1

NEXTOR

Normalization of Airport and Terminal Area Normalization of Airport and Terminal Area Operational Performance: A Case Study of Operational Performance: A Case Study of

Los Angeles International AirportLos Angeles International Airport

Mark Hansen and Tanja BolicNational Center of Excellence for Aviation Research

University of California at Berkeley

Fourth International Air Traffic Management R&D SeminarDecember 2001

2

NEXTOROutlineOutline

Background and MotivationMethodsResultsApplicationConclusions

3

NEXTORNormalizationNormalization

Analyze trends in aviation infrastructure performanceDetermine effects of deployments of new technology or infrastructureQuantify effects within and outside FAA control

4

NEXTORNEXTOR Normalization WorkNEXTOR Normalization Work

Sponsored by FAA Free Flight OfficeFocus on Delays and Time-in-System MetricsAnalysis at Daily Level

5

NEXTORNormalization and the Normalization and the

Modeling CycleModeling Cycle

Models

R&D and Deployment Decisions Introduction

of New Systems

Benefits and Impacts of New Systems

Normalization

6

NEXTORConceptual FrameworkConceptual Framework

Demand

Weather

Conditionsat otherAirports

Time at Origin

AirborneTime

Taxi-In Time

Total Flight Time

System Performance

•En Route

•Terminal Area

•ATM

•Aircraft

7

NEXTORDaily Flight Time Index (DFTI)Daily Flight Time Index (DFTI)

Daily weighted average of flight times to a given airport from a set of originsAnalogous to a Consumer Price IndexOrigins in “market basket” have at least one completed flight in each day of sampleWeights reflect origin share of flights to study airport over study period

8

NEXTORFlight Time and Its Flight Time and Its

ComponentsComponents

Origin Gate

Origin RWT

Dest. GateDest. RWT

Scheduled Departure Time

Actual Departure Time

Actual Arrival Time

Wheels-Off Time

Wheels-On Time

9

NEXTORDFTI Time Series for LAXDFTI Time Series for LAX

100

120

140

160

180

200

220

J-95 J-96 J-97 J-98 J-99 J-00 J-01

Date

DFT

I (m

inut

es)

10

NEXTOR3030--Day Moving AverageDay Moving Average

100

120

140

160

180

200

220

Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01

Date

DFT

I(min

utes

)

WinterSpringSummerFall

11

NEXTOR77--Day Moving Average with ComponentsDay Moving Average with Components

0

20

40

60

80

100

120

140

160

180

200

Jan-95 Jul-95 Jan-96 Jul-96 Jan-97 Jul-97 Jan-98 Jul-98 Jan-99 Jul-99 Jan-00

Date

Ave

rage

Tim

e pe

r Flig

ht

(min

utes

)

Taxi-InAirborneAt Origin

12

NEXTORVariance DecompositionVariance Decomposition

)],(),(),([2

)()()()(

INTAXIORIGINCOVINTAXIAIRBORNECOVAIRBORNEORIGINCOV

INTAXIVARAIRBORNEVARORIGINVARDFTIVARINTAXIAIRBORNEORIGINDFTI

−+−+⋅

+−++=−++=

13

NEXTORVariance DecompositionVariance Decomposition

-20

0

20

40

60

80

100

120

140

1995 1996 1997 1998 1999 2000

Year

Varia

nce

Com

pone

nt (m

in2 )

COV(Airborne,Taxi-In)COV(At Origin,Taxi-In)COV(At Origin,Airborne)VAR(Taxi-In)VAR(Airborne)VAR(At Origin)

14

NEXTORWeather NormalizationWeather Normalization

Based on CODAS hourly weather observations for LAXFactor analysis of weather data

Create small number of factors that capture variation in large number of variablesFactors are linear combinations of original variablesFactors correspond to principal axes of N-dimensional data elipse

15

NEXTORFactor Analysis with Two Factor Analysis with Two

VariablesVariables

X1

X2

First Factor

16

NEXTOR99--Factor Representation of Factor Representation of

LAX Daily WeatherLAX Daily WeatherFactor Interpretation

1 Warm temperatures throughout day. 2 VFR operations and absence of low cloud ceiling in the

morning. 3 VFR operations and absence of low cloud ceiling in the

afternoon. 4 High visibility throughout day. 5 Medium cloud ceiling throughout day. 6 High winds throughout day. 7 High ceiling cloud ceiling throughout day; evening

precipitation. 8 Precipitation in late morning and afternoon. 9 Precipitation in early morning.

17

NEXTORDemand NormalizationDemand Normalization

Deterministic Queuing AnalysisArrival Curve from Official Airline GuideDeparture Curves and Average Delays Calculated Assuming Range of Hypothetical CapacitiesFactor Analysis Applied to Obtain Reduced Set of Demand Factors

18

NEXTOR

0

200

400

600

800

1000

1200

0:00 3:00 6:00 9:00 12:00 15:00 18:00 21:00 0:00 3:00 6:00 9:00 12:00Time of Day

Cum

ulat

ive

Num

ber

ScheduledCompleted, Capacity=40Completed, Capacity=60Completed, Capacity=80

Queuing DiagramsQueuing Diagrams

19

NEXTORTrends in Values of HDD Parameters Trends in Values of HDD Parameters

and Scheduled Arrivals since 1997and Scheduled Arrivals since 1997

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

1/1/97 7/3/97 1/2/98 7/4/98 1/3/99 7/5/99 1/4/00

Date

HDD

(min

)

OAG Flights

HDD50

HDD70

HDD90

20

NEXTOR

0

200

400

600

800

1000

1200

0:00 3:00 6:00 9:00 12:00 15:00 18:00 21:00 0:00Time of Day

Cum

ula

tive

Nu

mbe

r

6/24/971/17/996/29/99

DAY HDD50 HDD60 HDD70 HDD80 HDD90 HDD100 HDD110 HDD1206/24/97 124.55 44.62 6.86 2.64 1.07 0.40 0.16 0.091/17/99 52.88 6.96 0.87 0.06 0.00 0.00 0.00 0.006/29/99 111.90 28.71 3.11 0.85 0.29 0.11 0.04 0.00

21

NEXTORFactor Analysis of HDD VariablesFactor Analysis of HDD Variables

22

NEXTORNormalization for Conditions Normalization for Conditions

at other Airportsat other AirportsConsider airports included in DFTI averageFor each compute daily average departure delay for flights notbound to LAX regionAverage airport departure delays using DFTI weights

23

NEXTOROrigin Airport Delay Time SeriesOrigin Airport Delay Time Series

0

5

10

15

20

25

30

35

40

45

50

Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01

Date

Ori

gin

Airp

ort D

elay

(min

)

24

NEXTORPerformance ModelsPerformance Models

ttttt ODELDMDWXfY ε+= ),,(Where:Yt is DFTI or DFTI component for day t;WXt is vector of weather factors for day t;DMDt is vector of demand factors for day t;ODELt is average origin departure delay for day t;εt is stochastic error term.

25

NEXTORFunctional Forms ConsideredFunctional Forms Considered

ParametricLinear (with 3, 6, 9, and 12 weather factors)Quadratic response surfaceNon-linear

Non-parametric9 clusters based on 3 weather factors12 clusters based on 9 weather factors

26

NEXTORLinear Model Estimation Results Linear Model Estimation Results

Variable Description Estimate T - statistic P - value

INTERCEPT Intercept 138.055 567.065 0.0001OAC Origin airport congestion 1.128 44.351 0.0001WX1 Warm daily temperatures -1.357 -12.101 0.0001WX2 VFR ops, no low cloud ceiling in the morning -0.988 -7.116 0.0001WX3 VFR ops, no low cloud ceiling in the afternoon -1.123 -7.583 0.0001WX4 High visibility throughout day -0.449 -3.575 0.0004WX5 Medium cloud ceiling throughout day 1.440 10.555 0.0001WX6 High winds throughout the day 0.512 4.531 0.0001WX7 High cloud ceiling throughout day 0.911 4.172 0.0001WX8 Precipitation in late morning and afternoon 1.871 8.324 0.0001WX9 Precipitation in early morning -0.379 -2.614 0.0091DMD1 Peak demand 0.075 0.725 0.4685DMD2 Base demand 0.440 4.574 0.0001ADJUSTED R2 0.743

27

NEXTORPredictedPredicted vsvs Actual ValuesActual Values

130.00140.00150.00160.00170.00180.00190.00200.00210.00

130.00 140.00 150.00 160.00 170.00 180.00 190.00 200.00

Observed Values (min)

Pre

dict

ed V

alu

es (m

in)

28

NEXTOROutliersOutliers

Used TMU logs to investigate days for which predictions have large errorsReasons for higher than predicted DFTI

East flowRadar outagesAir Force OneOver-stringent ground delay program

29

NEXTORModels for DFTI ComponentsModels for DFTI Components

30

NEXTORResponse Surface ModelResponse Surface Model

High demand increases the importance of visibilityEffects of precipitation and winds re-enforcingDiminishing marginal impact of winds

31

NEXTORImpacts of a Decision Support Impacts of a Decision Support

Tool at LAXTool at LAXFFP1 BackgroundThe ToolDeployment ExperienceImpacts

32

NEXTORFree Flight Phase IFree Flight Phase I

Deploy terminal area/en route decision support tools (CTAS, SMA, and URET) at selected sitesNormalization useful for assessing operational impacts and benefits

33

NEXTORFinal Approach Spacing ToolFinal Approach Spacing Tool

Decision support tool for TRACON P(assive)FAST advises on runway assignment and landing sequenceActive FAST provides speed and turn advisoriesAdvisories incorporated into ARTS displayPrior PFAST implementation at DFW

34

NEXTORFAST at LAXFAST at LAX

“Passive Passive FAST” (P2FAST) or “T-TMA”No advisoriesSeparate displays depict traffic up to 300 nm out using combination of HOST and ARTS dataA situation-awareness tool instead of a decision automation tool

35

NEXTORWhy TTMA for LAX?Why TTMA for LAX?

Significant “internal” operationsDepartures from within SOCAL TRACON and ZLA CenterNo acceptable “work-arounds”

Initial deployment until DS software can be adapted

36

NEXTOR

100

120

140

160

180

200

220

240

1/1/

00

2/1/

00

3/1/

00

4/1/

00

5/1/

00

6/1/

00

7/1/

00

8/1/

00

9/1/

00

10/1

/00

11/1

/00

12/1

/00

1/1/

01

2/1/

01

3/1/

01

4/1/

01

5/1/

01

6/1/

01

Date

DFTI

(in

min

utes

)

DFTI Before and After DFTI Before and After ImplementationImplementation

37

NEXTOR Variable Parameter Estimates DFTI At Origin Airborne Taxi-in intercept 139.29 15.23 115.9 8.11TTMA -1.99 -1.71 -0.19 -0.07OAC 1.39 1.29 0.03 0.06Peak Demand -0.35 -0.1 -0.24 -0.01Base Demand 0.97 0.74 0.04 0.19Weather Factor1 -3.37 -0.89 -2.63 0.14Weather Factor2 -2.66 -1.8 -0.75 -0.12Weather Factor3 -1.88 -1.36 -0.48 -0.04Weather Factor4 0.19 -0.24 0.49 -0.07Weather Factor5 1.48 0.73 0.79 -0.03Weather Factor6 0.46 0.12 0.31 0.02Weather Factor7 0.64 0.27 0.29 0.81 Adjusted R-Square 0.79 0.83 0.55 0.39

Significant at 5% level Significant at 10% level

TTMA Normalization ResultsTTMA Normalization Results

38

NEXTORCollaborating Evidence Collaborating Evidence

Evidence of increased throughput rates when system is under stressControllers love it

Anticipate overloads and slow planes down to avoid holdingBetter runway balancing

39

NEXTORConclusionsConclusions

70-80% DFTI variation “explained” statistically (in case of LAX)

Up-line delay is largest driverVariety of weather impactsModest gain from more complicated models

Normalization shows benefit from TTMA Implementation