Believe it or not?? Ghosts of Flights Crosswinds Delays Fuel

Heike Hofmann, Di Cook, Chris Kielion, Barret Schloerke, Jon Hobbs, Adam Loy,
Lawrence Mosley, David Rockoff, Yuanyuan Sun, Danielle Wrolstad, Tengfei Yin
Department of Statistics, Iowa State University
10
10
10
10
ARRIVAL & DEPARTURE
Most flights have 0 delay, with
fewer and fewer flights having
increasing delays. A secondary
peak occurs at 24 hours suggests a limit of 24 hours delay
is used by some carriers. Some
data is likely incorrect, eg flights
arriving 24 hours early.
6
4
2
0
-24
-12
0
12
24
36
Arrival Delay (hours)
Data
10
6
104
102
2000
2001
2002
Unique_Carrier
200
AA
AQ
150
AS
CO
100
DL
50
EV
Fuel Consumption (Mio Gallons)
Delayed, Cancelled, On Time, Boarding, ...... Flying in the USA
Fuel Efficiency
F9
0
2003
2004
FL
2005
HA
200
HP
150
MQ
NW
100
TW
50
TZ
UA
0
2006
2007
US
2008
WN
200
XE
150
B6
OH
100
DH
OO
50
The data are provided by Research and Innovative Technology Administration (RITA) and Bureau of Transportation Statistics (BTS).
Arrival and departure details for 123 million commercial flights throughout the United
States are recorded between October 1987
and December 2008, representing 29 commercial airlines and 3,376 airports. About 2.3 million flights were cancelled, 25 million flights
were at least 15 minutes late.
100
9E
0
-24
-12
0
12
24
36
10
Departure Delay (hours)
1988
1989
1990
BY SCHEDULED DEPARTURE
TIME & YEAR
Delays increase as day progresses. In 2001 they show an
overall decrease, likely to be
structural change, maybe FAA
policy. Delays deteriorated
again after 2003.
1991
20
15
10
5
0
-5
1992
1993
1994
1995
1996
Average Arrival Delay
20
15
10
5
0
-5
1997
1998
1999
2000
2001
Year
20
15
10
5
0
-5
1990
1995
2000
2005
2002
2003
2004
2005
2006
20
15
10
5
0
-5
40
50
60
70
10
20
30
40
50
60
70
10
20
30
40
50
60
YV
70
Carrier_Name
2007
NA
American Airlines Inc.
Aloha Airlines Inc.
1.2
Alaska Airlines Inc.
JetBlue Airways
Continental Air Lines Inc.
1.0
Delta Air Lines Inc.
Atlantic Southeast Airlines
Frontier Airlines Inc.
AirTran Airways Corporation
0.8
Hawaiian Airlines Inc.
America West Airlines Inc.
American Eagle Airlines Inc.
Northwest Airlines Inc.
0.6
Comair Inc.
Skywest Airlines Inc.
ATA Airlines d/b/a ATA
United Air Lines Inc.
0.4
US Airways Inc.
FUEL USE
Consumption versus distance flown. Three groups
of carriers are operating in
2000 - American Airlines is
moving into the lead,
Southwest is moving out of
the middle group, closing up
with AA and has overtaken
AA on distance flown and efficiency 2007-8.
1
100
NW
80
60
40 WNW
20
0 W
NNW N
2
NNE
WSW
SW
SSW
S
SSE
NW
NE
100
NW
80
60
40 WNW
20
0 W
All three runways in Phoenix are oriented East-West (see above). Do crosswinds affect air traffic?
On the right, arrival delays are plotted
against permutations of wind direction
(null plots). One plot shows the real
data of the observed wind directions.
FUEL EFFICIENCY
Smaller carriers more efficient, probably due to use of
smaller planes. Larger carriers American is one of the
least efficient, Southwest is
most efficient. Hawaiian Airlines very inefficient.
SSW
S
N
NNE
SSE
NNE
SSW
S
SSE
100
NW
80
60
40 WNW
20
0 W
E W
E W
ESE WSW
ESE WSW
ESE WSW
SW
SE
SSW
WSW
SW
SSW
S
S
SSE
SW
SE
SSW
NW
NE
NNW N
SSE
S
SSE
NNE
NW
NE
NNW N
NNE
ENE WNW
E W
E W
E W
ESE WSW
ESE WSW
ESE WSW
SSW
S
NW
NE
NNW N
SSE
SW
SE
SSW
S
SSE
SW
SE
NNE
NW
NE
NNW N
NNE
E W
E W
ESE WSW
ESE WSW
ESE WSW
S
SW
SE
SSW
NW
NE
NNW N
S
NNE
NW
NE
SSE
SW
SE
NNE
E W
E W
E W
ESE WSW
ESE WSW
ESE WSW
S
SSE
SE
SW
SSW
S
S
SSE
Arrival De
SE
NNW N
40
NNE
60
80
NE
ESE
SSW
S
SSE
SE
NNW N
SSE
SE
NNE
NE
ENE
E
SW
ESE
SSW
S
SSE
SE
Can you spot which one?
If you can differentiate the true plot from the null plots, we can reject the
null hypothesis of crosswinds being unrelated to arrival delays with a pvalue of less than 0.0625 (= 1/16).
Expressjet Airlines Inc.
2008
OO
20
15
10
5
0
-5
XE
MQ
YV
OH
EV
US
FL
WN
B6
CO
F9
AS
UA
AQ
DL
AA
NW
HA
Carrier
500 100015002000
500 100015002000
500 100015002000
500 100015002000
Mesa Airlines Inc.
Believe it or not??
500 100015002000
Scheduled Departure
BY DAY OF WEEK
Best days to travel and avoid delays are Saturdays, and Tuesdays
or Wednesdays. Fridays are bad
for delays.
Mo
10
8
We
6
Th
4
Sa
2
Su
1995
2000
2005
8
6.5
6
6.0
4
5.5
2
5.0
5
0
CO
1990
1995
2000
2005
2010
1985
1990
Year
1995
2000
2005
F9
EV
FL
15
10
5
0
HA
MQ
NW
OH
OO
UA
US
WN
XE
YV
1990 1995 2000 2005
1990 1995 2000 2005
1990 1995 2000 2005
1990 1995 2000 2005
1990 1995 2000 2005
15
10
5
0
0
1985
DL
2010
Year
15
10
Flight volume is on the increase - dramatically so
since 2000. Structural shifts (below) in the flight load
for airports lead to minimal average delays (above,
right) in `02 and `03. Delays have been on the increase
since.
5
0
1400
1200
Volume
1000
Sep '01 - W.T.C. Attack
Feb '91 - Eastern Airline
ceases all operations
March '89 - Eastern Airline
workers go on STRIKE
Mar 03' - Several carriers expand
their operations in Atlanta
Airport
?
ATL
800
ORD
Nov 03' - A.A. Buys T.W.A. and
cuts flights in St. Louis
600
SEA
STL
400
200
0
1985
1990
1995
2000
Time
2005
2010
Large changes in airports’
daily flight volume are triggered by different events,
including strikes, FAA order, and seasonal fluctuations,
Year
9E
AA
AQ
AS
B6
CO
Pinnacle
American
Aloha
Alaska
JetBlue
Continental
DL
Delta
EV
Atlantic
Number of flights (in 10.000)
Southeast
Frontier
2 F9
4
FL
AirTran
6
HA
Hawaiian
8
America West
10 HP
MQ American Eagle
NW
Northwest
OH
Comair
OO
Skywest
UA
United
US
US Airways
WN
Southwest
XE
Expressjet
YV
Mesa
I’m sorry, Sir, but your flight left 12h early’
According to the data, this was said to all passengers of
247 flights. Another 165 flights left at least 2h early.
Example: US Airways Aircraft N-881 - Ghostflight from PIT to RIC (222 miles)
Year Month Day DepTime
1995
3
8
1102
1995
3
8
1311
1995
3
8
1913
1995
3
8
2134
ArrTime Origin
1256
PIT
NA
CVG
2050
RIC
2300
PIT
Dest Diverted
CVG
0
PIT
1
PIT
0
MSY
0
``Within-City Hoppers’’- if you’re in a real time crunch
A total of 232,809 flights cover a distance of less than 50 miles. The shortest commercial flights occur between the New York airports La Guardia
(LGA) and John F. Kennedy (JFK). The distance is 11 miles, which according to google.maps can be covered in 18 min by car, and according to
data, takes 14 min of air time.
Ghost Flight Totals: over 1 million flights since 1995, with an average distance between airports of 1000 miles, corresponding to about 1.5 million
gallons of fuel. Since 2001 the number of ghost flights is cut, with major
airlines also decreasing the fuel consumption.
Smaller carriers are facing increasing costs
from ghost flights.
Number of Ghost Flights (in 1000)
100
30
20
10
0
0
2
4
6
8 10 12 14
5
10
15
Arrival Delay (min)
20
30
20
10
0
6
8
10
12
14
40
Average Wind Speed (mph)
Tools
80
30
60
40
20
0
CO
AA
DL
0
2000000
5
1995
30 40 50 60 70 80 90
BY AIRPORT
EWR (Newark) is the worst. ORD (Chicago O’Hare) is
not good, but also has high volume. DFW (Dallas-Fort
Worth) is relatively good - high traffic but relatively
small delay. Weather plays a huge role in delays - any
kind of preciptation, high winds, or reduced visibility
increases delays (scatterplots above)
Monthly Average Snow Fall (in)
0
40
Average Delay (in min)
40
Average Delay (in min)
Average Delay (in min)
Destination
Volume at Major Airports: Atlanta, Chicago, Seattle, St Louis
EWR
SFO
BOS
LGA
JFK
PHL
SEA
CMH
FLL
ORD
ATL
MIA
MCO
TPA
IND
MCI
PDX
SMF
LAX
LAS
MSY
DEN
SAT
CLE
SAN
ABQ
ONT
PIT
BWI
AUS
IAH
MDW
STL
DFW
HOU
IAD
PHX
RDU
OAK
DCA
CVG
DAL
SLC
SJC
DTW
SNA
HNL
CLT
BNA
MSP
MEM
Name
count
7.0
10
Carrier
Average % of Sunshine
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
mySQL database on fast server (thanks to Ted Peterson)
2007
Year
4
3
1500000
2
1
0
NW
UA
B6
FL
1000000
HA
1500000
5
500000
4
UniqueCarrier
2000000
Ghost Gallons
10
B6
Ghost Gallons
Average Arrival Delay (in min) by Flight
AS
CAN WE SEE WHAT IS NOT THERE?
Planes have, for reasons such as maintenance, weather, or schedule fly
empty between airports as so-called Ghosts. By tracking individual planes,
we reveal their paths, including situations, where a plane lands in a different airport than where it takes off later, i.e. a ghost:
Monthly Fuel use (in Mio Gallons) for Ghosts
# Flights (in Mio)
AQ
AA
15
average Arrival Delays (in min)
FLIGHTS OF `07
The maps above and below show all flights from 2007.
Southwest Airlines (WN) operates without a hub system, the
other airlines’ hubs are prominent. Small carriers tend to
operate locally.
9E
Three of American Airlines’ registered vehicles in the database are RAVEN hot air balloons. Based on the data, they
cruise at an impressive average speed of 430 miles an
hour. Fasten those hats!
Fr
Arrival Delay by Day of the week
DELAYS BY CARRIER, YEAR
Increasing delays
for small carriers,
except for Aloha.
Delta and US Airways are improving.
Racing Balloons?
Tu
1990
Hourly weather details for each airport from
Weather Underground at http://www.wunderground.com
Ghosts of Flights
DoW
12
Arrival Delay (min)
Additional BTS Data:
* Monthy fuel cost and consumption data by
carrier
* Fleet information by carrier
14
3
2
2003
2004
2005
2006
FL
1000000
HA
2007
2008
Google Earth
Year
1
500000
0
2000
2002
2004
2006
2000
2002
2004
Year
2006
2000
2002
2004
R and packages
- ggplot2 (Hadley Wickham)
- DBI, RMySQL (David James, Jeff Horner)
UniqueCarrier
B6
2006
2003
2004
2005
Year
2006
2007
2008
0
20
ENE
NW
NE
ENE WNW
SSW
NE
16
ENE WNW
SW
NNE
E
ENE WNW
SE
SSW
15
NNW N
NNW N
ESE
NW
NE
E W
SSE
SE
12
ENE WNW
SSW
S
SSE
ENE
11
ENE WNW
SW
SSW
E
ENE WNW
SE
ESE
NW
NE
ENE WNW
SW
NE
8
ENE WNW
SE
NNE
E
SW
SE
NNW N
ENE
7
14
NNE
NW
NE
E W
13
NNW N
NNE
ENE WNW
10
WSW
SW
NW
NE
4
ENE WNW
9
NNW
100
NW
80
60
40 WNW
20
0 W
NNE
6
WSW
SW
3
NNW N
ENE WNW
5
NNW N
NNW N
Southwest Airlines Co.
2007
Additional Sources
30
Distance (Mio Miles)
Miles Per Gallons
1987
20
Crosswinds
Arrival Delays (in min)
Delays
supported by NSF grant # 0706949
100