After expansion, impact of low
visibility on DFTI is essentially
eliminated.
( No other significant effects.
13
NEXTOR
Day-to-Day Variation in Impact
Frequency
150
100
50
0
-10 -9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
-1
0
1
2
3
Cumulative Frequency
DFTI Difference
600
500
400
300
200
100
0
-10 -9
-8
-7
-6
-5
-4
-3
-2
DFTI Difference
( Less than 1 minute
reduction on 60
percent of days.
( Greater than 4 minute
reduction on 10
percent of days.
( Largest reductions of
10 minutes.
( Slight increase on
about 25 percent of
days.
14
7
NEXTOR
Expected Reduction of DFTI
DFTI Impact vs DFTI Value
15
10
5
0
140.00 150.00 160.00 170.00 180.00 190.00 200.00 210.00
-5
Expected DFTI without Expansion
( Greatest estimated
reductions on days
with greatest
estimated delays.
( Implies increase in
reliability.
( May also have
masked benefit
since “bad” days
are still “bad.”
15
NEXTOR
Effect of Capacity Expansion
on DFTI Components
1
0.5
0.75
0
AIRBORNE TIME
-0.5
-1
TAXI TIME
-1.88
DEPARTURE
DELAY
-1.5
-2
-2.5
-0.21
( Sizable reduction in
departure delay due
to reduced ground
holding.
( Increase in taxi time
due to distance of
runway from
terminal.
( Small reduction in
airborne time.
16
8
NEXTOR
Sources and Components of DFTI
Change from Before to After
Expansion
Source
Departure
Delay
Taxi
Time
Airborne
Time
TOTAL
CONTRIBUTION
Demand
0.0771
0.3679
3.4032
3.8482
Origin Airport
Delay
Total Weather
0.0317
0.0043
-0.1437
-0.1077
0.1672
-0.1188
-0.2805
-0.2321
Origin Airport
Delay-Weather
Interactions
Capacity
Expansion
TOTAL
CHANGE
0.2028
0.0827
0.0341
0.3196
-1.8810
0.7549
-0.2091
-1.3353
-1.4022
1.0910
2.8040
2.4927
17
NEXTOR
Airline Adaptation Metrics
( Number of Operations
( Peaking
( Concentration
( Scheduled Layover Times
18
9
NEXTOR
Airline Departures from DFW
Departures from DFW
120000
Operations
100000
80000
TOT
60000
AA+DL
40000
EV+MQ
Others
20000
0
94,1
94,3
95,1
95,3
96,1
96,3
Quarter
97,1
97,3
( No significant
increase after
expansion.
( Remains 10
percent below preexpansion highs.
( No change in
AA/DL dominance.
19
NEXTOR
Schedule Peaking
( Find shortest time interval in which
given number of operations are
scheduled (50 operations in this
case).
( Convert to equivalent operations
per hour.
( Performed for arrivals, departures
and both.
20
10
NEXTOR
Schedule Peaking Results
(AA Only)
350
300
Operations
250
200
D
A
Both
150
100
50
0
01,96
07,96
01,97
07,97
01,98
Time
"Peak-hour" Operations
( Arrival peaking
stable.
( Departure peaking
decreases.
( Combined peaking
increases.
( Reduced gap
between departure
and subsequent
arrival banks.
21
NEXTOR
AA Cumulative Arrival and
Departures--Jan’ 96 and Jan’ 98
January 1996 operation
January 1998 operation
600
500
500
400
400
Operation
operation
600
300
A
300
D
200
200
100
100
0
0
0
5
10
time
15
20
0
5
10
15
20
time
22
11
NEXTOR
Hypothetical Deterministic Delay
(HDD)
HDD
250
cumulative number
200
150
demand
100
capacity
50
“area”= total delay
0
8
10
12
time
14
16
( Combined measure of
banking intensity and
duration
( Based on hypothetical
delay as a function of
• flight schedule
• assumed capacity level
(90 operations per hour
in this case)
( Delay calculated from
input-output diagram
23
NEXTOR
( Sharp jump in
summer 1996.
( Further increase after
expansion.
( Impact of expansion
is uncertain
Oct-97
Jan-98
Apr-97
Oct-96
Date
Jan-97
Apr-96
After
Jul-96
Oct-95
Jan-96
Apr-95
Jul-95
Before
Jul-97
3.50
3.00
2.50
2.00
1.50
1.00
0.50
0.00
Jan-95
Smoothed HDD
Trends in HDD
• Large values in early
1995.
• Fluctuations after
expansion.
• Role of confounding
factors (pilots union).
24
12
NEXTOR
Total Impact of Expansion on
Fight Time
(
Assumed
Direct Indirect Total
Impact on
DFTI
DFTI
Impact
Demand
Impact Impact
None
-1.39
0
-1.39
Comparison
-1.39
0.51
-0.88
with JulySep,1996
Comparison
-1.39
1.78
0.39
with AprilJune,1996
(
(
Direct delay reduction
offset by “Induced”
change in demand
(HDD).
Degree of induced
change uncertain.
Under most plausible
assumption, airlines
“traded in” about 40
percent of direct
reduction for schedule
modifications.
25
NEXTOR
Conclusion
( Expansion reduced delay, all else
equal, about 1.3 min/flight.
( Most reduction on bad weather
days.
( Some (~40 percent) of delay
reduction apparently traded for
schedule changes.
( Adaptations may have increased or
decreased benefit.
26
13