Operational Performance and Demand Management. Mark Hansen NEXTOR Short Course 10/14/04

Transcription

1 Operational Performance and Demand Management Mark Hansen NEXTOR Short Course 10/14/04 1

2 Outline Recent trends in NAS Operational Performance The Costs of Delay Operational Impacts of Supply and Demand Side Changes DFW Case Study LGA Case Study The Case for Demand Management 2

3 Recent Trends in NAS Operational Performance The Daily Flight Time Index Average Arrival Delay 3

4 Daily Flight Time Index Daily Flight Time Index (DFTI) is a NAS performance metric that reflects the flight time and its components for an average commercial passenger flight DFTI has been calculated for Key trends Increased 7 min from Decreased to 1995 levels by summer 2002 Subsequently increased 2 min, mainly due to increased airborne time 4

5 DFTI and its Components location DFTI DOTI DATI DDTI Dest. Gate Dest. Runway Origin Runway Origin Gate time Scheduled Departure Time Actual Departure Time Wheels-Off Time Wheels-On Time Actual Arrival Time 5

6 DFTI Trends: /11/2001 DFTI DFTI DATI DFTI 365 MA DATI 365 MA 90 Jan- 95 Jul- 95 Jan- 96 Jul- 96 Jan- 97 Jul- 97 Jan- 98 Jul- 98 Jan- 99 Jul- 99 Jan- 00 Jul- 00 Jan- 01 Jul- 01 Jan- 02 Jul- 02 Jan- 03 Jul- 03 Date 6

7 70 DFTI Trends: /11/ DFTI DOTI DDTI DOTI 365 MA DDTI 365 MA Jan- 95 Jul- 95 Jan- 96 Jul- 96 Jan- 97 Jul- 97 Jan- 98 Jul- 98 Jan- 99 Jul- 99 Jan- 00 Jul- 00 Jan- 01 Jul- 01 Jan- 02 Jul- 02 Jan- 03 Jul- 03 Date 7

8 Constructing the DFTI (New Method) Based on ASQP data Covers all flights by major pax carriers Provides out-off-on-in times for all domestic flights Weighted Average Set of city pairs identified and city-pair weights calculated Average flight time calculated for each city pair City-pair weights applied to determine overall average City pairs and their weights adjusted monthly Control for effects of re-weighting to maintain comparability 8

9 Steps in Constructing the DFTI Identify city-pairs Calculate city-pair weights Calculate unadjusted DFTI Calculate adjustment factors and adjusted DFTI 9

10 Identify City Pairs and Calculate Weights Identify city pairs for which there is at least one completed flight with valid data every day over a two-month period valid data: departure delay > -30 min and arrival delay<480 min Fi Wi = F Calculate weights as W i- Weight for city-pair i j CP F i - Flights for city-pair j during study period CP Set of city-pairs in the DFTI j 10

11 City Pair Daily Average Flight Time f Flight index i City-pair index d Day index DAFT id = f S id N FT id f S id Set of flights for citypair i on day d N id Number of flights in S id DAOT id = f S id OD N Daily Average Origin Time f id + TO f DAAT id = f S N AB id Daily Average Airborne Time id f DADT id = f S id N TI id Daily Destination Destination Time f 11

12 Daily Flight Time Index DFTI d = i cp W i DAFT id DOTI d = Wi[ DAOTid ] DATI d = Wi[ DAATid ] DDTI d = Wi[ DADT id ] i cp i cp Origin Time Airborne Time Destination Time i cp 12

13 Adjusted DFTI Allows DFTI to incorporate large and continually changing mix of city pairs (around 2000) Preserves comparability over time Based on comparing DFTI s for common month calculated with different weights 13

14 Alternative Weights for Month 2 City Pair Month 1 Month 2 Month 3 1 W 1 12 W W 2 12 W 2 12 W 2 23 W W 3 12 W 3 12 W 3 23 W W 4 23 W

15 Adjustment Factors Calculate unadjusted DFTI s for months and months 2-3: DFTI d and DFTI d Compare results for month 2 Calculate adjustment factors: Want: Solution: AVG( β + α DFTI VAR( β + α DFTI α = β VAR( DFTI VAR( DFTI = AVG( DFTI ) = AVG( DFTI ) = VAR( DFTI ) ) 23 ) α AVG( DFTI 2 23 ) 12 ) ) 15

16 Adjusted DFTI Determine baseline month (in our case this is January 2000) Calculate adjustment factors recursively forward and backward to beginning and end of time period Calculate adjusted DFTI 16

17 Trends in Arrival Delay Against Schedule Observed Delay Delay (min/flight) FY2000 FY2001 FY2002 FY2003 FY2004 Tim e Periods 17

18 Decomposition of Delay Difference by Causes (2004 vs. 2003) Delay difference (min/flt) Model 1 Model 2 Models EDCT Others Fiscal Year Storm Wind speed IFR ratio Operation Queuing 18

19 The Costs of Delay Not linear or additive these are accounting conventions, not empirically supported relationships Airline cost function study Cost= f(output, factor prices, ops metrics) Metrics included delay, irregularity, and disruption Only disruption had significant effect on costs Aggregate cost estimates of similar magnitude to those using standard cost factors: $2-4 billion in 1997 Does not include costs to passengers 19

20 Operational Impact of Demand and Supply Side Changes Case study of new runway at DTW Case study of Air-21 at LGA 20

21 Effect of New Runway at DFW 21

22 FIGURE 4 15-min Arrival and Departure Counts at DTW, VMC Conditions, Jan-June 2002 Arrivals Departures Number of Periods with Observed Count 22

23 FIGURE 5 Change in Distribution of Arrival and Departure Counts, VMC Conditions, Jan-June Arrivals Departures Change in Proportion of Periods with Observed Count (Shaded if Increase, Unshaded if Decrease) 23

24 FIGURE 6 Change in Distribution of Arrival and Departure Counts, IMC Conditions, Jan-June Arrivals Departures Change in Proportion of Periods with Observed Count (Shaded if Increase, Unshaded if Decrease) 24

25 FIGURE 7 Clearance Rates, DTW Arrivals, by Year and Visibility Condition Clearance Rate Demand IMC IMC VMC VMC

26 FIGURE 8 Clearance Rates, DTW Departures, by Year and Visibility Condition Clearance Rate Demand IMC IMC VMC VMC

27 Air-21 at LGA Effects of past policies on operational performance at LGA Interaction of LGA and the rest of the National Airspace System (NAS) 27

28 Epochs The HDR period: from January through August of The AIR-21 period: from September, 2000 through January of The Slottery period: from February 2001 through September 10, Post 9/11 period: through the end of Year Year The first half of Year

29 Average Weekday Scheduled Arrivals at LGA, by Month Arrivals per Day HDR AIR-21 Slottery Post-9/ Aug-99 Mar-00 Oct-00 Apr-01 Nov-01 May-02 Dec-02 Jun-03 Jan-04 Aug-04 Feb-05 Month 29

30 Operational Performance Metrics at LGA Average Arrival Delay Cancellation Rate Saturation Rate Arrival Count at saturation Arrival Demand at saturation Airport Acceptance Rate 30

31 31 Operational Performance of LGA Operational Performance of LGA Year Year Year Post 9/ Slottery AIR HDR IMC VMC IMC VMC IMC VMC IMC VMC IMC VMC IMC VMC AAR Arrival Demand* Arrival Count* Saturation Rate Cancellation Rate Average Delay Periods

32 Multivariate Model of LGA and NAS Delay Dependent variable Arrival Delay For LGA, arrival delay at the rest of the system For the rest of the system, arrival delay at LGA Explanatory variables Deterministic Queuing Delay Adverse Weather En-route (Thunderstorm ratio) Terminal (IFR ratio) Expected Departure Clearance Time (EDCT) Holding (EDCT ratio) Total Flight Operations 32

33 33 Model Specification Model Specification Model 1 (Arrival delay at LGA) Model 2 (Arrival delay at rest of Benchmark Airports) ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ˆ ) ( t t D t S t W t E t E t I t I LQ t t D t D j j jl i i il k k kl L L S L ν θ ω λ β β β β β β α = ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ˆ ) ( ) ( t u t D t S t W t I t I SQ t t D t OP t D n n ns m m ms l l ls S S L S = θ ω λ γ γ γ γ α

34 Estimation Results of Delay at LGA (1) Description Estimate Standard Error p-value Intercept Ds(t) Predicted arrival delay for NAS <.0001 LQ(t) Average queuing delay at LGA E(t) EDCT_ratio (count of EDCT holding arriving at LGA / total scheduled arrivals) <.0001 E(t) 2 Square of EDCT_ratio <.0001 I(t) IFR_ratio (Proportion of the day operated under IMC condition) <.0001 I(t) 2 Square of IFR_ratio <.0001 W 5 (t) Thunder storm ratio (number of stations reported thunderstorm / total amount of stations) in Region <.0001 R-Square

35 Estimation Results of Delay at LGA (2) Description Estimate Standard Error p-value D 1 (t) Dummy variable for the AIR-21period D 2 (t) Dummy variable for the Slottery period <.0001 D 3 (t) Dummy variable for the post 9/11 period D 4 (t) Dummy variable for Year <.0001 D 5 (t) Dummy variable for Year <.0001 D 6 (t) Dummy variable for Year <.0001 S 1 (t) Dummy variable for Quarter S 2 (t) Dummy variable for Quarter S 3 (t) Dummy variable for Quarter

36 Decomposition of LGA Delay Average Arrival Delay (Min per Flt) by Causes Demand Management Weather IFR_ratio EDCT_ratio Average arrival queuing delay Average observed delay at other airports HDR AIR-21 Slottery Post 9/11 Year2002 Year2003 Year

37 Estimation Results of NAS Delay Description Estimate Standard Error p-value Intercept OP(t) Total operations (Arrivals) in the system <.0001 D L (t) Predicted average arrival delay at LGA <.0001 SQ(t) Average arrival queuing delay of system <.0001 I(t) IFR_ratio (Proportion of the day operated under IMC condition) I(t) 2 Square of IFR_ratio W 1 (t) Thunderstorm ratio in Region W 2 (t) Thunderstorm ratio in Region <.0001 W 3 (t) Thunderstorm ratio in Region W 4 (t) Thunderstorm ratio in Region <.0001 W 5 (t) Thunderstorm ratio in Region <.0001 W 6 (t) Thunderstorm ratio in Region <.0001 R-S

38 Estimation Results of NAS Delay Description Estimate Standard Error p-value D 2 (t) Dummy variable for the AIR-21 period D 3 (t) Dummy variable for the Slottery period D 4 (t) Dummy variable for the post 9/11 period D 5 (t) Dummy variable for year <.0001 D 6 (t) Dummy variable for year <.0001 D 7 (t) Dummy variable for year 2004 (half of the year) S 1 (t) Dummy variable for quarter S 2 (t) Dummy variable for quarter <.0001 S 3 (t) Dummy variable for quarter <.0001 R-Square

39 Conclusion AIR-21 period witnessed operational improvements at LGA The entire delay impact of AIR-21 was in the form of increased EDCT-related delays 1 minute delay at LGA generates about 1.7 minutes delay for the rest of the system Traffic and delay at LGA are approaching pre-9/11 levels 39

40 The Case for Demand Management Microanalysis of Queuing Delay at LAX Demand-side Aspects of the Delay Problem Delay Management Altnernatives Final Thought 40

41 Example Interarrival Times for L=7nmTrailing Leading Embraer Jetstream Super Airbus Airbus BAe Boeing Boeing Douglas DC Douglas MD Douglas MD Saab Airbus Airbus Boeing 747 1* Boeing 747 2* Boeing Boeing Douglas DC Douglas MD Ilyushin II Lockheed L Boeing

42 Impact of Fleet Mix on IFR Arrival Capacity Number of Arrivals Hourly Arrival Capacity Heavy B757 Large Small Capacity Hour of Day 42

43 Delay Impacts Used deterministic queueing analysis to assess marginal delay impacts of individual flights First-cut analysis IFR Nominal Separations Two arrival runways No flight cancellations No traffic flow management 43

44 Queuing Diagram for LAX C umulative Num ber of Arrivals Scheduled Cap acity Co nstrained :00 10:00 12:00 1 4:00 Time of Day 44

45 Queuing Diagram II Cum ulative Num ber Queue length at 11:00 Delay for 310th Scheduled Arrival Scheduled Capacity C onstrained :00 10:30 11:00 11:30 12:00 Time of Day 45

46 Illustration of Procedure 260 Scheduled Arrivals 258 Completed Arrivals Cumulative Number Completed w/o 253 Delay Reduction to Flight 256 if Flight 253 did not occur :40 14:45 14:50 14:55 15:00 Time of Day 46

47 During Peak Periods, Flights Generate 3.50 Significant Incremental Delays Incremental Delay Impact (AC-Hrs) Small Large B757 Heavy :00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 0:00 Scheduled Arrival Time 47

48 Delay Impact Ratio (DIR) Weighs delay impact against convenience Numerator is congestion delay impact (CDI) of a flight (in seat-hrs) Denominator is extra schedule delay if flight did not occur, and passengers had to take previous flight from same origin on same airline (SDI) Any flight with DIR>1 is of dubious social value 48

49 Delay Impact Ratio (DIR) DIR = congestion delay caused by flight( seat hrs) schedule delay saved by flight( seat hrs) 49

50 Some Flights Have Very High DIRs Previous Flight Time of Flight Time of Flight Type Seats Origin Departure Number Departure SDI CDI DIR US J31 18 SAN 9: : US J31 18 FAT 9: : US J31 18 SAN 9: : US J31 18 SAN 9: : UA EM2 30 SAN 9: : UA EM2 30 SAN 9: : OE 7338 J31 18 OXR 9: : UA EM2 30 SAN 10: : OE 7017 J31 18 SNA 9: : UA EM2 30 SAN 10: : US J31 18 SAN 20: : UA EM2 30 PSP 9: : UA EM2 30 MRY 9: : A SF3 33 PSP 8: : UA EM2 30 SBA 10: : OO 5657 EM2 30 SAN 9: : UA SFO 8: : UA EM2 30 PSP 10: :

51 Demand-side Aspects of Delay Problem Schedule competition (frequency and flight times) Limited cost economies in aircraft size User charges geared toward cost recovery instead of capacity allocation 51

52 But, Because Pilot Cost Increases with Aircraft Size, Airlines Don t t Save from Upsizing Cost Per Seat-Mile ($) mi (Ex) 2400 mi (Ex) 400 mi (En) 2400 mi (En) Aircraft Seats 52

53 Demand Management Alternatives Demand Management Alternatives Auctions Currently under consideration for LGA Various forms Challenges What is appropriate number of slots Service to small communities Need to other resources (gate, curbside, baggage handling) Pricing Present pricing structure is obsolete Charge Congestion Surcharges During Peak Periods Significant Implementation Issues Administrative Alternatives 53

54 Flights During Peak Generate High Marginal Costs $3,500 $3,000 Surcharge per Landing $2,500 $2,000 $1,500 $1,000 $500 $ Hour of Day 54

55 Administrative Alternatives Slightly Modified HDR Slot Use Restrictions Performance-Based Allocation Industry Self-regulation with Government Facilitation 55

56 Alternative 1-Slightly 1 Modified HDR Grandfathered allocation with blind secondary market and use or lose provision Three slot categories: air carrier, small communities, non-scheduled ~3% of slots per year re-allocated to new entrants based on lottery 56

57 Alternative 2-Slot 2 Use Restrictions All slots re-allocated over 5 year period Staged re-allocation based on a/c size classes: 150+ seats, seats, <100 seats Restrictions carry over intro secondary market Possibly modify perimeter rule Possibly designate time windows for small aircraft slots Possibly allow joint operation of larger flights 57

58 Alternative 3-Performance 3 Based Allocation 5% of slots re-allocated every six months Formula-based withdrawal and re-allocation Withdraw more slots from airlines with low pax/slot ratios in previous six months Award more slots to airlines with high pax/slot ratios at LGA or pax/flight ratios elsewhere May also consider Higher weights for small community pax or separate categories for small communities Exemptions for minimum market presence slots On-time performance 58

59 Alternative 4-Self4 Self-regulation Turn over regulatory responsibility to airlines Form Responsible Scheduling Committee of all interested airlines (not just incumbents) Create principles, metrics, and criteria for responsible scheduling Create support tools and methods to enable airlines to schedule responsibly Scheduling conflict resolution mechanisms Graduated sanctioning for bad actors Circuit-breaker allows FAA to re-impose slot controls is ops situation degrades unacceptably 59

60 A Final Thought What is efficient use of LGA? Maximize pax throughput and thus time savings generated by the airport? Maximize WTP of those using LGA? Should we weight everyone s time equally of everyone s money equally? 60