m11111111E11EEI flllll fllll · Issue 2: Assessing Sensitivity to-Noise Model Thrust Assumptions The methodology for determining the IN's sensitivity to thrust involves using the

AD-AO8 426 MITRE CORP 0CLEAN VA METREK DIV F/6 1/3FAA INTEGRATED NOISE MODEL VALIDATION. PHASE 1. ANALYSIS OF INT-ETC(U)DC 79 R G GADS J M ALORED DOT-FASOVA-4370

UNCLASSIFIED NTR79WO00J95 FAA/EE-80-O-4 N

!llmmommommomIIflllll fllllm11111111E11EEI

*ulllllllllllum

m!lEEE//EE///hI

1=0 1111 .1.8

* IL25 .6

MICROCOPY RESOLUTION TEST CHAOTNATIONAL BUREAU Ol' STANDARDS-19634

Report FAA-EE-80-04 1 '.

FAA Integrated Noise Model Validation00 Phase 1: Analysis of Integrated Noise Model

Calculations for Air Carrier Flyovers

DR. RG. GADOSJ.M. ALDRED T C

cm ~MAR 3 1980DECEMBER 1979

Prepared For: F Apo~d~~U.S. DEPARTMENT OF TRANSPORTATION

Q. FEDERAL AVIATION ADMINISTRATION

Office of Envionment and EnergyL-W Washington, D.C. 20591

2 15.

Technical Report Documentation Page

. 2. Government Accession No. 3. Recipient's Catalog No.

f Tase 1. Analysis of Integrated Noise Model Calculation.:_for Air Carrier Flyovers_______________

8. Performing Organization Report No.

R. G. Gados 1 J. M. Aldred MTRS .' ar -g FganizMto a nd Address

The MITRE CorporationMetrek Division-'1820 Dolley Madison Boulevard (5 -A -47McLean, VA 22102 riod C15....Fd

12. Sponsoring Agency Name and Address

Federal Aviation AdministrationOffice of Environment and Energy1800 Independence Ave., S.W. 14. pnoigAec CodeWashington, D.C. 20591 __DOT/FAA15. Supplementary Notes

16.

e Federal Aviation Administration's Integrated Noise Model is aset of computer programs which is used to predict the noiseimpact of aircraft in the vicinity of an airport. Through useof extensive statistical analyses, this study investigates the

accuracy and suitability of the noise model in calculatingaircraft noise exposure by: (1T examining the agreement betweenthe noise model in calculating single noise events and theactual measurement of those events, (2-Y assessing the sensitivityand controllability of the noise model to aircraft thrust assump-tions, and (3-rYlrvestigating noise curves used in calculatingnoise exposure by testing variables for significance in estimatingnoise and by comparing the shape of empirical noise curves withthose already in the noise model. Data for the analysis wereobtained from field observations of noise from air carrier flightoperations over various noise monitoring sites near WashingtonNational and Dulles International Airports.

\U

17. Key Words 18. Distribution Statement

noise event relationship This document is available to thenoise monitoring public through the National Technicalnoise-thrust mapping Information Service, Springfield, Virginiastatistical comparison 22161

19. Security Clossif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price

Unclassified Unclassified 101

Form DOT F 1700.7 (8-72) Reproduction of completed page authorized

0 9 0

ACKNOWLEDGEMENT

I extend my appreciation to Monty Montgomery of the FAA NoiseMonitoring Facility for his cooperation, support, and assistancein noise data acquisition.

-J Cj j

EXECUTIVE SUMMARY

INTRODUCTION AND BACKGROUND

The Department of Transportation/Federal Aviation Administration(FAA) Integrated Noise Model (INM) is a set of computer programswhich are used to predict the noise impact of aircraft in theneighborhood of an airport. The purpose of the INM validationproject is to determine the accuracy of the FAA Integrated Noise

Model (Version I) by comparing INM noise exposure calculationswith actual measured noise exposure levels. The current phaseof the project centers upon statistical analyses of singleevents in which calculated and observed noise exposure levels

from air carrier flight operations are compared. This compari-son uses data from various noise monitoring sites around Wash-ington National and Dulles International Ai-ports.

The methodology employed in this analysis is a refinement ofthat initially presented in MTR-7913, "Analysis of IntegratedNoise Model Calculations for Concorde Flyovers" (Reference I).In that paper, statistical techniques were presented to quantifythe noise characteristics of Concorde operations at Dulles

International Airport. The same basic methodology together withcertain extensions are now applied to representatives of thefollowing types of aircraft. two and three engine narrow bodyjets, four engine narrow body jets, and wile body jets.

ISSUES IN THIS STUDY

Issue 1: Determining Agreement Between Calculations andMeasurements

In this validation study, INM calculations for various aircraftare analyzed separately. Statistical methods are used to quan-titatively check the agreement between calculations and measure-ments of noise exposure by using paired differences between theobserved noise and the noise model calculation for the sameflight condition. The paired difference is formed by using themeasured noise exposure from a single aircraft flyover andcomparing it with the calculated noise exposure resulting fromusing the measured slant-range distance, altitude, and velocityof the aircraft at its closest point of approach to the monitorsite. The resulting statistic portrays the average differencein noise exposure from aircraft flyovers abeam a monitor siteand the INM noise calculations for the same simulated flyovers.

ii

Issue 2: Assessing Sensitivity to-Noise Model Thrust Assumptions

The methodology for determining the IN's sensitivity to thrustinvolves using the noise-thrust mapping procedure developel inthis study and the computed difference between the observednoise and the INM calculation from the aircraft's flight charac-teristic. The noise-thrust mapping procedure allows the trans-lation of measured noise into equivalent thrust values. Thistranslation is based on the assumptions that the noise model isfunctionally correct and that the difference between measuredand calculated noise is primarily attributable to the assumedthrust profiles. This procedure enables the assessment of theinnate controllability (amount of change in INM calculated noisevalues corresponding to the allowable range of thrust values fora particular aircraft type) of the thrust profiles in the INMdata base to make calibrated adjustments.

Issue 3: Investigating Empirical Models for Noise Exposure

The third aspect of the comparison between the observed noiseand INM calculations involves looking at the observed noisevalues to answer three questions: (1) Which of the observedvariables associated with the noise event (measured noise forthe aircraft flyover) is most highly correlated with that noiseevent? (?) What is the mathematical form of the variable usedto describe the noise event relationship? (3) Using the "best"mathematical description of the noise event relationship, doesthe mathematical description agree with the noise curves used inIN! calculations?

RESULTS AND CONCLUSION

Over 6000 single-event noise measurements (measured noise for anaircraft flyover abeam a monitor site) were taken and pairedwith calculations from the INM for statistical comparison. Theobservations were categorized by aircraft type, monitor site,and type of flight operation, i.e., departure or arrival. Theevents for each airport, for each aircraft type, and each flightoperation were combined for grouped statistics. The aircrafttypes were then arranged in three groups: four engine narrowbody aircraft, two/three engine narrow body aircraft, and widebody aircraft.

The results of three separate analyses of aircraft departures,and arrivals their noise measurements, and their comparison withanalogous noise calculations by the FAA Integrated Noise Modelare summarized in Tables I and 1. Based on these results as

iii

00

~~ 0-T O 14 -4

ul w H -

LH > 3:

04 -4 L-

1-4. HO

0 000

wP. CA-4 NO -4 CIL) H HOH nC 0 1 HOHw

E-~ 0 0 a 0 0 0 0 0 0 0 z In

z Ln u I I H0~ H C

C-

0n 10f

I "I 4 H.

0 ON 0000 .~ 04

H 0 . r. r- r- 00 0

H)

14Hz0

u0H 0

L5~~ H O rI

0 3

c c)4C14

0 0

E-. pi 04 >

0 E

uAO E-E- o'..

00 C4 %N

zr.

tH HEn~ U 1- enCi n i N L n n p

C% cl 01 In Ha

6 -l n A .

MOO

04 Hzimd 0m" VP o4,8

well as others presented in this study, the following conclu-sions have been reached concerning the INM's performance inmodeling air carrier operations at Dulles International andWashington National Airports.

1. The criterion for "non-agreement" is defined as theaverage of the paired differences (between actual observednoise from field measurements and calculations of analogoussingle noise events from the INM) being greater than threedecibels (3 dB'). The 3 dB criterion was selected as acompromise between setting a narrow margin for agreement,but still allowing a large enough margin to account for thewide range of measured noise values resulting from fieldohservations of uncontrolled aircraft flight operations.Using this criterion, noise calculations derived from theuse of the INM do not agree with actual observed noisevalues for four engine narrow body aircraft for departureoperations, nor do they agree for most of the wide bodyaircraft types for departure operations. Because of thewide range of variability in measured values for arrivaloperations, the interpretation of the arrival data isdifficult. (The confidence intervals for the averageobserved noise differences as well as the confidence bandaround the regression estimates are large.) Again, usingthe three decible criterion for only the observed differ-ences, noise calculations derived from the use of the INMdo not agree with actual observed noise values for widebody aircraft for arrival operations.

9. I~N calculations of noise events can he changed orcalibrated most easily by using two methods: adjustingthrust profiles, and adjusting noise curves. The method ofadjusting thrust profiles alone does not have the inherentrange (controllability) to allow the calibration necessaryto make the results of the INN calculations for four enginenarrow body aircraft, for example, comparable to actualmeasurements (i.e., even when the INM thrust is set formaximum takeoff thrust, the average observed noise is stillgreater than the resulting INN calculation. This situationis unreasonable since the maximum takeoff thrust is alimiting value, and when set at this maximum thrust value,an aircraft would theoretically produce the loudest pos-sible noise value for that particular aircraft.) In orderfor the INM calculations to agree with the observed noise,the noise curves in the INM must be adjusted to reflect theactual measurements made in the field by redefining newnoise curves.

viEl

RECOMMENDATIONS

The objective of the FAA Integrated Noise Model is to calculatethe noise from aircraft operations in the vicinity of an airport(for an average day of the year in an operational environment).

The Noise Exposure Levels (NEWs used in the INM data base werederived mathematically from maximum sound level measurementswith duration corrections obtained from Effective PerceivedNoise Level (EPNL) measurements, and thus the NELs in the database are in part theoretical. From these theoretical values,the noise versus distance curves now residing in the data basewere obtained.

On the other hand, the empirical noise versus distance curvespresented in this report are a reflection of an actual day ofthe year in an actual operational environment. These curveswere derived from a cross-section sampling of aircraft opera-tions for eight months of data acquisition. The observationsare consistent and present a good means to satisfy the statedobjective through fine-tuning the INM.

The following steps are recommended in order to improve theaccuracy of the INM for aircraft types whose observed noisevalues do not agree with analogous INM calculations:

1. Adjust the noise curves in the INM for agreement byusing empirical noise curves resulting from regressionanalyses of observed noise values. The noise curves shouldbe adjusted to improve accuracy of noise calculations fortakeoff and climb flight operations only, since the actualthrust values used for these operations are procedurallyset to a relatively known and fixed value.

2. After adjustment of the noise curves, the noise-thrustmapping procedure described in this study should be used tofine tune or calibrate the thrust profiles for arrivaloperations. Certain assumptions will have to be madeconcerning what actual thrust value is being used abeam thevarious sites, as well as assumptions concerning the flightconfiguration (i.e., flap and gear extension). Theseassumptions are an integral part of the calibration process.

3. To insu;re that the calibration procedure is correct, acomplete set of noise observations should be taken at twoother airports and the statistical comparisons of observednoise versus INM calculations be repeated.

viii

.1

3. The results of the noise measurement comparisons basedon data taken at Dulles International Airport are supportedby those taken at Washington National Airport for two andthree engine narrow body aircraft. (Four engine narrowbody aircraft and wide body aircraft operations were ob-served only at Dulles Airport.)

vii

TABLE OF CONTENTS

1. INTRODUCTION 1-1

1.1 Background I-I1.2 Areas of Investigation 1-1

1.2.1 Determining Agreement Between Calculationsand Measurements 1-2

1.2.2 Assessing Sensitivity to Thrust Assumptions 1-21.2.3 Investigating Empirical Models for Noise

Exposure 1-2

1.3 Variation in Noise Observations 1-31.4 FAA Integrated Noise Model Computer Program 1-8

2. METHODOLOGY 2-1

2.1 Determining Agreement Between Calculations andMeasurements 2-1

2.1.1 Data Collection 2-22.1.2 INM Noise Exposure Calculation 2-62.1.3 Statistical Comparison 2-7

2.2 Assessing Sensitivity to Thrust Assumptions 2-10

2.2.1 Noise-Thrust Mapping Procedure 2-i

2.3 Investigating Empirical Models for Noise Exposure 2-11

2.3.1 Variables Tested for Significance in EstimatingNoise Values 2-16

2.3.2 Stepwise Regresssion Procedure 2-16?.3.3 Confidence Regions for the Empirical Noise Curves 2-162.3.4 Comparison of Empirical Noise Curves with the INK

Noise Curves 2-16

3. RESULTS 3-1

3.1 Four Engine Narrow Body Aircraft 3-2

3.1.1 Departures (Four Engine Narrow Body Aircraft) 3-33.1.2 Arrivals (Four Engine Narrow Body Aircraft) 3-8

3.2 Two/Three Engine Narrow Body Aircraft 3-8

ix

TABLE OF CONTENTS

(Continued)

Page

3.2.1 Departures (Two/Three Engine Narrow Body Aircraft) 3-83.2.2 Arrivals (Two/Three Engine Narrow Body Aircraft) 3-16

3.3 Three/Four Engine Wide Body Aircraft 3-21

3.3.1 Departures (Wide Body Aircraft) 3-213.3.2 Arrivals (Wide Body Aircraft) 3-21

3.4 Other Issues 3-27

3.4.1 Regression Model for Arrival Operations 3-273.4.? Comparison of Dul les and National Data 3-31

4. CONCLUSIONS 4-i

5. RECOMMENDATIONS 5-I

APPENDIX A: INM NOISE EXPOSURE CALCULATION A-i

APPENDIX B: STATISTICAL TOPICS FOR REGRESSION ANALYSIS B-i

APPENDIX C: EMPIRICAL NOISE CURVES FROM REGRESSION ANALYSIS C-i

APPENDIX D: REFERENCES D-1

APPENDIX E: DISTRIBUTION LIST E-1

x

LIST OF ILLUSTRATIONS

TABLE i-1: SOURCES OF VARIABILITY IN OBSERVATIONAND MODELING 1-4

TABLE 2-1: MONITOR SITES USED IN INN VALIDATION 2-4

TABLE 3-I: VARIABLES SELECTED FOR USE IN REGRESSIONMODELS FOR OBSERVED NOISE (DULLES) 3-30

TABLE 4-1: COMPARISON OF CONTROLLABILITY AND AGREEMENTFOR AIRCRAFT DEPARTURES (NEL OBSERVATIONS) 4-2

TABLE 4-2: COMPARISON OF AGREEMENT FOR AIRCRAFT ARRIVALS(NEL OBSERVATIONS) 4-3

TABLE A-I: AIRCRAFT DEFINITIONS STORED IN THE INM A-4

TABLE B-1: VARIABLES TESTED FOR SIGNIFICANCE B-2

TABLE B-2: STEPWISE CALCULATIONS FOR EMPIRICAL NELCURVES - 707-320 DEPARTURES AT DULLES AIRPORT B-4

TABLE C-i: THE EQUATIONS OF THE EMPIRICAL NEL CURVES,FOUR ENGINE, NARROW BODY DEPARTURES (DULLES) C-?

TABLE C-2: THE EQUATIONS OF THE EMPIRICAL NEL CURVESFOUR ENGINE NARROW BODY ARRIVALS (DULLES) C-2

TABLE C-3: TWO/THREE ENGINE NARROW BODY DEPARTURES(DULLES AND NATIONAL) C-3

TABLE C-4: TWO/THREE ENGINE NARROW BODY ARRIVALS(DULLES AND NATIONAL) C-4

TABLE C-5: EQUATIONS OF EMPIRICAL NOISE CURVES FORWIDE BODY DEPARTURES (DULLES) C-5

TABLE C-6: EQUATIONS OF EMPIRICAL NOISE CURVES FOR WIDEBODY ARRIVALS (DULLES) C-5

FIGURE 1-1: REPRESENTATIVE NOISE-TIME HISTORIES FORDEPARTURES OVER OLD TOWN (DCA) 1-6

FIGURE 1-2: REPRESENTATIVE NOISE-TIME HISTORIES FORARRIVALS OVER OLD TOWN (DCA) 1-7

xi

LIST OF ILLUSTRATIONS (CONT'D)

Page

FIGURE 1-3: DULLES INTERNATIONAL AIRPORT'S MONITOR SITESWITH TYPICAL SOUTH OPERATIONS 1-9

FIGURE 1-4: DULLES INTERNATIONAL AIRPORT'S MONITOR SITESWITH TYPICAL NORTH OPERATIONS 1-10

FIGURE 1-5: WASHINGTON NATIONAL AIRPORT'S MONITOR SITESWITH TYPICAL NORTH OPERATIONS 1-11

FIGURE 1-6: WASHINGTON NATIONAL AIRPORT'S MONITOR SITESWITH TYPICAL SOUTH OPERTIONS 1-12

FIGURE 2-1: VALIDATION METHODOLOGY 2-3

FIGURE 2-2: DC-8-60 DEPARTURES OVER CHANTILLY (IAD) 2-8

FIGURE 2-3: DC-9 ARRIVALS OVER FT. FOOTE (DCA) 2-9

FIGURE 2-4: THRUST SENSITIVITY METHODOLOGY 2-12

FIGURE 2-5: EXAMPLE OF CALCULATING INM THRUST REQUIREDTO AGREE WITH OBSERVED NOISE FOR 727 2-13

FIGURE 2-6: METHODOLOGY OF EMPIRICAL AND INN NOISECURVE COMPARISON 2-15

FIGURE 3-1: DIFFERENCES BETWEEN OBSERVED NOISE AND INKCALCULATIONS FOR FOUR ENGINE, NARROW BODYDEPARTURES 3-4

FIGURE 3-2: CALCULATED THRUST FOR FOUR ENGINE, NARROW BODYDEPARTURES 3-5

FIGURE 3-3: EMPIRICAL AND INN NOISE CURVES FOR FOURENGINE, NARROW BODY DEPARTURES 3-7

FIGURE 3-4: DIFFERENCES BETWEEN OBSERVED NOISE ANDINM CALCULATIONS FOR FOUR ENGINE, NARROWBODY ARRIVALS 3-9

FIGURE 3-5: CALCULATED THRUST FOR FOUR ENGINE, NARROWBODY ARRIVALS 3-10

xii

LIST OF ILLUSTRATIONS (CONT'D)

Page

FIGURE J-6: EMPIRICAL AND INM NOISE CURVES FOR FOUR ENGINE,NARROW BODY ARRIVALS 3-11

FIGURE 3-7: 95% CONFIDENCE INTERVALS FOR AVERAGE CPAVERTICAL VELOCITY 3-13

FIGURE 3-8: DIFFERENCES BETWEEN OBSERVED NOISE AND INMCALCULATIONS FOR TWO/THREE ENGINE, NARROWBODY DEPARTURES 3-14

FIGURE 3-9: CALCULATED THRUST FOR TWO/THREE ENGINE, NARROWBODY DEPARTURES 3-15

FIGURE 3-10: EMPIRICAL AND INM NOISE CURVES FOR TWO/THREEENGINE, NARROW BODY DEPARTURES 3-17

FIGURE 3-11: DIFFERENCES BETWEEN OBSERVED NOISE AND INKCALCULATIONS FOR TWO/THREE ENGINE, NARROWBODY ARRIVALS 3-19

FIGURE 3-12: CALCULATED THRUST FOR TWO/THREE ENGINE,NARROW BODY ARRIVALS 3-20

FIGURE 3-13: EMPIRICAL AND INM NOISE CURVES FOR TWO/THREEENGINE, NARROW BODY ARRIVALS 3-22

FIGURE 3-14: DIFFERENCES BETWEEN OBSERVED NOISE AND INKCALCULATIONS FOR WIDE BODY DEPARTURES 3-23

FIGURE 3-15: CALCULA-TED THRUST FOR WIDE BODY DEPARTURES 3-24

FIGURE 3-16: EMPIRICAL AND INM NOISE CURVES FORWIDE BODY DEPARTURES 3-25

FIGURE 3-17: DIFFERENCES BETWEEN OBSERVED NOISE AND INKCALCULATIONS FOR WIDE BODY ARRIVALS 3-26

FIGURE 3-18: CALCULATED THRUST FOR WIDE BODY ARRIVALS 3-28

FIGURE 3-19: EMPIRICAL AND INM NOISE CURVES FORWIDE BODY ARRIVALS 3-29

xiii

,, , un n lm l ll . . .. . . . .... . . . L .. . I . . .. . . . . . . . ' " .. . . .... ..

LIST OF ILLUSTRATIONS (CONT'D)

FIGURE A-1- VALIDATION 14ETHODOLOGY A-2

FIGURE A-2: EXAMPLE OF 727 FLIGHT PROFILES A-6

FIGURE A-3: EXAMPLE OF 727 NOISE CURVE DATA FOR INMDATA BASE A- 7

FIGURE A-4: INK NOISE EXPOSURE CALCULATION A-8

xiv

7 11

1. INTRODUCTION

The Department of Transportation/Federal Aviation Administration(FAA) Integrated Noise Model QINM) is a set of computer programswhich are used to predict the noise impact of aircraft in theneighborhood of an airport. The purpose of the INK validationproject is to determine the accuracy of the FAA Integrated NoiseModel by comparing INM noise exposure calculations with actualmeasured noise exposure levels. The current phase of the pro-ject centers upon statistical analyses of single events in whichcalculated and observed noise exposure levels from air carrierflight operations are compared. This comparison uses data fromvarious noise monitoring sites around Washington National andDulles International Airports.

1.1 Background

The methodology employed in this analysis is a refinement ofthat initially presented in MTR-7ql3, "Analysis of IntegratedNoise Model Calculations for Concorde Flyovers" (Reference 1).In that paper, statistical techniques were presented to quantifythe noise characteristics of Concorde operations at DullesInternational Airport. The same basic methodology together withcertain extensions are now applied to representatives of thefollowing types of aircraft- two and three engine narrow bodyjets, four engine narrow body jets, and wide body jets. It isassumed that the reader is familiar with aircraft noise des-criptors and aircraft noise calculation models. A sophisticatedknowledge of statistics is not required to understand the bodyof the paper; technical material relating to statistical tech-niques is presented in the appendices.

1.2 Areas of Investigation

A discussion of variability in observation and modeling pre-sented in Reference 1 postulated that considerable variabilitymight be anticipated between an INM calculation and an actualnoise level observation, even when model inputs are carefullyspecified to accurately reflect the characteristics of theflyover. This contention was substantiated by the observednoise data for Concorde aircraft.

No aircraft noise model can be expected to predict accuratelythe noise level of an individual event. A valid model, however,will correctly determine the average noise level of a largenumber of similar flyovers. This notion of correctly deter-mining the average noise level conceptually defines "validity."The following section enumerates the techniques which were

1-1

employed in the current study to put this notion on a quanti-tative basis. These techniques also examine the characteristicsof the relationship between the observed noise values and theINM calculations. The areas of investigation are those con-cerning agreement, thrust sensitivity and noise curves.

1.2.1 Determining Agreement Between Calculations andMeasurements

The first and most important aspect of the comparison betweenthe observed noise and INM calculations is the ability of themodel to portray the noise environment. In this study, INMcalculations for various aircraft types are analyzed sepa-rately. This approach is taken because the overall accuracy ofthe INM is dependent on the accuracy of its calculations foreach of the individual traffic mix entries presented to it as aninput. The accuracy of the cumulative noise metric is boundedby the accuracy of the individual aircraft type entry. Forexample, if the noise exposure from each aircraft contributingto the total aircraft noise at a point is accurate to within 1dB, the overall cumulative energy noise metric will also be

accurate to within I dB, as a worst case.

1.2.2 Assessing Sensitivity to Thrust Assumptions

The second aspect of the comparison between the observed noiseand INM calculations involves the sensitivity of the INM toassumed aircraft thrust profiles in the INM data base. Theassumed thrust for an INM calculation of a noise event is im-portant because it cannot be measured experimentally, but ne-vertheless must be estimated in order to perform the noisecalculation. The actual thrust for an observed aircraft eventmay be different from the postulated thrust because of dif-

ferences in individual pilot thrust management procedures ortechniques, or because of inaccurate assumptions concerning theaircraft's flight configuration (i.e., gear and flap positions).

1.2.3 Investigating Empirical Models for Noise Exposure

The third aspect of the comparison between the observed noiseand INM calculations involves looking at the observed noisevalues to answer three questions: (1) Which of the observedvariables associated with the noise event (measured noise forthe aircraft flyover) is most highly correlated with that noiseevent? (2) What is the mathematical form of the variable usedto describe the noise event relationship? (3) Using the "best"mathematical model of the noise event relationship, does themathematical description agree with the noise curves used in INMcalculations?

1-2

1.3 Variation in Noise Observations

A characteristic of field observations involving uncontrolledaircraft flight operations is that observations of aircraftnoise levels will inevitably exhibit a degree of variability.In tests conducted by NASA Wallops Fligt Center, significantvariation was observed among flyover noise levels for a singleaircraft, despite tightly controlled pilotage procedures andessentially constant environmental conditions (References 2,

3). It can be expected that an even greater variability will beobserved among operations of commercial aircraft using standardapproach and departure procedures at public airports, yet it isfrom such observations that the data to be used in the INMvalidation is derived.

Uncontrolled field observations may also be expected to exhibitfurther variability when considered in relation to INN calcu-lations. As Table 1-1 illustrates, the model is a greatlysimplified representation of the factors affecting aircraftnoise. Implicitly, the model assumes that the factors which arenot modeled will "average out" in the long run. One follow-ongoal of the INM validation project is to determine whether thisassumption is justified.

The primary factors governing noise generation at the source arethe aircraft type, engine type, and thrust. The thrust dependson the flight path, on how the pilot makes thrust corrections tocorrect his flight path (pilotage), and on aircraft configu-ration (flap settings and landing gear). Additional factors atthe source include the effect of engine shielding by the air-plane fuselage, variations in noise exposure values in a "lobe"pattern around the engine centerline (directivity), and fre-quency shifts because of velocity vector orientation (dopplereffect).

Propagation effects include the relative distance between thesource and receiver (spherical divergence), atmospheric at-tenuation (which varies as a function of the humidity, tem-perature, barometric pressure, and wind), and atmospheric turbu-lence involving temperature gradients and other atmosphericheterogeneities.

Receiver effects include ground attentuation, ground surfacereflections and additional attentuation because of ground coveror intervening structures between the source and receiver.

1-3

'-4 ~

0 Ix 9 w zz ~ 0.4 0- P4 U 0

'-4 Zz E-4 -

E-4 .f

ot

rZ4 0-4 0-

E-4 cn -4'

C.)~L iz. C~C

z zW- C-4

1.-4

z > H

C. -4. P W

1- -0 4

The INM makes certain simplifications to the real world effectsin modeling aircraft noise. The aircraft type and engine typecategories are combined. The pilotage variability is not ac-counted for; rominal thrust values for different flight regimeswere drawn from manufacturer's data. There are temperature andpressure corrections made only to the takeoff profile. Thelanding gear/flap configuration is divided into takeoff orlanding configuration only. Shielding and directivity effectsare combined. There is no correction for the doppler effect.Spherical divergence and atmospheric attenuation are modeled bya noise versus distance table. There is no correction foratmospheric turbulence, reflections or ground cover.

Figures 1-1 and 1-2 show an additional area of variability:variability in the shape of the aircraft's noise-time history.These examples of histories of aircraft noise versus time aretaken within the same hour over the Old Town monitor site atNational Airport: (DCA). Whereas the shape of noise time historyfor departures are more regular in that they increase in noiseto a recognizable peak and then decrease, the time histories forarrivals are much more irregular. The designation "MAX" repre-sents the maximum sound level, in decibels, for the particularnoise-time history. The abbreviation "NEL" stands for NoiseExposure Level, which is the level of sound accumulated during agiven event. More specifically, NE.,, in decibels, is the levelof the time-integrated A-weighted squared sound pressure for agiven event.

The time histories for departures have a distinct peak whichoccurs near the midpoint of the time during which the noisevalue is above 70 dB(A). Seventy dB(A) is the threshold valueof the sound level used at the Noise Monitor Facility to dis-tinguish between aircraft noise events and ambient noise. Thetime histories for the arrivals in some cases have distinctpeaks near the midpoint similar to those for departures; how-ever, there are also examples of histories in which that is notthe case, such as the bottom three graphs in Figure 1-2. Theshape of these curves are more flat and in some cases the peakvalue does not occur near the midpoint of time during which thenoise exceeds 70 dB(A). In some cases, the noise time historyis nearly cyclical in nature.

In the foregoing discussion, the variation in the shape of thetime histories for arrivals is much more substantial than thosefor departures. These inconsistencies are reflected in resultsof the analysis. Statistically, much less can be said concerning

1-5

90 MAX = 89

80 NEL = 99

70

cA 60

90 MAX = 8 9NEL = 99

80

S70

rA 60

90MAX = 85

80

z' 70

0i2 60

90 MAX = 88NEL = 99

80

41 70

60 TIME 401 1-0-10 SECONDSFIGURE 1-1

REPRESENTATIVE NOISE-TIME HISTORIES FORAIRCRAFT DEPARTURES OVER OLD TOWN (OCA)

1-6

80MAX = 7 8

-~ NEL = 87

S70

60

80

S 60 I80

MAX = 768-~ NEL = 85

S70

80

80A MA 787

NEL=8

70 - -

z P60

80

NEL =88

70.c

z m

60

RERSNTTV NOS-TM NITREL F 87

FIU1-2

REPESNTAIV NOSETIE-ISORESFO

noise observations for arrivals than departures because the timeintegration process for NEL values tends to mask or average-outspecific sound level characteristics of a particular aircraftapproach.

Figures 1-3 through 1-6 show flight patterns of typical northand south operations at Dulles and National Airports (thesefigures are direct extracts from Reference 4). There are avariety of ground tracks abeam the various sites depicted onthese airport maps. This graphically illustrates the fact thatthe observations used in this study are obtained under a varietyof flight conditions. In addition to the variation in flightconditions, meteorological conditions varied considerably fromMay 1978 to January 1979, the dates during which the noiseexposure levels were recorded. Consequently, one can expect acertain amount of variation in noise measurements even under thesame nominal flight conditions.

1.4 FAA Integrated Noise Model Computer Program

The Department of Transportation/Federa Aviation Administration(FAA) Integrated Noise Model (TNM) contains computer programswhich can be used to estimate the noise impacts of aircraft inthe neighborhood of an airport (Reference 5). The model esti-mates the noise impacts of aircraft operations using the fol-lowing metrics:

Noise Exposure Forecast (NEF) -- An energy summation ofthe noise from a series of events, expressed in EffectivePerceived Noise Level, weighted for a difference betweendaytime and nightime noise exposure, and adjusted with anarbitrary constant.

Equivalent Sound Level (L ) -- The level of a constant

sound, which in 24-hour time period has the same soundenergy as does a time-varying A-weighted sound level.

Day-Night Average Sound Level (Ldn) -- The 24-hour periodA-weighted equivalent sound level, which has a 10 dB pen-alty applied to nighttime levels (2200-0700 Local Time).

Community Noise Equivalent Level (CNEL) -- The 24-hourperiod A-weighted equivalent sound level, which has a 5 dBpenalty applied to evening levels (1900-2200), and a 10 dBpenalty applied to nighttime levels (2200-0700).

1-8

LLES.IjTERNATIONRL AlIRPORT

,/LEESBURG,

-B .U R

ii I~Ji~iT~ISTERLING .* ~NORTH PR

RESTONTYPICAL SOUTH OPERATIONS-y~OCTOBER 4 1978

r 69 ARRIVALS 71 DEPARTURES

CATLLY

-~ 7SHADED AREAS REPRESENT-~ - PERCENTAGE OF OVERFLIGHTS

* ~ LESS THAN 1 PERCENT

,K* ~ MANASAS U 1 to 10 PERCENT10 to 100 PERCENT

FIGURE 1-3DULLES INTER NATIONAL AIRPORT'S MONITOR SITES

WITH TYPICAL SOUTH OPERATIONS

.1-9

6JLLES INTERNATIONAL AIRPORT

LEESBURG

DULLES STERLINGNORTH PARK

RESTONTYPICAL NORTH OPERATIONS

-< .1 OCTOBER 17, 1978~'75 ARRIVALS 74 DEPARTURES

HANT ILLYI

1 CENTERVILLE

-~ -'- iSHADED AREAS REPRESE1NT-- PERCENTAGE OF OVERFLIGHTS

- ~-.-- * 7]LESS THAI'J 1 PERCENTMANASSAS I~ TO 10 PERCENT

N.. - U 10 TO 100 PERCENT0 1 2 -.3 4 5NAUTICA1. MILES

FIGURE 1-4DULLES INTERNATIONAL AIRPORT'S MONITOR SITES

WITH TYPICAL NORTH OPERATIONS

1-10

WASHINGTON NATIONAL AIRPORT

\ABIN JOHN

' ~ CHEVY CHASE

LA14GLEY FORESTI- TYPICAL NORTH OPERATIONSPOTOMAC OCTOBER 17, 1978

CHAINL PALISADESBRIDGE 310 ARRIVALS 300 DEPARTURESj

- ~ ~ GEORGETOWN /R OS SLY N!

BELLEVUE

OLD >,TOWN

MARLIN - FT. FOOTEFOREST

NAUT'ICAL MI1LESSHADED AREAS REPRESENT WYEODSPERCENTAGE OF OVERFLIGHTS OITANALLON3>-

LESS THAN1I PERCENT

1 TO 10 PERCENT -

* 0 TO 100 PERCENT. .

FIGURE 1-5WASHINGTON NATIONAL AIRPORT'S MONITOR SITES

WITH TYPICAL NORTH OPERATIONS

& GTO NRT IONRL AIRPORT

I Ir CHEVY CHASE

~4 ANLE FOES. - TYPICAL SOUTH OPERATIONSPOTOMAC OCTOBER 12 1978 7PAL ISADES

RIDGE283 ARRIVALS 276 DEPARTURES

>uGEORGETOWNROSSLYN3

SHDE AEA RERSN4WYE

PERCENTAGE OF OVERFLIGHTS

K LESS THAN 1 PERCENTfl 1 to 10 PERCENT* 10 to 100 PERCENT

FIGURE 1-6WASHINGTON NATIONAL AIRPORT'S MONITOR SITES

WITH TYPICAL SOUTH OPERATIONS

1-12

Time of exposure above a threshold of A-weighted soundlevel (TA)--the time duration, in minutes, for which theA-weighted sound level at a measurement location is above aspecified threshold.

The INM computer programs calculate the values of these metricsfor selected points on the ground or in terms of contours ofequal noise exposure.

The user of the INK provides the program with a description ofthe runways, ground tracks, aircraft types, operations and trackutilization, approach profiles and takeoff restrictions. Inthis analysis, however, only single noise events (NEL obser-vations) are considered. The ground track is defined by astraight line in the vicinity of a monitor site. Only oneoperation, either takeoff or landing, is assigned to the groundtrack. The values of the various noise metrics are calculatedfor the positions of the monitor sites relative to the groundtrack.

1-13

2. METHODOLOGY

This study investigates three areas to validate the accuracy ofthe FAA Integrated Noise Model (INM) in calculating aircraftnoise exposure. These are (l) determining the agreement betweenthe noise model in calculating single noise events and themeasurement of those events, (2) assessing the sensitivity andcontrollability of the noise model to thrust assumptions, and(3) investigating noise curves used in calculating noise metricsby using regression models from empirical noise data.

2.1 Determining Agreement Between Calculations and Measurements

In this validation study, INM calculations for various aircraftare analyzed separately. Statistical methods are used to quan-titatively check the agreement between calculations and measure-ments of noise exposure by using paired differences between theobserved noise and the noise model calculation for the sameflight condition. The paired difference is formed by using themeasured noise exposure from a single aircraft flyover andcomparing it with the calculated noise exposure resulting fromusing the measured slant-range distance, altitude, and velocityof the aircraft at its closest point of approach to the monitorsite. The resulting statistic portrays the average differencein noise exposure from aircraft flyovers abeam a monitor siteand the INM calculations for the same simulated flyovers. Theobserved noise measurements, when not corrected for distance,are usually not normally distributed, and thus, standard statis-tical techniques cannot be used to find the mean of the observednoise measurements. The method of paired differences, on theother hand, does make corrections for slantrange distance aswell as for altitude and velocity. The paired differences aremore normally distributed which allows employment of standardstatistical techniques for determining the confidence intervalfor the mean difference between the measurement and calcu-lation. The basis for the selection of this method is docu-mented in Reference 1.

The noise metric used for the statistical comparison is theNoise Exposure Level (NEL). The NEL metric is the basic singleevent unit used for Equivalent Sound Level (Leq), Day-NightAverage Sound Level (Ldn) and Community Noise Equivalent Level(CNEL). Effective Perceived Noise Level (EPNL) and Time AboveThreshold will be considered in a document to be published at alater date.

2-1

The analysis is performed in three stages: data collection, INMcalculations and statistical comparison. The methodology isdiagrammed in Figure 2-t.

2.1.1 Data Collection

Noise level observations for this study were obtained throughthe FAA's Metropolitan Washington Airports Noise MonitoringSystem (NMS). The NMS consists of a central computer complex,which is connected to a network of remote monitoring sites(RMSs) located in communities near Washington National andDulles International Airports. The function of the system is toprovide accurate data on aircraft noise in the communitiessurrounding the airports, to identify the aircraft responsiblefor specific noise events (by airline flight number and aircrafttype), and to determine the flight paths of those aircraft.

Noise levels recorded by the RMSs are transmitted to a centralcomputer system at Dulles. The computer keeps a record of eachnoise event for later source identification and analysis.Aircraft noise sources are identified by aircraft type andflight number, and flight paths are determined, by means of datafrom the Automated Radar Terminal System-Ill (ARTS-Il) airtraffic control computer system. This data is used to computethe time of closest point of approach (CPA) and associatedflight characteristics such as distance, altitude, velocity andvertical velocity.

Nine monitor sites were used in the INM validation study, fourof them located near Dulles International Airport and fivelocated near Washington National Airport. These sites wereselected because of their proximity to the airport and for theirposition near the centerline of departure or arrival groundtracks. The distances from the landing threshold and the pri-mary runway used for departures or arrivals are listed for thenine monitor sites in Table 2-1.

The noise observations, together with data resulting from theaircraft flight paths and flight plans, were correlated ormatched with each other as to aircraft type and then filtered.Conditions to accept the data for further processing were addedto help eliminate questionable data by accepting only that datawhich were most likely to be an aircraft on arrival or departureflight path. These conditions were necessary because dataconcerning the actual. flight path was unavailable and only thoseflight characteristics at the aircraft's closest point of ap-proach to the monitor site was entered into the data base.These filtering conditions were:

2-2

J ,,

14-~ 0

(n 0IAW

-4 C-

En 0

00-

00

2-3

'-4Z

z I0 w

ll ,-4 CCIlPOW 0'0 0 0 00 0.O 00 0

CD C')I

-4-4

W

Hn z -E-'-4 cO 0 0-

W 0 0

J-4 00

z~ 0 -O~

2-4

1. The distance of the aircraft to the monitor site at thepoint of closest approach must be less than 10,000 feet.(This condition eliminates operations which have low noisevalues close to the ambient level.).

2. The duration of the noise event during which the noiseexceeds 70 dB(A) must be greater than 6 seconds. (Thiscondition helps eliminate non-aircraft noise events).

3. The absolute value of the vertical velocity must hegreater than 200 feet per minute. (This condition elim-inates level-altitude flyovers such as those on down-windleg. .

4. The altitudes for departures must be greater than 1500feet above ground level. (This puts departures, from aprocedural standpoint, in the cutback or climb thrustregion).

In this analysis, in addition to engine thrust, two other vari-ables could not be quantified. These are aircraft weight andstatus of the aircraft engine nacelle/acoustic treatment. TheINM has different flight performance profiles for particularaircraft types according to the weight of the aircraft. Theseprofiles model thrust, altitude, and velocity as the aircraftproceeds along its flight path. For a particular position alongthis flight path, however, the INN noise calculation with refer-ence to a particular point on the ground is independent of theaircraft weight. The actual altitude and velocity calculatedfrom the aircraft's radar track are being used in the INN noisecalculation. Theoretically, the aggregation of noise eventswithout regard to weight should not affect the results of thecomparison.

The INM also has different noise curves for particular aircrafttypes depending on whether or not it has engine nacelle/acoustictreatment. Statistically, the two types of engines would causea bimodal distribution of noise measurements at the variousnoise monitor sites. The bimodal distribution would be es-pecially noticeable during arrival operations, resulting fromaircraft with the acoustic treatment measuring quieter thanaircraft with standard engines. If the quieter aircraft withacoustic treatment were to be removed from a sample containingboth types of engines, the average noise level of the samplecontaining aircraft with only standard engines would be higher.The actual effect of both the aircraft engine nacelle/acoustictreatment and aircraft weight on noise exposure levels is undercurrent investigation.

2-5

2.1.2 INM Noise Exposure Calculation

The noise model includes in its data bases descriptions of

flight profiles which define the flight characteristics along

the ground track and noise versus thrust and distance curves for

different aircraft types. Examples from this data base togetherwith the mechanics of the noise exposure calculation are givenin Appendix A.

Because the actual thrust setting data were not available,assumed thrust levels had to be input to the model before the

noise model could compute the noise exposure. (The sensitivity

of the noise exposure calculation to the thrust estimate isinvestigated in Section 2.2 of this study.) The thrust as-sumptions are complicated by the fact that Dulles and National

Airports have different operational procedures for takeoff.

Prior to October 17, 1978, aircraft departing from Dulles wereinstructed to use the standard takeoff procedure advocated by

the Air Transport Association. This procedure dictated that

aircraft use takeoff power until reaching 1500 feet above theground and then reducing power to maximum continuous limitingthrust for climb to en route altitude. Aircraft departures from

National Airport, on the other hand, were instructed to use amodified climb procedure. This procedure dictated using takeoff

thrust to an altitude of 1500 feet above ground level and then

reducing power to a thrust setting which will maintain 500 feetper minute climb rate under hot day conditions. On October 17,1978, the FAA issued an advisory circular (Rcference 6) whichchanged the altitude at which takeoff power is reduced from 1500

feet to 1000 feet.

For departures from Dulles and National Airports the thrust

assumptions were derived from the flight profiles included inthe INM computer package. For departures from National Airport,in particular, the thrust value was calculated by the INM com-

puter package so that the aircraft maintained a 500 feet perminute climb rate.

For arrivals the thrust assumption was obtained from an examplein the "INM User's Guide" (Reference 5) which gives the approach

thrust as a function of distance from runway threshold. For theDulles North, Chantilly, Arcola, and Old Town monitor sites, thethrust was assumed that for an approach using a 30 glide slopeand landing flap extension. For all the other sites, the thrustwas assumed that for maintaining level flight with an approachflap extension.

2-6

The distance from the aircraft to the Remote Monitor Station,altitude of the aircraft and velocity of the aircraft are alldirect inputs to the INM calculation of noise exposure. Theaircraft type, flight operation determination (takeoff or land-ing) and distance from the monitor site to the runway are usedto determine the thrust estimate which in turn is used to de-termine the noise curve relating noise versus distance from theaircraft to the monitor si.te. Once the noise exposure has beencalculated for the particular flight condition, the observed andcalculated noise levels are then compared.

2.1.3 Statistical Comparison

The statistical comparison presented analyzes the distributionof the differences between observed noise levels and cor-responding noise model calculations. The measured noise levelsare not normalized or adjusted for any variable. Since the INMhas algorithms which incorporate distance, altitude and velocityin its computation of the noise exposure level, the INM cal-culation for a given noise event accounts for these variables.The mean and standard deviation of the sample containing thepaired noise values was then used to determine the confidenceinterval for the average difference between the measured andcalculated values.

Examples of departure and arrival noise measurements and cal-culations are shown in Figures 2-2 and 2-3. In the upper leftscatterplots, each observed noise event has a corresponding INMcalculation. Each of the paired differences (observed noiseminus INM calculation) is then plotted in the lower left scat-terplot.

The probability density estimates for the observed noise and thedifferences are plotted on the right. The method by which thesedensities were derived as well as the calculation of the Kol-mogorov-Smirnov statistic ("K-S STAT") are described in Refe-rence 1. The Kolmogorov-Smirnov statistic tells whether or notthe sample could have come from a normal distribution. If thestatistic is less than 1.05, the sample can be said to be nor-mally distributed. If the statistic is greater than 1.05, it ishighly unlikely that the sample is normally distributed. Asseen in Figures 2-2 and 2-3, correcting the observed noisevalues for distance produces a distribution which is more normallooking.

The 95% confidence interval for the mean of the differencebetween the observed noise and INM calculation is shown in thelower right density estimate as a horizontal line joining twovertical lines as in a sidewise "I".

2-7

SN~-- aI

I.- C

Z .

0~~~~c 0' .' a ' a 'Za'~ ~~~ co a' a o ' a

0LS~( AL'l1VOI 0kINA 011I lgcC.)_ _ _ _ _ _ _ _ _ __z _ _ _ _ _ _ _ _ _ _ a

00 0 04z 0' 0 0 4

0 00 ow

u . w0 0 g Lolo44

0

a' 0 44

01 03T~AI 1/ N

2-84

-co

N ~ N-4'0 c~I ->

no z

L xs I

cc4 U. o'I)L CC,

'~~C us l -

0 0 .6.A1ISN~~~~~S ktLqg)a isa ~iig~10

'S U.

0 zn0~

'S

oCIa 113 Ol-Ml:rV WI0 3I

CI] 4a n al '12

- C..2-9

2.2 Assessing Sensitivity to Thrust Assumptions

A strong mathematical relationship exists between the NoiseExposure Level and the slant-range distance at closest point ofapproach. The noise level is also strongly related to thethrust level. The thrust level, unfortunately, cannot be mea-sured; however, the thrust level must he estimated in order toobtain a noise model calculation.

One problem of using noise data from Dulles and National Air-ports is that the aircraft use different thrust profiles fordepartures. Since thrust levels cannot he measured directly,this means that the absolute differences between the observednoise and INM calculations may be caused in part by differentthrust assumptions, namely, the assumed thrust for the INKcalculation can be a different thrust level from that actuallybeing used. Since thrust assumptions can influence the resultsof the comparison markedly, a method by which the comparisonscould be made to a common datum had to be found. This method isthe noise-thrust mapping procedure which translates the measurednoise into equivalent thrust values. This translation is basedon the premise that the noise model is functionally correct, andthat the difference between measured and calculated noise isprimarily attributable to the assumed thrust profile. The INMcalculation is corrected for slant-range distance, velocity andaltitude, leaving thrust as the only primary variable which isunaccounted for. Essentially, this method uses the INM as ayardstick to transpose from the noise regime to the thrustregime and back again.

Another problem arises from some of the monitor sites beingsituated near the point at which the power is to be reduced fromtakeoff power to a cutback power for climb. An example is theOld Town monitor site at National Airport. If an aircraft wereactually still using the takeoff power setting but a cutbackthrust setting was assumed for the INM calculation, the dif-ference between the two would be unusually large. However,using the noise-thrust mapping procedure to obtain an estimateof the thrust used by the aircraft abeam the monitor site,errors in the INM calculation because of an inaccurate thrustassumption can be found easily.

The problem of developing useful thrust assumptions is com-pounded for arrival aircraft. Included in a ten nautical mileapproach path are differing flight conditions such as whetherthe aircraft is in level flight or descending, whether or notthe landing gear is extended, and the degree of extension of the

2-10

landing flaps. Each combination of these flight conditions re-quires a different thrust setting to maintain a stabilizedflight path. For a typical approach profile, the IRM has fivedifferent thrust settings. By using the noise-thrust mapping

procedure, the relative location of these thrust settings are

found easily.

Another major advantage of using the noise-thrust mapping pro-cedure is that it enables an assessment of the innate con-trollability of the thrust curves in the INM data base to makecalibrated adjustments. If the estimated thrust resulting fromthe noise thrust mapping procedure is greater than the takeoffthrust, for example, then no adjustment is possible and thethrust curves may have to he redefined.

The methodology for determining the INM's sensitivity to thrustinvolves using the noise-thrust mapping procedure and the com-puted difference between the observed noise and the INM cal-culation of noise from the aircraft's flight characteristic.

2.2.1 Noise-Thrust Mapping Procedure

The noise-thrust mapping procedure begins with the thrust as-sumption used to compute the difference between observed noiseand TNM calculation, as shown in Figure 2-4. Using a linearinterpolation scheme based on the log average of the distancesat closest point of approach (CPA) for a particular sample, thenominal noise value is calculated from the thrust curve for thenominal thrust (initial estimate) and average distance as il-lustrated in Figure 2-5. The log average of the distance isused because observed noise is primarily a logarithmic functionof distance (spherical spreading). For this example, the cal-culated value corresponding to one side of the 95% confidenceinterval for the difference between the observed noise and INMcalculations is then added to the nominal noise value. Anexample of this confidence interval is shown in Figure 2-2 inthe lower right-hand figure as a sidewise "I." Using a linearinterpolation scheme again, the thrust corresponding to thenoise value resulting from the addition is calculated, againusing the log average of the CPA distance. The resulting valueis a calculated thrust (revised estimate) corresponding to theobserved noise as viewed by the INM. This procedure is repeatedfor the other side of the 95% confidence interval.

2.3 Investigating Empirical Models for Noise Exposure

Since the initial analysis of the noise exposure data and thecomputation of the INK sensitivity to thrust variations hadindicated problems with certain of the basic INM noise curves,

2-11

trj :D

w z

Ic z

-4 w

C, 1,

I 0I 00

LL>

-r zW U)

w Hn

u II U)W-,

m C4

2-1

M HUE-4

z~ zE-4-4 0 1

H

0

OH4

00

w I 00I~U.)I~~~ IeQ 'tAI Lnsaa SO

C4 >

E- 2-1c3

it was decided to investigate the mathematical values isolatedfrom the influence of INM form describing observed noise cal-culations by computing empirical noise curves from the observeddata. These empirical curves could then be compared with thecorresponding INM curves, and significant differences noted.

As illustrated in Figure 2-6, for each aircraft type and oper-ation there were four steps leading to the comparison of the INMcurve and the empirical curve. First, the variables for eachset of data were analyzed using a stepwise linear regressionprocedure to determine those variables statistically significantin estimating the noise level. Second, from the series of re-gression models output by the stepwise procedure, an appropriatemodel was chosen as the empirical curve. Third, its statisticalvalidity was examined to determine whether or not the sample wasnormally distributed. Fourth, whenever the sample was deter-mined to be normally distributed, a confidence region of atleast 95% was computed about the curve to portray the possiblevariation in the empirical log-linear relationship. Finally,the empirical curve was graphically compared with the cor-responding noise curve from the INM data base.

2.3.1 Variables Tested for Significance in Estimating NoiseValues

For each noise exposure measurement, five variables were alsocalculated at the aircraft's closest point of approach (CPA).These variables were: slant-range distance, altitude, velocityalong track, vertical velocity, and elevation angle. (Themethod by which these variables were computed is outlined inAppendix A.) Slant-range distance, altitude, velocity alongtrack, and elevation angle are variables which relate to thephysics of noise propagation and are used in the INN algorithmin calculating noise. Vertical velocity relates to either theclimb or the descent gradient and was included in the analysison an exploratory basis to determine whether or not the effectof vertical velocity would be significant. In addition, threetransformations of these quantities were included as variables:log1 0 (CPA distance), logi0 (CPA velocity), and logl0 (CPAelevation angle). Indicator variables for each monitor sitewere also included to check for possible differences i.n measure-ments between the sites. The use of indicator variables isexplained in Appendix B.

2-14

DATA FOR ONEAIRCRAFT TYPEAND OPERATION

ISTEPGRSEION MODELS

REGRESSION )ANALYSIS ," CORRELATION COEFFICIENTS

SELECTION OF

EMPIRICALCURVE

VALIDATION OF STATISTICALPROPERTIES OF EMPIRICALCURVE

COMPUTATION OF CONFIDENCE INM NOISEREGION FOR EMPIRICAL CURVE CURVE

COMPARISON

FIGURE 2-6METHODOLOGY OF EMPIRICAL AND INM NOISE CURVE COMPARISON

2-15

L

2.3.2 Stepwise Regression Procedure

In order to determine those variables significant in estimatingthe noise levels and the functional form of their relationships,a powerful statistical technique, stepwise linear regression,was employed.

Starting with the single most significant independent variable,these regression models are built, as the name implies, in astepwise fashion. At each step the procedure seeks to add avariable or to interchange a variable in the model with a vari-able not in the model in order to improve the least squares fitof the regression line. The procedure terminates when there areno more significant variables in the data. (The statisticaldetails are given in Appendix B.) Thus, the final output of theprocedure is a sequence of models, each being a better fit ofthe data than the ones preceding it. Various statistics abouteach model are also provided.

2.3.3 Confidence Regions for the Empirical Noise Curves

A confidence level region of at least 95% may be computed for aregression curve using the technique described in Appendix B, ifthe residuals are normally distributed (the Kolmogorov-Simirnovstatistic is used to test for normality. In a few cases, theresiduals departed slightly from normality, implying that whilethe estimate of the regression curve was still correct, con-fidence regions could not be computed in the usual fashion(Reference 5).) These confidence regions have the followinginterpretation: If data were collected 100 times while holdingthe values of the independent variables at the same levelsobserved in the samples, at least 95 of those 100 times onewould expect the empirical noise curve to lie within the boundsof the confidence region.

2.3.4 Comparison of Empirical Noise Curves with the INM NoiseCurves

Once the empirical noise curves had been computed, each wasgraphically compared with the appropriate INM curve(s) wheneverpossible. Since the INM curves express the noise level as afunction of the distance at closest point of approach, the onlyempirical noise curves that were graphically comparable werethose which included either CPA distance or log,0 (CPA dis-tance) as an independent variable.

2-16

When the empirical noise curve was a function of a set of vari-ables including one or more that were neither functions of theslant-range distance nor indicator variables, a graphicalcomparison was more difficult. In order to make any graphicalcomparison whatsoever, some constant value of the variables(other than slant-range distance) must be chosen for use in thegraph. In an attempt to select a generally representative valuewhile not overfitting the curve to the data at hand and losingthe desired generality, the median observed value of the vari-ables (other than slant-range distance) was selected. Forexample, the empirical NEL curve for the DC-8-55 arrivals re-sulting from a step-wise linear regression analysis was

NEL = 178 - 24 * LOGIo (CPA DISTANCE) + .006 *(CPA VERTICAL VELOCITY)

Using a CPA vertical velocity of -720 feet per minute, the curvegraphed for comparison was actually

NEL = 174 - 24 * LOG10 (CPA DISTANCE).

A further problem in comparing such curves occurs with theinterpretation of confidence regions. In some instances, theseconfidence regions are three or more dimensional, and as such,cannot be accurately graphed in two dimensions. The equationsof the empirical noise curves and the sample sizes for all setsof data examined may be found in Appendix C.

The stepwise linear regression analysis is a statistical tech-nique which focuses on the structure of simultaneous rela-tionships among three or more variables. The objective of thistechnique is to search for the best possible simultaneous re-lationship between the observed noise and other correlatedvariables. Mathematically, these relationship can be deter-mined. However, in simplifying these relationships, as inrelating observed noise to only slant-range distance, somesecondary relationships may be obscured. This makes inter-pretation of the data more difficult.

2-17

3. RESULTS

There are three independent methods of assessing the agreementbetween INM calculations and observed noise values. The firstmethod, the statistical treatment of paired differences, pro-vides an average difference over the range of CPA distancesinvolved. The second method, the noise-thrust mapping proceduredetermines the thrust needed to produce the observed noise, andprovides an indication of the capability of the model to simu-late the observed noise without changes to the noise curves.The third method, the regression analysis and comparison ofnoise curves, enumerates and quantifies the effect of the vari-ables used in the noise calculations and also determines themathematical form of the functional relationships. The regres-sion analysis provides a method for changing or redefining thenoise curves.

Over 6000 single-event noise measurements (measured noise for anaircraft flyover abeam a monitor site) were taken and pairedwith calculations from the INM for statistical comparison. Theobservations were categorized by aircraft type, monitor site,and type of flight operation, i.e., departure or arrival. Theevents for each airport, for each aircraft type and each flightoperation were combined for grouped statistics. The aircrafttypes were then arranged in three groups: four engine narrowbody aircraft, two/three engine narrow body aircraft, and widebody aircraft.

For each aircraft type, a noise-thrust mapping procedure wasperformed for each monitor site and flight operation. Theestimated values of the thrust were acquired from the statis-tical treatment of paired differences. For purposes of com-parison with the thrust assumptions, the INN thrust profileswere provided for takeoff, climb and arrival operations. Eachvalue was computed from the same sample from which the NELdifferences were computed.

For nineteen of the twenty-eight sets of data examined using themethodology discussed in Section 2.3, the empirical noise curveswere determined. The charts describing these curves are foundin Appendix C. Note that the curves are only compared for theobserved range of the distance. Since the empirical curves areregression lines and consequently are sensitive to extreme

points, the curves cannot be assumed to be valid outside of thisobserved range.

3-1

Prior to discussing the data case by case, two general ob-servations should be made. First, the data for arrivals wasless well behaved than that for departures, i.e., using thevariables tested, arrival noise levels were less predictable.

Of the fourteen sets of arrival data examined, empirical curvescould be found for only three. In those empirical curves notcomparable to the TNM curves, the consistently important vari-ables were altitude, velocity, and logl0 (velocity). Onepossible explanation for the more erratic behavior of the ar-rival data is the greater impact of ambient noise on the mea-

sured noise. During the approach or landing phase, the aircraftis using much less thrust than during the takeoff phase, andsubsequently, much lower noise values are observed during land-ing. The ambient noise, if correlated or labeled as aircraft

noise, could introduce a much higher and erroneous value thanthe actual aircraft noise value. Another possible explanationfor the more erratic behavior of the arrival data is that thepilots are varying the aircraft thrust during the landingphase. These thrust adjustments could be for a change inconfiguration (lowering landing gear or flaps) or for main-taining final approach airspeed.

A second general observation is that the data gathered at Na-tional Airport was significantly less well behaved than thatgathered at Dulles Airport for both departures and arrivals.Although some empirical noise curves could be determined fromthe National departure data, these curves (as measured by thecorrelation coefficient) did not fit their respective data aswell as did the Dulles empirical curves. Furthermore, no sta-tistically valid empirical noise curves whatsoever could bedetermined from the National arrival data. The statisticallyless well behaved data at National Airport may be caused in partby the greater dispersion in the ground tracks abeam the monitor

sites as depicted in Figures 1-5 and 1-6. (Monitor sites lo-

cated at greater distances from the runway threshold tend tohave greater dispersion associated with the ground tracks. Thedispersion in ground tracks, in turn, causes more variability inthe observed noise because of variations in thrust, descent orclimb rates, and shielding from different flight attitudes).

The results will now be presented by aircraft type groupings. Asummary of these results is provided in Section 4.

3.1 Four Engine Narrow Body Aircraft

Data on four different types of four engine narrow body aircraft

was gathered at Dulles Airport. Case by case synopses follow.

3-2

3.1.1 Departures (Four Engine Narrow Body Aircraft)

The Q5% confidence intervals for mean differences between ob-served noise and INM calculations are shown in Figure 3-I. Theaverage difference for monitor sites other than Centreville forthe different aircraft types is in the range from 3 to 7 rB.The sites all seem to be grouped together except for Cen-treville. The uniqueness of Centreville is discussed at the endof this section. Opposite each aircraft type labeled in Figure3-1 is given first the sample sizes (#) of the observations andthen the q5% confidence intervals for the mean differences foreither individual monitor sites (bars labeled "2" thru "5"), orthe combined differences of all the monitor sites (bars labeled"I"). Sample sizes of observations less than 10 are not shown.For example, for 707-120 aircraft, there were a total of 262observations for all the sites. Of these 63 observations werefor Dulles North. (The location of Dulles North was previouslyshown in Figures 1-3 and 1-4.) For Dulles North, there is a 95%chance that the mean difference between the observed noise andthe INM calculations of the same flyovers is in the range ofabout 3 to 5 decibels. In other words, this statistic says thatobservations at Dulles North are, on the average, 3 to 5 deci-bels higher than the INM would calculate for the same flyovers.

Figure 3-2 shows both the calculated thrust resulting from

application of the noise-thrust mapping procedure to the meandifferences of Figure 3-1, and also the takeoff and climb thrustprofiles resident in the INM data base. The family of curveslabeled "takeoff thrust" and "climb thrust" correspond to air-craft profiles of varying weights. The heaviest aircraft de-picted in the INM data base would have its takeoff thrust main-tained at the higher value for a longer distance from start oftakeoff roll than would the lightest aircraft. The verticalbars shown in Figure 3-2 correspond one-to-one to mapped valuesof the mean differences from Figure 3-1. For example, themapped differences for 707-120's at Dulles North equates tothrust values between about 15000 and 17000 pounds. Whereas theactual value of the thrust may not be important, the position ofthe mapped thrust value relative to the reference INM thrustcurve is important. In this case, the mapped thrust is locatedabove the takeoff thrust curve obtained from the INM data base.Since the takeoff thrust is theoretically the maximum thrustthat an aircraft can use for any operation, the location of themapped thrust above the takeoff curve indicates that there maybe a problem with the 1NM takeoff curves themselves. The pro-blem is that the INM cannot be adjusted to produce a calculatednoise analagous to the noise observed at the monitor sites. In

3-3

DEPARTURESAIRCRAFT TYPE NEL DIFFERENCE, DB

-3.0 -5 0 5 10

262 "

63 2707-120 128 3

50 4i

23 5

140

33 2

707-320 60 3

22 4

25 5

29 1DC-8-55

14 3

168

DC-8-60 68 3

49 4

47 5

ALTITUDES OF DEPARTURES >1500 FEET AGL

DATES: 3 MAY 78- 8 NOV 78

LEGEND: 1 ALL DULLES SITES2 DULLES NORTH

3 CHANTILLY4 ARCOLA5 CENTERVILLE

# SAMPLE SIZE

FIGURE3-1DIFFERENCES BETWEEN OBSERVED NOISE AND INM CALCULATIONS

FOR FOUR ENGINE, NARROW'BODY DEPARTURES

3-4

20g~qI 3g707-3202

600TAKEOFF THRU S

1200 - CLMB THRUST

18000'

80000 3 4 70 -2

16000 TAKEOFF THRUST

5

CLIMB THRUST-

12000

A~ CENMBTHRUSTL12000~ -02 04 0 6

D8 SANC FRMSATO AEFFRLX00FE

2000IGUR 3-2--6

1CACULTE THRUSTF FORUS

3-5

sites. In general, results from all four engine narrow bodyaircraft, when mapped into the thrust regime, indicate that therequired INM thrust necessary to produce the observed noiseexceeds their respective takeoff thrusts.

Figure 3-3 shows a comparison of the empirical and INM noisecurves. The INM noise versus distance curves (shown as solidlines) are obtained directly from the INM data base and have notbeen altered in any way. The empirical curves (shown as dash--dotted lines) are a result of an automated stepwise regressionprocedure. The samples used for the regression analysis werenot a priori separated into any groupings. The groupings thatevolved are a result solely of the automated statistical pro-cedure. The 95% confidence interval for the empirical curves(shown as dashed lines) are shown whenever the statsti-calproperties of the sample enabled its calculation. In general.,the empirical curves for Dulles North, Chantilly and Arcola (agrouping selected by the automated regression procedure) are 2to I decibels higher than the INM curve for takeoff. The em-pirical curves for Centreville are very near the INM curve forclimb.

The reason for the uniqueness of Centreville is suggested in

Figure 3-2, which depicts the calculated thrust which would be

required by the aircraft to produce the observed noise. InFigure 3-2, both the takeoff thrust and climb thrust are shownin relationship to the calculated thrusts as a function ofdistance from the start of the takeoff roll. Comparing thecalculated thrusts with the nominal thrust profiles from the INMsuggests that the aircraft are not cutting back to climb thrustuntil they are at downrange distance of more than 28,000 feet.An important observation to be made from Figure 3-2 is that thecalculated thrusts for the observed noise abeam the DullesNorth, Chantilly and Arcola sites all exceed the takeoff thrustsfor their respective aircraft types. The altitudes of all theaircraft in these samples are above 1500 feet, which means thatif the aircraft were following standard climbout procedures,they all should be at climb thrust.

The uniqueness of Centreville is substantiated by the comparisonof empirical noise curves and the INM carves in Figure 3-3. Thestepwise regression procedure selected Centreville as beingstatistically different from the other three sites for 707-120,707-320 and DC-8-60 aircraft. For the DC-8-55 sample, theobservations at Centreville were not separated out from the restof the sample. The message from three separate approaches saysthat Centreville is unique as a noise monitor site. A partialexplanation for this uniqueness is that aircraft are maintainingtakeoff thrust longer during climbout than assumed. Since

3-6

707- 20) FOR TAKEOFF

120 CURVE FOR 0-10 MCUVEMPIRICAL CURVE FOR

110 -IRiCAi CURVE FOR 10 DULLES NORTH,

IULLES ORI'iI. U F.' CHANTILLY, ARCOLA

(:iAN'II ELY. ARGOLA

1NMl CURVE ''." . 10 .M CURVE

FOR CLIMB FOR CLIMB

EMPIRICAL CURVE EMPIRICAL CURVE '>MP 1a 1CL CUVF FOR CENTERV ILLEFO)R LLNIERVILLE.

0

10 I l II!I l 70L ] I I I I500 I100 it000 5000 10,000 500 1000 2000 5000 [O,000

JISIANCE Al CPA (FEET) DISTANCE AT CPA (FEET)

L20 12U - -

INM CURVE FOR [AKEOFF DC-8-55 INM CURVE FOR TAKEOFF DC-8-60

EMPIRICAL CURVE FOR EMPIRICAL CURVE FOR

110 ALL SITES 110 DULLES NORTH, CHANTILLY

., ARCOLA

>100 - 100 S

2 ~ INM CURVEFMR CLIMB

11NM CURVE0

FOR CLIMB

80 EMPIRICAL CURVE80' I 0 - FOR CENTERVILLE

70 1 1---- - I I70 II 1oo ooo 2000 5000 10,000 500 1000 2000 5000 10,000

DISTANCE AT CPA (FEET) DISTANCE AT CPA (FEET)

FIGURE 3-3EMPIRICAL AND INM NOISE CURVES FOR

FOUR ENGINE, NARROW BODY DEPARTURES

3-7

Centreville is a greater distance from the start of takeoff roll(41000 feet versus approximately 23000 feet for the othersites), nearly all aircraft have reduced their thrust levels tothose specified for climbout, whereas at the closer-in sites,the thrust reduction has not been accomplished by most aircraft.

3.1.2 Arrivals (Four Engine Narrow Body Aircraft)

The average differences between the observed noise and INMcalculation shown in Figure 3-4 are in the range of I to 6 dB,with DC-8-60's having the best agreement and 707-320's havingthe worst agreement. In contrast to the departures, arrivaldata at Centreville are not markedly different from the otherthree sites.

Although the range of the differences shown in Figure 3-4 aregenerally 2 to 3 dB, when these intervals are transformed tocalculated thrust for the observed noise, the intervals becomemuch larger than those for departures as shown in Figure 3-9.These large intervals suggest that the range of approach thrustvaries'substantially. However, part of problem of interpretingthe arrival data may be in the INM curves used to calculate thethrusts. Since the calculated thrust are substantially higherthan the INM approach thrust curve, this suggests that the TNMcurves are not correct. Since the approach thrust values wereobtained from manufacturer specifications, the INM noise curvescorresponding to a particular thrust are suspect.

The results of the regression analysis (Figure 3-6) show that

the empirical curve for DC-8-55 arrivals is approximately 2 to 4dB higher than that calculated by the noise model. The empiri-

cal curves for the other three aircraft types are not shown be-cause they are not mathematical. functions of CPA distance. Thisproblem is treated in Section 3.4.

3.2 Two/Three Engine Narrow Body Aircraft

Data on three different types of two and three engine narrowbody aircraft was gathered at Dulles Airport, while data on foursuch aircraft types was gathered at National Airport. Case bycase synopses follow.

3.2.1 Departures (Two/Three Engine Narrow Body Aircraft)

Dulles and National departures are not comparable directly,because of the difference in takeoff procedures. At both air-ports, takeoff power should be maintained to an altitude of 1000feet. However, above 1000 feet, procedures at Dulles call fordepartures to maintain climb power, while procedures at Nationalcall for departures to reduce thrust to a setting required tomaintain 500 feet per minute climb.

3-8

LJ

ARRIVALSAIRCRAFT TYPE NEL DIFFERENCE, DB

-10 -5 0 5 10

288 U78 2M

707-120 98 3M58 4n

54 5

1

237 20"

60 3M

707-320 92 4i

25 51M

60

i

50 21 1

DC-8-55 20 313

5

1018

215 2 3

17rDC-8-60 66 4

72 560

DATES: 3 MAY 78- 8 NOV 78

LEGEND: I ALL DULLES SITES2 DULLES NORTH3 CHANTILLY4 ARCOLA5 CENTERVILLE

# SAMPLE SIZE

FIGURE 3-4DIFFERENCES BETWEEN OBSERVED NOISE AND INM CALCULATIONS

FOR FOUR ENGINE, NARROW BODY ARRIVALS

3-9

4.4

INM APPR Wc: Ii IIR1S'

22 -_ _ 0,0 73

22 [ 2I I~ N' PPROA Al 7W

800 [

8000 .- ji' KV Ii

1l '1 1 ,-O X D ' l ! 1 F

FIGURE 3-5CALCU)LATED THRUST FOR

FOUR ENGINE. NARROW BODY ARRIVALS

3-10

110

DC-8-55

EMPIRICAL CURVE FOR ALL SITES

100

> 90

: 80 INM CURVE FOR LEVEL FLIGHT

0z

70 INN CURVE FOR 3 SLOPE

60 I I I I 1500 1000 2000 5000 10,000

DISTANCE AT CPA (FEET)

FIGURE 3-6EMPIRICAL AND INM NOISE CURVES FORFOUR ENGINE, NARROW BODY ARRIVALS

3-11

Figure, 3-7 compares the average vetc al vI oiya P odeppartllresq at D'illeq and National Airports. Tine vertical veloc-it)es fo'- aircraft departures from National .1r) not liffer verymuch from those seen at Duller,. Tn order to maintain the sameclimb rate, a similar thrist setting for the same aircrafttvpe is required. This contention is also supported by thereqtults of the following three separate analyses.

In general, observPod noise levels it Dulles for DCl-q, 727 ani737 are lower than the INM calculations. One possible exTpla-nation for this observation is that a portion of the aircraft inthe sample are eciiipped with quieter, FAR 36 compliant engines.Tn this 1nalysiq, however, Pircraft which have tetrofit enginesare no,- distin2:uisbahlo from those which have stanilarl engines.Thus, all airc'-aft are assumed to be oquipped with standardengines.

The result- of the statistical r ompari*-on for the differencesbetween the observced noise and !NN4 calculations; for two andthree engine narrow b)odv aircraft are presented in Figure 3-8.'The 'euts for Milles and National Airporta are presentedseparately. The composite di'fferenres fo-r Dulles sites inidic:atethat the observed noise is 0 to 3 dB below the noise modelcalculations. Differences for Centreville are noticably lif-ferent from the other sites at Dulles. with the TNM calculationsbeing 3 to 7 d9 helow the observed noise. The composite dif-ferenres for National Airport indicate that the observed n-3ise

isto 6 dB hi zher than the TNM cn ul at ions . For National,differences for Old Town are noti.cabiv di fferenit from the othersites, with the observed4 noiehinv -7 to 10 4,3 higher thann theTNM calculations. Note that theP differences for thle two air-ports are meastired with respect to different thrust assumptions.

Fgure 3-9 shows the cairulated thrust corresponding to theobserved noise of Figure 3-8. The climb thrusts for Dullesdenartures Are assumed to be higher than those for Nationaldepartures. When the calculated thrusts for the two airportsare presented on the same graph as a function of distance fromstart of takeoff roll. the annarent- ininiieress of Centreville.sne 4 flV-! Town liacomes linderstandablep. The iistance': from start oftai'weo~f -oll for Old; Town is bet-ween the distances for Dull.esNr'-th An! C(Thntillv; similarly. Cpnterville- is between MarlinForpst ini1 Jaynewoodl. These relative positions suggest thatqir-rif- ri-Formncc near Old Town is comparable tn some Delles-sei I- Ad -ha ~int a irc-ra ft ner fonre near Centrceville is orn-nirall to- some National sites.

AIRCRAFI DULLES DEPARTURES AIRCRAFT NArLONAL DEPARTURES

TYPE VERFICAL. vELITiy. Fpm TYPE "ERTiCAS. VELOCITY, FPM~.0 1000 2000 3000 4000 so. 1000 2000 3000 4000707-120 212 1DC-9 305 U

63 on. 41 m126 c72C350 1267 cm23 75 U

50El707-320 140 0

33 cw727 322 1160 C310 m22 43 =25 123 m

0751

DC-8-55 29737 342

14 83 83160 m94 0

75IDC-9- 69 68 7

49 19 m47 1111

DC-9 344 1 BAC-111 260 050:262 0

200 194 E50 [a4638 m44

22 i727 325

78128 064 IU55=

737 31

15

747 86U23331014

DC-10-10O 93

44 C39 tm

L-1011 52

2119 m

LEI;END: ALL DULLIES SITES fIIII LEGEND: ALL NATIONAL SITESDULLES NnRTH OLD TOWNCHANTILLY ClFT. FOOTEARCOLA MARLIN FOREST gm(:E.TERV 11; WAYNEWOOD

TANTALLON ErSAMPLF 917 e

SAM4PLE SIZE 6.FIGURE 3-7

95% CONFIDENCE INTERVALS FORAVERAGE CPA VERTICAL VELOCITY

3-13

AIRCRAFT DEPARTURES

TYPE NEL DIFFERENCE. D8

0 -10 -5 10

344 1044 2

200 M

50 4

DC-9 38 5

305 6

41 7

72 8

67 9

75 1 m

50 1_______ _______

325 1

59 2 M

128 3

64 4

727

322 6E3

18

43 8

123 9

87 10

51 1

31 1

153M

737342 6

837

948

75 9

71 lo i19 Lim

268 6RAC- 111 62 7M

948

46 9

44 10

_______________22 ,1 ________

ALTITUDES OF DEPARTURES > 1500 FEET AGLDATES: 3 MAY 78 - 23 JAN 79

LEGEND: I ALL DULLES SITES 6 ALL NATIONAL SITES2 DULLES NORTH 7 OLD TOWJN3 CHANTILLY 8 FT. FOOTE4 ARCOLA 9 MARLIN FOREST5 CENTERVILLE 10 WAYNEWOOI)

11 TANTALLON

0 SAIPLE SIZEFIGURE 38

DIFFERENCES BETWEEN OBSERVED NOISE AND INM CALCULATIONSFOR TWO/THREE ENGINE. NARROW BODY DEPARTURES

3-14

1342 61llI

TAKEOFF THRUST12000

CLIM4B THRUST (DULLES DEPARTURES)

-8000 80 a o(NAT IONAL DEPARTURES)

r40001' - A -16000

2

TAKEOFF THRUST

3CIla THRUST (DULLES DEPARTURES)

S8000 1(NATIONAL DEPARTURES)

40 10 2 14 06

DISOFFTAHE ROU SATOFTSOFTOL 10FE

1200CALCLTED THRUST FORLETWO/HREEENGIE, ARRO BODDDE ARTURES)

I 3- 11 a

Figure 3-10 provides information concerning the statisticaluniqueness of Old Town and Centreville. As an example, considerthe empirical noise curves for DC-9 aircraft at Dulles and

National in Figure 3-10. As a result of the stepwise regressionanalysis, the empirical curve for Centreville was selected asbeing statistically different from the other sites at DullesAirport. Likewise, Old Town was selected as being differentfrom the other sites at National Airport. Figure 3-10 suggeststhat the empirical curves for Dulles North, Chantilly, Arcolaand Old Town should be compared with the INM curve for takeoffand Centreville, Ft. Foote, Marlin Forest, Waynewood and Tan-tallon should be compared with the INN curve for climb.

In some cases, as the DC-9 and 737 departures at National, thenoise versus distance relationships in the range from 5,000 to10,000 feet seemed abnormally high in relationship to the INNnoise curves. These observations may be the result of nearbyautomobile and truck traffic noise near the monitor site and arecurrently being investigated. High noise readings at thegreater distances would cause the decreased slope of the noisecurves shown for these two cases.

3M2 ? Arrivals (Two/Three Engine-Narrow-Body-Aircraft)

The statistical treatment of paired differences shown in Figure3-11 indicates that arrival procedures for the two airports aresimilar. These differences are measured relative to the samearrival profile. The DC9 and 727 differences are very similar,with the average difference in the range from -1 to 3 dB. Theaverage observed noise for 737s is about 2 to 3 dB below the INNcalculations. There is considerable variation in the calculatedthrusts, as shown in Figure 3-12. In an attempt to explain

these differences, one must consider typical arrival profiles asdiscussed in Section 2.1.1.

The following paragraph describes a scenario as a possibleexplanation to the thrust patterns as seen in Figure 3-12. Thethrust for Tantallon is low because aircraft are still des-

cending and have not stabilized at an approach altitude. Air-craft are in level flight abeam Waynewood, Centreville andMarlin Forest. Aircraft then begin descent for landing andreduce thrust abeam Ft. Foote to allow the airspeed to decreaseto final approach speed. Aircraft then adjust thrust abeam theremaining sites to maintain descent along the glide slope.

The arrival data from National Airport yielded no valid empir-ical noise curve for either DC-9s, 727s, 737s, or BAC-111s.Using the -egression analysis, the noise level could not be

predicted using slant-range distance.

3-16

120, 120 IDC-9 (DUILLES) DC-9 (NATIONAL)

110 - INM CURVE FOR TAXEOFF 110 INN CURVE FOR TAKEOFF

INN CURVE FOR CLIMBINCREFOCLM

100 >100

EMPIRICAL CURVE FORDULLES NORTH, CHANTILLS EMPIRICAL CURVE

- ~EMPIRICAL CURVE -N.',o FR CMTEVILE ~EMPIRICAL CURVE80 - 80FOR OTHER NATIONAL SITES

500 1000 2000 5000 10.000 500 1000 2000 5000 10,000DISTANCE AT CPA (FEET) DISTANCE AT CPA (FEET)

12 120,

1727 (DULLES) 1727 (NATIONAL)

LIMMN, CURVE FOR TAKEOFF INM CURVE FOR TAKEOFF

EMPIRICL CURVFF FOR CTNILAC

CUR VCURALL NTIONA SITE

FOR ENTEVILLU, KE

0 \ 80______________________I I I IOF I!

500 000 200 500 10000 500 100 200 500 1.00

DISTANCE ~ ~ ~ ~ ;M ATCA(ETLITAC TCA(ET

RIUE31

EMIIACN MNIECRE OTWO/HRE ENGNENARRW BDY DPARURE

3-1

120 120737 (DULLES) 737 (NATIONAL)

INM CURVE FOR TAKEOFFIN UREFOR TAKEOFF

110 110 -

100 " 100 EMPIRICAL CURVE FOR OLD TOWN

. INM CURVEF FOR CLIMB

INM CURVE

8 0-o 90 FOR CLIMB

EMPIRICAL CURVE FOR ALL ...

80DULLES SITES 80 EMPIRICAL CURVE FOR OTHERNATIONAL SITES?

500 1000 2000 5000 10,000 500 1000 2000 5000 10,000DISTANCE AT CPA (FEET) DISTANCE AT CPA (FEET)

120

BAC-111 (NATIONAL)

INM CURVE FOR TAKEOFF

110- /3100- EMPIRICAL CURVE FOR OLD TOWN

SIN CURVEFOR CLIMB\

,- EMPIRICAL CURVE FOR OTHER 'N80- NATIONAL SITES

70500 1000 2000 5000 10,000

DISTANCE AT CPA (FEET)

FIGURE 3-10EMPIRICAL AND INM NOISE CURVES FOR

TWO/THREE ENGINE, NARROW BODY DEPARTURES(CONTINUED)

3-18

AIRCRAFT ARRIVALSTYPE NEL DIFFERENCE, DB

-0 -5 0 5 10

417 1121 20160 30404 i81 5=

DC-9 356 6143 71 2 48

=

13 952 ioin24 ______11 _____ _____

599 103151 2 m200 3115 4C100 5

727 411 6 0

162 70140 8a26 948 10iiiii35 _____ 1 ____

35

233

737 204 6107 7

66 8 =

17 10 :

13AC-111 169 6E392 754 8111111

___________ 141 ______ ____ _ 1

DATES: 3 MAY 78 - 23 JAN 79

LEGEND: 1 ALL DULLES SITES 8 FT. FOOTE2 DULLES NORTH 9 MARLIN FOREST3 CHIANTILLY 10 WAYNEWOOD4 ARCOLA 11 TANTALLON5 CENTERVILLE6 ALL NATIONAL SITES # SAMPLE SIZE7 OLD TOWN

FIGURE 3-11DIFFERENCES BETWEEN OBSERVED NOISE AND INM CALCULATIONS

FOR TWO/THREE ENGINE, NARROW BODY ARRIVALS

3-19

1200 DC-

8000-3

5

2 4 APPROACH THRUST

4000 1U

8

0 I I I I

12000 LNF

3ACP HARRIO REST

3 I

4000 7 3 1

9 08

1000 2 3 40 B 60

DISTANEFRMLANDIN TARELD, FORES FEE

8 5 CENTERIGRE 3-12TALO

C A AC PLOT THRUSTT

4007 u8

TWO/THREE ENGINE, NARROW BODY ARRIVALS

3-20

7.. .. . .. WN

The arrival data from Dulles Airport was much better behavedalthough comparable noise curves were only computed for 727 and737 arrivals. As shown in Figure 3-13, the 727 INM curve is upto 2 dB lower than the empirical curve up to distances of about2000 feet and is then up to 2 dB higher than the empiricalcurve. The 737 INM curve is 3 to 4 dB lower than the empiricalcurve.

3.3 Three/Four Engine Wide Body Aircraft

Data on three different three and four engine wide body aircraft

was gathered at Dulles Airport. Case by case synopses follow.

3.3.1 Departures (Wide Body Aircraft)

The results of the statistical comparisons for wide body depar-tures are shown in Figure 3-14. The average observed noisevalue is between 2 and 5 dB higher than the INM calculation forwide body aircraft. The average observed value for 747s is 2 to3 dB, for DC-10s is 3 to 4 dB and for L-1011s is 4 to 5 dBhigher.

The calculated thrusts for departures resulting from the noise--thrust mapping procedure are shown in Figure 3-15. For the 747example, 747-200s and 747-100s could not be distinguished by themethods used in this study. The