A Macroscopic Tool for Measuring Delay Performance in the National Airspace System. Yu Zhang Nagesh Nayak

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "A Macroscopic Tool for Measuring Delay Performance in the National Airspace System. Yu Zhang Nagesh Nayak"

Transcription

1 A Macroscopic Tool for Measuring Delay Performance in the National Airspace System Yu Zhang Nagesh Nayak

2 Introduction US air transportation demand has increased since the advent of 20 th Century The Geographical extent of the country makes air transportation a viable option About 70% of airline operations are served by 35 OEP airports NextGen intends to reduce total flight delay by percent by 2018 when compared to do-nothing scenario The FAA Capital Investment Plan (CIP) intended to invest about $16.6 billion as of April 2009 for airport related projects From planning and policy point of view we need tools to test the system-wide effects of these investment and provide guidance for further investment 2

3 Operational Evolution Partnership (OEP) Airports Source: Federal Aviation Administration (faa.gov) 3

4 Methodology Estimate the interaction of flight delay at one single airport and delay of the rest of the National Airspace System (NAS) Predict system-wide impact of capacity improvement or demand management strategies at a single airport Case study: LaGuardia International Airport (LGA) Chicago O Hare International Airport (ORD) 4

5 SYSTEM-WIDE BENEFIT OF CAPACITY EXPNSION OF INDIVIDUAL AIRPORT Capacity LGA/ORD AAR Queuing Delay LGA/ORD Convective weather IMC Regression Estimates Baseline Observed Delay at LGA/ORD Predicted Arrival Delay at LGA/ORD Single Airport Percentage Change of Arrival Delay at LGA/ORD Total Operations (Arrivals) at NAS Convective Weather Regression Estimates Baseline Observed Delay at NAS Predicted NAS Arrival Delay Percentage Change of Arrival Delay at NAS System

6 LITERATURE REVIEW Beatty et al. developed the concept of a delay multiplier to understand the effect of initial flight delay on an airline s operating schedule. Their research concludes that the existence of a delay multiplier is due to the branching nature of crew and aircraft sequences. Schaefer and Millner show that the delay augments with increasing duration of IMC at the airport, while the propagation effect for the first leg was significant but diminished after each subsequent leg Schaefer et al. and Ahmad Beygi et al. analyzed the importance of schedule parameters on delay propagation in the NAS The model by Laskey et al. takes into consideration the dynamic aspects of flight delay, such as weather effects, wind speed, flight cancellations, and others, to estimate delay propagation in the NAS. Hansen and Zhang devised a macroscopic technique to study the delay propagation in the NAS. 6

7 Causal Factors for Delay at Individual Airport - Average Arrival Deterministic Queuing Delay - Average Observed Arrival delay at other airports - Adverse Weather Daily IMC Ratio Weather Index of different regions - Passenger Load Factor - Demand Management Strategies at different airports - Seasonal Dummy Variables 7

8 Causal Factors for Delay at National Airspace System - Average Arrival Deterministic Queuing Delay - Average Observed Arrival delay at individual airport - Adverse Weather Weather Index of different regions - Demand Management Strategies at different airports - Seasonal Dummy Variables - Total operations of system 8

9 Correlation between Average Observed Arrival and Departure Delay National Airspace System Arrival Departure Arrival Departure ORD Arrival Departure Arrival Departure

10 W k (t) LQ(t) LF(t) D (t) A Interactions between a Single Airport and the Rest of the NAS I (t) L ν(t) D (t ) A Dˆ S ( t ) LQ (t ) SQ(t) LF(t) Average observed arrival delay against schedule at LGA/ORD Average observed arrival delay against schedule at NAS Average arrival deterministic queuing delay at LGA/ORD Average arrival deterministic queuing delay of system Passenger Load Factor S i (t) I (t) L OP (t) Daily IMC ratio recorded at LGA/ORD Total operations (arrivals) of system D j (t) SQ(t) Dˆ S ( t) u(t) OP(t) W k (t) S (t ) i D (t) j ν (t), u(t) Weather index of different regions Seasonal dummy variable Demand Management regime dummy variables Stochastic error terms 10

11 Deterministic Queuing Delay Queuing Diagram of arrivals at ORD 11

12 Demand Management Regimes at ORD and LGA LGA Airport Slot Control AIR-21 Slottery High- Density Rule Apr 2000 Partial HDR Jan 2001 ORD Airport Slot Control Sep 2001 Post 9/11 Period Jan 2002 Over Scheduling Over Scheduling Jan 2004 FAA reduced AA and UA flights by 5 % Jun 2004 FAA reduced AA and UA flights by 2.5 % 12

13 1. LGA Airport The HDR period at LGA was characterized by limiting the hourly slots to 68 between 6:00 am and 12:00 midnight Later in 1986, the scheduling committee was replaced by use-it-orlose-it and buy-sell rules By 1997, 30 new entrant exemptions were approved for LGA In April 2000, a demand management strategy called AIR-21 was introduced to eliminate slot control. The terrorist attacks on September 11, 2001, affected airport operations in many ways. Beginning in 2002, air traffic increased each following year, leading to a period of overscheduling, and HDR completely expired by

14 2. ORD Airport The HDR strategy was also applied at ORD in 1968 In the 1990s, 53 new slot exemptions were created at ORD The HDR strategy was gradually reduced at ORD, and its complete elimination took place by 2002 The terrorist attacks on September 11, 2001, affected airport operations Since 2002, there has been a general increase in air traffic, creating a period of overscheduling, with more than 100 daily operations at ORD The FAA negotiating a 5% reduction in American Airlines (AA) and United Airlines (UA) flights in January The vacated slots were quickly taken up by Northwest Airlines and Independence Air There was further reduction of AA and UA flights in June 2004 by 2.5% to reduce delays 14

15 Epochs Period LGA Airport ORD Airport January 2000-April 2000 High Density Rule Period High Density Rule Period May 2000-January 2001 AIR - 21 Partial HDR February 2001-September 10,2001 Sep 11,2001 December 2001 Slottery Post 9/11 Period Partial HDR Post 9/11 Period Year 2002 Year 2002 Over Scheduling Year 2003 Year 2003 Over Scheduling January 2004 June 2004 Year % Reduction in UA and AA Flights 15

16 Convective Weather (IL (t)) Adverse Weather Conditions - Daily summary of En-Route Weather Information based on the hourly data - Each observation is a binary variable indicating thunderstorms during the day - The whole country divided into regions of 10º latitude by 10º longitude - Proportion of weather stations in each region reporting thunderstorms is computed IMC Ratio (WK (t)) - Proportion of day in which individual airport or the system is under IMC is calculated - Airports have lower capacity under IMC 16

17 USA Weather Regions

18 Causal Factor for Delay at Individual Airport Passenger Load Factor (LF(t)) - It is an ratio of number of passengers by number of seats available in the aircraft at the airport - Higher passenger load factor leads to longer average daily arrival delay since it causes uncertainty to smooth daily operations. Seasonal Dummy Variable (Sj(t)) - Dummy variables are introduced to indicate seasons 18

19 Causal Factor for Delay in the NAS Total Flight Operations (OP(t)) - The total flight operations to all 32 benchmark airports are included - This variable captures effects of traffic volume not reflected in other variables - It also captures the incidence of congestion at the airport Seasonal Dummy Variable (Sj(t)) - Dummy variables are introduced to indicate seasons 19

20 Multivariate Model of LGA/ORD and NAS Delay Model 1 (Delay at LGA or ORD) ) ( ) ( ) ( ) ( 2 ) ( 6 ) ( 5 ) ( 4 ) ( 2 3 ) ( 2 ) ( ˆ 1 ) ( t t j D j jl t i S i il k t k W kl t L I t L I t LF t LQ t LQ t S D t DA ν θ ω λ β β β β β β α = Model 1 (Delay at rest of OEP Airports) ) ( ) ( ) ( ) ( ) ( ) ( ˆ ) ( ) ( t u t D t S t W SQ t t D t OP t D n n ns m m ms l l ls A S S = + θ ω λ γ γ γ α 20

21 Two Stage Least Square Estimates for LGA/ORD LQ(t) LF(t) I (t) L D (t) A Dˆ ( t) S Average observed arrival delay against schedule at LGA/ORD Average observed arrival delay against schedule at NAS Pred_ D ˆ ( t) S Predicted average arrival delay against schedule at NAS W k (t) D (t) A ν(t) LQ(t) SQ(t) Average arrival deterministic queuing delay at LGA/ORD Average arrival deterministic queuing delay of system LF(t) Passenger Load Factor S i (t) D j (t) SQ(t) Dˆ S ( t) Pred_ Dˆ ( t) S u(t) OP(t) I (t) L OP(t) W k (t) S (t ) i D (t) j ν (t), u(t) Daily IMC ratio recorded at LGA/ORD Total operations (arrivals) of system Weather index of different regions Seasonal dummy variable Demand Management regime dummy variables Stochastic error terms 21

22 Two Stage Least Square Estimates for NAS W k (t) LQ(t) LF(t) D (t) A Pred_ D (t ) A I (t) L ν(t) D (t) A Dˆ ( t) S Pred_ D LQ(t) SQ(t) LF(t) (t ) A Average observed arrival delay against schedule at LGA/ORD Average observed arrival delay against schedule at NAS Predicted average arrival delay against schedule at LGA/ORD Average arrival deterministic queuing delay at LGA/ORD Average arrival deterministic queuing delay of system Passenger Load Factor S i (t) D j (t) Dˆ S ( t) u(t) I (t) L OP(t) W k (t) S (t ) i D (t) j Daily IMC ratio recorded at LGA/ORD Total operations (arrivals) of system Weather index of different regions Seasonal dummy variable Demand Management regime dummy variables SQ(t) OP(t) ν(t), u(t) Stochastic error terms 22

23 Regression Results of Arrival Delay at LGA and ORD LGA ORD Variable Estimate SE P-Value Estimate SE P-Value LQ(t) LQ 2 ( t ) Dˆ ( t) S I (t) L I (t) 2 L LF(t) W k (t) Average Queuing Delay < < Quadratic Average Queuing Delay < < Observed Arrival Delay for NAS < < IMC Ratio (Proportion of days operated in IMC conditions) < < Square of IMC Ratio Passenger Load Factor Thunderstorm Ratio (number of stations reported thunderstorm/ total amount of stations in region) Region < Region < Region <

24 LGA Variable Estimate SE P-Value ORD Estimate SE P-Value S (t) i D (t) j Seasonal Dummy Variables Dummy Variables for Demand Management Regimes Quarter < Quarter < Quarter < AIR < Slottery Partial HDR Post 9/11 Period < Year Year Year < Over Scheduling % Reduction in UA and AA < R-Square

25 Decomposition of LGA Delay by Causes 25

26 Decomposition of ORD Delay by Causes 26

27 Estimation Results of NAS Delay LGA Variable Estimate SE P-Value ORD Estimate SE P-Value SQ(t) ˆ D A ( t) OP(t) (t) W l Average Queuing Delay < < Observed Arrival Delay at LGA/ORD < < Total Operations (arrivals) in the system Thunderstorm Ratio < < Region < < Region < < Region < < Region < Region < <

28 LGA Variable Estimate SE P-Value ORD Estimate SE P-Value S m (t) Seasonal Dummy Variables Quarter Quarter < < Quarter < < D n (t) Dummy Variables for Demand Management Regimes AIR Slottery Partial HDR Post 9/11 Period Year Year Year Over Scheduling % Reduction in UA and AA R-Square

29 Decomposition of NAS Delay considering LGA by Causes 29

30 Decomposition of NAS Delay considering ORD by Causes 30

31 SYSTEM-WIDE BENEFIT OF CAPACITY EXPNSION OF INDIVIDUAL AIRPORT Capacity LGA/ORD AAR Queuing Delay LGA/ORD Convective weather IMC Regression Estimates Baseline Observed Delay at LGA/ORD Predicted Arrival Delay at LGA/ORD Single Airport Percentage Change of Arrival Delay at LGA/ORD Total Operations (Arrivals) at NAS Convective Weather Regression Estimates Baseline Observed Delay at NAS Predicted NAS Arrival Delay Percentage Change of Arrival Delay at NAS System 31

32 Single Airport Arrival Delay System Arrival Delay Baseline Queuing Delay Weather IFR ratio 10% capacity increase (AAR) Queuing Delay Weather IFR ratio Baseline Airport Observed Delay Weather Total Operations After increase in airport capacity (S2 ) Airport Observed Delay with Increased Capacity Weather Others Others Total Operations Percent Reduction in Airport Delay = O1 O2 *100 O1 Percent Reduction in NAS Delay = S1 S2 *100 S1 32

33 LGA Scenario Analysis Baseline 10% Capacity Increase 20% Capacity Increase 30% Capacity Increase (1) (2) (3) (4) LGA Delay (mins) Percent Reduction at LGA Base 1.83 % 4.93 % 7.90 % NAS Delay (mins) Percent Reduction at NAS Base 1.36 % 2.34 % 2.29 % 33

34 ORD Scenario Analysis Baseline 10% Capacity Increase 20% Capacity Increase 30% Capacity Increase (1) (2) (3) (4) ORD Delay (mins) Percent Reduction at ORD Base % % % NAS Delay (mins) Percent Reduction at NAS Base 4.40 % 6.02 % 6.67 % 34

35 Conclusion One-minute delay at other airports cause increase of minute and minute delays at LGA and ORD, respectively Region 11 (Northeastern part of the U.S.) is a major problem for LGA; similarly, Regions 12 and 13, (Uppermiddle regions of the U.S.) are problems for ORD. The summer seasonal effect shows the least amount of delay when compared to other seasons The lowest delay was reached post-9/11 Delay in the system and passenger load are the major factors affecting average arrival delay at LGA. 35

36 Average arrival queuing delay and delay in the system are the major contributing factors for the average arrival delay at ORD Capacity improvements at ORD show more system-wide impact than capacity improvements at LGA. A one-minute increase of delay at LGA causes a minute increase in average delay in the NAS, while a one-minute delay at ORD causes a minute average delay in the NAS 36

37 Future Work Refine the specification of the multivariate models Using more current data From single airport to regional airport system Build a multiple-equation model and apply 3SLS to estimate the parameters 37

38 Three Stage Least Squares Combination of Two Stage Least Squares (2SLS) method with Seemingly Unrelated Regression (SUR) SUR is a technique to analyze multiple equations with - Cross Equations Parameter Restrictions - Correlated Error Terms In the first step of 2SLS each endogenous covariate is regressed on all exogenous variable in the model. In the second stage of 2SLS the predicted value obtained from the first stage is regressed on the exogenous variables in the model Ultimately SUR is used for correlated error terms in the model. 38

39 Interactions between a Single Airport and rest of the NAS 39

40 Questions? Comments? Thanks! Nagesh Nayak Ph.D. Student, Civil and Environmental Engineering University of South Florida (USF) 4202 E. Fowler Ave. ENC 3300 Tampa, FL Tel: Fax: URL: 40