Distance-Based GDPs
Guillermo Lulli, Mike Ball (U of MD)Bob Hoffman (Metron)
Distance-Based GDP
Replaces Tier ApproachGDP options are defined by drawing a
circle around the airport and including all enclosed airports: All flights whose origin is within circle and
whose arrival time is within GDP time period are included.
Alternative programs obtained by increasing radius of circle.
Distance-based GDP vs Current Approach
Flexibility: a wider range of possible solutions is available.
Dominance: for each tier-based initiative, there is a corresponding distance-based initiative.
Simplicity: the range of programs can be varied in a simple well-defined manner.
Airports considered
Atl 2 Nov 99 ft 16.54 19.00 - 23.00 aar 76 Bos 2 Feb 00 ft 17.02 19.00 - 00.59 aar 40 Dfw 10 Oct 98 ft 12.48 14.00 - 19.59 aar 64 Dtw 6 Apr 00 ft 15.14 16.00 - 21.59 aar 36 Ewr 27 Feb 00 ft 15.19 18.00 - 23.59 aar 36 (3)/40 (3) Iah 24 Feb 00 ft 17.02 19.00 - 00.59 aar 40 Ord 3 Feb 00 ft 16.51 18.00 - 01.59 aar 76 (2)/80 (6) Phl 13 Aug 99 ft 18.04 20.00 - 01.59 aar 40 Sfo 18 Aug 99 ft 13.32 16.00 - 21.59 aar 30 (3)/40 (3) Stl 19 Mar 99 ft 12.58 16.00 - 21.59 aar 32
Statistics used in our analysis:
Total delay Airborne Holding delay Unrecoverable delay Affected flights Max delay Average delay Unnecessary delay: part of delay that is not necessary to
meet the increased AAR, if the program is cancelled
if OTD<CT<ETA min{ CT- OTD, ETD-OTD, ETA-CT},
else 0.
Bos 2 Feb 99
Distances vs Tiers
Aff flights
Max
Avg
Unrec
800
1300
1800
2300
2800
3300
3800
4300
4800
80 120 140 150 161 165 174 230 245 300 350 410 540 700 760 910 1000 1090 1600 >1600
0
100
200
300
400
500
600
700
Air Hold
Bos 2 Feb 99
Average delay vs Unrec. delay
Unrec
0
50
100
150
200
250
80 120 140 150 161 165 174 230 245 300 350 410 540 700 760 910 1000 1090 1600 >1600
Avg
Max
dela
ys
0
100
200
300
400
500
600
700
Unre
c.
dela
y
Avg.
Trading Off Avg Delay and Unrecov Delay
Choose weights for each and consider:wgt1 ave-delay + wgt2 unr-delay
Find distance that minimizes this weighted sum.
Nominal values for this analysis:wgt1 = 50wgt2 = 1
Bos 2 Feb 99
Weighted trade-off between Avg and Unrec. delay
300
350
400
450
500
550
600
650
300 350 410 540 700 760 910 1000 1090 1600 >1600
1100
1150
1200
1250
1300
1350
1400
1450
1500
Unrec
Trade-off
Avg
Bos 2 Feb 99
Varying Tradeoff Weights
400
500
600
700
800
900
1000
1100
1200
1300
700 760 910 1000 1090 1600 >1600
1
98
Bos 2 Feb 99
Avoiding Inefficient Solutions
0
50
100
150
200
250
300
350
400
80 120 140 150 161 165 174 230 245 300 350 410 540 700 760 910 1000 1090 1600 >1600
Avg M
ax d
ela
ys
0
100
200
300
400
500
600
700
Unre
c.
dela
y
Max
Avg
Unrec
Efficient solutions
Efficient Initiatives: numerical example
By varying distances inefficient initiatives can be avoided (Bos 2 Feb 99)
Distance 350 410 1000 1090Unrecov. Delay 407 390 525 507Avg Delay 54 49 32 31Max Delay 115 109 77 77
Bos 2 Feb 99
Consideration of Fairness: Delay Variability
Unrec
Avg.
0
100
200
300
400
500
600
700
80 120 140 150 161 165 174 230 245 300 350 410 540 700 760 910 1000 1090 1600 >1600
6
8
10
12
14
16
18
20
Delay Var
Stl 19 Mar 99
Distances vs Tiers
Air Hold Delay
Max
Avg
Aff flight
Unrec0
2000
4000
6000
8000
10000
12000
0
100
200
300
400
500
600
1th 2nd
Stl 19 Mar 99
Average delay vs Unrec. delay
0
50
100
150
200
250
300
350
400
0
500
1000
1500
2000
2500
3000
Unrec
Avg
Stl 19 Mar 99
Weighted trade-off, Avg and Unrec. delay
1500
1700
1900
2100
2300
2500
2700
625 680 760 810 910 1100 1400 >1400
55
60
65
70
75
80
85
90
Unrec
Avg
Trade-off
2nd Tier
Stl 19 Mar 99
Different Weighted trade-off
3200
3300
3400
3500
3600
3700
3800
3900
4000
680 760 810 910 1100 1400 >1400
20
68
Stl 19 Mar 99
Avg delay vs Unrec. Delay & Delay Variability
0
500
1000
1500
2000
2500
0
10
20
30
40
50
60
70
Avg
Delay Var
Unrec
Sfo 18 Aug 99
Distances vs 6 & 10 west
Avg
Max
Aff flights
Unrec
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
70 135 170 220 290 300 360 390 500 570 600 850 1400 1599 1600 1860 2200 2241 >2241
0
50
100
150
200
250
300
350
400
Air Hold
6W 10W
Sfo 18 Aug 99
Weighted trade-off, Avg and Unrec. delay
1000
1100
1200
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
500 570 600 850 1400 1599 1600 1860
0
10
20
30
40
50
60
70
80
Avg
Unrec
Trade-off
6W 10W
Unrecoverable Delay ==>Unnecessary Delay
unrecoverable delay = ground delay incurred by start time of program:if program canceled at start then all of this delay is
unnecessary but none is recoverableWhat if program is canceled earlier or later than start
time of program??unnecessary delay = ground delays applied to flights
that push their CTAs past the program cancellation time
expected unnecessary delay = expected value of unnecessary delay computed using distribution of possible cancellation time.
Sfo 18 Aug 99
Unnecessary delay (early cancellation time)
0
500
1000
1500
2000
2500
3000
3500
70 135 170 220 290 300 360 390 500 570 600 850 1400 1599 1600 1860 2200 2241 >2241
1502 1532 1600 1632 1703 1733 1804
Sfo 18 Aug 99
Unnecessary delay (middle cancellation time)
0
500
1000
1500
2000
2500
3000
3500
70 135 170 220 290 300 360 390 500 570 600 850 1400 1599 1600 1860 2200 2241 >2241
1804 1834 1904 1934 2004
Sfo 18 Aug 99
Unnecessary delay (late cancellation time)
0
200
400
600
800
1000
1200
1400
1600
1800
70 135 170 220 290 300 360 390 500 570 600 850 1400 1599 1600 1860 2200 2241 >2241
Un
ne
c. D
ela
y
2004 2034 2104 2134
Sfo 18 Aug 99
Unrecoverable vs expected Unnecessary delay
0
500
1000
1500
2000
2500
3000
70 135 170 220 290 300 360 390 500 570 600 850 1400 1599 1600 1860 2200 2241 >2241
Unre
c &
Ex
Unnec d
ela
y
Unrec
Unnec
Sfo 18 Aug 99
Unrec., Unnec. delay and weighted trade-off
3680
3780
3880
3980
4080
4180
570 600 850 1400 1599 1600 1860 2200 2241 >2241
950
1150
1350
1550
1750
1950
2150
weighted trade-off
Unnec
Unrec
Avg
local minimum
Conclusions:
Distance-based GDPs provider greater flexibility but are easy to understand and manipulate.
Flexible user interface could be developed to display and analyze options.