andi f/6 tcu) muuuuuuuuuuu …fad-a094 003 air force engineering andi services center tyndall....
TRANSCRIPT
FAD-A094 003 AIR FORCE ENGINEERING ANDI SERVICES CENTER TYNDALL. AF-fETC F/6 13/2ANALYSIS FOR THE ACCURACY DEFINITION OF THE AIR QUALIT ASS TCU)MAR 80 R J YAMARTINO. L A CONLEY, D M ROTE
UNCLASSIFIED AFESC/ESL-TR-B0-9VOL-2
l muuuuuuuuuuu-EhhEEc EEEinuummmmmmmmm-mmmmmmEEEmmEI-mEEEEmmmmmEE-mEmmmmmEmmEI
III I(I - LiESL.TR*8O4*19-
ANALYSIS FOR THE-ACCURACY DEFINITIONOF THE AIR QUALITY ASESM N MOE
f ~ (AQAM.) AT WILLIAMS AIR FORCE BASE,ARIZONAVOLUME -11. APPENDICES,R. J. YAMARTIND, L. A. CONLEY .0. M. ROTE
ARGONNE NATIONAL LABORATORY9100 SOUTH CASS AVENUE
ARGONNE, ILLINOIS 60439LEL~XARCN 4980
FINAL REPORT r cJULY 1916 -MARCH1980~ TEC
- / E
APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED
UnS ENGINEERING AND SERVICES LABORATORYp AIR FORCE ENGINEERING AND SERVICES CENTERU. TYNDALL AIR FORCE BASE, FLORIDA 32403
.1 2 - 2
NOTICE
Please do not request copies of this report from
HQ AFESC/RD (Engineering and Services Laboratory),
Additional copies may be purchased from:
National Technical Information Service
5295 Port Royal Road
Springfield, Virginia 22161
Federal Government agencies and their contractors
registered with Defense Technical Information Center
should direct requests for copies of this report to:
Defense Technical Information Center
Cameron Station
Alexandria, Virginia 22314
UNCLIASS I I 111-1* 1, A-I1 A1l111,4 0rF fW PA 1 I" ',1-
REPORT DOCUMENTATION PAGE .,.,,,m. : ,
4 T I L I I F,,hi t , R OR 5 PE RIC I L kF\NAIY, IS FOR THI ACCURACY DEFINITION OF THE A F nal Report:
QUALITY ASSESSMINT MODEL (AQAM) AT WILL.IAMS AIR ,lx 1 ', - Mrch !SOF0RCi: bASE, ARIZONA "-.V0l0.t'.7I II. APFN)ICES.
7 A U'HT R' 8 ; N A I-N -- A.~4
R... Yamartino, L.A. Conley, D.M. Rote, I I J7I 7F..1. L[;mich, V.P. Punplv and K.F. Zeller '
I FT R713 MIN , C)C ANI',ATION NM" AND AD )S - AN
Energy and Environmental Systems Division A HE. A 6 *'IRK UNI r N,,MEIArgonne National Laboratory PE: 626O1 F4700 South Cass Avenue 1ON: 1900-90-06
...Argnn Ilinois 60439ICON I ROLLING OFFICE NAME AND) ADD)[RESS ~" RE 9C10P DATE
Narlch 1980P (Al VI.'-c- .:inq neeriml and Services Center
T'mdni II Ailr Force Base, .lorida 32403259
14 MO-NITORIN3 AOENCY NAME A ADDRESS(if ditleAFnt frofF C,,.11,1 (fp ~j IS SEC jRI !Y CLASS. 0Ot thls,,,'1 .'r
Unc las sif ied15. OECLASSlICATION rCOWN-.F'AlN,o
SCHEDULE
16 [rI I RIFIITI(ON STATEMENT ,o, fhi. h'oroi
A:;,rovod for publ.ic release; distribution unlimited.
',7 I; - N T , MEN . ,h,. f ,r.h - ,,r ,nte,.ei , , R ck 20, if ,i,.,,, (rc,,, .po,,l
IE F '.',' I MI_-, TAR NOTE S
Av,. i.11iirv o this report Is specified on verso of front cover.
Iq 0 f Y OR[ ',,,, Iflm ,, .. , ,k U .' t ,,,hI .lfllI 14 h4 -- -)
Ai i uai i itv AccuracyAir P I hlt ion Aircraft Emissions
)is-~ii d OMel; ng
.,L , ,-,,r,, ' .. . ...... .. . ,, i .-.1 I ,dI ,, .0 h, E I ...h 1ie, Air ,ual itv Assessment Model (AQAM) overall predict ivte v ii ratv i
1is jag aCLIt ;A ;ir base ambient air quality measurements. ''tsc nIlitSG 5 ,T r , ,CO, NO, NO' , TIC, Ci14 , and visibility at Will tams Air F-.rc'- Base, Ar i z,,na, I r.r
Ttlt 197ob to June 1977 were compared with AQAM predit ted air pol I ut ionconcentr;it ions to deterine AQAM's predictive power. Tie AQAM acr'tra;cv ir:milv;red on an hour-by-hour basis and statistical hasis using acculil;i ivi.[rvqrenvv (Ifstribut ion. ''lh conclusions are that AQAM accuracy is wel witlii)the arc ur acy range expected for Gaussian urban dispersi(n1 mo1e(hl1s. VTI tI h11a .
DD , .: , 1473 EDI ION Or NO1 ?\6 IS O ,1S ILFETF UNC1.ASSI 1 E)"I .. I.T C .LASSI . _ ATI n,,r T PAIII I l.. ., .
__J ILASEL~* ~Si C.URITY CLASSIFICATION OF~ THIS PACE(W-., DI,. .,,,J ______________
an attempt was made t o~c a.-. iso-i~ed base from urban backg;:okind m-iins*the ba Aground concent ',ILiD11 had! to bh accounted for in the analysif; 0unectn-
trat ions in ttie air base vicinity wertc extremely low when compare71d Wit .1 ICk-ground concen--rat ions rt-SUIr.ing from iz-thz:n transport. Witho)ut.i :tk on
COuICL'tLIoLi on idjustn(nAs , the AQAM mode I tended to underpredi ct the Iu ut ntConcenhirat ionsM. The results also L a hat AQAN is especiallywvsimn1. .m-ing theu potent il worst case bc concentrations, ;Is-oc C at" wI
morning, hours, low wind speeds, stankI, atrxcspheric condition!;, ald~iiactivity.
I N ' 1 11 1)
I * ('it, I - All t I I
PREFACE
Thi s final report was prepared by tie Energy and Environmental SystemsDivision of Argonne National Laboratory (ANL) under contract to ISAF "ngineeringAnd Services Center, Tynda1l AFB, Florida. This voIume containl:; the supportinginlormat ion and data required by the analyses presente'd in Volume I. The workwas aCcomplished Under .lob Order Number 19009004. [t Col Peter A. Crowley,Majors Dennis F. Naugle and Joseph B. Hotten, arid Captai!! Harold A. Scott werethe Air Force project officers.
Contribut ions were made by many others including Mr Edward 1. Durphy,while ait the U1niversity ot New Mexico, Civil EngineCering Research Facility(CERF) and subsequently as a statistical consultant to ANI., and Mr Karl F.Zller, while, EPA project officer.
Northrop Environmental Services provided the ambient measurement data usedthroughotit this report.
The authors are also grateful to the following people for their technicalassistance: Mr John Connolly, EPA/EMSL; Lt Blair Thisted, USAF; Ms Polly Brovn,ANI1; Ms Lvn Deacon, ANI; Dr Kenneth Brubaker, ANL; and Mr Simon Bremer, ANL.
A special thanks is given to Ms Louise Benson, Ms Patricia Traczyk,Ms Sally Vargo, and Ms Linda Wulf for typing this and the numerous draftmanuscripts.
This report has been reviewed by the Public Affairs Office and is releasabi<to the National Technical Information Service (NTIS). At NTIS it will be avail-able to the general public and foreign nationals.
This report has been reviewed and is approved for publication.
IHAROLD ,A. SCOTT JR., Capt, USAF EMIL C. FREIN, Lt Col, USAFAir Oi.l itv Research Engineer Chief, Environics Division
N GS-I 5T7 :1 C- 441
Dep D)ir, Engrg & Services ,DJK TAB
!iborntorv Un e 1 c' dJuL t i;
Dis ., bspc." !1i
(The reverse of this page is blank)
TIAKIL; Of' CO)NTENTlS
Append ix T i tI
APPENDIX A: THlE WILLITAMS AFB AlIR QUALITY STPDY .............
I [NI'RODUCTION .....
2) MONITORING LOCATIONS AND STATION RATIONALE........... . . .. .. ..
3 AIR QUAITY MONITORING STATION DESCRIPTION........... . . .. .. ..
4 PRIMARY DATA FLOW....................................
aAerometric and Meteorological Data................ ..b Emissions Data......................... .. .
5 THE AEROMETRIC DATA........................................
3 Freque'ncy Distr ibu~tions. . . . . . . . . . . . . . . . . . . . . .
b Statistical Summaries.................................
c Cuinutat ivt' Di st ribut ions..............................................
APPENDIX 11: METEOROLOGICAL DATA: PROCESSING AND ASSUMPTIONS.........
I I NT ROTIUCTi (IN........................................... !
M I-ETEDVROLOG;I CAL DATA.......................................
1 TABIIlITY CLASS INDEX AS DETERMINED BY* EPA METHOD..............
SIABILITY CLASS INDEX AS DETERMINED BY ANL METHOD..........
SMIX ING DEPTH, DIRECT MEASUREMENT BY ACOUSTIC SOUNDER............
bDEVEL.OPMENT OF A MIXING DEPTH ALGORITHM............ . .. .. .. .. ..
Al' PENI' X C : SOURCE EMI ,SSIONS I NVENTORY PIROGRAM: I N I!T MOIILCAT IONS
I I NTRODUCIC 1ION.................................. ......
2 NAMEKLI SIT DATA, REASSIGNED) PROGRAM DATA....................3 DA'TA sEFT 4, AIRBASE AIRCRAFT AND RUNWAY TOTALS.......... .. .. .. ..
4 IDATA SET 5, AlIRCRAFT ACTIV ITY............................................
'~DATA SET h, AIRCRAFT PARKING AREAS.................... . .. .. .. .. ..6 DATA SET 7, AITRCRAFT TAX IWAY PATH SEG'(MENTS.....................
7 D)ATA SFT 8, AlIRCRAFT RUNWAY INFORMATION...................8 AEROSPACE GROUND EQlI IPMENT EMI1SSIONS.........................6 7
9~ DATA sEFT I0, AT RCRAFTr REFUEL [NC, SP'I1LAGE , AND VEtNT I N(; TOTALS... .....10' DA I-A SET 1'], TRAINING F IRE P'OINT SOURCES........... . ... .. .. ....
APPENDIX ID: Al RCRAFI O'PERATIIONS DATA.........................
I )I J I,"I IVII..........................................1. BASE OP'ERAT1 INS.....................................
I DATA A(-QUfISIT TON AND HIUICTION OF AlIRCRAF'T ' N; -OE........
TABLE OF CONTENTS (Cont 'd)
Ap pen dix T it le Page
4 DETAILED AIRCRAFT ACTIVITY AT WILLIAMS AFB FOR THEPERIOD 1 JUNE 1976-30 JUNE 1977. ................... 9,
a Aircraft Mix at Williams Air Force Base ............... 92'b Classification of Aircraft at Williams Air Force Base. .. .....c Definitions for Aircraft Activity Operations at WAFB..... .. .. ..
d Edits to the Aircraft Activity Data Base. . ...........C, Full Traffic Count on Airbase Runways ................fTouch-and-Co Training Pli-,hts.................. . . .. .. .. . . ..SSummary Report of Aircraft Activities WAFB for the 1Ptri.'dI June 1976-30 June 1917....................... 102
k %f, f. NCE S.............................. . . . . 1(1
APPENfIX E : ESSAY: VERIFICATION OF SHORT-TERM AIR Q UALITIY MO DELSUSING THE GAUSSIAN DISPERSION APPROACH BY KARL ZkALLER, EPA I 1
I INTRkoDUCTION.................................2 RECOMMENDED VERIFICATION APPROACH....................12()
a Discussion of Data Base ....................... tb Comparisons for Analysis................... . . ... . . . .. .. . ..
C Cumulative Frequency Distribuitons and Error Limits.... .. .. . ..3 Percent Error Distribution................... . . ..... . .. .. . ..
eEmission Submodel Adjustmient.................. . . .... . .. .. . ..
f Model Verification Discussion................... . .... . .. .. ....
3 CONCLUSIONS........................... . .. ..... . . ... .. .. . . ...
AC".JEDMNT..................... .. . . ..... . .. . . . . .........
kRE i<NhI-S . ........................... . .. ..... . . ... .. .. .. ...
APPI I- NuX : STATISTICAL ANALYSIS OF AIR POLLUTION DATA AND THE,VALIDATION OF THE AQAM.................. . . .... .. .. . . ..
I INTRODUCTION........................... .. .. .2 PAST AIRPORT AIR POLLUTION AND MODEL VALIDATION PROGRAMS .. .........
*3 DESCRIPTIVE STATISTICS AND DISTRIBUTIONAL CONSIDERATIONS. ,..... .....4 TIME SERIES METHODS....................... . .. ...... . . .. .. . . ...5 MULTIPLE LINEAR REGRESSION ANALYSIS............. . .. .. .. .. ....b RECOMMENDATIONS..............................0
. . . .. . . . . . . . . . . . . .. .
iv
TABIE OFI, CO)N'1'I.1NT; (Cmi 'dI
Appetid i x T i I ,p
APPFNDIX G: STATION-BY-STATION COMPARISON OF MIFASl:1E) AN)
COMPUTED CONCENTRATIONS . . . . . . . . . . . . . . . . . . . .7
APPENDIX H: ANALYSIS OF HIGHER REPETITION RATE DATA ........... 192
I IN'rRODUCTION ............. ............................. 192
2 INljEPENDENCE OF COMPUTED HOURLY QUANTITIES FROM SAMIPLING FREQ-LL,( Y. 1931 TI EFFECT OF SCAN RATE ON THE AVERAGE VERTICAL SNIND SPLED
ANP THE STANDARD DEVIATION ......... ...................... 198
4 THE STANDARD DEVIATION OF THE WIND DRJ(;TION AS A
FI'NCt''ION OF SAMPLING TIME .......... ...................... .5 MEAN SQUARE EDDY VELOCITIES AND FILTERING OF THE RAPID SCAN DAIA... 20) APPROXIMATE EQUATIONS RELATING VARIOUS COMPUTED PARAMETERS .......... 208
APPENI)IX I: PRELIMINARY TIME SERIES ANALYSIS ...... .............. 23i
I INTRODUCTION ............. ............................. ..2 CONCLI;S IONS AND RECOMMENDATIONS ........ ................... 2
REFI:REN FS ............... .................................
I
V|
LIST OF FIGURES
F uuTitlePat
A--I Locat ions of the Five Ambient Air Quality Monitoring Trai lers ...
A-2 Primary Data Flow....................... .. .
A-3 Frvqu'oncy 1)is t r i bu t ion for Hourly Average NO).. ........ .. .
A-4 Frequtency Distribution for Hourly Average NOX................
A-5 Frequency Distribution for Hourly Average C 4 . . . . . . . . . . . .. . .
:1 _ Frequtenc y Di st ribUt ion for Hourly Ave rage TC ....................
A-! F: eq1CI1C y Di st r ibut ion for Hou rlIy Average CO)...................
A-6 Frequency Distr ibut ion fo r Hou rlIy Ave rage bSt:A' . ...................
A - ) Frequticnc y Dist ribuition for Hou r l y Average N0 2. . . . . . . . . . . . . . . . . . . . . . . . . . . 1
A-lo Frequency Distribution for Hourly Average NMH . .................
A-;I 7requencv Distribution for Hourly Average Wind Speeds.......... ..
A-12 Frequency Distribution for Hourly Average Wind Directions..........l
A-13 Cumulative Frequency Distribution,, of Hourly Average NoConcentrations at Williams AFB: June 1976-June 1977... .. .. . . ...
A-14 Cumulative Frequency Distributions of Hourly Average No XConcentrations at Williams AFB; June 197b)-June 1977... .. .. . . ...
A-15 Cumulative Frequency Distributions of Hourly Averago NO-)Concentrations at Williams AFB: June 1976-Junle 1977... .. .. . . ...
A-16 Cumulaitive Frequency Distributions of Hourly Avecage COConcent rat ions at Williams AFH: Juno, 19 70-Jutnt 1971... .. ......
I Lunul 11kit iV1' "I t'qIIL'n1CY Distr ibut ions, of Ilot-ly Average "t14((jtiC'n t rat Ion.s at Willijamis AFC3: June 1976-.June 1977.... .. .. . ..
I -'tmiu I i t i v t F req tit itc y D istLr ibu titon s o f Hourl1y Ave rage 111Cn oiit r it c ) I:, a t W i I I I amus A FB: Jujie 1976-June 1977.... .. .. .. ....
->4 uimuL atLye Frequtency Di str ibut ions of Hourly Ave rage NMIIConcentratiens at Will iams AFB: J]une 19 76-.Jone 1977.... .. .. .. ...
A--2 hCUirnlat iVe Frequency Dist ribut ions of Hourivx Ave rageScate llg Cue ftit tents at Will i fins AFB...................
6-1 The Behavior of EPA Stability Index When Rosti icted toChanger of One Class per Hour........................................
t)-2 i7he 5~n tof LPA Stability Index for UnrestrictedHourlyv Ci ass Changes...........................................
rt3Q it reunk\ Ds tr i bu t ion o f Ac ousqt ic Sudr esreent y h% H )I
- :r&llocv D; stribut ion of Acoust ic Sounder Measuremenctts by>tn n.
f AcOr Lu. i (- S0111)(101 MVasu1 4d 1. 1d Cit i~ i1 L Vs,;)Z3 I Thf-lrt ii a I Mo)de.I .l i c IC ions lot- J).IytI Ill.- 11m)?...... .. .
Vi
ill)' () I - G1 i UIl ( (oil d)
F' i u re
B- 7 Average Dayt iaw Depth of Mixed Layeri by Nozak i.......
B-8 Mi x i ig, De pt h: A Compa r ison o f Obsetrvat- i ons- vs iTie .rct i alMi..
C- I Po in t Sourzce tLoc at ions.- Re lat ive t o ITax i wav \-, enwaivsand tAub ient Air Monitors................................................
C-2 Pe trolteum St orage Tank Loc at Ions 16-I at i ve t,, !ax i ., 1%,sRunwa's , and Amb ient A ir Mon itors.....................................
C-3 Eva porat ive Hydroc arbon Source Locat ions VkiTaxiways , Runways, and, Ambient Air Monitors..............
C-4 Ta nk Truc k Pa rk i ng Locat ions Re lat i ve to Ta -X i wa vs ,Runways, and Ambient Air Monitors................... . . .. .. .. ....
C-5 V~h ic le Parking Locations Relat ive to lax iways kui~wa\'sand Ambient Air Monitors...........................
C-6 Space Heating Source Locations Relative to Taxiways,Ruinways, anti Amient Air Monitors................... . . ... .. .. ...
C- 7 Non-Aircraft Line Souirce Locat ions Re lat ive to Ta-xi wais,Runways, and Ambient Air Monito-rs.................... ....
C-8 Environ Source Locations Relative to Williams, Air Fore, H.,X.
C _9 Aircraft Parking Areas Relat ive to Taxiways, Rnasandi Ambient Air Monitors..............................................
C- I) Airbase, Tax iway Segments Relative to Airbase Runwav or 1,! K:rs~ t
1)- I Aircraft Sqtiadrons at Willijams Air Force Base......... . .. .. .. ....
D--, 71It iOns of Tax iway Segments and Pad Areas................
-I Cumuul at Ive Fre~pionc y Distribution tor Site -:X dur iniluPeriods kit Y Wind Dire-ct ion............................................
-., lit I ta Conk en t r ot ioiln in th is ExamplIe , 7 7- 31 , o r 46
Time, the Di fference between Observed and Pred i cted waswith in -+ 20) hg/ . . . . . . . . . . . . . . . . . . .. . . . . . . . .
F - Fr rr Limnit D)i; agram..................................
1,-4 Percent [Error Distr ihoition.. ................. . .. .. .. . . ...
F -u P)e1rcent Error - (nmul at ive Frequency Diagram.................4
(;-I ii Cumulnat ive Frequency Distr ibut ions of CO at S)tat ion I..... .. .. ...
G- 11 Cuinlulat ive Frequency Distr ibut ions of CC) at Stat ion 2..... .. ....
G- Ick Ctimulat ive Frequency D)istribuitions o f co) at stat ion 3.... .. .. ....
G-1Id Cu1mil at ivt, Frequu'ncv Distrihit ion., o f CO ;it I t oion '. .. .. . .
0111111 (u at ive Y(eq('(I tw ltu is t ri I uti kDI S 0 a I ;l i ..... .. ..
Vii
L IST OF F IGURE;S (Conat 'd)
go!Ijre Tit Ic e' Pg It
t,- 2a Cu in Ij. a t iv u Frequency Distr ibut ions of THC at Stat ion I .. ....... 162
(,-211) Cumulative Frequency Distributions of THC at Stat ion 2. ....... 163
L,- 2c Cliu ula t iv e Frequency Distributions oif THC at St at ion 3 .. ...... 164
G-2d Cumulative Freqiuency Distributions of THC at Station 4 .. ...... 165
G-2e Cumulative Frequency Distributions of THC at Station 5 .. ....... 166
6~-3a Cumulative Frequency Distributions of NMHC at Station I ........ 167
'J-3b Cumulative Frequency Distributions of NMH-C at Station 2. ....... 168
-k Cumulative Frequency Distributions of NMHC at Station 3. ....... 19
u2-d ICumula t ive Frequency Distributions of NMHC at Station 4 . . ....... 1
(,, m~ CuIla t ive Frequency Distributions of NMHC at Station 5 ........ 171
1-- T- C11 aar iVe Frequte ncy Distributions o f NO, at Station I. ........ 172
C , - ., u ua t ive Frequency Distributions of NOx at Station 2. ........ 173
(,-4c Cumulative Frequency Distributions of NOx at Station 3. ........ 174
C -! - C umrulIa t ive Frequency Distributions of NOx at Station 4. ........ 175
G-4e Cumulative Frequency Distributions of NO, at Station 5. ......... 176
G-5a Frequency Distribution of Log (AQAM II/Observation) for the
Upper ' lD% of observed CO Concentration at Station I. ......... 177
G-51) Frequency Distribution of Log (AQAM fi/Observation) for Lhe
Upper %1O/ of Observed CO Concentration at Station 2 .. ........ 78
,,-)c Frequency Distribut ion of Log (AQAM II/Observation) ter the
Upper %10% of observed CO Concentration at Station 3... . . . . .....
C-d Friqiency Distribution of Log (AQAM ll/Observation) for theUpper -,10O% of Observed CO Concentration at Station 4.... .. .. . ...
t requency Distribution of Log (AQAM 11/Observation) for tile
Upper ,10/ of Observed CO Concentration at station 5. ........
-8ka Frequency Distribution of Log (AQAM IT/Observation) for the
tipper ,,10% of Observed NMHC Concentrations at Station I ........ 182
6-6b Frequency Distribution of Log (AQAM IL/Observation) for- the
Upper %lO0% of Observed NIIMC Concentrations at Station 2.... .. ..
6-c Frequency Distributicn of Log (AQAM TI/Observation) for theUpper '10% of Observed NH-IC Concentrations at Station 3. ...... 4
.. d -roquencv Dis tribut ion of Log (AQAM I 1/Observat ion) for 1_11<
::)er 4G'l o~f Observed NHMC ConICFentrat ions -it ')'[;it jion 4................5
;it r ribhit iii of 1,(o;' (AOAM I 1/fths.'rvalr I'm) I')? 1wi
r' N II -i t r I-i
..
...... 1
C-7 Frequenc y 1) i s r r ibu t i rn ot I o _ (AQAM i: .
'ppei s10 ot O)bs -rvn1d No. Colc , n! t.It ... . . .
(-Ic Vi--qit, I ,-v PIistributioni of Log (AQAM ,'C, b 'c.> , t
t pp r Jr1 (I. of Obs rved NO, Concentrat (ion at !t ' . . . . . . .
G- ;d tr,,q hlem 1)ist r ibl . nn i.,)g (AQA1 !I1 /fb , i . t t,:
Upper PlOZ of Obser-ved NOx (oncentrtia on:, Lt 5-. i . . . . . .. . !'Y)
G.-)h > Fi eqoe ncy Di str ibut ion t Log (AQAM : l I r:,Upper .f1);, of Observed Ni, (:ncet rat i1n .a _ ' .. . . . . .
H.- I J ltn _ion o C:ordinat .s . . . . . . . . . . . . . . . . . . . . . Z
H -2 Svn thts iztO Three-Stage, Biutterworth, low-fa.;,; Pr,A,>: 1 p,
Values tt Componts Chosen to Givo f: - 7 1.7o r T C U m I . . . . . . . . . . . . . . . . . . . . . . . . . . .
1l-3 Frequency ke'spolln-w of 3-Stage Low-P;aos tultni i, . Itr..........
;t-i Filtered Wind SprIk d Ifur Day 146, Hour t650-i/:;( . . . . . . . . . . .
H-5 F itered Wintd Dirct ion for Day 146, Hour 1650-1 7/ 1 ..... . ....... i
i-C Fi tred Wind .Spctd fnr Day 145, Hour 1200-13o ... ............ I
H-7 Filttred Wind Direct ion for Day 145, Hour 12.(0- I 300. ........ '.
i-8 ltlated Eq. (H-31) versus Observ, d on: -i H<l 7 V aiu-. ...... 2
H-9 Calculated !:q. (0-31) versus Observd (1 00 A,\.PI ai:
S i fillt -e V;I . .. . . . . . . . . . ........................... .
'-10 toI:a';oted Eq. (H-31) versis Observed 00: Militt V'In............
i-1I kalul ated Eq. (1-32) versus Observed 0: Av ,ra.:,
,! Ni ut " VlI I............ .............................
H - l .j I ul t >d El 1t1- 1 ) versus Observd 07 : A,v r Yan
t 1 I i e V ;! I i,, ..a. . . . . . . . . . . . . ..
- ' l k' Scric.. p,,r O t, rv,,I () (Snl retil r t - lt ),
,!! ,I ) t I,!",lu il t A u v -l 1> 14; I t)' . . . . . . . . . . . . . . . . . . . ..!
1-.' cq(wizet n V "Siw, I ruin to - thl: T imfl St. I- i'e s of bs,-rv- ,
i tit! i ' ; t <t t 1 2 ............ . . .................. ...
i-, uittgrot.tl S t .Ltra. Uensity for (tbsorved 0) (, ,. Crt'ins,t. 01 i. . . ............................
1-. Iime Str i es f i)r Obse_,rve d NOx Coren(,nt rat- i o , I t t 1tt ('n 2
Oiriny, Aug t. 1976 ... ... .. ... . . . ...................... .
I F roeuqui 'y Spe I.rum t,)i- tiwe Time Sori, of Obit rv, 0 ; ? ..
(nirelte rat inns at ta nti(,n 2 ..... . .. .. . . .................. .
[-fl C(IwiIre ic,' V ru! ; .I. i od ic ity o t th " Stil jbai :1 (1 -- I v, i (:,
nid N(Ix Iim, Sor, in . . . . . . . . . . . . . . . . . . . . . . . ..
i×x
iLisTr OF FIGURES (Cont'd)
,U r e Title Pare
1-7 Time Series for Aircraft CO tEmiss ion Rates on Al I Ground-Lef-vel Lines Based on Actual Hourly Aircraft Operat ionlS!or August 197b..........................................242
1-6 Fretquency Spectrum for the Time Series of Aircraft Co
Emi~ssion Rates . . . . . . . . . . . . . . . . . . . . . . . . . . 243
1 - 7 CoherkenCe VS Periodic ity of the Observed CO at Stajt ioll 2and the Aircraft CU Emission Rate Time SerieS ................. 24 4
Y.
LIST OF TABLES
T.bh Ic Titk'
A-I S i ng Rait imil . . . . . . . . . . . . . . . . . . . . . . . . . .
A-2 Aircraft Activitv Data Format . . . . .. . . . . . . . .
A-3 Mii nimum , Max mum , and Threshold Concentrat ion Values .... ........ 27
A-4 'Altiams AFB Data for the Period June 1, 1976 to June 30, 1977. . .2
A-5 Williams AFB Data Summary for the Month of June, 1976 .......... 2
A-6 Wi 11 iams AFR Data Summary for the Montl I July, ifJ........76 ......... 3)
A-7 Will iam s AFB Data Summary for tht Month of Au;,,o 13Th. ...... 3
A-8 Will iams AFb Data Summary for the Month of Sept.imbc r 1976.. . . .
A- t ) Wi I liams AFB Data Summary for the Month of Octobr, 1976........
A-10 Williams AFB Dat a S i immir y for the Month of Nov ember, 1976 ...... 3/,
A-I1 Williams AFB DaLa Summary for the Month of December, 1976 ...... 3
A-12 Will iaims AFB Data S1;ummax,' for the Month of January, 1977 . . . . .
A-13 Williams AFB Data Summary for the Month of February, 1977 ...... 7
A-14 Williams AFB Data Summary for the Month of March, 1977 ........ ..
-5 WIill:ains AFB Data Summary for the Month of April, 1977 .......
.\-,t tiilliams AFB Data Summary for the 'Nonth of May, 177 ... .
A-17 Williams AFB Data Summary for the Month of Junc,, 1977......... .-.
A-I8 Comiulative Frequency Percentile Concentrations (PPM) for
WA.PB Hou ly NO Data: June 1976-June 1977 ..... ............. .
A-19 ctumulative Frequency Percentile Concentrations (PPM) forWAFB Hourly NOx Data: June 1976-June 1977 ... ............. ..
A - '0 umo lal ive Froquency Percentile Con centrations (PPM) fur
WAFB horly NO12 Data: June 1976-June 1977 . ........ /
A-2I Cuimu lat ive Irequency Percentile Concentrat ions (PPM) for
WAFB HOUrl ' C( Dat a: June 1976-Jone 1977 . ............. .
.\-2'2 Cumulati.ve Frequency Percentile Concentrations (PPM) for
WAFB Hourly 14 Data : June 1976-June 1977 ... ............. ..
A -2 3 Cumu at iye Frequency Percentile Concentrat i ons ( PP) for
4AFB Hourly THC Data: June 1976-June 1977 ..... ..............
A\- Z4 ,KumIat v/e Frequency Percentile Concentrations ( I'M) f,,rWAFB Hotrly NMIIC Data: June 1976-June 1977 ... .............
A-25 Cumulat ive Frequency Percent ile Scattering Co.t f i ionts
(x 10- 11 for WAFB Hourly bSCAT Dat D ..) ..... ..................
I- I To rner Req ui rtment Relat ed to Sky Cover ..... ...................
?-2 Frequency Table ot Sky Cover Codes . . . . . . . . . . . . . . . . .
xI
. ..N. .
LIST OVA IABiES (Cont 'd
Tab! 'I it I " Page
-3 Insolation Classes as a Ftunct ion of Solar Alt Itido
*or Cloud Cover < 5/.0 . . . . . . . . . . . . . . . . . . . . . . .
i-4 Stability Classification Criteria ...... ................. ...
B-5 Acoustic Sounder Classification Codes tor Wil I Iam; AFB Study. . . . 8
C-I Engine Data (SRI) ............ ......................... /9
C-2 Engine Data (Update) ............ ........................ 79
C-j Air(raft Activity (SRI) .................................
,-4 Ai r raft Act iviLy (Update) ........ ..................... .. 1)
- A I icratt l-arking Are.. (comit r ies (SRI) ....... ............... 80
(-C Aircraft Parking Area Geometries (Update) ..... ............. ,
- Taxiway Segment Geometries (SRI) . . . . . . . . . . . . . . . . . .
('-s raxiway Sgment Geometries (Update) ........ ................ 82
Ainia. Aircraft Arrivals and Departures (SRI) 8........... 3
10 Annual Aircraft Arrivals and Departures (Update) ..... .......... 83
-1 NAFB Runway Geometries (SRI) ......... .................... 4
-2 ".'AFB Runway Geometries (Update) . . . . . . . . . . . . . . . . .. .
- Aircraft Movement over Taxiway Segments ..... .............. .
I' Service Vehicle Emissions in Kilograms per Operatim
(Arrival or Departure) ........... ....................... .
1 5 R,:fIeling InforIriat ion (JP4-Jet Fuel) .......................... nI
Ic oi o f- Aircraft Olprat ions ........ .................. 9)
, T i-ir -Mode Data ...................................... 9
-3 Aircraft Activity Data Format ........ ...................
Monthly Summary of Aircraft Mix at WAFB ...... .............. 94
0)-5 Classification of Transient Aircraft at WAFB ..... ............ .
D-6 Edits to Normal Operations Data at WAFB .... .............. .
L-7 Edits to Deviation Operations Data at WAFB ...... ............. 1)
[I-I Average Wind Direction (0 - WD) ...... ..................
H-2 Vector Mean Wind Direction (4 = VMWD) ........ ............... 223
H-J 3tandard Deviation of the Wind Direction ((P o 0 ) .......... '..24
H-A Average Wind Speed (4 = WS). ............................ 224
if- 11, , t ,1 " If,; 1 Ij I ,(,] d (€ VMW '1 ) ." . . . . . .-ftiflO r1 0,' it 4' d,,. 1w Wi rid )................................... .
t1 - 1 I . ly It ,.r I ,, 1,', ( '. :10 )1 . . . . . . . . . . . . . . ... .!.ii,
Al
T. , T I h I''u ,
11-8 Honurly Average N( x ( = Nt x) .. . . . . . . . . . . . . . . . '2
1t-9 Standard Deviat ion of the Vertical W i , S Ied ( ., . . . 22f,
H-I) Average Vertical Win edSpt, (e = . . . . . . . . . . .. . . . .. 226
fl-Il Calculat ion of p for Sampling Times o f 3 min and 60 rin ... ...... 227
11-12 Calcnlat ion of p for ;ampl ing Time:; , 1 10 wi, :1 d tr( m :,1i; .. . ...... . Z7
1H- 13 Calculat ion of p for Sampi ing 'lime,4 nt '3 mi .,n; D, mi n .. . . . ... 22
H{-14 Comparison of the Actual 1 0 to the o 0 Value i Il at-!
from Eq. (H-1l) using TS=3 min ......... .................. 2 2
li-15 Comparison of the Actiual 30 to the. no Valu- Ca, 1 cuI ait -d
from Eq. (:-I I ) usiu; t[ 10 mi n.. ........ .................. 22;-
Ii-I f' Mean Square Along Wind Eddy Ve lotc it y () .'7 . ... ........... 2 z
11-17 Mean Square Cross Win(d Eddy Velocity ( v-7T ) .... ........... ..
11-18 Mean Squart Along Wind Eddy Velocity (D = 'Z)
Computed Using a Fixed Mean ............................... 2
H-19 Mean Square Cross Wind Eddy Velocity ( -
Computed Using a Fixed Mean ........ .......................
ATTACHMENTS TO APPENDiX )
A Htourly Aircraft Activity Summary; A One-Day Example, 76-08-06......
B Monthly Aircraft Activity Summary; 1 June 1976-30 June 1977 ....... -
C (;rand 'rotals of Aircraft Activity, 1 June 1976-30 June 1977 ........ .
(T'he, r,,v ~rs, of th , i hl.lk)
..... i ... ............. • . .... , .. ..... , .......... .... A
AlPP.1ND X A
TilE W I ILI A !! APEB A I I? IJAJJ; 1 c 1 UI(
I I NTrRo DUCT IUN
Th is a ppend ix begins wi th a br ie f dec rip Ic p 01-1- ti h-i ham;- . ki \j.r imentt and conclIutdes wi th th e presentat i on o f s tlI is S :al I sulrlia rl e - f the dant~
acquired during the 13-month monitoring per i e big inning June 1976. A ( ),is ic( erabl1y more ext ens ive review of th, exper ince-n toL pe rat iens i S'i11 1 a I no.(in the paper of Sheesley et al . :A-1 a report is~c Sny III det r;lt~Services, Inc. and the IISEPA Environmental Monitortinp_ and 'SupportLabrty
2. MON ITORING LOCATIONS AND STATION SITING RATIONAL 1 -
The placement of the five ambient air (pialitv incnittriny, tra; V-rb;showni in Fig. A-i .* The rationale for stelecting, th-ese trailr h*
provided in Table A-I.*
3. AIR QUALITY MONITORING STATION IESCRIPTIONA-2
Each monitoring station consists of a mobile encl-sure4 (trailer) aprxi-miately 2.4 m wide, 4.3 m long, and 3.1 m from ground to roof. The V1i;i
Monitoring instruments are rack-mounted in each trailer:
1. Beckman Model 6800 Air Quality Gas Chromatograph for-analysis of carbon monoxide, methane, and total hvdro-carbons. The Beckman 6800 has an automatic calibrationcap.ability.
2. Mon it or Lab1,s Model 8440-RI ER/DA Dna 1-Gbamber NG / o xAnalyzer wi th te finn filter , flIOWmi-ter , rscybdon uin NO(2 - to-NO converter and 0.05 ppm full scale lowest rangef o~ptio)n.
3. B. M. Young Co. Model 35003 Gill Propellor Vane for winds peed and direction measurements.
4 . MR] In tegra t ing Nc'phelIotte ter that wil1 be used ais a me a ~cirkof 1 iglt scat ter ing due to part iculates.
Each trailer also has an Ester line Angeus 1,11-026 qt ri ( bar t 1a
rtecorder for cal ibrat ion and back-u p pur pa ksc'
The intake system for the above air qua lity v jst riim~eetcz inc 3c,".glass hallast chamber (no( in system at present ) t) cc puciice . ~I aril. iriresidenco rin, (ratio of volume to flow rate) (it at lattw[u,. III i.1ts VL tv(iii int-('rrogOt ion rate . The chnimbet w ill al I w het( avi :3;' un' I, tt cc; 1 ethii i g cooc en 1*o ions ofI po1 i t ant s that in iglit cct he iw i pa.s ccucdchSe rvc ci. 11add it i.,, the Bec kman 60(00 hla! a seJa rat e i~~~I1' iitii 3 to allo w 1,- c h
*F.igiircs andi tables appear consectit VIIV *c! the enTd ocl fil,, apl-e1ix.
-- %NN
five-minute instantaneous sampling rate. A separate sampling train is
provided for the MRI integrat iLig nephelometer so that the aerosol sizedistribution will not be distorted prior to entry into the nephelometer. A"T" was installed in the air quality sampling train so that, it necessary, air
cool] be sampled from heights other than the fixed entry port several It.etabov,, the roof of the trailer.
4. PRIMARY DATA FLOW
a. Aerometric and Meteorological Data
During the period June 1, 1976, to June 30, 1977, the U.S. EPA/Las Vegascollected reliable data at Williams AFB on five pollutants (CO, CH 4 , '[IC, No,NOX ) and b-scatt plus wind speed and direction iron the tivt-station netw,,rkdescribed briefly above. The raw data, collected at one-minute, intervals, wascorrected for span and zero drift before computation of the hourly averagequantities required for the AQAM evaluation effort. The procedure for produc-ing the hourly-average tape (DS Ill) is indicated schematically in Fig. A-2and described in detail in the report of Sheesley et al.A-I
Additional meteorological data, taken routinely by the U.S. Air Force atWilliams AFB, was recorded onto coding sheets, keypunched, and stored on
magnetic tape. This data set also includes mixing-depth measurements taken bythe EPA with an acoustic sounder. For logistical reasons, this data set wasnot merged (as erroneously suggested in Fig. A-2) with the aerometric d.,ta.
The format and information structure of the meteorological data tape, alongwith details of the acoustic sounder measurement program, are also given by
Sheesley et al.A-1 and further described in Appendix B.
b. Emissions Data
Figure A-2 also shows the essential steps taken in the creation of the
source inventory data set (DS I) and the detailed aircraft activity data set(OS 11). The source inventory data set was created by running the AQAM Source
Emissions Inventory Code on emissions data for Williams AFB, recently compiled
and coded by SRI and checked for consistency by the Air Quality ResearchDivision of USAF/CEEDO. Modifications to the emissions inventory are de-
scribed in Appendix C. The aircraft activity data set (DS 11) contains
hour-by-hour tallies of aircraft operations (i.e., takeoffs, landings, touchand goes, aborts, and transient aircraft arrivals and departures) broken down
according to aircraft type and runway number. This data set serves as inputto that version of AQAM (i.e., AQAM 1), incorporating the actual hourlyaircraft emissions as opposed to the version (AQAM I) that uses the tractional
,ipi'-rtionment of total yearly aircraft operations dictated by the AQAMSourze Emissions Inventory Code.
teneration of data set DS II was accomplished by processing intormationDr. eac. individiai aircraft operation. This information, extracted trom HSAF
Flig. t Forms ATC 355, Ac 90, and ['4 124 by USAF student pilots rod CVIt.)AS wa; -,,J, I and ., yprilh,.d ,rcotrdinw -, th.. h, tn. ' I,-" I
- /
I .i l I A 2 . I h l l r I t - , l t .1 y u l L I th , d 'j t a it ..,t i t , t o p r , t - ' ,. r o f
Ii)t;ttii.l in :iv:i lab l- ;l .AIt I I igl ,,fi ,,. I,. , I, i ni, .s 1,r"1;:? a !
po' i nt,,d (t l t by Ijun( all rid ln ph y , A-1 L ('qu d I l I ml, 1 -t i at on . t-
s impI i Iying assunipt ions; none of whi .:1 art prfesently tho'l4t t,; ni4 i&.:a
alter thie space-t imt distribution ot aircratt rmissi o s . These d *ic if-:11 e-
and approximations are discussed in Appendix P.
5. TH11E AFROMETRIC DATA
a. Frequency Distributions
As a first step, preceding more dtai e, ann yss, eli,. generallv :xaV Ii-.,
the frequency distributions of the meoasured quait tiea . V I..!l inspectii ,
skich distributions quicly gives one an .st imat: ,f the mt,.n, m o n<ian, 5tor,
deviation, range, and statistical nature (e.g., Ttml , log t u l) ,
measurables and occasional ly reveals systematic prob lIems ia the! data. ?-, r
example: an uncliaractorist ic bimodal snape in t he stt io0n 2, Oc_0) ofer t. UiI
tribution led to the discocery of an instrument calibration problem covering; a
two-sweek period. Though ti-e pollutant distributions for each stat ion ano
month were examined during the diagnostic phase, the distributions present,..C
in Figs. A-3 to A-8 combine all stations and months, as the subtle differnc,-.
between stations and months are shown more clearly in the statistical tabula-
tions. Distributions for the "calculated" pollutants NO2 (z NO X - NQ) an'
NMliC (OTHC - C.l 4 ) are given in Figs. A-9 and A-1O, respectively.
The meteorological variables, wind speed (Fig. A-11) and wind direction
(Fig. A-12) were also investigated in this manner. Of particular inter-st is
the greatly enhanced probability of winds from the ESE and West. Subsequent
analysis of this phenomenon revealed a strong correlation botweon wind dire,:-
tion and time of day. This strong diurnal dependenc, oi the wind dire.L: i:
(se, Fig. A-12) is characteristic of the mountain-valley flow feuild in o
Valley of the Sun, where Williams AFB is located.
liso t.ulness of the frequency distribution extends be'ond the ii -n;i
phase to tle analvsis phase. For example, a detailed !it of tue ,) 11l W, ;
frtquellcV distr ihit ions, assuming only a posit ive concentrat ion di!ir ib!t irni
corivitluted with a Caussia;n instrumental resolution funct i )n, shoul d di s, lo>se
the overal I accuracy, j , of the NO/NOx instrument near threshold. Visual
.xajlmin,:t ion of the "nOgat ive concent rat ion" tail of t,, u - ii stribut ions
su gge st ; a ,o o t a few ppb.
1 Lnt rument thre;ho 'd and repeatability noise mightr also h(, i nai atel f,:,
frequency distributions of the data. If a significant proble:mn, the r,,,eat-ability noise resolut ion function could bc folded into tie, experimental dat a
to produc, ireq tiency ideograms.
b . St a( ist ical Smmaries
B'I rt prI'sent i1); a ,;tatist C-e l s,,tiIarv (t t , I I t i i ' I, , t, I I
ml t d' t i no t hi ra ng . o I dat a v a Iiis t it at e 1 t pt - I T , r 1 .' '
3
allowable given the instrument range setting, instrument threshold and accu-
racy, as well as "data windows" used in the program that generated tile hourlyaverage data tape (DS III). While some discussion regarding the selection of
the "data windows" is given by Sheesley et al.,A - 1 selection of threshold
concentrations CT (such that CMIN < C < CT are set equal to CT) was based on acombination of conversations with the experimenters, instrument manufacturer's
specifications, and a strong desire to avoid taking the logarithm of a number
(n,-cessary in the computation of geometric means) too close to zero. The
a-minim'im, maximum, and threshold concentrations for measured, as well as
:onpttrd, pollutants are given in Table A-3.
Statistical properties of the entire 13-month data sample are given in
Table A-4, and for each of the 13 months individually in Tables A-5 to A-17.
In these tables, the geometric mean and geometric standard deviation should he
considered meaningless when the arithmetic mean is negative.
c. Cumulative Frequency Distributions
Cumulative frequency distributions for the 13-month data sample are
presented (Figs. A-13 to A-20) in logprobability format for each of the
pollutants. The interesting characteristic of the logprobability plot is that
a lognormal distribution transforms into a straight line that intercepts the
fifty percentile point at the geometric mean and has a slope proportional to
the geometric standard deviation.
These figures relate to equivalent tables (A-18 to A-25) at the end of
this appendix, giving the concentrations at several useful cumulative percen-
tile levels for each pollutant species.
REFERENCES
A-I. Sheesley, D.C., S.J. Gordon, and M.L. Ehlert, Williams Air Force 13a.,.r
Air QuaLity Study, Northrop Services, Inc., Report ESC-TR-79-26
(June 1979).
A-2. Zeller, K.F., and R.B. Evans, Airport Air Quality and Its ContreX,
J. APCA, 15 (Dec. 1965).
A-3. Duncan, D., and E.P. Dunphy, Acquisition, F,.>-iuotion, and Audu,:- .
Aircraft Activity Data from Williams Air Fornce liaec, Ari-zonm, letter
report EE-ll, final report to AFCEC/EVA (for period 8 March 197u-
30 June 1976).
4
-7
-.u
0n
Ko
00
E0
93 I 4)C
0 a'043 4, I)43C V
O0 +ra 0 0 0 ~
4-) J
o4'00 0 FZ-:c,
44,ac
a) ___ ____ 4_
'--4 -Rump
-oo
07f*
-z oizo~o
r.r'lf
4.
CDC
UMSA3C .3*O
01*0
60.0
-4-
4-44
E- -
00
t'0*0
000
-10,0-
SSLNSAH JIO 'ON
0 1:
___
z -
-('C
-0?
90
0l N _ HA0 O
raa
'0
101
Ufto
<CIO
cc1
z Vz
2ro
O'0
C)C', -:
C) -C--n 0
SINSA3dO *O
1-4
H a't
0t0
0
C)
C) )
s-4
00
do0 -a
0-0
UNRAS dO *O
-60,0
~J ~z80,0
90*0 Z
CL
00*
I 000
00*0
&LNHA3 dO 'ON
13
8.1
091
010
zo
ow
CL
-0,0
f,*o -
9,
SJNHAH 4 'IO0N
S -1 T t t 0'01
iz
-06
-,-/a- -0'8
Q'4-
r- -~0 -Ij_
z0 Ij
0.1.
I .'- I
0 __
-01'
-- of-
= 1 iI T1 . ..1 t -'0"0
SJJN3Aa JO "ON
~~i TT1r
H aN
4-J
0,00 hoof: (
0-og2!
> I
0~ 0.0oz1)
-0,001
0,0
&LNHA340 'O
TRAILER ONE TRAILER TWO
S1,o 10o1
10 -, 10 •01O
PROBABILITY " PROBABILITY
TRAILER THREE TRAILER FOUR
10 1071
3.10 .0$
t o I , t 1 I . . .I I I . .i . . . . ' t I I T II
,- J.u. • . . • • • • • .
PROBABILITY PROBABILITY
TRAILER FIVE0-1
Fig. A-13
Cmulative Frequency Distributions._ of Hourly Average NO Concentra-
3.10 tions at Williams AFB: June 1976 -June 1977
PROBABILrY .
17
TRAILER ONE TRAILER TWO
100,4
t1o
PROBABILI1TY PROBABILITY
TRAILER THREE TRAILER FOUR
10-' 10-4
PROBABILITY PROBABILITY
TRAILER FIVEFig. A-14
Cumiulative Frequency Distributions of10- Hourly Average NO xConcentrations at
Williams AFB: Time 1970 J ~une 1077
i L t -
PROBABILITY
TRAILER ONE TRAILER TWO
toIi -- r I~. to-I
!7
C10 a.10-7
10-9 ...... //o
I ,I./
/
PROBABILITY PROBABILITY
TRAILER THREE TRAILER FOUR
to:'
C. 10 01
//
~/
101
PROBABILITY PROBABILITY
TRAILER FIVE
I C~~:mnul1at ivc I'rc, l1ecc ~i~t rih~t,1 '
W~~i l iaxcmig XF): (;tz(lT l,,-.tn. r] t
; 10-I ,-, , . .
PROBABILITY
1 9
TRAILER ONE TlR A 11, ;IJ T~WO)
10 07
v-g gggog g 12 g2ggPROBAB[LITY PROBABILITY
TRAILER THREE TRAILER F4OU R
A. a.
0-
gggogg g 99 9 999PROBABILITY PROBABILITY
TRAILER FIVE
Flig. A-10h
10t
5100
101- TRAILER ONE T1R :\ I I 11PI1 TWO
PROBABILITY PROBABILITY
TRAILER THREE
10' rpj ) A i 1 1) [ -111,1
PROBABILITY
lo' TRAILER ONE ao- TRAI11I,.R 'I'W()
0. 0../
10* hh . .h . .. . . .10" 1"" itI i I ' ,
PROBABILITY PROBABILITY
10' TRAILER THREE TRAILER POUiR
0. . .. ' r • + 0... .. . -- -
Ng- g g aPROBABILITY PROBABILITY
TRAILER FIVE
Fi g. A- 18
Cumulative Frequency Distribut iolns of0. Hourly Average It(C Concentrations at
Williams AF: June 196-,June 9--
iO~ ~ TT T
PROBABILITY
2 ?
TRAILER ONE TRAILER TWO
t. 0 -t
10-A 010~
PROBBIMYPROBABILITY
TRAILER THREE TRAILER FOUR
100 10Cu I Itiv - n lc, i riI'I
10r 0
T10 -- 51
PROBABILITYPRBILY
TRAILER ONE T"RAILERi T1W0
X10 L 10
gggg gg- g gPROBABILITY PROBABILITY
11 TRAILER THREE lot TRAILER FOURI
lo-44_r__ 7
" gggggg g gggPROBABILITY PROBABILITYg
i'. TRAILER FIVE
Coo-CtUMuLat iVo 1requencv 1~ist r i bt I c"s kv'Hourly Average Scatter ing (oe H ic, i ,,I tat Willii ns AdiS: JUneC 19)-0 JUDOI.
PROBABILITY
-No
li, oh . A-I . Forl]r; en' .Ioj,
Sit e No. ia i r;ti I
a. Continuity with April o . I .
b. Upwi nd during (-arIy and I.- t i::,;T , ,. ,
c. Power already installed tur Apri i s lWicy
a. Downwind of T-38 takeoff an t,,ni . ,, V-cat ior
dur ing the morning pe riod L ,ai o 1s
and 11ndi ngs at Williams AI:i., 0, NW)
b. Downwind of T-37 and F t akcut s o-i te.-
downs and taxi route for ali aircra!t duringAafternoon
3 a. Downwind of T-37 qutuinig and tak': ff ar, aduring the morning period
b. Downwind of taxi segment for all aii.,-aftduring the afternoon
4 a. Downwind of T-38 apron area during the. m,,r-)jn;,
periods
b. Background to airfield operaLions ,hrIc.~ II.,
afternoon periods
5 a. Downwind of all airport activiti,'x during ti'-'morning period
h . Downwind of taxi to shutdowni tor 5
c. Background location for later a:ti ro,.onearly evening
Table A-2. Aircraft Activity Data Format
Header Card - One Per Day
75122SC 275 327 129 14 5 9
Col. Content
1-6 WML)) = Year, month, day
- C => Total aircraft counts
8-11 30L
12-15 30C
16-19 30R Full traffic count on these runways
20-23 12R Right justify integers
24-27 12C
28-31 12L
Aircraft Activity Cards
751225N30LF-S 0753,0804/0800,0843/0810,09O5/0812,0837/0845,0906/0915,1017751225130CC-130 1530,1S35,16107S1225030CC-130 1655,1702,1730751225\30LT-37 0824 ,0834/0838,0855/O8OS,O859/084s,OgO2/O8SO,0903/1250,1310751225A30LT-37 0620,0(20,1410,1600
Col. Content
1-6 YM D
7 N => Normal operations - times are local takeoff and landing times.Slashes separate sorties.
I => Transient inbound Times are in NF (i.e., :ulu)
0 => Transient outbound )V => Deviation - times are local engine start and takeoff times.
Slashes separate sorties.
A => Non-weather aborts - times are local engine start times
8-10 Runway
11-20 Aircraft type - left justified
21-2426-30 \ Local 24-hour times separated by coinis or slashes as
appropriate. (Zulu time for transient aircraft.)
76-79
2 f,
Ta'oh A-3I. lin i , Mai m mi, 1: d i r.2
M~ce t ~ lll rat in Va I o
Po 1. lutant tMlN ) (C.\X
NO -0.02 O.,.
NOx -0.02 0.3 0.002
C 4 1.25 6 ., I. J
THC 1.25 0.0 1 . 3
GO 0.0 8.0 o.0
b S CAT 0.0 9 9, 0(). j b
NO2 NOx - NO -0.2 u.5 0J.002
ptis CMIN and t-MAX , ,_s om N, 7 .:rnd ';(,
NMtIC '1 - Ct14 -0.2 8.0 0.05
pls CMI and CMAX cuts on ii(. and ( :
aValues given in ppm unless - tiierwiso indicated.
bvalues given in units of l0 4 m- 1 .
kd-,
It In In In InC I
'4: *'~~ ('' . ~ 4 JC ~ b.**)-* c.) C.) LI' I , I-
, j c. I j n- In I n 1)cI
'r U ... ,.ii U? iiL JL'JL r IJ -I
I'), (4 I -. 44140.lv C~II.4 L~CI'I
- CJ '- J -- c -l -I - I'
cLIj n I I ) c) rA I In Inllt. (,j)
I - - . I I
.fl )(JJ If'- ui4 nI n r i~l
LU C ' () I I.A UI .1 -I LI, ..l. In.L!4t.,) (..II' JL~l LL I
Lf Wt'1 fl'.t Cq c j $44)'Ccj aC -c s,ncj r. - In- j, In' j
InI-n j1 .01. InI' ''.) * - j-In l.C f.) (A)' 11I I 1) IncC' I n11 -4C -1 nI In'I. I. I-,0U1CIflc~~~~~i'l~~iC --t~. -tn 0n" J -.. - 4.'I A.'II.)( L (
I"-jIICJCI~cp. o~aicJ.o - -'C-C-r.- '-C-OcJ-I- c3--nC-7.-
LCC. . ' 1 l CC UL CS ) I)J ) U 0 . OCII., O OCL L L)U. -I ')(I.,.-
-- (z- -Inj .
- I 4I l - 1
I.~L4 I) . I~ l~IJI u) w U ~ .I) (A Iii ,r. nLU f I r IIL.,DLl w u1 7 >-4 I l cl Lu TJLI 7
;- 7-l.L Lf '1. 1J.J D '. . ) 4) J.0 '4(U: C--WO4.4.~ 1 I I0> >I'C L >>(I.C- 1 > > C~J- ') C .1cC I
I, InQ z m Xo L In 1. 'I : .1. 4.Ir mo l
(n L - In (I-CC H 2 '-I'-
S
* - '. '~5,**~r ,' . .. ,.
- .. . '5- C-'5 .5.>!,'- C.. V ~ -.2'
-'5* 5
C-'5..' (5.
(35 CS
5'55555 C''5
C, Cs I- . 5.
S * (5
'-''7. .5'' '' **,555 .5
cc .-- 5 C'' . $((('5.251 (.*fl '5 . .
'4, 7'" 5
- InJ'~ ('jc-''~ ''S -'s's:st'~'5 ''5 ('1.5 -5 - '.
"-5.5(55~1 5.55 -o . '.5"' 5!', --. -
5*''- 5~.55.5'~55~ .55'' - -.
.5 frt(5''5-. I t'~ ~ .'' I-., '5'-' 5.' . . -'.2 *fl*
-r
Sn
'C In
-c(13 '5-4.0 .~J
H
(I I' '* . 7-. . -. - ' 5
'1 . 'I .57-- P. 7- I 5.5.5 ('55 (0 55' 5
', (4
-e
rr
, I n4 .14. 4o4.4 44.4. f.'o444444 C444444
r ' 4, 4 ' ' 4 r I .,4. . . 4-. 4.,, 4 '4 . ,~. .. 4 4 L-* ~ ~ 11*~ 4',* . .
4.... . .) 4 . .. l 4 . 444 , , -.
C~~~~.(.JC ~ ~ ~ ~ ~ ~ 6 t'P 4C 44 44. 4 C . -4 4 4 4 .
c4444 1. ~ I4 .~ I 4o v, .r c4 I4. j r r
4' ~eS'4 "5,,.~,r' C44.44.5~4 C4'L
- -l L44. 4'4 4.44..4'.
74 4 7 .. .
ul$
- *. ....... ... 4- _ , ...
. . .. . .. . .. . .. ,
, I I),.., C s
Sr * c . . . .
•~ I .
C C.. . .... ' : ', *' 4" -,. "- ....-Mal
C C C'C'-' C... . .. ..CC- CC ' 'c'C
C' C '. C .' .". .O fl O.C
- . - -.... -'------c, '-'--,'.*. -c .."-
C 'OC l' '''*''C-
Il ' l-'"C '3 '.*- " "[JCC "- C J ...- ]~ r t. j (
.C. . '.. .C . . -.. r. ... _," C- : : -J
'0.* .C '''- C . . .- , . - .. .'" ' -'C , -C 'C . - : .. : '
... . . .CC, '.,C ' '', . .. . C., . . . , " ,
i i i i le • . . . . . . . ..C-- 1ii
*~~ ~ ~ ~~~~~ r W,4 4 V44 4 4 . 44 4 .. ,, ... . 4 - 4 .,
4 4 ~ ~ o .44 4 4 44.'
In444 44-4' 1)4,. '1 4,4 4
4 I44 4444 4., (44 I4 4
44~4*'*fl
i$
Lr
.C .. . . . . . .. . .. . .. .
* C C C . .. . c ,CC. . . -
S- . .
-clH ,
- C..C. ' C,, - . , : ' C - ,. C . .
• . . . ' " • ! * " ' ' I " ' '- ' " : .. .
o )Iii i ~ o n. i i ii. i ( ( i i
-- -- - --- I e.- - ,.., i - -
i o- 4 4 ' i -i c i iJ i ' J i- I A i - -T I i
- - - - iiCi .CYii .---
7ni .j5-i,,C 7' i e '
'r5 s -Iii~~ I iii- i
c+
• )' " ' . . . . ." "
C J , ), C L
tL *~ ~ . . . . . . .. .. . ... .
C) C, • CC C' ( . ...
: ,o- -, ,- .n...C C' ",
CCCJ
j r .-5C. C
o- - +C C - CC C . .
,.CC .C ( . CC . C _ '', C . CC.IC. ., .CC cJ ... •- C: . . . .
'-4
H -" , CL :'. -J- -- --- :::,l -- C
-
.5C
. . ." , . . . .C.. C C C A .. CD ". .. .. .C'-...• '.C • .. .. .- . -
C. C- , ,
(C,. Cf,, ,, . . . .
15¢, * { }. ¢ (J( I , ( CC . ..
11 '~C C' I* .I C .
C i (2 cS .. .CC.... C -... 4.4 1.. i .. C.. C , c. C)' C I.
0C .. I'llC L C .. 1 C-C
.C, C-$
I3c ' C 3.>C
. 0C . C. . . . . . . ..CC a ' C . .C. . . . . .
OCC .C.r Cs CCO Co CCC1C.C.C C.'. C'' r '.II<nCCClj , C
C..CcC....C...C.C.,CIL:~ ~~~~~~~ j~'C .I.L.C4U 0 C C CC~,C 4 I ,.4 C .'C,.J.
CN7,
-l i C-r 'I I- "I "" L- I
'C C-CCC' cC-
-' LCC)DC aCC r
ci >. :L-
L.JC-CCJZ ~ ~ ~ ~ c toT . .. J - CiC .C C
L: ... na tI~ C -n
30
C.4( 4
C'44 4- '44 4 44*44.4. <4 4j
L'
017
C13 j4J 3,' It N'3 l
C)~~~~~I In c. j13 01 1t .2 ., 3 L L j*
3' ~~~ ~ c "-I.. . tit
L., '1OCW CC0 2' -.
3,-r r-
ct I r3 *I.- .3)0 I''141
01 .l C111 q- .a, W. 01 3 32 <
-- (2 )I..
04.-CC C.0 -.. Ot 1. Lr- >
,r -j I) rj c14 C -'.0~~~t t0 k2-' t0 ).)-
Z.~ 4-~-~J'0
t~fl"r( C3 t,-'fl'- '- C C'-7:L3CC LW 344 Ct. . 000)- - .. rt4
U. Z71 >W -1 3.- ol I
-qwx
C 4~*
.. t'
Ci " '.. ... .. ..
... ' ( ~ " ' ' " " -C . . " ' -'" - - "
r I.0.. . 0 ',. . . .C, '. , , rC,, .. . , '.
c j c C Cr I1 11 1 j- I - '-. .I I 1 -.c j- j. __ c I r j r * r -
F w : - i.- .. . i .... . ...., hi-' . .. . . .. .... ..
al r ))c 0 LIki , , tI I LI ¢ , l J ltl t J f , l ,rn cg J , ,
C. -I 1 7
.0
co
". (0 _. Ifod e
- , z . - e , : . : - , . -
. .L i.. . ). .i p)C i .i. .. i C ~ . .. . .t .L J L, L U .
. .JC. . 2. . .) ... C . .. . .*ij *.)i.- .. ') . '%
OC~~~)fl...~cl -C -)4 c c c( -.C C .i)C -l . -( -- c)
ofl) C, r i) *LL a
In ri0~ c1
-iC .0 rn r, ri (j .0 j. (n.YcJ .
-~a .. ... C C U
In In InC ) * . I.) )(ln .. 11' L))n j Ip<1!c- 11
.5(0 C) ) 2''ti) .&5iCLL.,.L7
OWCC~'C el o'- C~ 00C'r O)
C.Ci)L.C.OC' C)CC~it~fl~c' O CLC'~ ni) C)))'If .. C Ii i....lC C-. LA V. .I,,. ... C.) cA .O.Li,)A ili N.ctCU .L.L.i .. C.C),i*
al., C C2 o'C~'fC r 'o 'P C.) r-o' '(C ri ct~. fl. "1)c
- ~ ~ ~ c z 0 3. iC C C C ~ o.' C''~ C (C CI. CCn C C. CLi CC
- iL 4) IT) U) ii ) ) )- - N .. t)AL L )CiJLC.L..)L L t CL C ).. .Ci' )'A C>C
r5c. ::L j . ) '-.),' C C') L L ' .Qc. CO oi) ' t ... '. LJ ~ S i
C.n f)u m~ I ~ C~
M.t O~f C?' C. LO ' r:t.U ' 'C).' (LCC C ' 7 C-CCl 7 ,b2.0~ O. J) a030.~o.r-..O~C c<C a a' tflj'..~oz
U'c ~)C4 CC)C.'.o'PLA Uf)fC- v, C C3l" C' 1')'.V
Li- --------- C-.) O C.'CCCCiOCi
4.2. Cl ....----
.. CC, C--o. ' Cj,7.
iC
- .C. ' I.n¢
., .
=' S i
'
I (
In .. ,C:
. ',
: ; .. .o,
. . ...... r CC* (C;C , ' C .;
. -
i -o
('F- C. - C , ' . C . ''h.ClK
( C ''
Table A-18. Cumulative Frequency Percentile Concentrations(PPM) for Williams AFB Hourly NO Data:June 1976-June 1977
Station No.
Percentile 1 2 3 4 5
50. 0.002 0.003 0.002 0.003 0.002
80. 0.004 0.004 0.004 0.008 0.005
90. 0.006 0.006 0.005 0.012 0.008
95. 0.008 0.007 0.007 0.018 0.010
98. 0.012 0.011 0.011 0.028 0.013
99. 0.016 0.016 0.015 0.038 0.016
99.5 0.023 0.021 0.020 0.048 0.022
99.9 0.052 0.045 0.037 0.075 0.036
99.95 0.060 0.050 0.045 0.096 0.045
99.99 0.106 0.115 0.094 0.131 0.059
Table A-19. Cumulative Frequency Percentile Concencrations(PPM) for Williams AFB Hourly NOx Data,June 1976-June 1977
Station No.
Percentile 1 2 3 4 5
50. 0.008 0.009 0.009 0.014 0.008
80, 0.014 0.016 0.017 0.027 0.015
90. 0.020 0.022 0.024 0.040 0.023
95. 0.028 0.030 0.033 0.055 0.032
98. 0.039 0.041 0.048 0.080 0.047
99. 0.053 0.052 0.061 0.100 0.061
99.5 0.071 0.060 0.076 0.120 0.073
99.9 0.106 0.094 0.106 0.158 0.096
99.95 0.125 0.107 0.114 0.170 0.102
99.99 0.206 0.228 0.198 0.259 0.108
42
-leftj
Table A-20. Cumu Iat ive Frequency Percit iIt 'm Cent rat io.s
(IPM) for Williams AFB Hofi, l NI iaIta:
June 1976-June L977
_ta[ ion N-,.
Percent ile 1 2 35
50. 0.005 0.O00 0 0070 0t.1 o. uO ,
80. 0.011 0.013 0. 014 0 . 2 0 12
90. 0.016 0.019 0 020, 0. 00
95. 0.022 0.020 0.02, (1 .4. 0. 02>-
98. 0.034 0.036 0.041 0.o6l 0.041
99. 0.047 0.045 0.053 U.074 0.051
99.5 0.058 0.052 0.065 o.084 0.061
99.9 0.099 0.077 0.088 U.106 0.t74
99.95 0.103 0.082 0.099 0.!28 0.077
99.99 0.108 0.113 0.108 0.147 0.097
Table A-21. Cumulative Frequency Percentile Concentrations(PPM) for Williams AFB Hourly CO f'ata:
June 1976-June 1977
Station N,.
i',,icentile 1 2 3 5
50. 0.096 0.124 0.081 U. ( .1 51
80. 0.168 0.221 0.15- .514 0.21')
90. 0.262 0.318 0.252 1.758 0.397
95. 0.386 0.444 0.4)() i 0.562
98, 0.637 0.64t 0.60)) 1 .404 0 .84
99. 0.856 0.632 0.16' 1.775 1.050
99.5 0.988 1.029 0.968 2.182 i.261
99.9 1.553 1.548 1.454 1/.40 1 .810
99.95 1.810 1.667 1.684 .fb7 l .97
99.99 2.026 2.471 2.261 1.14,2 2.17?
4-a
Table A-22. Cumulative Frequency Percentile Concentrations(PPM) for Williams AFB Hourly CH4 Data:June 1976-June 1977
Station No.
Percentile 1 2 3 4 5
50. 1.615 1.603 1.581 1.623 1.624
80. 1.675 1.686 1.654 1.735 1.b97
90. 1.722 1.743 1.712 1.861 1.779
95. 1.792 1.800 1.780 2.010 1.882
98. 1.891 1.893 1.890 2.246 2.082
99. 1.966 1.966 1.966 2.420 2.237
99.5 2.071 2.043 2.069 2.655 2.415
99.9 2.735 2.422 2.392 3.446 3.120
99.95 3.135 2.995 2.825 3.689 3. 199
99.99 3.339 3.100 2.987 3.980 3.440
Table A-23. Cumulative Frequency Percentile Concentrations(PPM) for Williams AFB Hourly THC Data:June 1976-June 1977
Station No.
Percentile 1 2 3 4 5
50. 1.651 1.704 1.637 1.775 1.685
80. 1.749 1.839 1.760 1.993 1.811
90. 1.829 1.917 1.856 2.218 1.938
95. 1.951 2.028 1.974 2.495 2.097
98. 2.131 2.161 2.119 2.866 2.309
99. 2.244 2.290 2.253 3.175 2.522
99.5 2.364 2.369 2.424 3.550 2.693
99.9 3.027 3.265 3.152 4.836 3.343
99.95 3.203 3.390 3.406 5.355 3.516
99.99 3.898 4.328 4.384 6.238 4.006
44
Table A-24. Cumulat ive Frequency otc.nt i, tlientrations(PPM) for Williams AFO ,Uurv NM1ht, Data:
June 1976-June 1972'
St [d lOl N".
Percentile 1 2 3 u
50. 0.048 0.079 0.U(),4 I1 1 1.058
80. 0.107 0.176 0.137 2 ° 0.1 0)
90. 0.155 0.229 0.i94 . , .418
95. 0.224 0.288 0.2D7 0.521 (.234
98. 0.322 0.378 0.331 , (7ib L2 u
99. 0.397 0.435 0.417 1.0/
99.5 0.559 0.515 0.516 L.297
99.9 0.990 0.727 1.222 2.541 0.642
99.95 1.152 0.842 1.372 3.075 1.23)
99.99 1.554 1.330 2.002 3.98-i
Table A-25. Cumulat iv Frequency Percent i Ic Scait Lr in
Coefficients (x 10 - m- I ) for WiIl jams AFBHourly bSCAT Data: June PQ76-June 1977
Station No.
P rcent i le 13 4 5
50. 0.469 0.49) 0.51 li.4 /) .Vi9
80. (). 11 0.809 0 771 71, 1 J.
90. 0.945 1.051 0.982 I .0", 0.998
95. 1.207 1.352 1 .243 1 1 . 2,.9
98. 1.669 1.894 1.800 1.114 1.74
99. 2.030 2.333 2.453 2.117 2.133
99.5 2.499 3.136 3. 254 2.69o f) ,
99.9 3.861 3.974 5. 896 .7()1 1 4 ,1
99.95 4.408 4.550 5 .962 4 19/ 4. i)1
99.99 ).007 5.828 7.00{} 5.686 4. 48
"no
' I i n .. . . . . . il In . " I . . . . I . ... .... . i il t
.. . . . . . .
APPENDIX B
METEOROLOGICAL DATA: PROCESSING AND ASSUMPTIONS
1. INTRODUCTION
The prime objectives of this re'search effort were to evaluate andassess the accuracy of AQAM, using only the meteorological input data thatis routinely collected at U.S. Air Force bases. Initial time constraintsdictated that the EPA collect and process this meteorological data in a mannersimilar to ETAC; however, the standard ETAC hourly meteorological data basefor Williams AFB ultimately was used. These data, in the DATSAV-SURFACEformat, were then merged together with the EPA provided acoustic sounder data.Additional EPA/NOAA acquired data on u, v, w wind velocities, solar radiation,and vertical temperature gradient exists for several months of this study butHas yet to be integrated with the hourly average meteorological data base.
The method for determining Turner stability class from the USAF/ETACdata is discussed in this appendix. In addition, the acoustic sounder datacoding is interpreted in the context of determination of an houriy mixingd-ptM. ThE mixing depth is a key AQAM model input provided neither directlyby ETAC or indirectly by a Short-Term AQAM calculation. Finally, a model formixing depths is presented, based on ETAC-measured, surface-level meteorologi-
cal parameters.
2. METEOROLOGICAL DATA
Meteorological parameters used to determine stability class by theTurnerB- 1 method are wind speed, cloud ceiling height, and amount of skycover. All parameters were supplied by USAF/ETAC but only wind speed valuescould be used directly. A comprehensive table of codes for ceiling heightsprvidvd oasy interpretation of this data field but the sky cover data field,also presented in code, was interpretable only with the aid of ETAC personnel(Capt. 1. Clark, private communication, February 1979).
Sky cover denotes the amount (to the nearest tenth) cf sky covered byclou'ls or hidden by surface-based obscuring phenomena. In order to completely,r,e;,'ive the sky cover data field, three steps were Laken: M) a description
,[ th,- coded field was examined; (2) it was then compared to the Turnerrequirement (see Table B-I); and (3) an estimated sky cover (tenths) withinbo:nis .jf the re.quirement was assigned. For example, the ETAC given code torsky rover is -8, indicating "broken" conditions havin, a sky cover ot 0.6-0.9,in,clusive; but the Turner method requires the range 0.5-1.0, inclusive, todfine the same broken condition; therefore, wherever a code -8 is supplied,the value 0.6 is assigned as the sky cover. In the case of "partly obscured"tonditions, we indicate only that the amount of cover is less than completethough the possibilities span the interval 0.1-0.9. Table B-2 gives thefrequency of observed sky cover codes for the 13-month study period. 11efollowing table, is used to illustrate the assigunent procedure for the sixdifferent sky c vv, r code q suppl ied among 13 months ol hour lv dat a.
46
E'rAC C;ode Tcn(NS) l~sytinRpirretSky cover
0 CLEAR. The state oi tu nV 0 .4sky whon it is eitherclIoud le ss o r the sk y c oveii s l e-s t han 0 .1I.
-3 SCATTERED. A sky cover ot cmV (K 5 Wav) 0.40.1 through 0.5 at and( ClV n'' (igt
between the lovei, of i I aveialoft.
-8 BROKEN. A sky cover ot 0.6 I V .u 0Kbthrough 0. 9 atL and be Iow t Ielevel of a laver aloft.
9 OVERCAST. A sky cover of WoV 1.0 LU01.0 (ten tenths) at andbelow the level of a layeralIo f t.
8 OBSCURED. The condition when Cov = 1 .0 1 .0sky is completely hidden hy
stirface-based obscuringphenomena; e.g., fog, smoke,precipitation formns, etc.
10 PARTLY OBSCURED. The con- COV < 1.0 .9dit ion when 0.1 or more but
not all of the sky is hiddenby surface-based ob~scuringphenomena .
3. STABILITY CLASS INDEX AS DETERMINED BY EPA METH0OH
Ani exami nat ion of the Single Source (CRSTER) >1odo I8- re\val id t ha tone, of seven stabhiI i ty c lasses is deter'iied f rom me te,,r o Io icalI d.ai a f oreach hour by i t preprocessor . The first six Of t h 0eS (. a L e g or i , ( I-(correspond to l'asqu i II's clIas s i f ic at i ons (A-F) . B Thet seventh (atk gor vcorresponds to the "'dashes"' in Pasquil I 's original rla si ficat ion and rtipri-sent,; the ex istenice of a strong, ground-base-d nor t irnaii teinperatu~e i nve-rs ionand nondef inable wind flow condi t ion- . Thie CKS'IER niodil -,repro(_si (S"tIets tr ic ts changes in - tahi I i ty to one c lass pe r houir a: sh11 WT i n Fi g. R- I. Th istype of restriction allows undesirable daytime stability classes such a- 5, 6,and 7. Figure B-2 shows the result, of niirestrictod hour lv class changes.
In ititally, the preprocessor deteri-n res the hoiir aug 1 e of tilie sun aird tiliet ime s of sunr iseo and sun set from thle day nuimbe r I onigiti'de, and t ile Zonel topermit different .at ion of daytime and nightt ime cases hv the Wool 1 3- method.For dayt ime cases, the appropriate insolat ion class, is solected by movans ofthet Turner objec t ivo iii. Ihod witsing c louid cover, cveil) hu le ight , :cil solIa rv levat ionl as i rid I ca tors . Th i S me t hod asqi gn s nett rad i ait i 'ni i nd i ' s , ii gthecriteria shown ini I'ahle 8-3, for cases, where tilt total i. laid rijvor I l.tilt cloud cover ) 5/10, hut less than 10/10 (overcaSt), tho insolat ii la-is reduced hy one category when the -vilIing hevight is i trthan I' ( W !
/47
R-
[ RR ROODOD T RR
25.0 500.0
450.0 -
lO.- - 40 .0 ,
15.0- 30.0
z 250.0X
n10.0- 2 ,o.0
150.01(n
5.0- 100.050.
5D. 0
0 0 0 a O 0i- 00 0.0
a A &A A & a A aA
S V V 9V V 9 V A V 7 7 X V V V V
-4 YY A Y A A
A AEL A
2379
0- 1V0.................................. ........... £1301
01 12 1 '4S 6 13'R GZHOUR
Fig. B-I. The Behavior of EPA Stability Index (A) WhenRestricted to Changes of One Class per Hour,
December 10, 1976
(See legend followigy,-)
48
I RR R 0D000D T RR
150.0
1-
15.0- 3m.000250.0
10.0- 200.0 -C.,
150.00 I....
5.0 100.0 =:C)
50.0 "
00 0 0 3 0 o o L 0.00*
7 A A& A A A
6 A a A A
s- V Vv V v l45 V V A 7V 3 V
-
3 A lEWA A
2 V VV 2
1 1
0 . . . . . . . . . .. . .. . . . . . . . . + 0HOUR
Fig. B-2. The Behavior of EPA Stability Index (A) for
Unrestricted Hourly Class Changes,
Dec'mber 10, 1976
(Sce I egevd f I , 0Wi n,.
49
LL 'M
LEGEND FOR THE STABILITY CLASS INDEX ANALYSIS
TOP GRAPH
fil Acoustic Sounder determined mixing layer height (meters).
U Undefined layers.
R Return due to noise, possible turbulence, rain, etc.
D Daytime noise due to airbase activity.
T Transition period.
M Mechanical turbulence superimposed.
X An expected height value was not supplied.
I Inoperative equipment.
Windspeed (knots)
BOTTOM GRAPH
A Stability class index determined by EPA method.
V Stability class index determined by ANL method.
Cloud amount in tenths of total cloud cover.
HORIZONTAL AXIS
The computed times of sunrise and sunset are
indicated by lowered numerals.
50
and by two categories f[r ceil ings betwk!.n 7,000 ind 16, O it. For ckiling:;
below 7000 ft and 10/1) cloud cover ( i .. , overcasL), a n.t zadiat ion of 0 isdefined and neutral stability is spec ifid , With Lb,. 'x opt ii, o t tihe 1 0i1
low cloud cases, the net radiation indox during daytinc boors is n,.ver rdu'-
below 1, or "weak." The final sthility catego,-, is s.A1tel from lableand Turner's insolation classes.
Table B-i. Turner Requirement Related to Skv Covtr a
The net radiation index used with win,! spee, t o .bt; ;n stabilItv cla~sis determined by the following procedure:
1) If the total clond cover is 10/10 and the cdliing is
less than 7000 ft, use net radiation index cooual to
0 (whether day or night).
2) For nighttime (between sunset and suonrise):
a) If total cloud cover < 4/10, use net radiation
index equal to -2.b) If total cloud cover > 4/10, use nec radiation
index equal to -1.
3) For daytime:
a) Determine the insolation class number as a
function of solar altitude from Tatle A-2.
b) If total cloud cover < 5/10, use the net radiation $index in Table A-i corresponding to the insolation
class number.
c) If cloud cover > 5/10, modify the insolation class
number by following six steps.
1) Ceiling < 7000 ft, subtract 2.
2) Ceiling > 7000 ft but < 16,000 ft, subtracL I.
3) Total cloud cover equal 10/10, subtract 1.
(This will only apply to ceilings > 7000 itsince cases with 10/10 coverage below 7000 ftare considered in item I above.)
4) If insolation class number has not been
modified by steps (1), (2), or (3) above, Issume
modified class number equal to insolation classnun her.
5) I f mod i fied inso I at ion c a ss number is less than1, let it equal 1.
6) Use the net rad i at ion index in Table A-I corre-pending to the modified insolation class ntmber.
aTaken from Ref. B-I.
51
Table B-2. Frequency Table of Sky Cover Codes
0 -3 -8 8 9 10
1976 June 545 138 28 9
July 54 594 87 9
August 191 512 29 8 1 3
September 122 494 78 22 4
October 300 341 85 17 1
November 464 t95 35 26
December 391 232 70 51
1977 January 127 267 80 28 2
February 256 300 92 18 6March 330 281 82 44 7
April 273 359 64 23 1
May 258 387 84 15
June 130 457 82 24 3
Total 3441 4557 896 294 1 27
% of Total (9216)a 37.3 49.4 9.7 3.2 .01 .29
aExcludes II days for which pollutant observations are missing, 1977
January 2, 12, 15, 19, 20, 21, 22, 27, 28, 29, and 1977 June 30.
Table B-3. Insolation Classes as a Function of Solar
Altitude for Cloud Cover < 5/10 a
Solar Elevation Insolation Net Radiation
Angle (a) Class Index
0 ° < a < 150 Weak 1
150 < a K 350 Slight 235* < a < 600 Moderate 3
600 < a Strong 4
aFor > 5/10 cloud cover, see Section 3.
4. STABILITY CLASS INDEX AS DETERMINED BY ANL METHOD
This section is included mainly to document the method used to determine
stability class indices during the initial phase of the Air Quality Assessment
Mudel validation effort.
In the initial phase of the validation effort, pollutant conceotrationswere ostimated at nine receptor locations. Subsequent data analysis on the
acoustic sounder codes revealed that certain discrepanciep existed at and
around the tim- )4 sunset. One obvinus discrepancy wa ; the measurement of afinite lid-height occurring several hours after the prt-stned sunset time
thus a possible unstablb condition wa,; in'dicated ai a ,table ,)ndlion.
Table B-4. Stability Classification Critieria
SurfaceNightt ineSurdfspeed Davtime Insolation - 5 liK 5me
wind speed ... 5/10 < 5/10
(knots) Strong Moderate Slight Weak :.vrcn i Cloud Cloud
< I 1 1 2 3 6 72 1 2 2 3 4 6 73 1 2 2 3 4 6 74 1 2 3 4 4 5 65 1 2 3 4 4 5 66 2 2 3 4 4 5 67 2 2 3 4 4 48 2 3 3 4 4 4 5
9 2 3 3 4 4 4 510 3 3 4 4 4 4 5
1 3 3 4 4 4 4
>12 3 4 4 4 4 4 4
A comparison of the ANL method t th Ei'A .a,,hod showd that both are the
same in every essential except that the ANL ,:utho,' il-sumed tie following:
1. Designed for urban conditions, only the fir-t five sta-
bility categories were allowed.
2. The times of sunrise and sunset were estiaated from charts
presented in the Smithsonian Meteorological ;ils.
3. Fract ional wind speeds were rounded w,', :; oq ii t
Turner's Stability-Net Radiation table.
Of the two methods, EPA's method was selected forc in A,)A-
its more accurate determination of sunrise and suns, t imes U.t,
which can be -een in Fig. B-2 where tne lid heigiv a, -,,,n v. t ,- . '. , ,- ic
sounder is approximately 130 m at the hour of sunset , 0 hr and ' .,ht 1Vunstable atmospheric condition (class 3) is correctly d, termin-,i tn , - 11A
method. On the other hand, a slightly stable atmospheric condir i2 , !-l iss 5)
is incorrectly determined by the ANL method.
5. MIXING DEPTH, DIRECT MEASUREMENT BY ACOUS'IIC SO'NDIEH
The Acous.tic Sounder information is presentd in t od- as a six- ligit
number of the form ABmDEn, where AB and DE are av--rA1, hii ht val ,,P; ( rn-,
of meters) from which echoes were received, "m" is ;i (oclo used to cxplai aspecial local effect, and "n" is a code used to des ina- a sp-,,i I if atnoe-
spheric phenomena. In all cases of inoperative eqaiipinant the odo is 999999.
The average height values, AB a.id DE, fall within the 20-500 -n vertica]
range of the recording chart paper.
53
The special effects, m, cause graying of the sound pulse as indicated bythe following four values:
1. Superimposed return due to noise, turbulence, or precipitation.
2. Background acoustical noise due to airbase activity.
3. Uncertain acoustical return due to unstable-stable atmospheric
transit ion.
4. Apparent mechanical turbulence due to high wind speed.
The specific atmospheric phenomena, n, are conditions applied to theWilliams Air Force Base study. The values of "n" (1-10) are described inTable B-5, which gives the AFB Acoustic Sounder Classification Codes.
[he table below shows the assignment of mixing depth values basedon these codes. They are used only in the development of a mixing depthalgorithm (Section 6).
Acoustic Sounder Class Mixing Depth (meters)
I Top of radiation inversion 0.02 Drainage winds 0.0
3 Layered returna. Top of radiation inversion 0.0b. Top of stratified echoes 0.0
c. Surface based 0.0
d. Undefined layers 9990.e. Transition period 9993.f. Mechanical turbulence ABO. or 9994 if missing
4 Inversion base forced aloft ABO. or 9998 if missing5 Subsidence inversion ABO. or 9998 if missing
6 Frontal inversion ABO. or 9998 if missing7 Daytime noise 9992.8 Return due to noise 9991.9 Top of mixing layer
a. Top known DEO. or 9998 if missingb. Top unknown 9998.
c. Transition period 9993.d. Mechanical turbulence DEO. or 9998 if missing
10 Instrument inoperative 9999.
Figures B-3 and B-4 present frequency distributions of acoustic soundermeasurements based on the hour of occurrence and the month of occurrence,respectively. These two figures supply the bases for the amount of presentdata deemed valid using the above interpretation of the Classification Codes.
6. DEVELOPMENT OF A MIXING DEPTH ALGORITHM
For dayt ime ho,jr;, the acoust i sounder provi d id mixing dihpths 1-) 996hours of tht ;3 -mlrTu l h exporiment. '[11v palin ity ()I 'in t is ( II ul tluu r-1illIdayt ii. de.vv iOlm t f ilhi, mi, id i I jy r aut flit -h I i d r .utin j i, i u' air i f i ,u
( i.c ., i)0 hm ) rul I t. u i ,n ,ln r I , + n, u u, l nr' , III ;u I I I .1 itnl+n l , I is,. It- ')
14 .
C\)
-4
C)
EZZIZIZZIIL ,33H.
Cqox
II AX0
u0
o "l
1-4J
Lr
0
u
.,
J4p J%-Jc
I.,,
(/~
+2 aLU
II II0
c! ~
* . --
* .. . ~ 5;. . z ~
.~ IS S SC - -
- . - . . . . 0..... j. .. . .
* . . . ..
* . . S * C ***
* -. .. S ~ ~ .. 0* ~* C.) ~
* .. . . .
* * S ~ - * S~ * -. N
- . S S ~
* S ~ * * C - * * ~ . C
-. S.. - S ** CS .
* S. *~ .. S. .. fl
_____ _ F. a . . ... ... . S. .5 CSflC *** ~ ~ S Cr,
* S *~ em.. ** 5 . 1
I I -- -1.0
(sK~~) 1V~IJJ3UO2HJJ AX'
57
Table B-5. Acoustic Sounder Classification Codesfor Williams AFB Study
1. Top of radiation inversion: ABO 001
AB is top in 10s of meters
This type of inversion is typical of night time desert locations.It is ground based around sunrise hours but aloft during late evening
hours; stable to very stable atmospheric conditions. Much of the turbu-lent mixing occurs only in a shallow layer near the ground. The thick-
ness of this shallow layer varies over a very wide range, from meters tohundreds of meters making it impracticable to define the layer heightfrom this code.
2. Drainage winds: with or without stratified layers above: ABO DE2
AB is top of Drainage in 10s of meters
DE is top of stratified layers in 10s of meters
a. AB: top of drainage ABO 002b. DE: top of stratified echoes ABO DE2
c. top of stratified echoes unknown ABO 002
Drainage winds are defined for the dense air that flows downhill in
mountain-valley terrain. They were observed, specifically during wintermonths, to change directionally during their westward onset at approxi-mately 2100 hr, to calm winds, and then eastward after 2200 hr.(*)
There may or may not be stratified layers above the winds but the layerbelow the winds is assumed to be stable. The mixed layer as a whole may
span several hundred meters in height making it impracticable to definethe layer height from this code.
3. Layered return: use when there are layers in addition to radiation
inversion: ABO DE3
a. AB: top of radiation inversion ABO 003b. DE: top of stratified echoes ABO DE3c. surface based -- no radiation inversion 000 DE3d. undefined layers 000 003 or
ABO 003 if radiationinversion isdeterminable
e. transition (unstable to stable)AB is top of transition layer AB3 003
f. Mechancial turbulence superimposedAB is top of echoes AB4 003
A layered return is produced when sounding echoes are stratifiedindicating the presence of multiple layers. In addition to the radiationinversion as defined in 1 above, the formation of multiple layers isstrongly influenced by the onset of drainage winds.(*) The mixed layer,during these nighttime turbulence episodes, is defined by category (f)and will extend to the top of the echoes.
58
Table B-5. (Cont'd)
4. Inversion base forced aloft by surface heatin.g: ABo DE4
a. AB: base of echo ABO 1)14b. DE: top of stratified echoes A'O0 DE4c. top of stratified echoes unknown AB) 004
This type of inversion is due to morning solar heating arid islimited to one or two hours at Williams Air Force Ba!ce. The mix,.d laverin this case will extend to the base of the echo.
5. Subsidence inversion: ABO DE5
a. AB: base of echo ABO 04J5b. DE: top of stratified echoes AB DE5c. top of stratified echoes unknown ABO 005
A subsidence inversion is usually caused by synoptic sale highpressure over the region. Since this condition did not occur throughout13 months of records, the assignment of mixing depths posed no problem.
However, mixing depths would have been treated in 4 above.
6. Frontal inversion: ABO DE6
a. AB: base of echo AB() DE6b. DE: top of stratified echoes ABO l)E6c. top of stratified echoes unknown ABO 006
Frontal inversion (same as 4 and 5 above except for cause).
7. Daytime noise: 002 000
Daytime noise is an undetermined graying of the return (bar kg ouniacoustical noise) due to air base activity.
8. Return due to noise, possible turbulence, rain, etc. 001 000
9. Top of mixing layer (thermal spikes), normal daytime, situation ABC UE2
a. DE: top known 000 DE9
b. top unknown 000 009
c. transition period 003 1)E9
DE is top of transition layer return (NOTE: this i; similar tocase 3d above; however, thermal spikes are mor, priminmInt.)
d . apparent mechanical turbulence due to hi ght
wind speed 00/4 M..9
DE is top of echoes
During normal daytime situations, the mixing depth is unobtai nable onIlyif the top is uinknown (9b) or if the sound irrg retutiri recold is uiniert.1a1r
(9k) .
10. Instrument inoperative or unable to r,,(l chart 999 999
(*)Bowen J., EPA Las Vegas, private communicat ron, Mjrch 1979.
S q
with the theory of Nozaki,B - 5 as corrected by ETAC, (J.B. Hooten, privatecommunication, October, 1978), which gives the mixing depth L in meters as;
121 (6-J) (DPT) J(0.087) (UZ + 0.5)
6 6 f in (Z/Z o )
where:
J = the stability class index,
DPT- the dew point depression (°C) i.e., the differencebetween surface temperature and dew point temperature,
UZ - the wind speed in knots at anemometer height(Z = 373.4 cm),
Zo = 5.0 cm is the roughness length characteristic of theanemometer exposure, and
f - 0.0001 is the Coriolis parameter.
The comparison shows that the theory overpredicts the observed mixing depthsubstantially. In addition, comparison of Figs. B-6 and B-7 show the inabil-ity of this model to duplicate the observed lid height time dependence. Oneis cautioned, however, that the average height of the lid during afternoonhours is certainly underpredicted by these acoustic sounder data as theinstrumental sensitivity cuts off at 500 m.
A substantially better fit to the acoustic sounder data has been obtainedby chisquared optimization of the equation and choosing for lid height themaximum of the two expressions as follows:
t-tmax
HC + C2 (T - TMIN e C5 + CE + C dT1 2 G IN3 4 dt
or
H C 6 * WSS
where:
C1 - 66.452 meters
C2 6.9644
C3 1 1.4402
C4 - -14.941
C5 - 6.1866
C6 - 8.4085
t - hour of day (0-23)
T a teynpratiire (K) at hour t,
WSS - wind Rp,.Pd (knotsq) at holir t,
0 - th. anylr, of aliti ,,ds )f th,. mkJn ;it hwir t,
(p (4
42
F---- -i-- - ---- 1 -~
I -
H----- - --- A-
- V.
F-
,~ '*~J
n
'p-I
0 0 0 0 0 00 0 0 ifl
N N -~ -~
(IA~i) Lull DUll!
Fi
IV
N
il
H---
I- -- ----- U-II
I'I
?.
; REGRESSION PARAM4ETERS
v = mx + bm = 0.36b = 85.7 meters
=0.69ave. dev. = + 41%
L7
.I
IN
Fig B-8 .Miin Deth A Cmaino-)be o s versu
" I_ L ' t it .
i6 3
LI -$
, , c, ~~~~~I . '? ll} !,
Fig.1^,-1 ! B- . M iin De t : A Co p r s n of O s r!io S v r u
aTeort.cl Mde
( k the highest temperature attained up to hour t,
MLN the lowest temperature attained up to hour t,
tMA( the hour at which TG occurred.
Results of this comparison are presented in Fig. B-8. Although based entirely
on surface observations, one obtains a correlation coefficient of 0.69, whichis comparable to results obtained using vertical temperature soundings.B-6
This expression for mixing depth was then used for both AQAM I and II predic-Lions throughout the 13-month period.
REFERENCES
B-1. Turner, D.B., A Diffusion Model for an Urban Area, J. of Applied
Meteorology, 3:83-91 (Feb. 1964).
B-2. User's Mannual for Single-Source (CRSTER) Model, EPA PublicationNo. EPA-450/2-77-013, Environmental Protection Agency, Research Triangle
Park, N.C. (July 1977).
B-3. Pasquill, F., Atmospheric Diffusion, D. Van Nostrand Co., Ltd.,London, 2nd Ed. (1974).
B-4. Woolf, H.M., On the Computation of Solar Elevation Angles and theDetermination of Sunrise and Sunset Times, NASA Technical Memorandum
NASA TM X-1646, Washington (Sept. 1968).
B-5. Nozaki, K.Y., Mixing Depth Model Vsing Hourly Surfact Observations,Part 1, USAF/ETAC report 7053 (Nov. 1973).
B-b. Benkley, C.W., and L.L. Schulman, Estimating Hourly Mixing Depths fromHistorical Meteorological Data, J. of Applied Meteorology, 18:722-780
(June 1979).
64
APPENDIX C
SOURCE EMISSIONS INVENTORY PROGRAM: INPUT MODIFICATIONS
I. INTRODUCTION
The Source Inventory ProgramC -l computes the total annual pollutantemission for several source categories at a given airbase, thus it must be runsuccessfully before the Short-Term Dispersion CodeC -2 can be used.
The required input to the Source Inventory Prop ram was prepared atWilliams Airforce Base, Ariz., by the Stanford Research Institute (SRI).C-AThe input consists of operational information related to aircratt and nonair-craft source activity (see Figs. C-1 to C-7),- as well ,is inform;tion abn,!tthe airbase environ sources (Fig. C-8), such as the Phoenix area. Mod iti ca-tions to some of the airbase operational information were requested by th'-U.S. Air Force and involved the addition of aircraft names, engine names,engine identification numbers, engine fuel rates, and engine-pollutant-emission rates. These items were modified as were the number of airbaseaircraft parking areas and their center coordinates, the number of taxiwaysegments and their end point coordinates, and the annual account of eachaircraft type and runway usage. In addition, data related to aerospace groundequipment and aircraft refueling operations were extended to include thkreeadditional aircraft types. Finally, all training-fire point sources wert!deleted from the input, though times of actual training fires were noted.
The requested modifications are detailed in this appenlix.
2. NAMELIST DATA, REASSIGNED PROGRAMMED DATA
The Williams Air Force Base (WAFB) AQAM Source Inventory Program inputas supplied by Stanford Research Institute (SRI) has undergone several addi-t ions and updates per Air Force requests.
The namelist input to the Source Inventory Program consists of threenamelist group names: (I) EGDATA, engine data; (2) ACDATA, cycle- data;and (3) DSDATA, temporal distribution data.
These data are considered to be good overall averages of aircraftengine emission factors, aircraft landing and takeoff parameters, and thetemporal distribution of aircraft and airbase activiti es; thus tho' areautomatically used by the program when the user has not input other data.
EGDATA is the only group name considered in the update process for whichthe reassignments involve the addition of aircraft names, engine names, cen gin
ID numbers, fuel rates, and engine pollutant emission ciata.
Two tables, Tables C-I and C-2,* are present,,d to record the chango,
from the SRI supplied input to the updated input. Notv that the extende d list
*Figures and tables appear consecutively at the end of !his appendix.
65
of aircraft types results from actual observations at WAFB and from guidelinessubmitted by the Air Force dated 12 October 1978.
3. DATA SET 4, AIRBASE AIRCRAFT AND RUNWAY TOTALS
[he information coded in this data set defines the total number o!aircraft types, runways, parking areas, a special-case wind condition, andtaxiway segments. The following table shows the value of these parameters forthe SRI and updated inventories.
SRI Update
Aircraft Types 4 6Runways 6 6Parking Areas 4 3Special-Case Wind 1 1Taxiway Segments 16 24
4. DATA SET 5, AIRCRAFT ACTIVITY
Activity for aircraft is defined as the total annual number of arrivals,departures, and touch-and-go operations. Each arrival and each departure ofan aircraft is considered as an operation in the landing and takeoff cycle. Atouch-and-go, however, is a complete cycle and occurs during a training flightwhen an aircraft approaches and lands on a runway, travels down the runway forseveral seconds, accelerates, and lifts off.
EstimaLes of activity for aircraft were made from actual counts at WAFBtaken during the period 1 June 1976 to 30 T,'n 1977. A review of counts ofJune and July 1976 shows that the information needed to determine touch-and-go operations is presented as "daily estimated averages of total runwayactivity." These daily estimated averages could not be used realistically inthe su mation methods required for an annual number of operations. Therefore,
the activity for aircraft is redefined to reflect twice the activity informa-tion presented for the 6 month period August 1976-January 1977.
Tables C-3 and C-4 are presented to record the change from the SRIsupplied activity to the updated activity, respectively. Note the newT ANSIENT aircraft category consists of tho F4, Cl30H, and C140 aircrafttypes. In regard to touch-and-go cycles, an account is made for both normaian,1 transient operations by F5, T37, and T38 aircraft.
5. DATA SET 6, AIRCRAFT PARKING AREAS
Information in this data set must describe the geometries of the aircraftparking areas. The center coordinate of each square in a series of squaresmakifng up the parking area is defined along with the length of a side of eachsquare. The model assumes the square is situated so that a line drawn par:il-let with its right or left side wil' 'i directed north-south.
I
Tables C-5 and C-6 rtecord the change from the- S<I suppl ied geometries to
the updated geometries. Figure C-9 show.; the location of aircraft parking
areas relative to taxiways, runways, and ambient e-ir monitors.
6. DATA SET 7, AIRCRAFT TAXIWAY PATH SEGMENIS
Each taxiway path used by aircraft at: WAFB is d.fined as a series ofconnected straight line segments. Information in this data set des;ribes thegeometries of these straight line segments and assignq to each an identifyingline number. The line number is used for defining the Parti,-:2ar segments
that will be used to make up a complete taxiway peth.
Tables C-7 and C-8 record the change from the ';fo1 supplied georne tr i,-a In,;
the updated geometries. FigorH C-10 shows th, i (b- e I ta. wa-.,-, mnt
relative to the airbase runway configuration.
7. DATA SET 8, AIRCRAFT RUNWAY INFORMATION
All information concerning airbase runways is defined in this data set.
Annual arrivals and departures are defined for each runway (see Tables C-9 andC-la) as well as length and direction, Its length is the physical length ofthe runway pavement; its direction is determined by its orientation in rela-tion to true north (see Fables C-11 and C-12).
Each individual aiicraft type may use several different runways forlanding and takeoff operations. Therefore, inbound and outbound taxiway pathsare defined to/from each aircraft parking area to/from every airbase runway by $a sequence of taxiway segment numbers.
Table C-13 records the update of aircraft movement over the taxiwavsegment configuration (see Fig. C-10).
8. AEROSPACE GROUMD EQUIPMENT EMISSIONS
Aerospace Cround Equipment (AGE) consists of all motorizd v.( i:,except refueling tanks used to support incoming and outgoing aircraft. Tl.n'esupport vehicles generally consist of coolers, power generators, heate-rs, andhydraut i test stands. The emissions for this eqripment must be deterwline dand inpo, directly into the model. The model assumes that all AGE activities(,, It in the aircraft parking areas, but the emissions are calculated sepa-
.. roma hoe' Pmi ss ions result ing from aircraft park ing activities. Since-t this data ;et involved only the replacement of the single
.', rv with V4, C130H, and C141 aircraft types, Table 14I ' ll, 'Ipated input.
_V'111 Vf !N ; IN 1()IAIS
. , , ItIIIR r tl
fuel drained froti the aircraft fuel lines. Since the update of this data set
involved only the replacement of the single TRANSIENT category with F4, C1301,and C141 aircraft types, Table C-15 is presented to record the updated input.
L0. DATA SET 13, TRAINING FIRE POINT SOURCES
Training fire point sources are defined as shallow ground level sites on
the airbase that are filled with fuel and ignited for the purpose of training
airbase personnel in the art of fire fighting.
The update to this data set involved the deletion of all sites (a total
of 4), designated by SRI, from the inventory. Thus, training fire pointsour'es are not considered in the WAFB modeling effort.
REFERENCES
(:-I1. Menicucci, D., AQAM Data Reduction and Operations Guide, Air Force
Weapons Laboratory Report No. AFWL-TR-75-307 (Oct. 1976).
(-2 Bingaman, D.J., Air Quality Assessment Model for Air 'orce Oper,'tticn --
Short Term Emission/bispersion Computer Code Documentation, Air ForceCivil Engineering Center Report No. AFCEC-TR-76-34 (Apr. 1977).
C-3. Viezee, M., Source Emission Inventory Data Collection for Wiltixn8 Ar,
Force Base, Final Technical Report, Stanford Research Institute, Menlo
Park, Calif. (Sept. 1976).
08
1kj
op3
IPz
b-o'
69 ~
-- Gap
a
F
7
~v)
a ri
2'
.1
j
0
a C~Q
1 0)
Ii s-LwCO H
/
COc~
I H
~c >,
F LI cn~
&
0)-'CO0:3)-~ --44-JX
C
cc-4IL.
+d
a - I
a+
/a
,--. -
g F
-.
- 0 °
p - -
Lu aM-J H 0
71
jd
iLIL
-L
-I I £
J x
a, H 9
-4.
r0
a-i V. Ii-
4-
7 L''JJ .. . .. .. ' . ... ;a" 11
I
C.
('C
-j
*1
C.
r
CI.
- QLC?
.~ ~GJ~V
U
73
x
C r
.- CU- ~
C': 0
* m
.,-4 C
I~-~ c00
U,'-C ~'-~0<
0C -4
F:0<U)I-. 1~
In
OLrC>-4 ~
C
U4
*** .-
~ KU.~ >.
.. I. --4* .
* , z ~-.
r-.
I U
-'.4r..
75
V,)u0
-C -, C
(C,)V
.44
i ~i
0 -.4-- 4
x <'-,.
Xx E
76
T
ct(2
EB W_0
Lt
(1 t 4
* Hz- A 2-
c
-41-
FI77
ao
11
p- 44J
,J
'N
4~~ ~ -- 2r4
',,, ,XC
x 0
, U
" "; 00'•J• r ,.
"1"8" J,~ I
Table C-1. EAWI"ine D;ita (SRI)
Pollutant Eminss-n )aLa In
(ACNAME) (EGNAME) (IDACEF) (EGFF) Pounds per 1000 las of F.e1
AIRCRAFT AIRCRAFT ENGINE ENGINE THRUST FUEL RATE EGF2EC)
NAME ID NAME ID SETTING 1000 LB /11R wu HC NOX PM
Transient 50 J79-G15 I Idle 1.131 56.70 1D.70 2.47 0.50
Normal 2.720 1.-0 1.33 4.25 2.22
Military 8.921 2.28 0.22 8.94 2.36
After Br 32.240 4,3)0 0.01 3.11 0.15
FS 13 J85 6 Idle 0.453 180.00 29.90 1.26 0.013
T38 32 Normal 1.462 4 .3.30 1.37 2.32 0.O017
Military 2.630 29.50 ,1.64 2.68 0.018
After Br 8.323 26.00 0.07 1.99 0.008
T37 31 .69 14 Idle 0.231 127.00 29.50 1.51 0.729
Normal 0.698 49.10 1.29 2.07 1.017
Military 1.095 31.30 0.50 s.,O 0.020
Table C-2. Engine Data (Update)
Pollutant Emission Data in
(ACNAME3 (EGNAME) (IDACEF) (EGFF) Founds per 1000 lbs of F el
%IRCRAFT AIRCRAFT ENGINE ENGINE THRUST FUEL RATE (EG&NIFC)
NA.iE ID NAME ID SETTING 1000 LB /HR CO HC NOX 7>'
F4 12 J79-G15 1 Idle 1.130 57.00 12.00 2.50 ).50
Normal 3.500 9.40 1.10 4.80 i.eu
Military 8.929 2.20 0.20 8.90
After Br 32.240 4.0u O.01 1.11 0.I"'
(141 14 TF33-P3 4 Idle 0.900 84 .o0 I01.0 1.8 0.21
Normal 3.797 6.10 2.6 5. 0.99
Military i.436 1.Ao 0.6 10.0 1.73
F5 13 J85 6 Idle 0.469 178.o J0.0 1.30 0.0oi
T38 32 Normal 1.000 73.6 6.40 1.8(1 0.007
Military 2.527 29.o (.8) 2.60 0.0i8
After Br 8.323 26.) 0.07 1.99 0.00f
T37 31 J69 14 Idle 0.231 129.0 19.0 1.5 O.355
Normal 0.288 107.o 11.1 1.7 0.28
Military 1.095 37.() 0.5 .b 0. 2
CIOH 44 T56-AI5 21 Idle 0.493 18.1 15.1 2.4 0.)8
Normal 0.827 8.5 3.4 1.7 0.47
Military 2.392 1.6 0.2 11.7 U.71
79
Table C-3. Aircraft Activity (SRI)
Aircraft - Annual Number of-
Name Arrivals Departures T/G Cycles
1. F5 8391 8391 1682
2. T37 39662 39662 63357
3. T38 77281 77281 55963
4. TRAN 8112 8112 0
Table C-4. Aircraft Activity (Update)
Aircraft Annual Number of
Name Arrivals Departures T/G Cycles
1. F5 6566 6666 15493
2. T37 25394 25904 34614
3. T38 27180 27686 64125
4. F4 1310 1320 0
5. C130H 66 62 0
6. C140 20 22 0
Table C-5. Aircraft Parking Area Geometries (SRI)
Center Coordinate (UT)
Area Square X Y Length (km)
1 1 438.10 1685.57 0.10
2 1 437.70 3686.10 0.20
2 437.90 3685.90 0.20
3 438.00 3685.70 0.20
3 1 437.95 3684.70 0.20
2 438.1o 1684.00 0.20
3 438.11 1685.01 0. 14
4 4 18. I'i.!438.1, 0'i8 . l.
A - .1 003 AIR FORCE ENGINEERING AND SERVICES CENTER TYNDALL AF-SYTC F/6 13/2ANALYSIS FOR THE ACCURACY DEFINITIONl OF THE AIR QUALITY ASSESS4-ETCCU)MAR 80 Rt J YANARTINO, L A CONLEY. 0 14 ROE
UNCLASSIFIED AFESC/ESLTRA80-19VOL-2
21
-IIIIEEmhIIIE
IuIIIIIIIIIII-IIIIIIIIIIIu-IIEEEEIIEEEE-IIIIIIIIIIIu
Table C-6. Aircraft Parking Area Geometries (Update)
Center Coordinate (UTM)
Area Square X Y Length (kin)
1 1 437.41 3686.10 0.132 437.54 3685.97 0.133 437.67 3685.84 0.13
2 1 437.83 3685.50 0.132 437.83 3685.30 0.133 437.83 3685.10 0.13
3 1 437.71 3684.81 0.132 437.58 3684.68 0.13
Table C-7. Taxiway Segment Geometries (SRI)
Ground Level Coordinates Ground Level Coordinates
Line of One End of Line at Opposite End of Line SegmentNo. X (1) Y (1) X (2) Y (2) Length (km)
1 440.800 3684.430 440.560 3684.180 0.347
2 440.560 3684.180 440.050 3683.750 0.667
3 439.760 3683.980 440.050 3683.750 0.370
4 439.760 3683.980 439.940 3684.140 0.241
5 439.760 3683.980 438.530 3685.230 1.754
6 438.530 3685.230 438.230 3685.580 0.461
7 438.530 3685.230 438.130 3685.250 0.400
8 438.130 3685.250 438.080 3684.820 0.433
9 438.130 3685.250 438.130 3685.500 0.250
10 438.130 3685.500 438.230 3685.580 0.128
11 438.230 3685.580 437.870 3685.900 0.482
12 438.230 3685.580 437.520 3686.380 1.070
13 437.870 3685.900 437.520 3686.380 0.594
14 437.520 3686.380 437.730 3686.380 0.210
15 437.730 3686.380 438.400 3686.380 0.670
16 438.400 3686.380 438.830 3686.380 0.430
81
Li --
Table C-8. Taxiway Segment Geometries (Update)
Ground Level Coordinates Ground Level Coordinates
Line of One End of Line at Opposite End of Line SegmentNo. X (1) Y (1) X (2) Y (2) Length (km)
1 437.570 3684.600 437.900 3684.920 0.460
2 437.900 3684.920 438.040 3684.790 0.191
3 437.900 3684.920 437.910 3685.330 0.410
4 437.910 3685.330 438.050 3685.330 0.140
5 437.910 3685.330 437.920 3685.720 0.390
6 437.920 3685.720 437.600 3686.050 0.460
7 437.600 3686.050 437.220 3686.460 0.559
8 437.220 3686.460 437.460 3686.460 0.240
9 437.460 3686.460 438.100 3686.460 0.640
10 438.100 3686.460 438.530 3686.460 0.430
11 439.100 3685.870 438.900 3685.650 0.297
12 438.900 3685.650 438.530 3685.330 0.489
13 438.530 3685.330 438.270 3685.330 0.260
14 438.530 3685.330 438.410 3685.150 0.216
15 438.410 3685.150 438.040 3684.790 0.516
16 438.410 3685.150 438.270 3685.330 0.228
17 438.270 3685.330 438.050 3685.330 0.220
18 438.050 3685.600 438.050 3685.330 0.270
19 437.920 3685.720 438.050 3685.600 0.177
20 438.410 3685.150 439.470 3684.090 1.499
21 439.470 3684.090 439.730 3683.850 0.354
22 439.730 3683.850 440.220 3684.300 0.665
23 440.220 3684.300 440.470 3684.500 0.320
24 439.470 3684.090 439.580 3684.230 0.178
Table C-9. Annual Aircraft Arrivals and Departures (SRI)
Runway F5 T37 T38 TRA\
301, Arrivals 0 3451 0 0
Departures 0 34515 0 0
30C Arrivals 2960 4353 24917 7950
Departures 2960 4353 24917 7950
30R Arrivals 5263 0 50818 0
Departures 5263 0 50818 0
12K Arrivals 0 704 0 0
Departures 0 704 0 0
12C Arrivals 60 704 508 324Departures 60 704 508 324
12L Arrivals 107 0 1037 0Departures 107 0 1037 0
Table C-10. Annual Aircraft Arrivals and Departures (Update)
Runway F5 T37 T38 F4 C130H C141
30L Arrivals 0 24608 0 0 66 18Departures 0 25140 0 0 60 18
30C Arrivals 5084 0 21962 1270 0 0
Departures 6356 0 26884 1286 0 0
30R Arrivals 1180 0 4396 0 0 0
Departures 0 0 0 0 0 0
12R Arrivals 0 786 0 0 0 2
Departures 0 764 0 0 2 4
12C Arrivals 260 0 654 40 0 0)epnrtures 310 0 802 34 0 0
12L Arrivals 42 0 168 0 0 0Departures 0 0 0 0 0 0
83
,4
Table C-Il. WAFB Runway Geometries (SRI)
Coordinate (UTM) Angle Length
Runway X Y (Deg) (km)
30L 439.94 3684.14 315 3.17
30C 440.56 3684.18 315 3.12
30R 440.80 3684.43 315 2.85
12R 437.73 3686.38 135 3.17
12C 438.40 3686.38 135 3.12
12L 438.83 3686.38 135 2.85
Table C-12. WAFB Runway Geometries (Update)
Coordinate (UTM) Angle Length
Runway X Y (Deg) (km)
30L 439.58 3684.23 316 3.08
30C 440.22 3684.30 316 3.03
30R 440.47 3684.50 316 2.76
12R 437.46 3686.46 136 3.08
12C 438.10 3686.46 136 3.03
121, 438.53 3686.46 136 2.76
184
Table C-13. Aircraft MovemenL over Taxiway Segments
Aircraft Annual Inbound Parking Annual OutboundRunway Name Arrivals Taxi Segments Area Departures Taxi Segments
30L T37 24608 8,7 1 25140 6,19,18,17,16,
20,24
C130H 66 8,7,6,5 2 60 4,17,16,20,24
C140 18 8,7,6,5 2 18 4,17,16,20,24
30C F5 5084 9,8,7.6,5 2 6356 4,17,16,20,21,22
T38 10981 9,8,7,6,5,3 2 13442 3,4,17,16,20,21,22
T38 10981 9,8,7,6,5,3,1 3 13442 1,3,4,17,16,20,21,22
F4 1270 9,8,7,6,5 2 1286 4,16,16,20,21,22
30R F5 1180 10,9,8,7,6,5 2 0
T38 2198 10,9,8,7,6,5,3 2 0
T38 2198 10,9.8,7,6,5,3, 3 0
1
12R T37 786 24,20,16,17,18, 1 764 7,819,6
C130H1 0 2 2 5,6,7,8
C141 2 24,20,16,17,4 2 4 5,6,7,8
12C F5 260 22,21,20,16,17, 2 310 5,6,7,8,94
T38 318 22,21,20,16,17 2 401 3,5,6,7,8,94,3
T38 318 22,21,20,16,17, 3 401 1,3,5,6,7,8,94,3,1
F4 40 22,21,20,16,17, 2 34 5,6,7,8,9
12L F5 42 23,22,21,20,16, 2 017,4
T38 84 23,22,21,20,16, 2 017,4,3
T38 84 23,22,21,20,16, 3 017,4,3,1
85
Table C-14. Service Vehicle Emissions inKilograms per Operation(Arrival or Departure)
CO THC NOX PM SOX
Gasoline:
F5 0.674 0.040 0.002 0.005 0.001T37 0.133 0.008 0.001 0.001 0.001T38 0.674 0.040 0.002 0.005 0.001F4 0.674 0.040 0.002 0.005 0.001
C130H 0.674 0.040 0.002 0.005 0.001C140 0.674 0.040 0.002 0.005 0.001
JP4:
FS 0.072 0.007 0.014 0.007 0.002T37 0.003 0.001 0.001 0.0001 0.001T38 0.072 0.007 0.014 0.007 0.002F4 0.072 0.007 0.014 0.007 0.002
C130H 0.072 0.007 0.014 0.007 0.002C141 0.072 0.007 0.014 0.007 0.002
Table C-15. Refueling Information
(JP4-Jet Fuel) $
F5 T37 T38 F4 C130H C141
Refueling Value 2146 587 1105 1893 1893 1893(liters per fillup)
Fuel Spillage 4 10 4 4 4 4(Liters per fillup)
Fuel Venting 2 0 2 2 2 2(Liters per arrival)
Fuel Venting 0 2 0 2 2 2(Liters per departure)
APPENDIX D
AIRCRAFT OPERATIONS DATA
1. OBJECTIVE
The objectives of this appendix are: first, to develop plans andprocedures for the acquicition and reduction of aircraft time-in-mode andaircraft activity data; second, to present a summary of an aircraft time-in-mode study performed at Williams AFB; and third, to present a summary cfaircraft activity at Williams for the 13-month period of the EPA air qualitymonitoring experiment.D- 1
2. BASE OPERATIONS
Williams Air Force Base is a relatively remote, air-training command basenear Phoenix, Arizona. The main mission of the base is the training of jetfighter pilots. There are three squadrons at this airbase, each with its owntype of aircraft (Fig. D-l), and these aircraft account for most of the airtraffic: approximately 2100 T-37 flights, 2300 T-38 flights, and 550 F-5flights per month, depending on the number of students in training at aparticular time.
The layout of runways and parking areas for squadron aircraft is illus-trated in Fig. D-2. Only three runways are used at any giver time. Runwaysprefixed by a 12 are used only when the northwesterly wind exceeds 10 knots.The suffixes L, C, and R denote left, center, and right, respectively. Runway30L is used almost exclusively by T-37 aircraft. Runways 30C and 30R areused jointly by T-38 and F-5 aircraft. Transient aircraft operations arenormally performed on runway 30C. A schedule of aircraft operations byaircraft type and runway is provided in Table D-1.
3. DATA ACQUISITION AND REDUCTION OF AIRCRAFT TIME-IN-MODE
Time-in-mode is the time spent in modes of ground operations. A mode issimply a subdivision of an aircraft ground (or air) operation. These timescan be characteri.zed statistically as an average for each group of time-in-mode observations. NaugleD-2 has collected aircraft time-in-mode data and hasgiven estimates of time-in-mode for several aircraft types at five Air Forcebases. This present effort was made because of the need for accuracy insource inventory data to be used for Williams.
The results of a six-day study of aircraft time-in-mode at Williams AirForce Base are presented in Table D-2. Only time-in-mode observations forsquadron aircraft ground operations were made. Observations of idle atstart-up times were made from the parking areas; all other time-in-modeobservations were made from the control tower. Approximately 1200 separateaircraft mode observations were made during the study.
The time-in-mode observations were stratified according to aircraft, run-way, taxiway segment, pad area, and aircraft mode. A diagnostic investigation
87
f.
T-37 Aircraft
T-38 Aircraft
F-5 Aircraft
Fig. D-1. Aircraft Squadrons at Williams Air Force Base
88
12R 1C 12Loaton o Txia
Segm Ent an Fa ra
TAXJW AY S5ME T
PAD AEAS 3L 39
Table D-1. Schedule of Aircraft Operations
Schedule Runway Aircraft Operation
I 30L T-37 Takeoff2 30L T-37 Landing3 30L T-37 Touch-and-Go4 12R T-37 Takeoff5 12R T-37 Landing6 12R T-37 Touch-and-Go
7 30R T-38 Takeoff8 30R T-38 Landing9 30R T-38 Touch-and-Go
10 30C T-38 Takeoff11 30C T-38 Landing12 30C T-38 Touch-and-Go13 12L T-38 Takeoff14 12L T-38 Landing15 12L T-38 Touch-and-Go16 12C T-38 Takeoff17 12C T-38 Landing18 12C T-38 Touch-and-Go
19 30R F-5 Takeoff20 30R F-5 Landing21 30R F-5 Touch-and-Go22 30C F-5 Takeoff23 30C F-5 Landing24 30C F-5 Touch-and-Go25 12L F-5 Takeoff26 12L F-5 Landing27 12L F-5 Touch-and-Go28 12C F-5 Takeoff29 12C F-5 Landing30 12C F-5 Touch-and-Co
31 30C Transient Takeoff
32 30C Transient Landing33 12C Transient Takeoff34 12C Transient Landing
of the observations for each aircraft mode was performed to provide insightinto the distributional characteristics. The median, mean, standard devia-tion, and the number of observations were computed. Observations of time-in-mode were contaminated by queueing situations, aborted operations, andemergency situations.
An examination of the size of the standard deviation relative to themean reveals that the inbound modes immediately preceding takeoff and tiheoutbound modes immediately following landings, with the' exception ot taxiing
90
Table D-2. Time-in-Mode Data
Numberl1rt t I, of V t iv a rI
O)perat n'n Arcraft Mloda Setting Observations Md I an M
.;n l1- it i nn ix[an, t ',n
F-17 Aircraf t
Outbound IdL at Star-up 46 2 25 26 1.1O 2 hl.,Faxi 9 13 Iq1 .00 1? 1.61 "l l F emra! I x t Cin
taxi sS 16 1,2.00 160.',9) 3 ,1? Ite lefrli : t Sc ment 5raxt 55 Ill 189.0' 940.491 L. ,in- S frn Pa" 1,
TIxi 9 17 130.00) 134.4r x ' r. t ) -'r i ad
1a t ,, 41 215 n( 21f.1) .' 11 11 1 1 T Pt. 9Last (hance55 s -- I6,, . 2 4h ," ,' I n t
Lant Chance 40 1 1) .00 fn 0, .01 1.'ref light (ibt-l' 45 l '. 9 5n 101. 3: ! ,I'wxt 4 21 )) 22.4 i. I xi ,P.1Itigine Cleck 1)0 - 00 1 . 1 2.05 rvRuiway Roll )on) 41 2) .00 2. rv
I i nib l...t ))I .. .. -- .%ppr,,,w h 81.... . .
)1 ,..I I.indilg Rot 60 5l 1 .01 [21. 14 m q1,a x I )1 ',. 5') 7" . nP'nrl lght ".on 5 64.0 7). ,.lIx 4) llq.50 12 .1' I l , " v: . . . . ;n
aXi 9 6. 164.00 14..7 av I I' " At-X191 35 29 1i-1.1)1 4 .0. ]. IxL t, 1'! 1
rIaxi 5s 14 155.50 1S6. 3" 32. i, '!ax t, 'd C'Sut dilow 40 A 5.1 0 1 A IilI 19),
T- 36 Aircraft
-mat ,,''n Idle -it Startup 46 26 70.(10 282..06 -60 40 120. 00 11.h4 c6 ,eraI Taxi t" P'ont
.axt 60 15 431.)', 421.80 . 3' - -: t ;- n, c' ) ,anr,
r a x n l 4 - - 7 9 .0 0 1 t 1 X I ( P -,'I n 'Taxi 60) 1. 102.56 105.79 . , 1 " 1 , in, , c ,
lxI I0 15 137.00 144.27 3?. L r,ata (tO 7 -- 1,9.29 .I: fr -
Taxi 60 22 265. 00 260.7 26., a,' I r. Pt . -Last Chain-.a; t Chiance 50 49 - '.10, 7, .13 12. 1
Taxi 60 3Q 3O00) 34.03 ]t.>) Fax. .tt ' 1'XPrefl itghr 50 29 46 .00 106. 16 75.'.>
63xiB 41' 25.13' 24.3:' . I-.' 11, ,<1t, S,a.
Engine Check IO0 46 27.00 3-.SI t.1 MIit . rvRmilwly Poll lt0O+Afterburner 38 In fin 19.5 . 2 ] t !r , , ,, lh.q.
Inflisht 'I innb 't 10
+Afterburner --.. .. .. 'lil- ,r, 9:11 A!t ,-.: r
)pprach 66 .... .. ..In,,- Ind .cnding Ro 1i 55 32 89.00 67.50 6 0 6, 76*'axi ( R)fo} I I} onBn RL). QI I m. fo Iax I, P.tt,ant 1)101 6)) 11 90 .00 51.9 0.4
3:i 1-1- ilc;'int (S)lt) 601 25 5I.1 5.7
,' 2. - 1 1.]>: c',P' f l
I-nfl) 0 40 4.0 .(0 46. 60 ' 416ax) 5n. 4. 407 30 3)(13 79,hA1 It, I Inhod . ,...'.I
Fan) 2)) 19 1,10.00 16,. 7 "(l ", 1 i I " P."i 1.
raxi 20 7 -- 347.0) 1a, I Fat t.0 l
Taxi '10 -- 361 .SO 21 .51 laxi t' PIdTax 10 -- 411.0 1 4. 17 7.W,' Taxi I P 0Taxi 50 -- 4,6.) 44 1.2 an a ad I
F-5 At rolrft
0,;1 ,1d !die at 'it:irtl4)l 50 16 74>10 72. 16 'Q"q IIX 1 65 1} 1 .*.00) 1 P'. 8- - 1,~l t~ ,*>l f,~t,
'.ini 65 21 7'im 4219% 7).' FaIxi '.t,- r i t
,' , ," ,':
1ii6-, 4 71 15 O)n 2,1.7$ In f- '1-gx-i t i.,'ap"n Ar-/v /
u-c Chance Min 11 11 )) 1,61, ''"0lI -If 65 13 i.fl)() V)' 7 C 7ii,
6m~
I, Ii r7 I I t 0i q, no 71 97 X I)n9
',0t ,m t I l t0 k -6.- , .'-- '",1mw, I V) R,r I IN -lA ftfrbt..rn~i 12 16.50 1 , P. 7
ll I Ohl I Ill)t I ..) -.I lp r,,;i l . . ... ( ..
In'xit 1 14 77.f 1 1'i.61 t'S? 1.49, I,- ) . -
l', i f tehc /Ite.irmlnt 9) 0 12 i,' 1 0 5 no l ' 0'' ) II, (Ill"0O 14 2.-1. O p If'. 7,1 41)"1
operations, are the most highly contaminated distributujonis. Suc ([n rail2I
tiated di stribut ions create prob lem,; in choos nig :1 spatral d st- ibut ion ,r fiost
typical value ro assiciate with the time-in-mode.
Freflighit and postflight modes for thre F-5 aircratt exh'ibited the I.Irgeststandard deviations. 'Me extent ,! contaminat ion Of the observiitlin-; oftime-in-mode depends on the number or aircraft in a ft ii?,It , the igrouind servi 0:
cr-w, pilot proficiency, and whether arming or dearming is~ t- bep-ron..This range il lust rates the fact that there mnay be more th an on i ueeu
mie. ban ism funct ion ing in an aircraft opera t on il mude.
~.DErAILED AIRCRAFT ACTIVITY AT WI1ALIAMS AIR FORCE BIASE. FOR III!%PERIOD 1 JUNE 1976-30 JUNE 1977
The aircraft act ivity information Lr-ns~ ribed front zl ght forin, A!~3 ,
ATC 90, and TW!24~ Consists primarily ot the dates ai time-s oi a.cttmal takeot.,
and' landings of all aircraft making use of tke Williams Air > orce hase (;WAn?
facilities. Form ATC 3S5, the Runway Supervisory Unit (WI) ieo pertain!operations involving successful takeoff or landing ot locl 1 '-,.aft; rol-m.: A';90, the Aircraft Traffic Log, pertains to the activitmie, oi trauisieiL itcreuir; and Form "FW124, the Flight Deviation Report, F-erLt1I to ) Perati .-
wh ich involve deviat ions from scheduled takeoffI. TFhesp A r Force, f V1eLf orms , used dur ing the a ir polIlut ion mon itor inrg progr ini, w~e e~ cl-~reliable source on which to base hommil- _-.nissrons Lictivltv (lota needud for tlw
Air Quality Assessment Model (AQAM) evaluation effort.
The completed transcription process; pr-,;idd th', ! railt activi-,base, which consists of the daily rcjrds dsrh n 1-t-3 whert edcOC
day's act ivity begins with one li-ad m I C';; giinj thu- vr, , m ath, mliv, andI
fiull tra ffic count on each airhase runiway. Any ikm~er o f Act ivi tv Card.- -na
fotllow the Header but each one must re-state the date, spec Ifyi oiv C. If v, po s s ib I P ope ri t 2-ens l de no te Lthe riiniway i n ue, gve t -ir rt vpe
i.-- Ivol , and I stL tne iit rary t imepo the otei at i ua.
6. Aircraft Mix at Williamis Air F'orce 1dice
A review of thre activity data base for the pe~r tod I Tune 1976-30 Juine1977 shows that aircraft at WAFB involvils scheduled aiid nionscheduiled fI i _htopprations of approximately 50 different airumaft types. 1-shile mutst aircraft-ate small twin-engine trainer types, ethers are he-licoptker types and iarigt"
arg,_ types equipped with either tuirbo1 rep eng, ines or jet unZ
Table 0-4 pr ov id es a ai onth u suminar -T V a, i r :I i t V; e 0 S ' 1: g I.,A lwJ3
Jting tho study period. 'This intoriution also ipptirs asn .-r 05 rt
tne Monthily Actilvity Summary round n Attairmont. 1.
5. Cla-;ifineation of Aircraift *iyp(-- at ',illliams Air VI~ot, F-,(
For sumrriary z epcrt purposr , s;-i c rn . ! i riu a ~ '-mr, wi-n
Table D-3. Aircraft Activity Data Format
Header Card -One Per Day
Col. Content
1-6 YYMMDD - Year, month, day
7 C => Total aircraft counts
8-11 30L
12-15 30C
16-19 30R Full traffic count on these
20-23 12R Right justify integers
24-27 12C
28-31 12L
Aircraft Activity Cards
Col. Content
1-6 YYMMDD
7 N => Normal operations - times arelocal takeoff and landing times.
Slashes separate sorties.
I => Transient inbound ) Times are in GMT
0 => Transient outbound (i.e., Zulu)
V > Deviation - times are local enginestart and takeoff times.
Slashes separate sorties.
A => Non-weather aborts - times are local
engine start times
8-10 Runway
11-20 Aircraft type - left justified
21-24
26-3O30 Local 24-hour times separated by commas or
slashes as appropriate. (Zulu time for- transient aircraft.)
76-79)
93
Table D-4. Monthly Summary of Aircraft Mix at WAFB
Month Aircraft Type
June 1976 F5 F4 137 T38 RF4P C140
A4 A7 139 A6P FIOO c135
A6 Hi AHI SH5 OH58H3 UHI HHI
July F5 F4 T37 T38 C130 C130A
A4 HI UHI H46 FI01 VCI30
A6 T2 T39 T33 C135 KC135
C4 H3P TA4 CH46
TC4 F14
August F5 F4 T37 T38 C130 FIOOP
A4 A3 A37 T33 F105
T2 E2 UHI T39 C135
C9
September F5 F4 T37 T38 C130 A21P
A4 T2 T33 T39 FIOI FI0
C9 HI UHI H53 C1146 CH53
A7 H46 T28 OH58
October F5 F4 T37 T38 T39 C130
C9 A7 UHI TA4 RF4 C140
H2 A4 A37 T33 T28 VC140
HI T2 A6 A48
November F5 T37 T38 TRANHEL TRANTRN TRANRCP
December F5 F4 T37 T38 300C BEECH55
A7 A4 IAI TA4 C135 BARONH3 Hi T39 CH3 OVIO F1I
A6 T33 AV8 OUIO FIOO
F14 UHI OH58
January 1977 F5 T37 T38 TRANHEL TRANTRN TRANRCP
A4 T39
February T37 T38 T39 C130 F5
A4 F100 HI F4 UHI
T33 A3 OVIO A7 T43C140 1357 T28 T2 H53AV8 CF5
March T37 T38 F5 T39 A4 F4
F106 A7 T33 T2 UHI OVIOF100 C130 C4 TC4
April T37 T38 F5 A4 AV8
F4 H58 T39 OH58 H3
A37 HI A6 C130 T2
UHI H46 C135 BEECH 55 FIOL
OVIO T33 C140
94
Table D-4. (Cont'd)
Month Aircraft Type
May 1977 T37 T38 F5 (130 U21
T33 OVIO Fl01 A/4 HIUHI F4 FI00 C9 T41
VC140 H53 T39 A7 C135
H46
June T37 T38 F5 1146 A4T39 HI CESSNAl T33 F4C130 C140 H3 OVIO A6
Classes 1, 2, and 3 will be used to designate aircraft types F5, T37, and T38,respectively. Classes 4, 5, and 6 will be used to designate three classes of
transient aircraft, comprising aircraft types that are similar to F4, C130,and C141, respectively, where F4 represents a fighter type class, C13() aturboprop engine class, and C141 a multiengine jet class. Class 7 will b,reserved for transient aircraft of the helicopter type.
For AQAM evaluation purposes, only the first six classes of aircraft willbe considered; class 7, the helicopter types, will be omitted.
Table D-5 shows a listing of transient aircraft types and their respec-tive classification. Note that all aircraft-type notations are as given after
the information transcription process.
c. Definitions for Aircraft Activity Operations at WAFB
Aircraft activity data compiled for WAFB during the period 1 June 1976-5()Juno 1977 provides an accounting for aircraft operations that were normal,
transient, and aborted or otherwise deviated from takeoff.
(1) Normal Operations
Nonnal operations are defined for aircraft types F5, T37, anl T38only. Local military time was recorded at the moment of liftoff (dur intakeoff mode) and again at the moment of touchdown (during the landing mode).For landing operations involving touch-and-go maneuvers, the military timerecorded was the moment of last touchdown. Whether some Air Force flights didor did not originate at WAFB, their takeoff or landing time was reported inthis normal group.
(2) Transient Operations
Transient operations are defined for all aircraft. Greenwich
Mean Time (Zulu) was recorded for each inbound and each outbound transient
95
Table D-5. Classification of TransientAircraft at WAFB
Class 4 Class 5 Class 6 Class 7
A4 A3 C4 AH7
A6 Beech 55 C9 CH3
A7 Beech Baron C135 CH46
A37 C130 C140 CH56
AV8 C130A C141 HI
B57 Cessna I CF5 H2
F4 Cessna 300 KC135 H3
F14 E-2 T2 H46
F100 OV-10 Tc4 H153
FloI T-28 VC140 H58
F105 T-41 Hill
F106 T-43 OH58
RF4 TA-1 SH5
Tee TRANRCP UIO
T39 300C* U21
TA4 UH 7f
TRANTRN TRANHEL
A48* H3P*
A6P*
A2IP*
F1OOP*
RF4P*
*This aircraft type is questionable: possible data
recording error.
operation. These operations at WAFB account for nonscheduled aircraftarrivals and departures.
(3) Abort Operations
Abort operations are defined for aircraft types F5, T37, and T38only. Local time was recorded at the moment of aircraft engine-start. Thlis
is the only recorded information in regard to nonweather abort operations.
96
(4) Devi at ion Op-rat i ons
Deviation from scheduled takeoff is an operation defined foraircraft types F5, T37, and T38 only. Local time was recorded if the time atengine-start deviated more than 5 minutes, from scheduled start time and againif the time at takeoff deviated more than 10 minutes from scheduled takeoff.Upon completion of the takeoff procedure, the deviation operation becomesidentical to a normal operation; only exceeding it in duration by severalminutes of ground-time. The total ground-time being the full time intervalbetween engine-start and final liftoff,
d. Edits to the Aircraft Activity Data Base
Some of the entries resulting from Lthe trainscription process were foundto be erroneous and were resolved by making a best-guess estimate of actualtime values. For example, some normal operation time entries (sorties) wouldextend beyond the hour of midnight, while other takeoff/landing times forsorties seemed transposed.
In the cases where time entries would extend beyond the midnight hour,the sortie was assumed to begin and then end on consecutive days. This wasachieved by substituting the time -1 (i.e., unspecified) as the landing time,then adding, for the next day, a sortie whose takeoff time was -1 and landingtime was used from the previous record.
In the cases where sortie times seemed to be transposed, the time entrieswere interchanged.
(1) Edits to Normal Operations Data
Table D-6 provides the list of edits made to the normal operationsportion of aircraft activity data for WAFB. The date, runway, and aircrafttype are given in columns 1, 2, and 3 respectively. Uncorrected takeoff andlanding times are shown in columns 4 and 5 beside the corrected times shown incolumns 6 and 7. The indicator, -1, shows a nonoperation so these arenot included when the operations are tallied.
(2) Edits to Deviation Operations Data
Table D-7 provides the list of edits made to the deviation opera-tions portion of aircraft activity data. The date, runway, and aircraft typeare given in columns 1, 2, and 3, respectively. Uncorrected aircraft engine-start and takeoff times are shown in columns 4 and 5 and their corrected timesappear in columns 6 and 7.
e. Full Traffic Count on Airbase Runways
The layout of airba.,' runways is shown in Fig. D-2. Runways preli×.d by
a 12 are used only when the northwesterly wind exceeds 10 knots. The letters
97
Table D-6. Edits to Normal Operations Data at WAFB
Aircraft Uncorrected Corr'c tedDate Rnwy Type Ta Lb T L
760630 30C F5 t600 1554 1600 1654760702 30C T38 -1 0043760702 30C T38 -1 0058760702 30C T38 0005 0057760701 30C T38 2343 0043 2343 -1
760701 30C T38 2350 0058 2350 -1760701 30C T38 0005 0057760702 30C T38 -1 0008760702 30C T38 -1 0018760702 30C T38 -1 0030
760702 30C T38 -1 0024760701 30C T38 2308 0008 2308 -1760701 30C T38 2318 0018 2318 -1
760701 30C T38 2327 0030 2327 -1760701 30C T38 2330 0024 2330 -1
760720 30C T38 2311 0007 2311 -1760721 30C T38 -1 0007760720 30C T38 2306 0005 2306 -1760720 30C T38 2308 0007 2308 -1760721 30C T38 -1 0005
760721 30C T38 -1 0007760712 30C T38 2322 0019 2322 -1760713 30C T38 -1 0019760712 30C T38 2311 0006 2311 -i760713 30C T38 -1 0006
760712 30C T38 2308 0004 2308 01760713 30C T38 -1 0004760712 30C T38 2342 0032 2342 -1760713 30C T38 --1 0032760712 30C T38 2314 0001 2314 -1
760713 30C T38 -1 0001760712 30C T38 2335 0036 2335 -1760713 30C T38 -1 0036760819 30C T38 2312 0006 2312 -1760819 30C T38 2314 0005 2314 -I
760820 30C T38 -1 0006760820 30C T38 -1 0005
760818 30C T38 2322 0004 2,322 -1760818 30C r38 23 30 001 3 130 -1760818 30C T38 2330 0013 2330 -1
760819 30C T38 -1 0004760819 30C T38 -1 0013760819 30C T38 -1 0013
98
°Tab 1 D 0-6. (C(,nt'd)
Aircraft un ) ir ec t sd . r edDa t e Rnwy Type Ta Lb L
760812 30C T38 231? 0024 231] -1760812 30(, T38 2329 0026 2329 -1760812 30C T38 2334 0026 2334 -1760813 30C T38 -1 00247b0813 30C T38 -i 0026
760813 30C T38 -1 00 26760811 30C T38 1909 1621 1509 121760802 30C T38 23G4 0004 2 04 -1760803 30C T38 -1 0004761108 30R T38 1311 1252 1211 1252
761108 30L T37 1120 024? 1i20 1247
761201 30L T37 1836 1728 1836 1928761203 301, T37 1640 1451 1340 1451761210 30C F38 1509 1410 1309 1410770112 30C T38 1617 1533 1617 1733
770120 30C F5 1513 131/ 113 1317770222 30c T38 2224 1907 2224 -1
770301 30C F5 1303 0901 1303 1401770301 30C F5 1750 0849 1750 1849
770307 30C T38 1403 1237 1203 1237
770308 30C T38 1329 0442 1329 1442.770308 30C T38 1249 0620 1249 1620770314 30C P5 1620 1408 1620 1708770331 30C T38 1805 1351 i805 1851770408 30C T38 1233 1125 1033 1125
770408 30C T38 1233 1125 10133 1125770415 30L T37 1939 1058 09 0O 1058770504 30C T38 1744 1003 0844 1001
770523 30c F5 1132 101" 1132 1213
770607 3OC '138 1259 10,2 1259 1402
770614 30c F5 0822 (0( 0 (1822 -1770620 30C T38 234 3 0047 34 3 -I776020 30C T38 ?343 0026 2343 -I776020 30C T38 000 3 0049
770620 30C T38 0003 0036
770621 30C T38 -1 004/
710621 30c r38 -1 0026770621 30C T38 0003 0049770621 30C T38 0003 003b770629 30C T38 2228 0000 2228 -1
aT = Takeoff Time
bL = Landing Time
99
- L , .. ..... ....J ! .. ... ... .. . In ... . . . .= 'm -' " -= - -- -7 -Y -_ Z _. 7 , ... . ... . . .... .. _ _ .
Table D-7. Edits to Deviation Operations Data at WAFB
Aircraft Uncorrected CorrectedDate Rnwy Type ESa Tb ES T
760621 30C T38 0654 0619 0604 0619760818 30C T38 1658 1620 1658 1720
760804 30C T38 1723 1540 1723 1740760901 30C T38 1439 1352 1339 1352760902 30C T38 1545 1401 1545 1601
760929 30C T38 1656 1620 1606 1620761027 30C T38 1630 1605 1605 1630761027 30C T38 1334 1251 1251 1334
761103 30R T38 1555 1514 1455 1514761109 30R T38 1510 1447 1410 1447
761122 30R T38 1545 1403 1345 1403761123 30L T37 1558 1402 1558 1602
761215 30C T38 1156 1116 1056 1116770201 30C T38 1147 1101 1107 1121770519 30L T37 1717 1628 1617 1628
770523 30C T38 0750 0728 0728 0750770601 30C T38 -- 1332 1312 1322
770614 30L T37 -- 1556 1546 1556
770620 30C T38 2130 1249 2130 2149
770624 30L T37 1327 0357 1327 1357770628 30C T38 1630 1045 1630 1645
aES = Takeoff Time
bL - Landing Time
L, C, and R denote left, center, and right, respectively. Runway 30L is used
almost exclusively by T37 aircraft; runways 30C and 30R are used jointly byT38 and F5 aircraft.
The full traffic count on a runway is determined as the sum of landingplus takeoff operations, plus twice the number of touch-and-go operations on
that runway. It is unfortunate that these daily runway traffic counts,gathered during the WAFB experimental period, were inaccurate because of the
several methods used to record aircraft operations. Specifically, three
methods exist by which runway traffic counts were recorded, thus the data
base, in effect, is partitioned into three subsets, in regard to daily
traffic counts: (1) the actual count or the daily average based on estimated
monthly activity, depending on the runway, (2) the actual count on each
runway, and (3) the actual count or the daily average based on actual monthly
activity, depending on the runway.
100
(1) Full Traffic Count for June and July 1976
The methods used to record June and July 197b full traffic count"
are runway dependent. For runway 30L (and 12R) , which are used alr.ost
exclusively by T37 aircraft, the actual daily activity count is recorded.
Traffic counts for runways 30C (and 12C) and 30R (and 12L), used jointly by
T38 and F5 aircraft, are average daily traffic counts derived from estimated
monthly activity.
(2) Full Traffic Count for August 1976-January 1977
For the 6-month period, August 1976-January 1977, the traffic
count for each ruaway is given as the actual daily activity count.
(3) Full Traffic Count for February-June 1977
The methods used to record February-June 1977 full traffic counts
are runway dependent. For runways 30L and 12R, used almost exclusively by T37
aircraft, the counts were supplied by the Flight Records Branch, WAFB, and
represent, for most days, the actual daily activity count. However, there are
several days for which the traffic count is an estimated value.
The traffic counts for runway 30C, used jointly by T38 and F5
aircraft, were recorded on AFCS Form 5a (Control Tower and GCA logs). The use
of the Tower and (CA logs provided an estimate of daily traffic counts since
the Tower, in addition to controlling the center runway (30C), always assumes
control of other runways whenever shutdowns of respective RSUs occur. During
shutdown periods, no distinction is made between runway use by aircraft.
The traffic counts for runway 30R, used jointly by T38 and F5
aircraft, were supplied by the Air Traffic Control Superintendent at WAFF.
Daily traffic counts were derived by equally distributing the supplied monthl;
totals of RSU station name GASSER among the weekdays of the month, i.e., bcfl
weekends and holidays were excluded.
f. Touch-and-go Training Flights
Touch-and-go training flights are carried out by student pilots using
T37, T38, and F5 aircraft. In performing these maneuvers, an aircraft
approaches and lands on a runway, travels down the runway for several seconds,
accelerates, and lifts off. The runway in use depends on the aircraft
involved. Runway 30L is used exclusively by T37 aircraft and runways 30C and
30R are shared by 'r38 and F5 aircraft.
Unlike the hour of landing and hour of t ake-oft describd for normalopt-rations, we detine an hourly, discrete time history of touch-and-go opera-t ions specific to runway and aircraft type as a function of the total number
of operations (landing and takeoff) on that runway by the relation:
101
I*' HA. / N 24TG 2 Nf24ull Traffic Count - (HA..j + HT.)1
ii 2 ~ HA j * j=l i-Iii
where:
HAij = the number of landing operations on the runway by
aircraft type j during hour i,
HTij = the number of takeoff operations on the runway by
aircraft type j during hour i, and
FTC = the daily total of landings, takeoffs, plus twice
the touch-and-go operations on the runway.
In consideration of the Full Traffic Count as derived by the severalmethods stated above, it is realized that the imbedded touch-and-go operations
specific to runway 30C or runway 30R do not distinguish between the aircraft
types F5 or T38. Thus, all or none of these operations could be performed byeither aircraft type. Therefore, we note that both F5 and T38 aircraft are
equipped with two J85 aircraft engines and that all significant operationalparameters, such as their altitude at final phase of approach; their speed atpoint of touchdown, etc., are much the same. It can be concluded that theabove relationship is sufficient for this application of the AQAM, though itmay not suffice for more detailed studies, such as air quality impact analysisof individual aircraft types.
g. Summary Report of Aircraft Activity at WAFB for the Period1 June 1976-30 June 1977
In order to demonstrate the type of activity data available for input to
the Air Quality Assessment Model, a comFuterized program was developed thatsummarizes the Aircraft Activity Information collected at WAFB, Arizona. The
input data to the program is formatted as shown in Table D-3; resulting sum-maries, formatted with different headings but identical bodies, are producedfor the following:
(i) Hourly Aircraft Activity (A one day example summary
is presented in Attachment A),
(ii) Daily Aircraft Activity (not presented here),
(iii) Monthly Aircraft Activity (Attachment B),
(iv) Grand Totals of Aircraft Activity (Attachment C).
Summaries (i) and (ii) are headed by each date having a day-label suchas, wekday (for Monday through Friday), wxekomi daij (for Saturday or Sunday),
zxekda -holiday, or ixekend-holidaij. It is instructivt, to note that a hoIdav
pattern is realized when very low aircraft activity periods are E'xam nvd. Thefollowing table illustrates the observed pattern tor the dates that rpported
no aircraft activity -.t all.
1 0:
Dates 1_ol iday Day of Week
3 Ju l y 1976
4 July 1976 Independence bay Sunday5 July 19766 Sept. 1916 Labor Day Monday
11 Oct. 1976 Columbus Day Monday
25 Oct. 1976 Veterans Day Monday24 Dec. 1976
25 Dec. 1976 Christmas Day Saturday26 Dec. 1976
31 Dec. 1976
1 Jan. 1977 New Year's Day Saturday2 Jan. 1977
29 May 1977
30 May 1977 Memorial Day Monday
Thanksgiving Day, Thursday, 25 November 1976, although a holiday, wasfound to have one outbound transient fighter )peration.
In the body of each summary we see that the column "Full Traffic Count"is an aggregate of all aircraft types, but that each of the other columns have
aircraft type indicators with the corresponding meaning as follows:
F5 Class 1
T37 Class 2
T38 Class 3
F Class 4 (F4 - Fighter Class)T Class 5 (C130 - Turboprop engine clsss)
M Class 6 (C141 - Multiengine jet class)
H Class 7 (Helicopter class)
The repeated indicators are then grouped by the mode of the operationI and ing-i nbund, takeoff-outbound, ground time, touch-and-go), and then,
where appropriate, by the previously defined activity operations NormalDeviaLion, Abort, or Transient.
The frequency of occurrence of each of the activity operations, in modeof the aircraft type, is given for each airbase runway. The ground time ofdeviation operations, however, is given in minutes on the runway.
Six runways :.re identified from which all airbase aircraft operations are
performed. The computer program is coded so as to assure that the followingassumptions, regarding runway use, are preserved:
(1) All T37 operations (landing, takeoff, and touch-and-go) are performed on the inside runway (301, or 12R).
(2) All C13, and all C141 operations (transient) areperformed on 301, (n.b., only 301 can handle the
heavv aircraft).
(3) All 1138 and F5 operations are performed on either thecenter runway (30C or 12C) or the outside runway (30Ror 12) with the shared pattern as follows:
103
i) All takeoffs (T38 and FS) occur from the centerrunway (30C or 12C).
(ii) 88% of T38 landings occur on the outside runway(30R or 12L), the remainder occur on the centerrunway (30C or 12C).
(iii) 80% of F5 landings occur on the outside runway(3CR or 12L), the remainder occur on the centerrunway (30C or 12C).
Just as was stated above about the holiday pattern of aircraft activity,we note here a weekend pattern of aircraft activity. In general, there arecases of weekend airbase aircraft activity, bct these events are not recordedas ar,: the weekday events. Some specially requested weekend data weresupplied covering the weekends of December 1976 and January 1977 and thesedata, though limited, served to define the weekend pattern of events atWAFB.
REFERENCES
D-1. Zeller, K.F., and R.B. Evans, Airport ,uality and Its Control, J. ofthe Air Pollution Control Association, 16:12 (Dec. 1965).
D-2. Naugle, D.F., and S.R. Nelson, USAF Aircraft Pollution Emission Factorsand Landing and Takeoff (LTO) Cycles, AFWL-TR-74-303, Air Force WeaponsLaboratory, Kirkland Air Force Base, N.M. (1975).
104
onn
Cl).
C, , ' I
C) '.. 0C)~~~~~~CC' C. ' uC.) > C .S. ' (C C C ' C ; C
C '0 L- C3 ) '.. C -) - ' 0r
C,. cC.),C ' c O ''C d j- .O~''.C 1 c c'C e .,
'Cll
'C ~ ~ r rt - ))).CC ) CC CCJ,, C 0.) C ) CC CC C C
c yD
ol
C' OO CC 0 ) C) OO CO 0 C)CC )) C CCC CC CC S
L LC)JY(.C ' CCO C'C J C).CC (C )C '
* ' OCC'.'..C; CI C0 )C .C) C. C).)C.C.C C) C.2).) C)1 1C'.P, (
3 '1 ~ ~ ~ ~ ~ C C) OCCC) C C, CX., C) 0 C2C" 0 C1 ()...C( 2 2, 11C,'. 1
-P IO a'( C C . C ZO CCC Cc CC'I>CC 14 C;,'' 3 2rC C
C> .lr c c, c - ; c, c - -
c- 0Cz C) Ca I0.1 0 0 Cc 0C C C OOC cY L' ~O.C.C
(3j CCC71c
c )Dc CCa0 ( 0) 0 0 0 0 0 0
"O O. c C) CS'C 0,C c .CZ C c
C). OC..00C.C) C)LS OC C3 C-r . o C) -,o~ C , 0 C
1) C7 0( ( 11 c 2)ZlCC 0 00 )0 .3 C0 CC O ,C C c cl cl I I C-
cI c> CD- C.)Cec c .9 0.D aO O .4) 0CC Cl .c, C, c)C )' ~ CO CC
9' j-- C .
0 306
C C I 0 C' ' C C CC C C, C -1 1. 41 C , C C.. CI I -.. C C s .C 5' ')
>2
CC 1C 1; 1 9 I clC ;C C:I* I; I I
C,'
* C' : O C C C , C) CO2CCC C C) O " C CC i O C C C4
CC' I -CIC'C>)CC C, CCC5-'>C 0 C.CC 'CC C '
C CC") 1 1C.,'~ "C> C.CC C.)CCfC'C, Cl I'> CC'C C.) C.C)CCS 0 C
rC I' C I'C'CZ. C C, C , i'C C) CC CS wCic t CC) IC') ' ., C C
C.. CCC C 'CC) IC C C( C2 ' C t::CC JIC ) C) C) '. V'C ) C
.C, " CCCCCC'C C) , 0 0C 'OC 00 1 00 0 CC CC0 .Cn '
I-> ~ CJ O C C, lC C ,)CC~ C) 11 C) C 0 00 01
CC)1'C-' .) I~ CCI' C. 00 IC 1'O C'C "" CC'C'' 004 C1 I
11 )
IC I C' C I.cfI C OO C '7, C C C , C OC C L, '''2 3 C C
I CC CC 'CC C, r CC.)"CC C3 C CL 0 CJC C:, C" C' C'C) C") C
J,. I-- I" I- I C
107
C' C Cl (4 Cl r
I.- C' C t
C -;
I ~ ~ ~ - 11 c C tC. C. C.CCI) CCOC I" C. C,))C.)) I) *)C '
C C C-C; L CZ C ) ; CC.) C()C.IC C;) 1 CCC4 .0) C.', l).
IC C) C. C, C) (C C' CC '( C C*.) C C- C -C ; ''C .' Cl 1 C I C
* C- C ;C;- C
* 0C) '.CC C; fl'C .3 C I4 ) C ))J C C. C) -0 4 1"0 I I I
I4 ;
C ; VC C2 CC. CLC'4)00f Lc oc" c' C . c 1 C)CQI4)C'C'(I I.
I
CC C C *~ 0 0 C) 0 0(1 00 CD C2.3'O DCC C 'C ICC CD1;
C; j
D) -t
100
11 Cl£ P- .- VJ r.C) I fI
'r L
I- C, CC T c, C) C.,.fC..0 C . CI., 0 D Cc C) O C C O. .00 0 . .~~0 ' . ..C C ~ t . .C 0 0 . C.,.C. C. C. .
0 C( c.C C OC )C 0 Ci ClC, m -, C) C 0t O .0 C
cil
.4) CI,.,.I
C'- ((2 n*J~l C 04 OC O rC 1 a; (0 00 0 0 0(.C <,C, CO C C CC
C) ,l Cr, CcClCrr C0 O I0C-0C 0r 00 0 00 0 0 CC C) 0 C-3C 1
1") 00(0000'' c C 0 G.(C.00C C ;1 I 00000( C CZ O c)C I" C-iC, C.
_j u r, j I
n. t.*1~~~~~i :-' CC ,,) C C OC C ) C CC 0 C C) ' 0 - tr C'C
C10
C)
t-. 0'<, C. . 11. '0 c) . C...:)) I. i C,. C) C)
7z.1i C l C CDCC0C9C C, C .O4OC9CO) ') 1" C t CO 0 C)C (00. .
n~ C 0 C CD O CC) C, (D0(9 (M ).~ C')
* cl CO C4 1 C) cl) 00000'' C C) 000000 C C C C)9( C-) C, C'
I in 000 C 0 (7, tJO ,;cS c)C) C'00 C) 0 0.7. CC C .. CD
i*. -: C
CD C)r ,II C C. C 1 C C C )1 C ,c (1
CD) C,)m..( - Z C C n - lC :)C D C
CC -C CLA 00 00 Ci 0CJ >00C-) 0 0.7 00 C 7 OrC b C
ci j i cCC:)
99 000C: CE. 00000 0 00 00 FY00 C CJ:i.
LAAV
'i ) . 3' -r
N . . 3-LI .
33- i 3-, ..- 3 -,, , .
- .- -t 3)., ., o [ t
o : r .3-
(3-. (3(.' 3'3-CF
'-0
_ )-. ,.. I
-3, 3.'¢; .,:<) • ,) - , ' , (3
3- .-, 3
- ) )-
, ' .;
I-IF- 3393
H1
- ~ .1
r-.< if)., C) If c'it5
t)) C C 1- 1'
I Ceo ) v
L) I r-11 n -'C) C
E. C- In c, C, t)V '
C,>
c- C3c, c' ci~
(A.ic)C)C C3
17-4 C)
C 7)0 C)C3 C) Ci CDC 3IDL
0j
CinI I .0) 11
C) C)
In ci si, J 7 I
IV Cr In
I~ LC C
I .7j
I C, :1IL
(C IC~Ci5C,)C, ) C- ~C I r.~CCC,,O
I."
, - "l ,t i ( < , ( i,
L,4>
4 14 C 4,')Cri I Iil
, -, , , -, . , ,.... .. ..-
4.- 4:
i* I
LII
i , . . -
l.~
Ii l.e
P ' II .4
-* I l f I I ,( I ,I
4 -i-1
.1 44) I ( I 41 44'4
.. ....4 4 I '. 4')t1 i 1 " . . .. . . . n i l I . . . . .. . . . I - - I -
c- t c) > I > I C I I
I1 r 4 () -
LI-~~ )b I. .* 1 I
.CL C-'*
I I 1r-- C)> CC > C)
1- rcJ .. 0' c C) '4 LC,. CID.) *
I n 47D ~ Z)(> <i Cn 1>
1If) CD 0>- i: C)CDC C
L -> -
I1 t'1 In Ill t71C-f) C,
2~I trI~ : 1r (>
0) C
-l C'
-. ~~- U,'If O ~ f O
IN0 I In)r Ci C> r .
Zr I-
C3 ) 1 t-- It> 1-) C7 C-C3
Lb -. C
-- No
c I I I C > f
9 Ct -9I .. t C
J:~ I "'I'I 9 '9)
I C) 0 9 ' ~ 1 ' U. C
c, 1
9.n ~ o I- I9,. ') (9(9) 1
In - C ' C' 9
19~~ C I- I -
II I' I' ' In
-@up,
4(4 4 .J . 4..-CC"(.) 4 (3C.41;4.404(4c3 .4
4.) 4 CC.4<, 44)..-*C> 444'~ 4
~.44 4.444 4.4. * 444 . .4
I 44.. (.4. *'i4 (4 4 4(4
*44* 44. ~ 4,
4 C. C) C) 1.4 4 4) 4') 444,444 CC* 444 (4 42
4. 44
414 44 (41 4 4. 444 444
4 - .4) 4 -, C) C . Cl '(4 4 4 4 .. , 4
4-4'4 4,
4,. .44 rc.; 4444444(4 4
424-10 4.4 4
44) 4~4.
C).
4 )c)4'-42,, 444444) 4
1c..44r
4r-.C442)4' 4)4 4
- 4'. cr. 44- 4. 4 4.-
4 ~-Ccnr <44.444 44
.)Cn<,:)4fl0 C.> 4 -.
'4.. 44444 4.4 44
C'. C.4.4' C'C44.44444.4 44 4) 4-44 4- 4- I
44 4CSC4C.fl4.C4 Cl 4 44 (44 4 4)OC4C'4'C.40 (4
4 4 >44404.C4C4C40 4-- 444)4 t.0C42)
44flC> C)
4 4fl ID ~~444C.44- Cl "4
C4C(C)OO4J 44 444 4flC'.4jC4444. C)
4 tC44 '44.4C44t4 $(44)4, <4444414 4..
4 4 44- 4.- 4
44~ 4" 4
4.ic.)4144r-.4.. Cl I 4 -*C~ 4C.4c' (4-t 444 .444 4~3 44 44 444
I .. 14- (4 414 4 I
4 ,4 4 .C44?4C44) Cl 4 4 4- 4~'). )C. 4)44 4')4 (.4)444 4~.
41.. ~. 4- 4 4- '1 4.4 (44 4 44
4-4 L(4044C4044'4() (444.4-C 2C44V4Z4 4-'. 4-4 44.. (-4~ *4~~ 4 '4 $4
44) t4 7 '4 4 7)
4.ID
I Y)t.)4.0C4040 .4) 4 4'4'.C' 4C'C)C 44 .14 4i. 4 4.. 4 444
'4 I 4- (4 4') 4 4' 44)cu 4
C U.. 4-44 4. 4447)14-14)4) C)-~' (IC) - 4 4-C5c,4)flOC) C4 .4I L.44'4~" .44) 44 4444 44~' 4 4- 4.1
1 C) (.2 4 4.)I- 44.
4 LThOtICd44-~ 4) Cl I 44)444.4(44004) Cl 4-4 4.4. (U' (4 (4 4 14444
.4) '44 4.
.44.4) .J 44
4.44 4 44.4 444') 4 444 4 4444444 4 4 .,'444 4 4 4
C 4 '4
I if.
- '''-Ni ~ C -~'
-. '.ON~I
I iICI 9J' I
CI . CJ1 ILIIC e
.. LI- rC,
'Cl-
(. i(C ii
II I"'
. . .. .. . ... ..'";)i [,I it--,
1r- I,
Li.*C '' I I
II. . - JI* l
--
1 ) LI L-. I
~ ~'--~ C)) 'ii
i0
40 0c r'., C> In,
F--I LI,
> -. Ff, r Cjoc- c, 1 ,0 'i. ZC,'
CD f1C) 10 : C) CC) j F C . ) - c, " c
In (' r- 0 0 . 0 1 f . If)( 0 ) C C .
C, F") :I Ca)C
1L F- '13 I=I C)j rJ D)L
U) CI. I. C)Ci50>.').04 C
* i 0 0 0 0 C 0 -) (.i-i I C!
<'C' .-. 9C'C' I .. CD.i
01
I) ' Ci
- ii~ (ifl'~CflCi.) C)
*tl I'1
2(tflflC9C)C 2 - .-I -
rn . . - .I--f I (4(7)1) 9CtiCi CS C)
I C-,*1 I-'")
lnr-,C.icii~c-lC. C"
Cr-
'SC'-' F'-)* - C) CC- C' 'SCC' C)
i-li-' (I LI I- *l CCI I -. -, ii
ii) LflCflIDI.i~C(CI ii
I'-I Cm
CCI I'i'IC' C - ' C')'I', -
III i.. LI -'tilL.. 119101'2C') Ci I '2
LII I tic,' ICI1CiC~CC)I i)~ .&J
I-- I
- I C Ci' ~) ~-t- ,i C'I Cc;-..', c~ Cl '9 IF'2<' "9 (3
I 1-- tim it) -
'2 Vit'-.'.oCl'~.Ci'C2 I"- CI I LI'ii,-.nC'C~9'~C) I-
I- . I- '- '- Cr' ''9
LI I'-''-. CI $I' S iiThiiCiC'OC)C'li C) I f(~CtIOCOC9 Cii. i1 '7
C'II
C-I I cCIcirI('CICCI ('C 1919CC' C'SLII , - 4C' I I I I-- -) I -
I"I I
nI Ciii? iCC) Ci I I V.flciC'FsrZII) .4)
C&I I
A I- ' ' -.
-I -9 ItC' 'I
CI
I Ij 'C'-. 11- Cl I") ii-- *- .- 9-- I I
.sIW '.iI't.I 4) I NI?- (I'"U)C~ Ci .9I'''- iL' CI IF~)'- C')I -' I-U' (~J'. IC-- r.i
-I U:
I i(iC('C iI.)C'C.Th'" I UIIII.C.iII'Ci 00 C-ii- C) i-I I Cc
I "'I
'1-I -J U
Ci) Ii ('CC) C-'- ('''Fl Il-Ill I-- IFC iI-iII'. C') 'iliFI' I~ C) ii
C- I-- -
119
-ie
r e
CC..) C C CCCCC'It) C ' () C -1 CO '
12CC.:- C--iC I .r-C-<, C CD I
C- C-, CD.C
CC
C C 0' C'C C) cI =
0 M C : C C-1 C) -CJ-C)
0 ) CD C C. C
CCC CCC n c)
C j CC CDCC-I' IC I-)
'C, n- "-C CZ F)-)
CC CI I 2f)
.
S (St
(CS . I
- C :5 - (VP C -,
I C C - - C) C5C- - , . ~*S - -
C-S - -I- IC - '- -
oc~rS C -,
I ~tu'')vCC)L5 C
-, 'I P~~C p .-. ~ Cp PS pp - -
p., C-p p -
I1-JC
(P (P Ci 1 C.(II CC ~ p p - p
C I-I
VI C> 115
p I CDp . -
CflSCtPCCI .SCCCVCCCPl PC(4
CC
-, I
C C;..2C)(I (P 1' p 'F-Pp CC OC-.CC2
Pfl --
* C-
UP' cPC)5 r--CC-zCC', p..
CC p 'SC-C-f CPZsOIC C-CICC5)O() C,C-C PCI P - - C
.1(1 1-2CC-C
(4 (J I p
VCCV)ICCC C-)C4ppPPp ifl
CS C p
Cl P1
-- <CCC PC) PCI, CCC- (CISC-S-sU . IS ,
.4-- P
CC P-. pCp.5CC5CIpI5Cp PC-p -c
Ci (J P --
V)CCCSCCC) Cl PP. P'5CPCPC)CCI) PC-C Lip 'CC (4
-. 'S 'I pP C-:.~ (I ti IC--
P IPPCSPPSCPCDCC C-. I pflrp&C'PCCIC, c~C-- P-- I
"4 pp..4CC ('I I p Cl *jPCSCPC(Ij -( P.C
CCi C CC.PSS;CPP- p P-
I-- C-- -i
12 1.
.CCCC5 'ICTC C
C..C). l~CC C 'C
cn 7,. C7, .-.- C-
CC.J~C - 17))Ci
C- C, C,
CC cl-)'CJ C) CI
C--)
(jC)C C ZCDS)C3- c: ' r, kC " t.. CC.1
C C I cI : 1 C.J (.
I jr. n )1 7CClfl 7S .-. 13 C.) C...c C-
c:, C3 (, cC..C. III .
cCC) I. .' Cr ( <I - c .
r' C) C)c)) C) Cc) C), I -7 , IAC l
LIc- n 3--c cj
r.0
j U,'.
CD (
LO > .. J IiL'CrCI i
. . . . . .I . . . -
--t
I cr
-~ ~ r -. r -
i. i. •i ,+
', C.I 1 I ,In C
0' ,C "
, -
C'
1,2
¢,.) I(+¢) .l(.) CJ I .+( (i C.) 'j 0 j C' )
' - .4), I - ( e '
9 iiii I-
in I
1233
or (35
G IS
tic I55l) S -f IS IfJ S.
L - r* C
C- C, 5)CS CX, CU
I-
0C C--fl-
-12
APPENDIX E
ESSAY: VERIFICATION OF SHORT-'TtERM AIP QUALITY MODELSUSING THE GAUSSIAN DISPERSION APPROACH
by Karl ZcIler, EPA*
(January 1977)
1. INTRODUCTION
A need exists to define an acceptable approach for the verification oft short-term Gaussian air quality simulation models for single and multiple
source complexes. The demand for answers to air quality questions is sostrong that some models have been formulated and applied extensively prior toproper evaluation. It is hoped that proposed models would be subjected to averification procedure to enable the model developer either to defend aparticular use of his model or to indicate possible situations in which hismodel produces invalid or questionable results.
Each complex Gaussian air quality model is comprised of submodel'- todescribe source emissions, pollutant dispersion and transport, plume rise,source-receptor geometry, and meteorology. In the modeling of complex gi.o-metric situations, it is currently necessary to use empirical dispersionparameter values (standard deviations of plume spread as a function of down-wind distance or time and atmospheric stability) derived from continuous pointsources. Sometimes these values are applied with subjective modifications.For example, models used for airports or highways use the same dispersionparameters as models used for elevated area or point sources.
One logical approach would involve the separate verification of each ofthe submodels based on specific experiments; however, expenses involvedusually dictate that accuracy of the overall model be examined.
There are many opinions as to how a Gaussian model that us-s s') -tive modifications to handle the various (onfigurajtions crncountered sli:validated, verified, or calibrated. In this discussion, a specific approa(hifor Gaussian models is suggested. The data requirements ftor model verifica-tion include meteorology and emission data (the necessary information termodel concentration calculations), and measured air quality concentrations(including background information) with which to compare the calculations.
Previous verification programs comparing short-term (1-3 hours) cal-culation, (Koch 1971; Rote, et al., 1973) with observations have demonstratedthe difficulty in coming to specific conclusions using any one comparisontechnique or statistic. The scatter diagram approach prevalent in theliterature has not been particularly useful for verification because of
wide scatter and low correlation between measured and calculated concenitra-tions. The reason direct comparison of calculations to observations does notprovide good results is the statistical nature of the complex Gaussian model.For instance, the Gaussian approach assumes ste'ady-state conditions over aspeci fic period of time, usial ly one hour. In reality a cont inuous p( int
*Present address: Environmental Research and Technology, Inc.,
Fort Collins, M,
12')
source plume is subjected to many variations of wind velocity during any givenhour. When modeling intermittent or moving sources such as automobiles or
aircraft, simplifying assumptions are necessarily made; for instance, taxiingaircraft are usually modeled as line sources considered to be continuous over
periods relatively long compared to individual aircraft taxi time. In real-
ity, there is not a continuous Gaussian plume stretched out downwind of suchsources but intermittent ones which are locally distorted by variable wind
speed and direction (turbulence). Data collection for such comparisons ispresently accomplished by a network of monitors at fixed horizontal andvertical Locations. The total number of monitors or stations in most studies
is usually less than five due to monetary restraints. Considering these
limitations, it is not surprising that short-term calculations and obst-rva-tions do not correlate well. The comparison of cumulative frequency distribu-
tions on the other hand has enabled some modelers to make general statements
about the overall performance of their specific models.
2. RECOMMENDED VERIFICATION APPROACH
It is assumed that a large data base is taken over a period long enough(at least one year) to provide information under the varying conditions of
71eteorology and emission modes. Assuming such a data base is available, thefollowing procedures for model verification are proposed.
a. Discussion of Data Base
A discussion of the data base should include efforts made to providebackground information and quality assurance during the actual collection of
data. A discussion of the adequacy of the numbers and locations of actual airquality monitoring sites should be included, together with specific geometry
relating sources and receptors to the general meteorological conditions during
the data collection period.
b. Comparisons for Analysis
A thorough verification requires that data and concentration calcula-
tions (model predictions) be compared under a number of different circum-
stances. Because of the conglomerate nature of the complex Gaussian model,
certain meteorological conditions or receptor locations may give better or
worse results compared to others in the same model application. Each monitor-ing location should therefore be evaluated separately in addition to overall(valuation of the model. Evaluating each monitoring location separately will
give the modeler an indication of the performance of the dispersion submodeltinder different situations. This is important because at the present time the
dispersion submodels of most Gaussian complex models have been altered and are
subjective in nature and not based on applicable dispersion experim.,.ts.There is reason to believe that a great deal of the wide scaitter in observed
and predicted values already discussed is due to the dispersion submodel
as well as faulty ernission estimates. The pollution data should be stratified
in at least the following catcgories;:
120
T
(I) High and low emission density periods
(2) Periods of the day (can include nijre rhou one hour)dependent upon emission opiritions and ,e teori.
(3) Seasons (this category is opt ional - deptnd.s on the
nature of the data and problems).
(4) Meteorological categories:
wind direction - increments depend on situation.
wind speed - increments depend on situation,.
stability category - Pasquill categoris oi in some
cases stabl , neutral , and
unsttable will h sufficient.
mixing height - two or three categories based
on a chosen mixing heightdependent upon the scaleof the model applicat ion. F(rinstance, in the case ofairport models: mixing heightbelow 100 m; mixing heightabove 100 m; and unlimitedmixing.
Note that the above categories are not meant to be mutually excluive.Also, only data that are above the noise level of the particular pollutant-monitoring instruments used should be evaluated. Although there is a specific
interest in pollutant levels approaching National Ambient Air QualityStandards, all levels should be evaluated.
c. C'umitlativo Frequency Distributions and Error Limits
The data should first be sorted according to tile above scheme; ftrinstance, all cases in a specific wind direction category, al1 cas..s durin,the' morning hours, etc . (approximately 26 separate cate)ories). The . data
shouid then be displayed on cunmlative frequency distribution diarams,similar to the on,, in Fig. E-], if at least 30 observations within a giv I
cat,'gorv are available. If there are less than 30 r)bservations, a scatterdi a)ram of observed versus calculated concent rat ion should b, preparf.dLon,,er term averages, for instance , month ly or yearl v, wi I I inherent ly hav,.
fewer dat a points for comparisons and therfore wi I require tile scat tecrdiagram approach for verification. The use of the scatter diagram willg iv' a q,,al itat ive feel for the performance of the model even though ill somecases resultant coefficients for calibration purposes may be lacking in
s ign if ic anc e.
It is desirable to have a number measure of model performance for spe-cific cases. A c'ini.ilative frequency distribution of the calculation errorshould he constricted to) al low the model hr to report the perctnt age of
trials for whiich his estimate was within a specitic range of the observation.A similar approach was used by Koch and 'rhayer (1971); trequency distributionplots of absolute value of overpredictions and underpredictions similar to
127
--u
7 /
1000
0
2% 20 50 so 98%
CUMULATIVE FREQUEtNCY DISTRIBUTION(CUMULATIVE PERCENT)
Fig. E-1.. ct,-illative Frequency m~strilut ion lor Sit i., dulrtn. e-,,o t" "Vind Direction.
128
I.I
.. . . . .. . .. .. , .... , - .. _ . ../
those in Fig. E-2 should be prepared . The appro i ', 1, ted in this discus-
sion is to subtract each observation from .ich correspov idig model prediction.
This wi 11 provide information on tl di st r i b-t ion L1i eve- rpred i ct ions and
underpredictions in addition to the overall tendcncv of thc model to over- or
under-predict in that specific cate1,orv . Suhtr jct g th cumulative percent-
age of overpredictions from the cumu at ive -,rcent:,' I -f und erpred ict ion
(see Fig. E-2) at a specific concentrat ion di fference wit i give th percent of
comparisons that fall within that specitic enc,'ntration oror limit.
The error limits and corresponding percentages cair th,-n be plotted oi a
frequency distribution diagram similar t" Fig. 1-3. fii. modeler will therr be
able to report that in the given situation hi i redi't iors were within
x (Pa g/m 3 ) y percent of the timen for that prt i,_ul or d .ta categorry using the
verification data set. It must be recognizel that th, percenitae,, of error is
not displayed in the above technique; i .. , the rirnber di f frrrce between a
predicted value of 2 :ind measured value of 4 is thr siame as bctween that of
100 and 102 but the percent error is quite different.
d. Percent Error Distribution
In order to evaluate the percentage error ind makii tire verificat ion
results usable for all ranges of concentrations, th,, following proced ir,
should be followed:
Using the same categories previously discussed, prepare frequer,v
diagrams of the percent error, E;, as shown in Fig. E-4:
xc ( b)
F; = *o -( X )o b
wh re
o; pertcent er ror per cas, ,
0 mode I calcutation without backgrorind,
o = observed air quality, and
b = estimate of background air qua1 litV.
The display of Fig. E-4 will al,;o enable the model evaluator to discuss mrodrelb i :i;, E , rand om n e s s , u E2 , a nd ov e ra ll v a r i nice , t ,r ,,A l L a t ,g O r,:
wh rt.
In
F X E. (bias),
i=
" : "- . ( - ) r-pr.-;,.nta th rarl'nrmress, ;idL1 .n
i,-oft
C)
Lu,
W
LJ*)coC)
_ 100
2 ----
COMPARISON-~ DIFFERENC
S20
(10
m
C)
c ,
u-JC)
Cj)
I PERCENT OF OVER AND UNDERPREDICTIONS
2% 20 31 50 7 7 80 981/1
CUMULATIVE PERCENTFig. E-2. Delta Co~ncentrat ion: in T i- xamrpl e, 77- ,) , (-r 40 ( Ptr,(nto, the Tfme, tb Di 1 f (vr nc I-t 'I~ TI r(1' on Pred it cw"; withir, ,20 .gm
I 10
TO0
r:L
+1
If)t---
-iEXAMPLE FROM FiGURE' E2.cr-_ 46% OF THE TIME THE MODEL
Lr~~ PREDICTS WITHIN +T010
6 - QUESTION. HOW OFTEN DOES THE MODELPREDICT WITHIN + 6 g/M32
L ANSWER 20 PERCENT
2 20 50 80 98
CUMULATIVE PERCENT
Fiy. i - . r rr 1. i iit i Nia ,rau,,
~1 -1
BI-AS
I E: BIAS0r2E= RANDOMNESS
C) r2T = OVERALL VARIANCEC-
0
I-r
-E
-20 -10 0 10 20
PERCENT ERROR
Fig,. E-4. Percent Error Distribution
u_ 132
2 the ovtraIl vat i. : .
i= I
This inlthrmat ion ouIld also be displayed on a po!r, et *-rror -- .nemriwat ivefrequency diagram similar to Fig. E-5. In this dia,'iam, the slope, curvature,
and position from 50 percent frequency of the lite woil indicate randomness,normality, and bias, respectively.
In Fig. E-5 -i horizontal line through zer., percent error represents a
perfect model. A very good model with no bias, allowing :o)r Lhe random natureof the model input parameters would possibly ](olk like the distribution
represented by the dashed line. An example appliCatia:j of this diagra. wouldbe as follows: suppose the modeler was trying to substantiate a specificmodel calculation of 8. He could say that in the category represented by
Fig. E-5, solid Line, 95 percent of the time his model has an error (f 2.jpercent or less. Therefore, only in one out 0t twenty cases when 8 is pre-dicted (by the model) will the actual value x.o be 10 (i.e. -2((8- )/x) or
greater, or there is a 95 confidence in the value being, I0 or less.
e. Emission Sub-model Adjustment
Beca,,se each of the inputs to the model have uncertainties expressed
as a standard deviation ( a), the model output cannot be expected to matchperfectly with the measured concentrations. However, if the actual variationsof the input parameters are independent of each other, and the input para-meters are the true mean values of their distributions, then the percentageerror assuming Xb = 0, E = (Xc model - X0 observation), should be normallydistributed with mean zero and a total standard deviation, o . If the meanE is significantly different from zero, as determined by a statistical teqt,then the question, What is causing the error? should be asked. Assuming thatadequate quality assurance is maintained for the measured pollution concentra-tion and meteorological data, the largest uncertainties will probably beeither in the emission inventory or within the model itself.
Although in theory, the measured product of the concentrat. ion, Xo,
observation and the wind velocity, i, is a flux (mass/area-time) and can be
used with a model to compute the. emissions Q (gm/sec), this inverse of thestandard modeling technique is not practical. A possible technique to test
the emission inventory would be to break the data sets into two significantlydifterent meteorological conditions such as stable and unstable. If the
emission inven tory is in error and t he model is consist ent , E slou I d showsimilar bias frotn zero for both periods.
If thet F values are, significantly different tr both conditions the
em issiF inven tory and the model may both be in error,
Such an aialys is can be usd to ju;tify moditying the emissi innv l)orvQ to set E to zeto for beth data set; as a whole. If this adjustment ofQ i ; valid, tien ;I third data 'et , independent of the, two used as abov, at:
be used to detennin, E. If E is not significantly di fereot from zero forthis new data set, then the adjisted Q values wouildr appear to be the, valitones to use for further sqttdv of the model.
133
CATEGORYEVALUATION
+20 ~-RESULTS
10 i
BIAS
JIDEAL DISTRIBUTION ALLOWING-(0 LFOR RANDOMNESS OF INPUT
,PARAMETERS
550 95
CUMULATIVE FREQUENry
Fig. E-5. Percent Error - Cumulat ive Frequenc\' P tigrirn
I M de I V r i I i( at loll Dii 5( 5 il
A subject ivv d i scuss ion shluuldo acconipainy -, .-h -)mpa r 1i, n , proi nt ii P o -: t
various aspects spec ific to the modt 1 and data thzio he inr ulSkd. Fue diiscus-s ton w ill depenld Upon the interprotat ion cf the investil-atr.Nx the datashouild be mult istrat ified using the abovke single L'11; I Z inst 'ince , on,
stratum may be all cases (luring winter MOrning period. w.i th wind tflow f roma northerly direction, atmospheric conditions stab>t, and 1,w .,;,-,a~~d Inthis case the categories would be mutually cxvlusive. 1ibis data comparisonshould also be presented in both frequency distrih)utjion anid scatt.-r diagramsof observed versus predicted if there ar- fmare thaoi .30 i-x-impl es,; ndo scar. ter
diagrams, only, if there are less than 30) examples. Ecu i , ' istribu. ion And/orscatter diagram should be evaluated and sub ject ive I y dis( is s ,,
It is recogo zed that there is a great expense- involved in extpnsive,model calculations, to fulfill the above requirements. Perhaps, in practice,models can be evaluated generally in the broad single category groups using asubset of the total data set to evaluate cumulative frequency distributions.A few of the multistrati f ed categories should then be selected for specificmodel performanct., preferably those with the most data. When the modelis then used in a spe-cific- application and that application is in question andnot previously verified, the data base would be reevaluated for the specificsituation in question.
3 . CONCLUSIONS
Tlpon completion ot the above analysis, assuming a good air quality, database, the modeler would have a good idea of how valid his model really is. Inaddition, an indepenldent model user or Environmental Impact Statement (EIS)reviewer, would be- able to scan the yenificat ion report and obtain ag quic;.
[eel of liow the model per forms compared to a real data set.
Th., reco~mmended yenificat ion procedure discussed above does not reLquirespec itic model correlat ion coefficients or any other number on which todei',rinn a pass/ fail grade for the model.
'In( final decision, therefore, as to whether the model is valid, will nutdepen"d upon any specific perfurmance limitation; I .t will depend upon theuser's ability ta support the use of the model in ainy instance, bas;ed onthe verilication analysis.
A CKNOWLEDUG ME'NT
Thanks are OUe Dr . Dayvid Mage, EPA, Research Tr i aug 1- Park, N.C., orhel pful suggest ions tised in the preparatioun of this disciss inn.
Koch, Robert C. ., and Scott 1). I'h.iyvr , Vql f!ctinn cmqn Ot~Ot.iiS
the Gviir0tan /T7 ur'?'e.io j/'§ou'e~" Irar i)('3 uri- n 1"o!,,/ , t,enmo't Rep. No.EF-60 NTI S No. P13-206-951I EPA N,). APTI)-093 5 (197!)
1 3'
136
Rote, D.M., I.T. Wang, L.E. Wangen, R.W. Hecht, R.R. Cirillo, and J. Prat apas,,4-,pont Vicinity Air Pollution Study, Argonne National Laboratory Report toFederal Aviation Administration, Report No. FAA-RiD-73-113 (DOec. 1973).
136
S I]I ';'1 1(.AL ANAt, YS I'; lt A 1i 1('OP IJI t, 1)A1AAN) liE VALIDlAtI 'UN 01' II, AI'AMl
1 . INIRC)DUCTIi)N
This sect ion first discusses several past airport- a tr pol lut ion programsand pcevious attempts to validate air pol lut ion nnodels. 'tatisti(:ai mrethocifor testing several. hypotheses related to both the analy s is of air JolIlut ionimeasurements ind the problem of model valiidat i-t are, proposed. I :-c C luoro (the discussion art. descriptive stat istics anld dttrtu oa ocdr~v
time series analysis, and mult iple linear regression zinolyrtis, al i of ic
are proposed in the diagnios tic- investigation aind aoo vysis )f observec. -3r,,
predicted air pol lut ion concent rations for use i n de f in ing the l i 1its L taccuracy of the AQAM. Emphasis has been placed onl thle details of peneralinterest and just ificat ion of these techniques rather than onf thle presentation
of tithe r igoron s mathematical formalism beb jul each one. The os sontlia 1 math-ematiceal detailis aind computatijonal methods employed in these techniqlues. ark-hothI lengthy ana quite compliicat ed, and should be described subseqoer. nt to
their impl1emen tat ion in another report by Argonne Nat ional Laboratory. f, Jrthle present., however , the reader should consult refereiices Lo the. Ii tbratlirethat are provided for specific mathematical details o~f the techniqlues de-scribed here.
2. PAST AIRPt.RT AIR POLLUJTION AND MODEL VALIDATION lPRoCHAMS
Past airport air quaIit y mon itor in stud ies i nclukd ing Thay r ,CP IaLt, F-2 Rot e, F-3 and NlacWaIt ers et a- . thave failed to observe LC~n-1trations of air pellutants significantly above local. background. In order tounderst:3jtt~t thle complexity of val idating or verifying ort ciining the( 1limiits otaccu-raicy Of airport air pollution models, one must appr-c iaie thle difficultie4involved in monitoring ambient air quality levels at air1 )(Jrts and in atte-fmpt-ing, to relate them to) aircraft activity. An investigation of recent airpo.Jrtair p0 1 it ion mon itotring programs was therefore performed.
to thle t ina report to the EPA-NERC ,* "Air Q ual ity Measured at Atlantalritt-ritat-i ona I Airport be fore and dur ing Expe rimntal o Aircraft. Taxi ing Operaj-tious , ' Mac-Wa It ers et a .F- haVe, cone )1(ded that:
it is not possible to detect. statist ically significantreduction in air quality levels which couild be, attributedto the omiss ions control procedures . . . a nd thle ana lysis ofair quality measurements failed to s~how a statisticallysignificant improvement as a result oJf tilk app1 itat ionof aircraft emissions control procedures ... I T) view otthe limite-d application of emi ssions controls ind the result-
anit stital I I Ii e it emiss ions du ringp any given houlr , asutstant ia I I y it'rsaml Lug, pe itt! wool d be requ"i red todo to{t a st it 1st 1 J'Illy s yi ('I i c alt -I f tfb
*Env i rnnmoetoI 1r ot v ct iont Agenit v , NotI iotna I Liiv i ronnient alI Rese-arch cent (-ILas Vegas, Ne-vada
1 17
Plume rise effect and masking (if aircraft aerometric signals by back-
gruund were advanced as possible explanations for the restilts of tile At lanta
st udy.
One of the earliest attempts to study the sensitivity and accuracyof an air quality computer model for airports was performed by P'latt et al.l-k
on a model called "AIREC." Meteorological parameters were subjected L) asensitivity analysis using the average and maximum value (the maximum ,btain-d
as the average of the three highest values) as the two levels for each factorbeing considered. The resulting changes in the calculated concentrations were
then evaluated as a percentage of the data values:
The results of the sensitivity analysis for the meteorologi-
cal parameters have a direct bearing on the selection of
short-term periods used to calculate maxima concentrations in
the impact analysis. Hence, time periods that exhibit low
wind speeds, stable atmospheric conditions, and high wind
persistence in the prevailing direction should be selected.
The accuracy of the model was investigated indirectly in the analysis
of various control techniques. Source parameters including strength and
location of emissions sources with respect to a receptor were studied. These
investigations reinforced the following limitations to tile validity of the
model:
(I) the model is limited to time periods that are much
longer than the characteristic times of individual
aircraft activity; and
(2) the model is limited to receptors that are not in prox-imity to the emission sources (closer than 100 meters).
Such limitations of spatial and temporal resolution of any air qualitycomputer model need to be established in defining the accuracy of the model.
Under EPA contract, GEOMET has completed a program of air quality
measurement, model development, and model validation at Washington National
Airport.
Statistical analysis techniques for model validation oft the NREC and
modified NREC models F - 5 are developed in the Geomet Task I Program Plan
Model Verification -- Aircraft Emisions Impact on Air Q2Walfty. The val ida-
tion of the model involves:
(1) Comparison of versions of the NREC model -- performed bycomparing frequency distributions for observations and
predictions for specified pollutant-location-averaging
time combinations. Other methods for analyzing the error
of the predictions, such as the Wilcoxen test or the
t-test for paired comparisons are suggested, based on an
examination of the data.
(2) In comparing predicted and observed annual m.ans, th.
bias of predictions are proposed to be t 'st,.d ,' ls n ' the,
138
)i n i om i I t est to (et o rilli ne iI thQei , is a !; i ii I icant
dv i at ion f rom tle exlette va lIe.
(3) Com,ar i son o tIe number o I t ime s t tie pr ed jted andobserved values exceed air quality s taindr d , . T1 i srequired Iit ting lognoriial di stribut ions to the avail-
able values of predicted and observed da-ita, and coui.-paring these to the standa rd in order to, caIc latt- eie
number of times during the year that the standard is
exceeded for each distribution.
(4) Determination of the conditions under whi ie the NRECmodel operated best and worst by stratificatiuL, of
the data into subsets of speci tic treteorolog cal and/oremi ss ions cond it ions and us ing the Ko I oi', og,ur ov-Sm. i r ii ovtest (referenced below) to compare frequency distribution:.of observed and predicted concentrations.
In the NREC and th- mod ified NRIL (GEOMI.T)
mode l ...
The steady-state Gaussian plume model , which i s t,kernl of the concentrat ion calcul at i o ;iod asstmes
steady conditions during the period 01 calculat ion(in this case one hour), is cha act ri st ical lynot expected to give good results when compaied to
observations on a paired-comparisons, hour-by-hourbasis. Such a model, when performin,, well, will
reproduce distributions and means to a reasonabledegree, and this will be highly useful for most
purposes, including those of making comarisons with
standards, and studying the impact of various types of
coatributing sources ...
Seve ra l measures (descriptive statistics) and statistical tests, in-clud ing the t-test for paired means (Graybi I; F - 6 Brown lee, F-7 imple i i near
regression ( Rao) ,F8 the Spearman rank correlat ion coe t I ci ent and mat - hfdpair sign rank test (Kendall),F - 9 and the Koliinuigre v-Sniirlov g-oodu, .ss ,'
fit test (Koloogotv), F - 10 - 1 2 were discussed. These techniqjes wore- r,,pl ,Afor the statistical treatment of the data, but are n,,t related to tilk t,t i,,'
of spec i fic hypotheses concerning the model Vl 1 ildat ion. The 0!i scuss
each technique includes the statement of the nul 1 hypithi.is and iahbr-,vi/it d
disc,,ssion of the underlying assumpt iolls and metlidoloyv, along wi th th,I ormat ot tie pr intot for each test. Wi i le it i re. g( )niZd toat siR I,techniques are the subject of numer,-us textbooks on sLat ist ics, no eecswere provided for further examination of dii.scussions of these mettuods.
nTh approprlateness of d it erent masures and tests is not adequatl vdeveloped and nwither are the inferences made sub:equent to the aplia it lca l (ithe' tests. 'Tloe fact that paired comparisons were "not gim iiallv had," ,ur that
the t-tests we're. " dul" te lls the reade.r very little. lh. j-v.- ,s, ctI l,.v, Iof statistical olufidt, i.,. it whi(h th.i reversal of tle nil iypiti.'si'. !.'-espl;tea, were jlt even meln iotd. 'hs issllllimp ions o! normalit, and inllepo nl .
underlying the- appl icat in of correlat ion and regression alnalvsls inl part icu-lar appeared ti, place , w supportfd by pieliminmatv invest i ,ations.
1 39
At Washington National Airport the contributions of aircraft and airport-related sources make up only a very small fraction of the total air pollutionburden in the airport vicinity. Since the aircraft and airport contributions,which are the most important elements in the model study, represent such smalltractions of the measured concentrations and since no correlation betwuvn themwas demonstrated, one cannot be sure that the model as it relates to thesefactors has been validated.
Only a few air quality computer models have been developed specificallyfor simulating the impact of the airport complex on ambient air quality(Thayer;F -l Platt et al.;F - 2 and RoteF- 3,13,14 ). A recent study, A SurVey "
omputer Models for Predicting Air Pollution from Airports, prepared for theFAA by Haber,F - 1 5 concludes that no properly controlled verification ofairport models has been performed to date:
Airport air pollution models, recognizing the uncertaintiesinherent in the transport and diffusion process, use theGaussian type description. These have, however, not beenproperly validated... Adequate validation has been hamperedby the difficulties in obtaining reliable data and by thecomplex nature of the problem.
Among the defects in airport air pollution computer programs, 1laberF- 1 5
notes the large volumes of complex input, the need for large computers torthe comprehensive air quality computer programs, the lack of appropriateattention to the relative significance of different pollutants, and theprogram's failure to recognize large uncertainties in both the input data andthe models implemented in the programs. He also concludes that:
Since meteorological variables can be observed but notcontrolled emissions data should be substantially under thecontrol of the investigator.
By conducting a controlled experiment at a relatively remote, highvolume military airport where accurate statistics on aircraft type mix andactivity are available, HaberF- 15 notes that:
The basic Gaussian transport and diffusion equations can bevalidated and any shortcomings identified...questions regard-ing the accuracy of sets of values for diffusion coefficientsand their appropriateness for line sources (as has beensuggested by ANL) should be validated empirically.
The problems of evaluation and validation of air pol lut ion models arediscussed by BrierF- 16 in Statistical Questions telatin.; to the Valib,.t ion/;" Air Quaiitp SimuZation Models, a report to EPA-RTP.* (Haher F - I 5 has de-
scribed model validation as "reproducibly generating adequate estimates.")Brier's discussion enlarges as well ul,.n the appropriateness of stat mst caland analysis techniques such as the computation of the mean squar' error (.MSE)and regression analysis (including robust linear regression). He fails t,,
APnvi ronmental Prot fe: t ion A .nc , Research Irian iI'o Park, N(, i t h ;mrl rn, .
140
i-
meni oul , iiowev r , t hiat tit( MSE is a d i sper s ion paramlet ior wlj rt-as tthe men : isa Ioc atL i)on pa ramte tevr. Iit d isc uss ing the, e t ct. of tu (, ie1r po itts onl t hr usulIt s o f reyre ss ion and correlIat lifnL tecaittq ue s, 11 gr al 111 4 i 11 v i 1 lit str t t--t 1-c fact that a great deal of thought must gk: I I"() i s lc t ij: o f a Lt_ xh-
Br jers ' generalI guidelines arid sug-i'e t Lions for :i 'al i oat ioan pr -- -drinc lude a sens it ivi tv analysi s to indicatc anyv int £ :ii tnQnltP < i
model and to under-itand parameters that do.minate the process, feil (-d !by*Sfand regress ion ana lyses suipported by appropriate diuk)!oi c iflv,-sLiSzit I s oll
input-output errors and of the assumptions concerninug ircui
As Koopmans (personal communicat ion) I se e alIso N-H pci: t. lt:
The use of robust methods to tit regression i linos 0.-re wce11,conspire Co make the model look mnuc h better Ili.an i t re.al I \
i s. A scat tpr diagram May well b e useful fot displayingthis data, hut fitting :j regression 1ine, Wi Lb a-l L of theli
correspond jug probl1ems, aprpear,; a very pool-r choice ,t methboi,since much of the important information lies inl the( outlierts.Possibly, a combinat ion of' repression lintes; withi indepoendenit
explanat ion of the, nutlIit-is would he(- USEtl
3. DESlCN r r IVL S-ATI STICIc ANODIPSTRHIWf bu tAL. CONS [DLi-,A 1,10N"
A -reaL deal of attent ion has been focusedl on the i '-er- )fdiiLterrlnn nthe distribuirjonal characteristics ot abiietit air polliition conctentrait iilru
Most researchers w-il1 agree, th-it ;tit polluition d ata apfpear to have a skewedurindoiviing cistribiit ion. There are several distribut ioni thal-t COUld t iSL-for exprossing this type of skewed data,. The actual or empirical diqt rib iipof air -el lii inn concent rat oris do not live tieyat ive valii- auii aii-, -- I-
as;vmmetr i~al with the, lonwer toil s extending zibove ti iiwt-han voiicnrt i-it I
T b I, 1)niirma I di stribut ion has beeii proposed as auccrateiyv s~r i rpo I titLanlt c onc en tra t ion s by Z/immir antd La rsen) it 1 7 1 ;1 r s "U , F'- 5I VI
Mare Du r iDurin the EPA Svripisiuui onl Stat ist iral Asp.u-its of A ir (Ia ti v
D a ta (t-ob cr 19 74) *~ ftIinr ieI Kniiox , Po L I ; (k Tuirnte r, a i d Iv i i all1 I, -m,, 'ic rt, i
the vi ust- fu I ne ss o f t Iin( f reoq tienc / d i st r ih ut i on1 in th ie i i-iV' I sI! o t a ir 1, o i i i tc encen lt r at ions. St andard s tat i s t i ca 1 cd leou I at ions mmiv hie app!I i el i t we f i rst
t rans torni thev conc (iut rat ionst., t o tlie i t lI mr i t hms base-11) Y t ii i' geIlnet r j mtiic~
i ; alIc i Iat (.0 as t 1:, n' - i I oiIl rtlitill ofI te 1 s1ait l--ir in . 11i- st aI'd ari
g j rt'm t r Ii ( 14v iamt -o, i s I It-- all i I oga I ithi ll it f i h t - i dlI i -i nt i c o i thi.
l og a r it hils. It Ill, data fnollow i ln 'iiirinaI ilisti ibtit nil, thenl (ihe ciiiulat iv.
percent igi's liii 1 ipproxiately ,in a :tatighit ]i- oni 1. lfidail Iitvy jae
The 50t pe ent i t' point i a g iv.- i hy the geuotin'tti( c i-.1 ill il- I he %;I ope t' o t h
l ino i, v iven fiy thu gerinut t 1 t andmird ut-v iat ion.
Lmr.i(,onI , I ha-s coneliuhed that c n e t a i'l ar lp T-Xmj. v 0 111
ma! ly tlintriujtted for all polltitant, tin all cities lir -!i avt-raj'inc i-a-,based on at-tirdi1 air iqualityv data. Lairseii ciiitt'is that flit,' loyni-urnna di"i i -.bit ionl proi)V s a !iflIt 1Ilils I oi prt-di i on plir po~ , i-pt .i a I v ist-Iiil ill
*Ffvir nifon al 'r llo-cf iiii Av*- ( v, t-' ; n it I I i allp I- P- '?t11 1,iit;
141
air pollution control applications. Mage F - 2 0 by treating diftusiot as a
stochastic process has obtained the censored three-parame ter lognormal model,
of which Larsen's standard two-parameter model is a special case. 'ie new
model has been compared with actual air pollutant concontrat ion data andappears superior to the standard lognormal model, especially in its apprxim;1-
Lion for the upper percentile points of the cumulative distribution ililct i11.It is probable that this improvement resilts iot just from :iddition of anot!i(,rparameter but from tile fact that the new model is more appropriate for tli,
underlying physical processes that generat(, atmos pher ic pollutant coInCn-
trat ions.
The method for estimating the parameters of the censored three-param,ter
lognornal model is rather complicated in comparison to that for the two-parameter model and involves numerical optimization for determining a best-fitting, cumulative distribution function, using a weighted least squaIsapproach. The model is "censored", meaning that it will predict ne pat iveconctntrations that are simply taken to be zero. The censored three-parametrt.model proposed by Mage appears to agree or fit the actual data in the uplertail of the distribution more satisfactorily than does Larsen's standardtwo-parameter model. While Larsen's modelF - l may be adequate ftor making.assumptions about control strategies, the place for more complicated distribu-tions like the one proposed by Mage, F - 20 is in fitting large bodies of data soas to produce a better estimate of the behavior of the upper tail of thedistribution.
An important aspect of any sampling program is to determine the fractionof time in which a certain concentration is exceeded. The empirical cumula-t ive frequency distribution plot can be used to get the percent of thesample population exceeding any specified concentration. However, one mustkeep in mind the accuracy and resolution obtained in estimates associated withthe tail of the distribution. A very few observations, sometimes referred toas extreme values, are usually available on which to bast, the extreme perc,-x
tile points. Since the standards promulgated by the Enviroimental ProtectionAgency are stated in terms of exceedance, the fre1 ency (,t occurrence of highconcentrations are of major importance. For this r ,ison, tie straight I itii ormore complex fitted cumulative frequn,.cy ,.riiat ions nve,! to be criticallyexamined and interpreted, and supporte-d by dIiagnost ic invst ihat ions of thebas ic assumpt ions concerning normal i V a1d i I- nde.i . Ai tcIiso a d
Brown F -22 provide a general discussion it h,, logn(,rmal th,,,rv and dist r ito-tion methods, and should be consulted for ,,neral discul:sion , of lognorialdistributions.
Another techni4 ue known as the boxplo)t has been su,,gksted v J. W,
rukey F -23 for examining di stributional characterisLics, The hoYvrlot technliquec,-n ists of building a stem-and-leaf data strtctir, similal to a iiistgram,
in which stems correspond to higher order digits and leaves to lower orderdigits in the ob,,ervations. The structore is used to compute plott ing
posit ions for outli er points anui extreme values, and a box itidicat ing tiltinterquartile range and tile median. The boxplot produces a horizontalschematic display that indicates the distributional character of th, observa-tions. Un ,.r t ).- assumption of a rrrtain i , r -,ut ,ag, of the,.:b s,rvations f.iA I ' , l, ut tO, int r i,,; rt , rI, i ,. . 1t1i1r .ini extr.u.,'j a l ,,.- , a r ,.s . o r, n t ',, r , .,, il l , Il t ! - i t , I- l
a111; !1 . 1--11, l l * .'ll, , 1
range and the number ot such data values exceed in the prcpor ,nt ioln of ubse-rva--
t ions normally found in thk, tails of the normal d st ibu io .. The appi i. it ion
of the boxplot technique to th, origin,il dlt.t nc to tI. 1c 19 traiistr .1 11the data will provide insight into the ai-su'up ,,n that t!1 i,t.i I.rc intrtit-
sitallv normal, that is to say, loenca l l. '. is t Cir, i, t:,- It c r s new r,, r , .... .
tives on the distributional character ot the data, i, at ,s ,nt Iier-, and
provides estimates of several descriptive statistics.
SaltzinanF- 2 4 has suggested a nomoralphit itlprese,tat inn for estimnating
confidence intervals for mean values atn d teritiatiot. of conl idonce intervals
for proportions ol a populat ion ex 'ed i ju,. a spec i fied ':,ent rat ion iI tI,d:ta follow a normal or lognormal distribl it t . .Jo ,loso a;.i ,oe - ( :;r<v ide
a discussion of confideace-interval fst imat io , th,. popolat lon-t um al t IV-
distribution function. Slich methods may prove osef,!l w,. co r iogrm two ,r
more cumulative frequency distributions for signi i cant. d it terences, and also
shoild provide insight as to the accuracy of speciti,( percntie points that
are based on observed or predicted data being compared to air qual i ty
standards.
Most natural phenomena ar, observed as statistical distributions and
can be represented by a trequency curve, the parameters of which can be esti-mated from the obsrvations. As R. McCammonF - 2 6 has suggested:
Natural eteints are the product of compound effects re.sultingfrom multiple causes and the observed distributions are not
always representative of homogeneous populations,but rather a
mixture of two or more subpopulations each of whicti posses.esunique parameters. An observation therefore can represent a
sample from one of several populations ... The statistical
problem that arises; is the dissection of a mixed frequencydistribution into its individual compotents...
Background concentrations for various pollutants superimposed on concen-
tr,it ions for which modeled emissions activity are known to occur can olt en
caus,, ei,t. empirical frequency d i stribut ion to appear to ) b,. bimodal . f' ; ic
norma I I v re I Ioc ted in the colmilat ive distrib t ion I inct ion's depar r tr
from .j straight line when plotted on lop, paper. 'lho inability to relat,
pollt:tion levels to emissions activity can nsually b explained by the maskin',effect tie background concn r at ions have on the pol Ii Lun signal . A
graphical-analytic techniq.u, for dissectin g normal or I ognorma I dist r i but ion,
by g,,raphical est imat ion of the parame tors for the two (i stribut ions has ;been proposed bv McCammon. - 6 ,,issect ing the diSt ibhi, i ns is accomplished
thrigh the use of graphical first est imates of the lo,nortna I components
l lowed by ;a regres s ion analIys is that provides final s imates. A eti-squar
test e'rl bh, s,, ttf test the hyoth es that the da;i i -present a mixt ure of
tw,) l ol,)rmal I, sI ibl t ioan;. Such a; n a pproach coili h,lIp to ident i Iv '. I n
d i st r ibit io ol i Fi 'ak roind C(ol i-t rnt inis and at the s.'le,, t illi would profi( -- ;I
bt t e r s tra ig t L ia, te I t o ,r (t hi g ,h r pe-rcent i I - ,t t ,. c iltllilat ivi, Ii,-qiiuncv fi I st r ibui iou that is of ,;rti ulir intor est ilt the liti i ,,f
acctlira( v ,t air p',llIit ion prod ii I ion,., Havinp indent it id ,' ba(kprioll gl 'o;
us i ng so wt ;i techtnique, des.i i pt v st i st ic f ( fi hjokgroinld siit ra t (,,Iaerome t ric q Ig als Lall be uiIltut ed V.: i 1 V
14
Robust techniques for the estimation of the mean, median, and confi-dence-interval estimates have been explored and developed by Dixon and"'k.y ,F - 2 ~Tukey and McLaughlin,F-28 and Gross. F - 29 The computation ofdescriptive statistics are often made under the assuimpt ion that the datato low a normal or intrinsically normal distribution. In practice thisassumption is often false. A procedure called the wave-interval, propose~d byGrossF- 2 9 for the construction of a confidence interval that is not dependenton the normality assumption, has been found to perform very well and is recon-mended for general use in place of the t-interval for sample sizes of eight ormore. It is assumed that the distribution of the data is unimodal and sym-metric and is most likely long-tailed. This technique is recommended for theestimation of descriptive statistics such as the mean, median, and confidencelimits for small sample sizes as a means of compensating for extreme pointsand outliers. Small sample sizes will undoubtedly arise in the stratificationof available data on factors such as level of emissions activity, time of day,
etc.
It is important to emphasize the role of graphics related to the previousdiscussions. Plots of actual and fitted cumulative frequency distributions,boxplots, and other descriptive statistics are essential aids in the analysisand interpretation of air pollution data. Graphics, as we shall see in thenext section are also extremely important in time series analysis and inplotting the data itself.
4. TIME SERIES METHODS
This section describes time series analysis as it relates to the statis-
tical analysis of air pollution and meteorological data. The time dependentnature of such data and the liklihood of serial correlation found in suchmeasurements make obvious the need for analyzing such data in terms ol
its frequency composition. This section provides background on time seriesconcepts, attempts to justify the use of time series analysis, and discussesspecific analysis techniques, providing the reader with references ftt furtherreading.
ClevelandF- 3 0 has proposed the use of the inverse autocorrelation of aime series in the analysis of air pollution data. Two different methods tor
estimating the inverse autocorrelation arise from two ditferent methods ofestimating the spectral density -- autoregressive and periodogram smoothing.The estimates of the inverse autocorrelations can be used to assist inidentifying a parsimonious, moving average, autoregressive model tor theseries. The feasibility of the autoregressive-moving averige models has beendemonstrated by Lee.F - 3 1 The use of time series in the ainalysi of air pollu-tion monitoring data has also been suggested by Barlow and Singpurwalla, F - 3 2
Saltzman,F- 3 3 Marcus,F - 34 and Tiao and Hamming.F- 3 5
The value of a time series at any given time is called the amplitudecomponent at that time. A useful measure of the "activity" of the timeseries is its power. The amplitude is a positive number measured In theamplitude urits of the problem at hand. The phase is a dimeisionlt.ss para-meter that measures the displacement of the sinusoid relative to tho given
t imc; origin.
144
Ant oxtenii;ve treatment of the spectral analys: oln t imre series ha.
been wivten by Koopians.1 '3 6 Among his views on times s*rip, are:
In the samte sense that various cc lots ot I igh- are cetrp.sed
of a blend of monoch romrat ic c omrponen t ,arid rI, ui cl tones, a rt-
formed hy a super impos it ion of pure haronik , tX imp sr)escan be const ructed by composing a number of hiarmoni f unc-t ions wi th varying frequencies, anrpI itudes and pha1ses . fvc-ryt ime series has an expl icit spectral reprear, nitat ion in Lterusoif frequency ,ampl i tude and phiase.
TiOP series (dita ana! ys is crnd heory a]lso is hi t~~ i eYxtensively 0"
Br illinger, F3 1 and tire ftir ier analIys is of time series by Bi omf icnf-
provides considerable dM'ail rrn spectrum analysis arold (iscuqiitot of thegeneral framework of time series analysis methods. Mid i r innal iv, a bibli-ographry of time scrip, antalys is papers has been given 'VWo~.F3
The pr intc ipal piaI o I sp-y t r um analyvsis is to rocorpoise the pov'-r ofthe' given time seriet; inrto its arrtoc cr~ponients, tha. is, to estimnte thepower spectrurm fromt the ava ilaP le data. 'The est fimatod SpNr'Cmir can then beused to gain in format ion ahout the mechan ism that generated the data. 1heSt at is tical theory of spectral nalvs is has been based on certain hypotheses,that the underlying process is statin narv and Gauss ian and that the pr 'cvssz=~an is zero and th~e spec trums is cootinuois. Some of the pri process ingoperat ions performed on tihe dat a he for a spec trum anal vs is is performid ar.
nt ended to br ing the dat a in to reasonahmi conform ity v wi h thes, hvpothesos.
A non-zero mean and possible trend in the mean can fi Vc an id icat ir-n ofnoit-ntat ionarity or of a drift in the intinina insstemis . inn ectimat ion
and removal of linear or qiuadrat ic ti rds that be"t fit the data in Mne least
squares sense cart be subt ractvd from the original data with the ru-srli=residual con fortiing more close ly to the stat ionary hypothesis.
S, Ie oral imrporitant rvason4 for the use of t itle %et ics aita 1 's areo I I o e I oWIi)t W
M As a di agnes ti c antd 'xp I 'rato iv tool in checkinog tin'
assumrpti otis of i ode ondenc anid ntornmality otdd tot
irtht'' Atatis!rcal t test s, ptart i ulari y those rela~tedl to
sign if i':ant test inig, an,] frequency distnihut i n conistrun-
t o'n technii que s
(2 ) As a means of analyzing the serial. correlat inns in L edata;
1) As a1 rietlod (of examining frequency dependrir to()r i a'lt i
(cal le': coht eenee between two t imes series) ; "mt teMar iotnmetthords in t he t ime doma in -- for examp 1 , ' Saltinu ini( "?T el at ion by risiny, stra ight I in' MIithods, -- ai 1 - very'rr it rv*- to the assunrpt ion of iridnpendeo' ant i rnva Iti-
dar-1t by thre presener' ttl set il corielat ion;
'r) As rut -xyp1 orat ''y I ,ol ini searchiiing for i i rnt io it1 vand gro~'. var ia inn "t power wi th f rvqin v tn pt o id,i ns i p-Itit i i i I I it f t-mi clIi a n i siTn It' i inil the dat a,; antd
(5) That it is frequently possible to describe complexphysical mechanisms better in the frequency domain than in
the time domain.
By examining periodicities in the data, we have anothor way of lookingat some of the fundamental mechanisms that produce diurnal pattern, andpo l l ution episodes. The correspondence between aircraft eli ssins aid l( llj-tion concentrations should be readily observble throuigh an examination (,J fill.spectral parameters.
Two models of spectral analysis are now described -- tie univariatespectral analysis and the bivariate spectral analysis ot time series. hult i-variate spectral analysis has not been considered for the present.
Univariate spectral analysis involve the, vo1m1putat ion of the p(werspectrum or, as it is also called, the ,stimated spectral density. Ihe p,)werspectrum is the Fourier transform of the autoc,,r.lat ion iiinct ion. 'The
Fourier transform of a sequence of antocorriIat io s contiins precis,.ly tilesame information found in the original autocorrelat ions themselves.
Many statistical analysis methods that operate on the data in the timedomain make no use of the time dependency in the data. For example, in theestimation of the frequency or cumulative frequency distributions, the dataare sorted into non-decreasing order, and information in the time dependency
of the data is no longer available. An important property of t ir.e seriesanalysis is that much of the time dependence in the data is preserve .. ThIeaccumulated spectrum is essentially the first integral of the spectrum and can
be used to indicate a range of frequencies that contain a substantial propor-
tion of the total power of the time series.
Univariate time series analysis is also an important tool in tht, designof digital filters. By examining the spectrimi,the etfect of 1i Itering can beseen in terms of the frequency composition of tile tim' series. In this
fashion, digital filters can be developed that rmove irequency ciniponreuts
that might be regarded as noise prior to the analysis of trends. For example,if the original data are sampled every five iinutes, but only hourly averagesare required, then the frequency components in the data due to nose in thefive-minute data can be removed or filtered out,resiilting in a smoothed
five-minute series that can be used to calculate the one-hour data. In asimilar fashion, the effect of the intake-manifold, averaying bottles can beinvestigated by examining the spectrum of already smootheo pollution concen-
trations. Similarities between the spectra of two series, such as peaks atsimilar frequencies, may raise the possibility that the series are related.To investigate such possibilities, we compute estimates of the cross spectrumof the two series. This is an extension of the definition of the spectrim andis usually estimated by smoothing the cross-periodogram, and is performed inthe bivariate spectral analysis of the two series.
Bivariate spectral analysis for the spectral decomposition ii twoseries into various spectral parameters is equivalent to descriptive statis-tics. These include the estimated spectral density of each series, the phasefunction that indicates the relative shift of harmon il components of the' two
time series at the same frequency, the transer uinction that indicates thelinear relationship be'twen th,. two seri,.s in the se wav r,,ressiii
146
c(,et icito-ts rel at ' the linear ti at io lil kp betw-.en two data sets, arid tr.
coherence funct ion tha: rovi t, a iI. -s t imate of th k f r-.q ency-depf-ndo-nt
correlation between the tlj., seri,-.
Recause of the vast amounts ,f n11t r ical informat ioi cont ained in the
estimates of spectral naramiters, graphs should be ol L..iillt-d for the spectral
parameters of inlierest, ,:,r tho univarizito spectral a , I -s i t h,:se ic t:Co
the estimated spectrum and the ic c uui ated spect rim. In the ivar iat . stec-
tral analvsis are included the spectral densities fur -a h sriis, ds wtI
the phase, transfer and coherence functions. Plt- f the nri liinml and
corrected data traces shorild also he obtained
A general oul Iine of the appl icat ion of time se-i':s is given below:
(I) Collect all data tor times series analysis, making
efforts to ensure that the data are feasible and if
necessary- adjust the data fort comparability.
Obtain a ,raph of the t ime series.
(3) Perform trend analv.;is.
(4) Accommodate and adjust for ste as ona I variat ions if
required.
(5) Adjust the data for trend.
(6) Compute and graph al I spectral parameters kf interest.
It is proposed that spectral analysis he applir i t, tne air polluti:
and meteorological data traces for e:.ch station. Miss data can ,.
sated for by sinple linear interjolation, which does not affect the esti ated
spectral parameters. The purposes of these time series analvsis tech.iquie,
have been discussted In th , sect ion of this report deal in, with st ate,-.t ct-h1Yp111)t h , .s
5. Vqlt T ll. I .1NEAR REGRESSION ',AlYSIS
'There art. Iwo main t,s 1i, molt ipr ' I ir :itir -i' r ssion t thlir111 .
'lh f i-rst prov ides an arina Ivs i. : ic ) irIroir' fo r th11. eva l tin n t th It-qI :it I '
eIt,'ctl& and s i nj f 1 iiicn 1) 1 1 i ;t o -." at i t~c ti ti t h ' d o , d. nd. ,t var i ab 1 e . II th 1 iscas', the de rnderi vtari.ibh , ic I Hit, tr,pl iiee r -,r .stiody and t hr io etPr u.t
variab lts art, the' -ikt, riil va li '-, )It ..iramrters le lt( d to lie response . Thte
set-( ,nd ll sa - pi c v l 'a n - i -oril t raniewor k for plann i rig and ana 1 vs i, of
sen., t vi ty exrrit'n? s. Ti s te'chniqme is especial iy ,isu.fill in si t iat ion ill
wh ih h a i rIt.r',( , ,or cip'cnld'rit variable may he' exp,.nsivr or dift icilt to
lii I .i n .
('1Cliri id t I it, r c, 'rtr s iii r) rr q ;i 1 io t tihe f(orill:
' 0 - 0 + 6I XI + 5';X 2 + + I..
whe rf:
Y tire It pendoi v;r i ab It' wh i ch is I he re srrii;. r Iali! it v
m1d,(it , on: id, rat i , "-
147
X i = independent variables,
5 i = regression coefficients (sometinies -jiled partial
coefficients), and
E = the random error term.
A discuss ion of the assumptions underlyi in, the application of re. ressionmethods can be found in most statistics texts, such as Fitt .iuno !'q'r ori to:..zt' by Daniel and Wood,F - 40 which discuss the least squares method. '111seassumptions concern a knowledge of the correct form of the regression ,qua-tion, use of typical data in the analysis, and assirance that the dependentvariables are statistically uncorrelated (or, more generally, are said to beindependent). Less important assumptions arc: that the dependent variableshave the same (though unknown) variance, that the values of the independentvariables are known without error, and that the random error is normally
'!i stributed.
The use of regression methods and experimental designs in the contextof performing sensitivity analysis is to estimate the effects of the independ-ent variables of interest (and possibly their interactions) in a reasonableriamber of experimental runs or values of the dependent variable. Fractional
fact,_,rial designs, discussed extensively by Cochran and CoxF- 4 1 can be used.or this purpose. These designs provide estimates of the main effects and, ifdesired, some low-order interactions with a minimal number of computer runs or
values of the independent variable.
The model using two levels of the factors is called a first-order modeland, hence, contains no higher order terms. If for some factor a quadraticcomponent is to be estimated, the factorial designs must be rerlanned to take
that factor at three or more levels. In practice, however, the quadraticeffects are often neglected in the earlier stages of experimentation in orderto make decisions about future experiments on the basis of the first-ordermodel. Orthogonal polynomials are suggested when quadratit effects are to beestimated.
The fitted equation could fail to be statistically signiticant because(1) the assumed -orm of the equation is not the true or (2) the variation of,,bserved points, though random, is so large that the fitted equation couldhave arisen by random sampling from a population with all partial regressioncoefficients equal to zero.
The hypothesis that all true partial regression coefficients,6 1 's, equal,ero, can be tested by an F-test of the variance accounted for by regressionrelative to the error variance. The F-value calculated from the data can be
compared 4ith the tabulated F-value,F - 4 2 and the hypothes~s that all truepartial regression coefficients equal zero can be rejected if the calculatedF-value is larger.F - 4 3 This same hypothesis, if rejected, also allows testingto determine whether the regression accounts for a significant amount of thevariation observed in Y, the response quantity.
It is re-rmminded that the modified Gram-Schmidt analysis of varianceho, usfed to perforn, I no least squares solut ion in the factoi ization process in
,,hvir1, the regre.,sion equation. In the modi tied ,ram-Schmidt procedure, all
148
_ : ..... _ .. .. . . ... . .. . . .. .. ... . . .. .. ... . ... . ... . ..... ..o w
co Illins ot I Ot dvi i it a!t ix i r. - ot' Oyri:l I i /-d uc'~ s c v, I v wi t i r s c to
theit pr '(mi oi r 't nfla I<laimn i iI t te 'er/Ia I p r f)c e '. r)o a'iI!r a g I;i',,'0i s f in ea i ng wi Lit iI I -- oi '(I i i etI0 t'Oi ! - i I c o il co a t ,* !.fi !T,3 o d I
(;rani-Schmidt mt thoul works 1dooiii 'y ''i-4''-#
b. ICOMMENDA'I ONS
It is reotmended t ha L thi one -h oulr a ircra I t IL t v 1 ta1pe bfe r<,iCJif rom th e d isc:re te TfIi gh Lt tinm- d at a and tha t e xp La nat- io ns o f t Isu .'t .nsunder lV i lip the es t ima t ion o f s' cit quan t it ie s as tife i;.imber of t ouch-aind -g,
operat ions per hour be provided.
Since the ac tualI aircra ft modes d isciisseu iter w ill very Ii kel' I,- cd i tt erent from Cte modes to h)ec usedk i n Lte AQAM, some explan~a tion of ;rat lonal (- behind the choice( of AQAM modes should be deve lped
Since a vas t amiounitt o f uetir ircalI i ntormat ion wll bte involved ill L tob. e, rv c ( and proed i c t ed do a a sott !; a rid wJi I I alIs o be go no ra ted dui r i ng t lieAtAM
mo del validat ion, i t is recomniided Chat computer graphics capabilitices beutilized extensivoely in all pbis-. oif thle invest igat ion. Graphic rout int.-s for
gene.-ralI two-cl ime ns ioat1 p 1 t t iug (scatter diagrams, Lime h istorijes , etc . ),wind and polluttion rose, e ife t ive -source mapping, and two-dimensional con-
touring will be required. Thie importance of graphics capabilities cannotbe overstressed . Any development costs will be substantially offset by, the
contributions tbest. techniques will make in the analysis process.
It is recommended that all statistical analysis of data from the 1ijlliarnsAF13 air pollution monitoring program, itic-uding the reduction )f the raw datatapes and the computation of cumulative frequency distributions, be coo-rdi-nated with the AQAXM model validation program performed by Argonne NationalLabor at-'rv.
Fliiiocc. r 'rs analIys i c is recommended for use i n thle i nvest igat 1 0, ,.a ssuiimp t ions and hypotheses, for the analyvsis of ser ja corre lations irF t a.da jt a, and for thle determination of frequency dlependlent corre laot ets ( (oh-
relic, hietween two time Si(ri s.
MulIt iple reg Test; on mia I y;. i " i s reoriimend, d as "; :1oonas o f a:a I vs i lig,exporinental des itis ill the dhft 'ml oat ion of the relative signi fi canoe andiniportanct, of olis-iEos (inicluding r1t least total otn-buse;L sources, aircraft,an bi ackgroundl) aol, meteonrology ( inclIudiing dat a st rat it ied on wind dirct i 01,winil speed , atmpis plIeric sta;b ilIit v, and d iurnal1 pat to is) onl thle accuracy ofAQAII predictions. Th e Ito I lowiniig oensiderat ions are recomnided in) the ove-ralldoftin it len of arcciii icy of tl-e AQAM:
I ) Id''ne i t iloat ion of 1)01 I otant (s ) , time period'; atel condj-t i )is nod(e r wht i (-it th(- mode I poer t o rms hi'at iino woil st
(2) comntar iSln o t i id it d at)nd 1 osi'rv e i'xc 'i'etainc(.' S I t ita ir quia I i L v s t;ituha rd s.
149
3) General comparison of predictions and obsc ivatL ionsdur ing worst case air pollution opisodes. These I.pi-sodes should include time periods of substant iaI air-craft activity and periods of persisternt atMosphi icstability by an analysis of data Itrom the Wi l Jams AFBair pollution monitoring program.
REFERENCES
F-I. Thayer, S.D., et al., Model Verification - Aircrtaft /,'rntfl:lon nqmpat okiAir Quality, Geomet Report No. EF-262, EPA Contract No. 68-02-0665(May 1974).
b-2. Platt, M., et al., The Potential Irri)a(,f i' Aircraft on Air d2ualitiNorthern Research and Engineering Corp. , Report No. 1 167-1, NTI S No.PB208-950 (Dec. 1971).
r--3. Rote, D.M. , et al. , Airport Vicirnity A 'Y Pollution Study, ArgonneNational Laboratory Report to Federal Aviation Administration, ReportNo. FAA-RD-73- 13 (Dec. 1973).
F-4. MacWalters, J.T., R.C. Koch, and S.D. 1hayer, Air Qualitp .'!earuried aAtlanta InternatiornL .4irport be"ore ao,2 dur"9'r7 ExperimenZ AircraftTaxiing Operutions, Geomet Report No. EF-330, EPA Contract No. 68-03-0358, Gecmet, Inc., Rockville, Md. (June 1974).
F-5. Zimmerman, J., Model Verification - Aircraft !,ies cu Impact on Air
Quality, No. E-152, Environmental Protection Agency, Research TrianglePark, N.C. (Oct. 1972).
F-6. Graybill, F.A., Linear Statiotical Models, McGraw-Hill, N.Y. 1961).
;-7. Brownlee, K.A., Statistical Tzheor anti 41ethodol ojy 1, Ac. 7CO ,
PAngineerinj, John Wiley & Sons, Inc., N.Y. (1965).
F-8. Rao, C.R., Linear Statistical Infjtrencea and Ite, ART/iatins, JohnWiley & Sons, N.Y. (1965).
F-9. Kendall, M.C., Rank Correlation Method, C. Griffin, L.ondon (19148).
F-I. Kolotnogorov, A.N. , and S.V. Fomin, ELei.'aents of the T'oeor., ," unotieniiAnalysis, translated from the 1st Russian Edition, by Leo F. Boron,Graylock Press, Rochester, N.Y. (1957).
F-Il. Kolomogorov, A.N. , Foundations of the I'eor., of if'+ilf'1 , transla-tion edited by Nathan Morrison, 2nd English Ed., Chelsea 'ub I icat Ions
Co., N.Y. (1965).
F-12. Gnedenko, B.V., and A.N. Kolomogorov, Limit [)istrihut ion, for 'url; )."
Inde,.,-i'ivrt -znd.om Varizhles, translated from Russian and annot ted byK.L. Chung with appendix by J.I.L. Boob, Addi ,son-We ,slv Ptubi icat ion C,.(;ambridye, Mas,. (1954).
150
F- I ROte, 1). M . , and L. E-. Wangen , CU.f .4 :r .: .. i, 'JA~g,
.Cr ? Ai r t'or'e 'Putzc~orz, AFWL--R-74-JQ4 , A rpg'einc ,.a Ional Labor at ory',Argonne, Ill. (I'eh. 1975).
F-14 Warigen, L.E. ,and D.M. Rote , A Geriev'e7,y1,ize /' ir I i r o(s c~rie r.>SAti' i' ()pe at ionfs - An 're'tr ' ,AFWI.-T!'- 74-54, Argonne
Nat ionat Laboratory , Argonne, It I. (Jlily I9i-1
F-IS5. Hlaber, J .M., A4 :j ci Conijutref !H) r' 1 ~.~ i"iirsc'nAl 'r e , R-i'- 1231 , 3 .1. Wiggins C o. kedondo Bfeacli, Ia i i
(May 197,).
t." t ~~NTi 13- Pli-24 1 -88 b, L vi r,,1imIiLtalI Pi ,It c t
Aj7 en cY I Na t io na 1 1,nv i r om n LnalI Re s ea IchI Ce I te( t, t 0ore oo ty !.oib,-,rat-.
Research riangle Park, N. C. (March 1975).
F-17. Zimimer, C.E. , and R.1 .Larsen, Cal.cul.ating Air Q-Lu-'ity andl It '8.'ontrol , J. of the Air Pol lution Control Assrz., 1102) (Dec. 1 9f5).
F-18. L~ar stn , R .I . t. a I A Ana 7. q z ri A ir P~oI I ut o n. Co w,-( izt Pi. t ion (11 .&17' ;C
3),. oJ. olf the Air- Vol lit ioni Control Assn . , ];'(2) (1967)
F-I19 . Larsen, R .1 . , A! V,-Zzceknt",-a1 A'ocdel forAn-t n A ~r .,ua.e2? . G r<p,!74
1;,.i!i ;,,Pidr'~ Air Programs Vnhib icat im-., No. AP-84, E~nv~ ron-
mntal Prot ec , on Agency , Re-search TFr iang 10( Park , N . C . ( Nov . 1971) .
.onc'n/ rz el t t7 n '.z , U .S . ,lov i rilleil ~ -t~t 1741 AIg'ntc v , Las V'epas,
Nevada , tor presentat 101 it- tin ottn AnnualI Meet ing (0 the A i VoIlI lo-t ion (;ont rolI Assoc iation , liostmin, Mass., pp. 3-2 7 (Jtml, 1 '-20, 19>1 .
F-2 . n ks I.. I., A4 N,,,,, Mth"''I'~~i.o o:, '' *o::' '7,.n
MZ Id: 11 1 ' i* 17.. viI .7 , .0) , J1 . oi thti A i r Poi I Ilit i On (071t ro I As 7-17.
1 l)24-30 J an . 1969 ).
F - '2 . A it (I i 1 T i, I1. n nd J .A.JC'. Brow no in. L :r, e tlI! ! ':tn C 1 , .ri h ri
l ve I II V I' s, 111 bhi jc I o d 9
F-2 F. ikey, J .W. ,'x o ;Liri tor0 rito'j Ml I I I I ir.,te t a 1mItim (A.
Mass-,: . (c I I .11)
H Ito7 7;? .''o, .. ot t 1w Ai Pl' Iu i ~jon Cent rol As!;1 :" 2 ( 191
-i . I-dinrsoe~, N7.1 . tild - C. I.-m t it t 1,(-r ,' t X e'I " 'k7? .ra :Pt
In-. , N.Y. IV64)
1 -21'. M( Cami n, N k h., , A o( 77 (P;c e r 7!7. /1 'i'
1::7 I * .. ' cI A I'm'i,' 1 ' it/ o' 'l[ 1, 4.t riHn ",1 on ?. Ii j.s.
I oyg, ve .11 1 V !,It ionm. I I-t IIc 9,12, Re st on, Va . 1I 97 ;h (Cmimt inr
171 iph i ( , / ( :1 , i,. . I f rh 6 'I' l I Anmma I (>m t mv i 'n, in ('sMipumt m'r
Gr a ph i I4, 1 111 , m i i ;w t ji m[. m .ilfs andmm I maym Pr O ' c *-s' m rca - Sr nm :3pIIIIIVtV I I Vm I IIm! v I Van 1 1 , ', P i I 1,lm1 11ll I Il i Im V 1i t14 16 , I i )
F-2 7. Dixon, W.J. , and J.W. Tukey, Approximate /lc/iaoiur oJ the .' triL, LF,,oj" Winorized ( Trimming/Winsoriza,. ion :) , Technometr ics, 1 (1) ( 196).
F-28 . Tukey, J.W. , and D.H. McLaughlin, Less Vulnercabtc CoijLdrwne arniSignijicance Procedures for Location Bta.;ed on a sinj,? aplu :Primming/Winsoization, The Indian Journal ol Statist its, Series A,25(Pt. 4), Reprint, Sankhya, Statistical Publishing, Society, Calcutta,
India (Sept. 1963).
F-29. Gross, A.M., A Robust Confidence Interwal for Location j'ov ymrfte-tiC,Longtailed Distributions, Proc., National Academy of Science,70:1995-1997 (July 1973).
F-30. Cleveland, W.S., The Inverse Autocorrelutiona oj a Time S;eries aoTheir Applications, Technometr ics, 14 (2):277-293 (May 1972).
F-31. Lee, T. S., Feas ibi ity S't cdy of" Artria (jI q) 6*1,J' ,ipjtiea to Aaat ur,Phenomen, Symposium on Air Force Applications of Modern Control Theoryat the Air Force Institute of Technology, Wright-Patternson Air ForceBase, Ohio (1975).
F-32. Barlow, R.E. , and N.D. Singpurwalla, Awe rajginj Tici aned -laxima 'orDependent Observations, from Proc. of the Symposium on StatisticalAspects of Air Quality Data, EPA-650/4-74-038 (Oct. 1974).
F-33. Saltzman, B.E., Fourier Analysis of Air I, onitoring Data, from Proc .ofthe Symposium on Statistical Aspects of Air Quality Data, EPA-b50/4-74-038 (Oct. 1974).
F-34. Marcus, A.H., A Stochastic Model f'or Lstimatin& Pollutant Exposure b:,Nleans of Air Quality Data, from Proc. of the Symposium on StatisticalAspects of Air Quality Data, EPA-650/4-74-038 (Oct. 1974).
F-35. Tiao, G.C., and W.J. Hamming, A Statistical Anaciv o," the Los An :i,'sAmbient Carbon Monoxide Data 19b6-1972, J. of the Air Pollution Contiol
Assn., 2(11) (Nov. 1975).
F-36. Koopmans, L.H., ',he Spectral Analysis of the .),ncs, Academic Press,N.Y. (1974).
F-37, Brillinger, D.R., Time Series Data Analy'sis V,l O _ , Holt, Rinehartand Winston, Inc. (1975).
F-38. Bloomfield, P., Fourier Analysis of" Time deni o: An lntoaucti( l , JohnWiley & Sons, Inc., N.Y. (1976).
F-39. Wold, H.O.A., &ibliography on Jime Sen ces and .stojactart c uIc dCic,
Oliver and Boyd, London (1965).
F-40. Daniel, C., and F.F. Wood, Fittinj Equations to Ixta , Wilv-Interscience, John Wiley & Sons, Inc., N.Y. (]971).
15Z/
F-4 1 .Cochr an , W.(,. , and (;.M. Coxe Pi Axorb Z .. oc, Jb WilIvy & Sol)',Irc. , N. Y (n. d. )
F-4 2. A b ram ow i r z ,M . , i nd I . Se gu , harzu:j, i OJ . Ca"1: .17 iwi C t "U~ , LDa orPub licat icons, Inc ., N.Y. ( 1964).
-F-4 3. Crow, E. , F. Dav is, anrd M. Max ti jeIl, -izt t k R f-': , - a r orhPe par twen t , U1 .S. NavalI Or dnanc:e To!;t S tat io r, Dov er PubIi c at Ion S , I nc
N.Y. (1900).
F-44. I saac son , E . a and HI. B. Ke 1 ler, At,. z :'e Nu-,# 'K . 7 i (AJin W ilI& Soils, [t:c.,N.Y. (19M1.
F-45. Pa rIe t t , K N . h' ALj, vz :ar~a, in Mai ii nait i c Nle thod s ftor1) i i ta I COM)-ute r se i t ed b y R I s ton and W ilIt (J,, lv 19 64)
K I BL I (ICRAPIIY
Ab r arowi t z , M. , -iiit I. Se(I in , i/looibool (-J i i''l 7( I r tet D~W ioverPu tb I i ( ot i on,, ,t In,- N.Y. (1964).
Air Po 1ltit ion Coot ro I District, Counllty of Los An (It-, *.,z~~.
Air Pollut ion Control District, County of Los Angeles (I 971)
H ingiiamcn ,D..1J. , 1.1%.. Wangen , and S .D. /e I I rAt . .4 it.yi. ". 1 . ai : 71 '-.t , ArgLonne, Nat i oal 1
I aborat or v , Ar,, onn(, ,II I . ( June 1975).
Beni .it S .8 -wJi A'.. Pit-rson , .; ; i u I at o o ... ~ r,!i l y I trst toti e John Ni W1Ii & -Olts, 111L 1 . 971
C iI(1 I I ; 14toii r ku. , Cl-i land (19/4)
B1:j i. j itan, R ., a .i ,u e a nt,< r it . Iq. . I , ItiPih I i, it ions , I iitl , N. Y . ( 1 9')8
hr&'htmet , -. A. , it ;I ., l.itiwio)d-Shaarv 1iliriry (No. 36(1 i,-l 1,6 .008) ,CUNMI(;S,
11ii ivvr-sitlN ol tilra, Atihens , tttI V I m libI)raty (No. (,2-.,;i 1-1, inwood) So ft war(,1) 1 .; r i hvit w it It pt .,C i , '. Al t o, C il i I . (11t . )
b ro w at , . , " '' ' r ; ""' !t ilo;:k t 7 titocij .r
-u- r~i.-, A'\-kD-/4fi77, Systrms, Consiltants, Ink,. (1914) .
Birown ;t t K . P 14 t' .V' n. :t loa c) to- ' i. Ptl i : etc - oI, AA-fi-I-7--Iit, Svstems (raistiltants, late. (1974).
C(r.toh ... 1. 1~ d-. r J.- re& Ki! tj ,Pj ic~,' AFWL- I P-
71- itt Kitt a a la Ait r -,a t kase, N .M. (May 19/72)
[)atil F. W-S j-oi, -u .1idIW. (,)irmao, /t 7 Ih. 'i il
Cut irs I. tie .,, N.Y. I 9/i 1
Deming, W.E., Statistical Adjustment oj" Data, John Wiley & Sons, Inc.,N.Y. (1948).
Dflaiuney, B.T., Air Quality Assessment Model (AQAM) Yield ],ita C.'oltlectonk74iri, AFWL-TR-75-220, Kirtland Air Force Base, N.M. (Aug. 1975).
Env ironment al Studies Center, Uuide L Znes f'or A Ir QeIitq, !1aintenan(£, i 2 ,ii,.tri. Aal y is, Plan Preparation, Report No. PB-237-581, Vol. 2, Environamnt Il
Studies Center, Research Triangle Park, N.C. (July 1974).
Gifford, F.A., Modelinj Urban Air Pollution, Atmospheric Environment,7:131-136, Great Britain (1973).
Gold, B., and C.M. Rader, Digital Processing of Signals, McGraw-Hill BookCo., N.Y. (1969).
Goldberg, S., Introduction to Difference Equations, John Wiley & Sons, Inc.,N.Y. (1967).
Grenauder, U., and M. Roseblatt, StatisticaZ AnaiyLs of' sti Zonar: U.
o'er'le-, John Wiley & Sons, Inc., N.Y. (1957).
Hahn, G.J., and S.S. Shapiro, Statistical M4od(e in Lnginccoity, John Wiley &Sons, Inc., N.Y. (1967).
Hillier, F.S., and G.J. Lieberman, Introduction to operati',,:s koearc.,Holden-Day, Inc., San Francisco (1968).
Hocking, R.R., and W.B. Smith, Optimum in.om.Zete .I-Mtinonz/. Uaryls,
Technometrics, 14(2) (1972). 4
Jon, P.W.M., Statistical Decign and Anal ei3 o, "xJ,'m~OtO, The MacMillanCo., N.Y. (n.d.).
Kahn, H.D., Note on the Distibution of Air Jbilutante, j. ,f the AirPollution Control Assn., 3(1l):973 (Nov. 1973).
Kalpasanov, Y., and G. Kurchatova, Investigation on the dtatistz'ia Ko:,'r "Lbstnibution of" Certain Chemical Pollutants in Air, submitted for publicationto the J. of the Air Pollution Control Assn. (1974).
Kapteyn, J.C., 5'kew Frequency Curves in -,ioon an' :;tatistice, publishod byNoordhoff, Astronomical Laboratory, Groningen, Netherlands (1903).
Kirkwood, J.G., Statistical Mechanice oJ." Trannprort Proe e.: Jcne ,,;:- -C,, J. of Chemical Physics, 24:180 (1946).
Kramer, C.Y. , in the Analysia of" Vanrir<% o, "o-Wa . :2- t £on i,;-.neqwzd Su7-ass !ihrbersa, Biometrics, 11:441-452 (Dec. 1955).
Land, C.E., Ar 'val uation of Approxiwt Co,'> , i? .ntcou ,i;. t iria t,,,pi"ltklo' ,o:norwa '!,'zns, rechnomtrics, /4(1) (19;2).
154
Lange, R., ADI't1 - A TIree Dimensional T''napo)rt-L:4uic, ,'.,Z o' theDisvou ra. o" Atmospeic Pot'Zutants and its Vahi&ztion Aaiowt Aenal..' 7 'vacerVtz !iee, USRL-76170, Preprint, Lawrence Livermore Laboratory, University ofCalifornia, Livermore (1975).
Langan, L. , Lfomobt, ' enwinj Aeaswremrenti oj" qioru . aie!ouv utOn, AIAAPaper No. 71-1060, Joint Conference on Sensin g of Et)vironmental Pollutants,Palo Alto, Calif. (Nov. 1971).
Libby, P.A., ,d, The FluL.7 Dynamics Aspects of .iir PolHution Related zoAircraf't Opera.ion ,,., AGARD-AR-55, Technical Editing and Reproduction, Ltd.,Harford House, London (1974).
Los Angeles County Air Pollution Control District, St,1 o ct AItrova,':Emssions and Air 2talit0 in the Vicinit 9 of the Los Anjeles inten .Airpopt, NTIS No. PB-198-b99 (April 1971).
Lucke, J.W., and G.W. Brown, Fatigue Data Processing ith a 2mall i;a' aComt uter, ISA Transact ions, 11(2):128-134 (1971).
Marcus, A.1l. , Air 1'olZutant Averiing Times: Notes on a Statistical Modl,Atmospheric Environment, 7:265-270, Pergamon Press, Great Britain (1973).
Mendenhall, W., lntro,liv. 7':1n to Linean Models and the Design and AnaZyris of"Ea',imots ,Wadsworth Publishing Co., Inc., Belmont, Calif. (1968).
Mendenhal I, W, , Tho o;r and Analy ui o," Exper'iments, luxburg Press,Wadsworth Publishing Co., Inc., Belmont, Calif. (1968).
Menicucci , D.F., Air Q2a.it Aserssment Model (AAM) Data Reduction, andOpcr'ation Uuido., AFWL-TR-75-307, Air Force Weapons Laboratory, Kirtland AirForce B.ise, N.M. (Dec. 1975).
Naugle, 9.F. , Do . 1ut io Lw,'"o a ion Ana. o is na U. . Ai Foo-urce AiI ,'t" i'Ain<:. , 75-93.1 for presentation at the 68th Annual Meting of the AirPollution Control Assn., Boston (June 15-20, 1975).
Nanigi-., D.F., and S.R. Nelson, U.l'AF Aircoaf't PoZlution A,-i:si-ion Fa<it e an,Lanzdiuq *nd 7%Ek(c j' (LT,)) ' ye-Zcbr, AFWL,-TR-74-303, Air Force Weapons Labora-tory, Kirtland Air Force Base, N.M. (1975).
ostle, B., Stat i iet,7e, " f i,.,,,a ch, the Iowa State University Press, Ames(1903).
Panrz,iu, 1., l iScussi on No. 2, h cus,,wn oj" 7'he n ,',. , t.',o'r'e 1 21 o& ,,5oe r i'7! .1ho, or, AitLiatizz , by William ,; . Cl. voland, Tochno-
metri, ;, 14(2) (1972).
Ra 1st (o , A., and HI.S. Wil f, f at mt tu i, fit boi" ' , tVol . I I , Johl Wi I ev , Soiis, Inc . , N.Y. ( 190)
RHo ,, O.M., ',' t:17.',1 ," o;4','*j i:(,Iz j~r'r)S A ' , o sz 4 , , Aronno, Nat iuo1,;it,,r;tory, Argmn,, I I. (April 1976).
155
'. ], - .
Sclove, S.L., (Y vS X) or (logY vs X)?, Technometrics, 14(2) (1972).
Sprunger, P.R., and L.E. Wangen, Numerical Studiet; on the Seneitivity oj 0h.Air 4uality Assessment Model to Potential Errors in Model Parameterv, ArgonneNational Laboratory, Argonne, Ill. (1975).
Soong, F.F., Random Differential Equations in Science and Engineerinj,Academic Press, N.Y. (1973).
Wyzga, R.E., Method to Estimate Missing Air Pollution Data, J. of the AirPollution Control Assn., 23(3) (March 1973).
1 '
F~iI.~I[X :)TIl 11;1-IY-T;1ATJI ON COMPAR IS (J0 MEASUkElU U .fM~IrFED CON CE NJ I f
~N ~-666f
~66
086
(I.B
K09 2 .
01
O.2
o 0o0-
I (Y
Ndd
I15)7
66366
(666
6'66
V6'866
00
- 14
X 'P4
>1 09 >
C'5
cz 9,0
c' a,
.0,
WI
-I- - - T'
Wdd.
T I iI r ~
U6(
'866
------------- ','
.16
OiAl
oil
-op, E.
'- "07 -'-
CL
>
0 <
11 11 Vi-4
4-
I IOO
111
.. 1.. 100
Wdd
66,66
666
'66 0-J
06 0
' -09 ci
'oiN.'
'N.00
M'5 crZO j
-. *0
1'04
C~i CC
Wdd
TTT T------ F---------
6614.4
'06
- 4-
0O2
I--
WI-
L ii ii
66-66
(3,66
4'3
4'9
.06
J
*01
-0 : (4-09
~ 0'0 -ci '
It II II1iz C-14
> 0- l 0
I 62
6,66
Wfb
6b
-08
10
T- -101)
i 00*
I II I II T T r -- 6E3 66
V L
(.66
4 J
-- 4
FF
N 1-0
I I - -r -T r-- --- - 10.0
Wdd
16
L I r-
I 6H(366
V 666
OC
N -- o
1656
ioi
00
.,4
(oou
06 0-U)
r~r.
.0~
00 c
Noil
1664
U-66
- 66
x N NN~ 86
N 'Oi
-K0.
(Y --
*o
t- O~
10,0
--- T-T T--r-
Q ~.t ~CiCb
WddC
167
. ~ ~ rr F T
V 8'66
V66
0
-68 0'
0i
N'O
uu
tr.o
Wdd*
T TnT rT T ~ ~ --1-- 8
.F T Ir T
[ , r- 8"66 r
."66
'6b
- 86b-
K ...-
Ci 'ou
01
ol
c cr
C.,,
F u
I 1 IId C)
PWdd
169
. .. I . . . .. . . . . - . . .. .. - q lr . ..--. .. . . . . . ..m p an ' ' '
-6666
8'66
I66
06 11
'5 008 .,
Nr
N')-o
tu U
I- 0
0 11 11 - TO*
Wdd
170j
- - - - -- -- -- - - - -- -- -- - - -
(.C)
-
>*1-
0?'
woo
INdd
ill
B66
-4-6
-66
4
F-8
-08 4
U)
h '~ ~- Co *
o 0 0
NdN
172
- I-
V -~-66
66
. 8 6
-06
'N0
')
r-, a' . o9
SI
1731
60.(
7- T7T - F T Tr-Tr TT - - - 1 '
173
66
866
8966
-66
44
0 0'.
00 4
-Aa)
~ oc
-4 c r ~ o
100
-o
Wdd
174
--r - -- -1 1 I
(;f366~
-866 Z
- I ~86 t
r44
10,
F N,,
175'
I ;U66I6
R'66
- 866
'66
E-4
N-4
N 0N
N' 0
L C
'2'
I o.4 CJ Mg~
TT- 1',0
Wdd
1761
AD-A094 003 AIR FORCE ENGINEERING AND SERVICES CENTER TYNDALL AF--ETC F/6 13/2ANALYSIS FOR THE ACCURACY DEFINITION OF THE AIR GUALITY ASSESSMN-ETC(U)MAR 8O R J YANARTINO. L A CONLEY, D N ROTE
UNCLASSIFIED AFESC/ESL-TR-80-9VOL-2 L
INDc-E
UZ~
____-4M
1--,
01-
114
02
cQ N
SINHARL40 *O
1773
46J
4j
0
-I
0 L)
4j0
>J 44
.0 C)
Lo 9,0-o4j
0
Z 01- 4j'
j4J
o-
SLLNHAR dO'O
I .
1 ,
I '
C
14-
- : ;I 44
/ U
." N
dSN~ d r- 0 .
_179.....r_ _E--, ] =
I.-
d)
179I I
41I
0,?
0
U
0 (
01 1
0,0
0u
SU A 40 'ON
I fi
A1
-4
C)C
>4-4
0 'o
E. 0-0
00
.010
'-4
I -
-OLnj
11
'a
SINHALH JO 'O
181.I
OZ
,r'l
00
-- 0
$4
Z
wC)
01
0 0
,.. t4.
i 00
[--- j 0'2- ,
4
O 0.00N
I~~ Ai----
C
*
z U0
o F--41Q
Z I- C
0
coC-
CY
CQ.
SLLNHAa dO *ON
183
CY)
Z" a4 0
00.q
4J 4
0*0 4
0.
'00
z 4
___ 01
LIZ-
CD_ In1
SINHARdO *O
184 t
4.'
O-Z
4.'
0
C;))
.9C 0 --
ca'
I c)
0*1
0-C
SINHARC AO *O
185 3
IV I
0i 0
a 0
rn -
-01w
00
~ 4j
.4 -4
SLN'4A',i 01
r -!
-01
00
CL
* 4j
F---- iV
-r~~ vs-I
187
i A
-1
H -40 Cl
wC)
.0 cH 00 C00
c01 -4
-4
* * N NQ -4
SLLNaAa do 'OK
inil .'- r r --..- -- 1 ;T -=
(C -
0
Y)
ol z
00
t 4
-(,0i-g l au
d [-o2
Q)C
00g / ~.- '
-"-
"L -I... J ++' La.
L .u n0 - -4
[- -- - -'4
0 0 - 0"0--
SiNdAa do 'ON
189
-"Nor
4.J
C"))
0)
00
a0
a)
oi0 0
4-J 4
zv -
190
0.04
4
rI)r
109
APPENDIX H
ANALYSIS OF HIGHER REPETITION RATE DATA
I. INTRODUCTION
D~ur ing normal ope rat ions of the me teorol1oi cal and a (romn,' tic i ini tor 1I 4'facility at Williams AFB, each station and thus~ each e xperimental lv reau'qua'ntity was sampled once per minute. This repet it ion ratt was, at I irstconsidered questionably low, (,specially in view of thet slinit t ime s;cales
a ssoc iated with a ircra ft operat ions or atmosphuer ic ttruirhn ec-- Thufs, thquest ion was invest igated of how hourlIy averaiges otI 0(in1wpr i not t counpitcwi th hourl1y averages c omputed on the ha si i; of 1 liglut rtope- I it iouu ratt dat 'Ia
The data sample for th is compar isonl was hbt ai id (Iil. i g May, 25-27, 1,)? /,by ope rat ing the Monitor 'labs 9400 Dat a Acqui ; it ioni System at a i gher salin-p1 ing rate at the expense of number of channoe Is sanitt ed . Thie samuples chosenfor analysis consist of 17 hr of wind speed (hswind direction (WD), NO anld
NO5 data taken at 6-sec intervals, and 5 hr oit vert ical wind speed data takenl
at sampling rates as high as once per 2 sec.
The basic analysis approach was to:
(1) calculate hourly averages using rapid scan data;
(2) subdivide this rapid scan data into subsets containingdata at 1-mmn intervals;
(3) calculate hourly averages using the once-a-minutedata in these subsets; and
(4) compare the two calculations.
These data suiggest that the following uncertainties relited to samplila,ate- nav be associated with the once-per-minuite Wi I I jams AEI3 measuretmnts.
Pollutants. NC)X < 2 /,
NOC) 1 2; (usually 3'/3%)
Mteorological quantities: WS < 3%
-Z <8 (of e
a@ < 12% (ulisutiall Iy K8)
ow 1 7'/
In the last sect ion of this appendix, tho "one-fifth" power law isinve .t igated . Thlis law, describing the time-dependent increase in o, wi thaveraging t ime raisod to the power 0.2, is anl assumption inherenvt nt mall" 4itquality dispersion algorithms, incltiding AQAM. These data sugggst thaitbetwt-.n avera.,in. times, of 3-60 min thi!s power is 0.27 -+ 0.12 in wr-mi t
wit!- 1heonefiV vjtu
192
INIEINIANCK OF coMmaI'I) IIOURLY QUANTITI;S FROM SAMPLI NU FREQUENCY
From the rapid scan data taken on days 145, 146, and 147 ot the experi-ment, 17 hours of windspeed and wind direction measurements are availabl,..TIi s data, t aken everv six seconds, can be used to compute var ious5 houry1 Iquantities such as hourly average windspeed, hourly 0(, etc. It all the
6-sec data values for I hr are used in these calculations, the sampling tr,.-quency " ), the data is once per 6 see, whereas the sampling time (0) is I hi,since hourly quantities are computed.
If, however, one only wanted data with a sampling frequency )l s=1miuutc - 1, one coild simply pick out every tenth data point from the S=(#s,*c) - data. However, because there are 10 of these data points -.v,'rvminute, there are 10 different times on which to start the first minute of til,.S=1 inin- data. So, for the subsequent analyses, the S=(b sec) - 1 data is
broken uip into 1,) independent subsets of data for each of the 17 hr. Eactsubset of data therefore has a sampling frequency of once per minute.
Le t
= the data points that make up the complete raid scandata set, with sampling frequency S-(6 see) -
. i is
an index.
a nid
$_ik = the data points that make up the jth subset of theS= min data. k is an index while j indicates the subset.
Then the process whereby the 10 subsets of S=(l min) - I data are ex-
tracted from the S=(6 sec)-1 data can be represented by:
4jS.] = P#j +2 !
3k : *j+lO(k- )'
wh ich can be repeated for each j, j 1,2,... 10. The asterisk indicates thatthi, data has .i ;ampl ing frequency of 1 in -
The object of this part of the analysis is to compare the values ofvarious hour (v quant it ies obtained by using the S=(6 sec) - 1 data (all ito the vatues obt ained by using the S=1 min - 1 data (the k ) .
193
Let St (Xi ) be some stat ist ical process that operates on tit set of dat apoints Xi for the sampling period 't For examphlie, S.- (Xi ) couldt represent -InhlourlIy average of thle Xi, in which case:
I x. hr (Xi) N X1
where:
N =the number of Xi in 1 hr.
Then define the following operations on the meteorological data, and letthle variables indicated represent the desired hourly quantities:
S1~ hr (00 = value obtained from the statisticalprocess ST~ using thle S=(6 sec)-1 dataduring a 1-hr sampling interval Cr = I hr).
=S 1 hr (''0= value obtained frm the statistical processJ I ~ST using the S=1 min-' data in the jthl
subset during a 1-hr sampling interval.
To show the independence of Sl hr(Xi) Onl sampling frequency S, one mt;compare the value 0 (thle hourly quantity computed from S=(6 sec)-1 data) tonthe 01values (the hourly quantity com~ued from each of the S=1 min-, datisubsets). But because there are ten Ojvalues (one for each subset ef Smin-l data), it is convenient to define some method for comparin thtv.c Val!estwithout having to look at each 0.Hwvr ic ahsbe j .the same chance ot being selected as a unique set of minute-averaged data(assuming a random start time for data collection), all values should he'considered when comparisons with 0 are made. Thus it is re-asonable to deltea "stanldard deviation"' that measures thle distr ibut ion ot thle "t value's a reuuid
tht2 %alue:-
1 10 - ] 112 (
wh ere : i
oC3M the standard deviat ion of the u ' aroiin id wilu %h. x
A low valuec for r3(40) would inpl v t hat tilh houi , wk~li it ;' . t hr!th ~ S1l minl data (4*) and the S-(f sec )-I li t (1 .. ue,Iv tie lt, i Alrdtherefore, that the operat ion SI hr(Xi ) Is idpirih'te avcra ' ni,! t !of the data Xi.
one may further define the mean value lt thc A
194
10
(H-2)
j=1
As a measure of how well the value 0* represents the individual quantities 0>*
use the standard deviation:
o(1) = K 0 -(H-3)
j=1
In cases where S defines an arithmetic average, t will be exactly i-cqalto 0*. For other processes, however T* can be used as a representative value
for the individual Il's and may differ from the value ).
As an example of this approach, suppose that ST(X i ) represents the hourly
average of Xi, and that the Oi are the rapid scan measurements of wind speed.
Then
Oi = wsi
jk = WSjk
600
1
O W = S hr(0 E -(vsW.
ji~l
60o=ws = s (wsk)- V ws.k,- hr j 60 and
k= I
-- [' 1 0 -. ~ 1/o(0) = o(WS) = " X E (W - -S-)2 j/
=l
() O(WS*) l[~+ ]
Oi supposo., that S.E(X i ) represents the hourly standard deviation of the
wi nd i rection, ( . Then:
195
*i = WDi,
1k = WDjk,
=Oe = .600 2 1/2
600D(WD.i=1
* = o. 6_k0 - - 211/2
8ij 0 k1(wD j WD 6
j=l
6 10 100 andjl
S11/2
j~el
If a computed quantity * (such as WS, WD, oe) is independent of thesampling frequency S (at least to the extent of S being either (6 sec) -1 orI min-l), then one would expect that o(O) will have a low value, indicatingthat the O's are very clost, in vali. to 0.
3
Let Vi represent each individual wind-speed measurement and let 0 i repre-sent each individual wind-direction measurement. Then, the equations actiallyused to compute each of the hourly qtiant it ies are as follows:
average wind speeds:
N
ave rage wI rd d I 1,'11
WD IWI' = t,, (H- b)N
vector mean wind speed:
VmWS = - V. sinO. + V cos 1 , and (H-o)
vect, mean wind dire.-tion:
N
I Vi sin 6 i
VMWD = tan - 1 i= -7)
Vicose i
i=1
The standard deviation is defined by:
0 --- (X i - X)2
In computing the standard deviations of the above quantities one should notethat for wind directions the smallest angle between X i and X is used; that is,X i - X is always less than or equal to n radians.
The results of these calculations for the 17 hr of rapid scan meteoro-logical data are shown in Tables H-1 through 11-6.* The important thing to
notice in these tables is the low value of o(o) in most cases -- implying goodagreement between the 0 and ** values. In the case of the hourly averagewind direction WI-), the largest value of o(W-D) one encounters is 2.40. Forhourlv VMWD, the maximum o(VMWD) is 2.20. Both values are well within therang of instrument error. For WS the largest percentage deviation (a(;4)/ x100L,) is 3% and for VMWS is 4%, but most values of o(1) are less than or eqzialto of the value of 0 itself. In the case of 06, the relative values ofo&( ) are larger, ranging between I and 12% but with over 1/2 of the values ator H,,Iow 6%. Therefore, the value one computes for a quantity during a givenhour will not depend greatly on whether the data used for the computations was
taken at six-second intervals or one-min intervals during that hour.
In addition, from the values of the standard deviation of o(D*) it isapparent that the values computed from the S=1 min - 1 data subsets (f) do notditfor widely from their mean. This lack of spread implies that t* is ropr,-sentative of the individual 0 and that the starting time of the first minuteis not important --. any set oi minute-averaged data recorded during an hour isjust as valid for computational purposes as any other set that starts at adifferent time.
So it is apparent that any minute--averaged data, regardless of thesar t ing point during an hour, is just as acceptable for use in computing
*Tables and figutres appear consecutively at the end of this appendix.
197
hourly quantities as data taken with a higher scan rate, at least down toS = (6 sec) - l. Hence, if computations are to be done only for I-hr samplingp1 riods, nothing substantial is gained by using a scan interval of 6 sec
instead of I min.
In the same manner, hourly average pollutant concentrations can bmhecked for dependence on the scan interval S. Along with the meteorologic.ildata, NO and NOx were recorded at 6-sec intervals during the 17 hr on days145, 146 and 147.
Tables H-7 and H-8 contain the results of the same calculations usingthis NO and NOX data. Again, notice the good agreement indicated by the lowvalues of o(o) for most hours. For NOx, the value of o(NOx ) is never morethan 2% of the value for NO itself. The values for o(NO) are relativelylarger, ranging from I to 12% of the value of NO. For most hours, however,the- ratio is under 3/.
Therefore, as in the case for the hourly meteorological quantities, dataaken with a scan interval of one minute as opposed to six seconds seems to.ake v~rv little difference in the values computed for the hourly average
"- - nt rat ion.
i. THE EFFECT OF SCAN RATE ON THE AVERAGE VERTICAL WIND SPEED AND THESTANDARD DEVIATION
Let W represent the vertical velocity of the air as measured by a verti-cal propeller. The data used to compute W (the average vertical wind speed)and OW (the standard deviation of the vertical wind speed) comes from five $hours of measurements taken at various sampling frequencies. The notationused here is the same as used in the previous section, with the exception thatnow there is more than one rapid sampling rate. However, the object still isto compare the results obtained using data with a short-scan interval (the,mpuced quantity denoted by 0) to the results obtained using S= min - l data
(the quantities denoted by 0*).
As before, the S=l min - l data is obtained by selecting points from tterapid-scan data at 1-min intervals. Because the shortest scan interval variesfor different hours there will be a different number of subsets of minute-interval data in each case. In general, for this minute-interval data k,
j=1, 2, ... , 60s, where S is the sampling frequency. The use of the standarddeviation o(o) to compare the 0* values to 0 is very helpful here because
,) accounts for the different numbers of O values.
Tables H-9 and H-I0 show the values calculated for o W and W. A samplingperiod of I hr is used here, and, as in the last section, the relation between
and the 0 is emphasized. In this case, however, notice that the agree-ment between the 0 values and the O values is very poor, as indicated bythe large values of o(O). For aW. the values of 0(cW) are on the order ot10 of the 0W value itself. For W, the relative values of 0(W) are muthhigher, ranging between 13 and 73% of the value for W. Thus, it is obviousthat the scan rate for vertical wind speed data does infI,,n', the (;alclated
,,rly quantities. Thin influence could be diie to the large d .viat i,)ns
198
tillullt.rtd in th. v,.r icI wind s ,d --- nt ice ti;t (,W is always Iarge.r than
W. A ,-rn ilt -sc an 1n trva I may not provide enough detail of the W traceI)r use in calculat ing an Iiouriv quant it y that is representative (i.e., onethat is independent of th, starting time of the first minute). So for verti-cal wind speeds, a -can rate higher than once per minute is desirable.
One other interesting point to notice is the diurnal variations of W4.Of course, no definite conclusions can be drawn from five hourly averages, butit is reassuring to see that tihe highest values of W occur in the afternoon,when the temperatnre lapse rate is maximum, whereas at night and in the morn-
ing, when conditions are more stable, the W values are lower.
4. THE STANDARD DEVIATION OF THE WIND DIRECTION AS A FUNCTION OF
SAMPLING TIME
It has been suggested that the standard deviation of the wind direction,o 0 , depends on the sampling time in the following way:
aT) 0.2_0( __ )- (H-8)
where :
TO = a reference sampling time,
r0(T o ) = standard deviation of the wind direction calculated
with sampling time to, and
o0 (t) = standard deviation of the wind direction calculated
with sampling time '.
The 17 hr of rapid-scan meteorological data provide a data base with which to
tesL this equation, since, even for short sampling times, there will still bea suitable number of data points available for the calculations of 08.
i-quation (H-8) will be tested for three sampling times: r = 3 min,10 min, and 1 hr. In all comparisons, t is taken as the shorter samplingtime. However, to compare o 0 (M) and o8 (To), the average of the a0(t)'sduring the longer sampling period To is used as the value to be compared
with oe(to).
Assuming that the form of Eq. (H-8) is correct (i.e., that it is a powerlaw relation), the exponent can be optimized with respect to our data in orderto check the agreement. Generalize Eq. (H-8) by using a variable p as theexponent:
a (r) P0 )(H-9)
o
Taking tie log of both sides, rewrite Eq. (H-9) as:
199 I.
log (T75)(H0p = Y; \ 7)O
log TO
For each set of calculated 06's J0e(T) and oe(To)J a value for p can bedetermined. The mean of these values for p, denoted as P, can be used as thevalue for the exponent in Eq. (H-9). The standard deviation is then an indi-cation of how representative is for different conditions.
In Tables H-11 to H-13 a value for p is calculated for each of the 17 hrof rapid-scan data using different combinations of sampling times. Note thatthe mean values for p determined from data are higher than the exponent valueused in Eq. (H-8). For T=3 min and 1o=60 min, =0.265 with a standard devia-tion of 0.116. For t=10 min and 10=60 min, P-0.227 with a standard deviationof 0.125. For T=3 min and To=l0 min, P=0.322 with a standard deviation of0.134. This last comparison, however, is between the averages of the 0
(3 min) values and o0e0 min) values during each hour.
With such relatively large standard deviations, it is apparent that Eq.(H-9) is only a first approximation of the relationship between Do and
sampling time.
Instead of this power law relation, an exact identity can be derivedthat relates a o0 calculated for a given sampling period to the G0's calcu-lated for a set of shorter sampling periods that lie within the longer period.Specifically, suppose one has a set of wind direction measurements taken witha scan interval of s seconds. If the longer sampling interval is denoted byT L and the shorter sampling interval by TS, then there are exactly J=TL/TSshort sampling periods contained in the long interval.
If j is an index that represents the position of the short samplingperiod during the longer one and i is an index that represents the positionof a wind direction measurement during the shorter sampling period, then theindividual wind direction measurements can be represented by Oij , where i=l,2, ... , I and j-1, 2, ... , J. I is the number of measurements taken duringthe short sampling interval TS . The mean wind direction during the longersampling period is
80 ) .- :2L ) 1 i j ij
and the mean wind direction during each short sampling period is
6 (!I . .j S I i ij
rhe standard deviation of the wind direction ior the long and short satp
puriods is defined as
2 C i
a T TX(6i L)-ij)21/,
0 0S (0~ j TS
By squaring and expanding tile expression for O(_TL), adding and stl:tract-ing the second term, on the right hand side of Eq. (H-11), and substitutingfrom the other equations, the following relationship can be obtained:
2 I ) + (-(- ) - )) 2
0 a (- L = 1 " j ' S + J j .L S (H - Il)
This is an exact equation that relates the 9 's calculated, over several shortsampling periods to a 0 that would be calculated if these siort periods werecombined into one long sampling period.
Using one hour as the l mg sampling period IL and 3 and 10 minutes as twodifforent short sampling perids TS, values for 0j(tS) and o 0 (i S ) can be com-puted from the rapid-scan, wind direction data using the same methods as inSec. 2. Equation (H-1l) can then be used to calculate a value for ce(I hr).Tables H-14 and H-15 compare the actual hourly 00's and the os's predicted by
Eq. (H-l) for short sampling times of 3 and 10 min.
Since Eq. (h-1l) is exact, the close agreement between the calculati-d andobserved oe's are not unexpected. With one exception, the two values tr ue
are within 1.5 degrees of each other for the case S=3 min and within 0.6degrees, for the case tsl t min. The exception in both cases is the hour
starting at 0930 on day 147 where relatively large differences between the c evalue s occur.
All these differences in the 0e values are probably due to the fact thata is calculated using Eq. (H-5) of Sec. 2 instead of the linear mean, for 0used in this section to derive Eq. (H-It). The larger deviation occurs duringan hour with low wind speed and a meandering wind direction.
5. MLAN SQUARE EDDY VEIOCITIES AND FILTERING OF THE RAPID SCAN DATA
In order to define what one means by mean square along-wind and crosswindeddy velocities or tluctuations, one must first define the mean velocititsfron which the instantaneous quantities deviate. In the case of the along-wi nd component, the instantaneous Ve locity, u i, is lust the sum u6 + ui , where
6 is te mean or low frequency wind component and u i is the eddy or highfrequency wind component.
At the time of the ith sampling, let the instantaneous wind speed anddirection be given as Vi_and 6i, respectively. (See Fig. H-l)* If the "mean"quantities are given as V and 9i, then the along-wind fluctuation is just
Ii = Vi - Vic soi (H-12a)
*Figtres and tables appear consecutively at the end of this appendix.
201
I. it tI h' cross-wind f Inct uat ion is just
V i = -visill i (11-121))
l , re
6i = 6i -e
The mean square eddy velocities are then given by the expressions
u12 , (U') 2 (H-13a)
N,2 1 2vo = N (v, ) (H-13b)
t i ere:
N = the number of events available during
the sampling period of interest.
To obtain the "mean" quantities Vi and 6 i, a number ot averaging (orriltering) techniques can be considered. One method, )f course, is astraightforward linear average of the vi and 6i during the sampling period.
The advantages of this scheme are:
1) it sets an upper bound on mean square eddy velocities;
2) it may be most indicative of what is actual!y happening
to the pollutant; and
3) it is a standard statistical procedure and is easy toimplement and interpret.
Some of the disadvantages, however, include:
i) The mean value depends strongly on the duration of thesampling period, which most likely is of arbitrarily
selected length.
2) Poor connection with u v w results, especially under
light, variable winds.
To overcome these problems, it is necessary to consider a moving mean for
7i and Ui. In this way, large, low frequency fluctuations during the sampl ingptiod do not introdue large deviations in the along-wiud and cross-windVc,cities ( i and v i ) because these low frequency compotent s bec-ome part otthe moving mean.
One way of obtaining such a moving mean is to borrow the techniqutis ot,l,--tronic signal filtering and actually construct a low-pass filter. When
202
the individual 0; and v i are run through the filter, the output is a timeseries consisting of the frequency components below the cut-off frequency ofthe filter. The higher frequency components will have been removed, andtherefore the output signal is suitable as a "moving mean" in that when it issubtracted from the complete signal, the remaining values are the true highfrequency turbulence components of the wind speed or wind direction.
Consider the three stage low-pass Butterworth filter of Fig. H-2, wherethe output is taken across resistor R5 . From analysis of the circuit,given an input voltage Vin(t), the output current 1 through resistor R5 isgoverned by the third order differential equation:
a I + b I + C I d I Vin(t) (H-14)
where:
the dot denotes a derivative with respect to time, and the coefficients havethe following values:
a = RlR 5C 2 C4 L 3
b = RlC 2 L3 + R5C4 L3 (H-15)
c = RIR 5 C2 + RIR 5 C6 + L3
d = R1 + R5
To make this filter have a cut-off angular frequency of wc (rad/s) thevalues of the components should be set to:*
RI
L3 2
1(4
-
R5 I
Actually, wc is the half-power point on the frequency response curve of thefilter, which means that at wc, the output voltage is 71Z of its value in thepass band. Because of the construction of the Butterworth filter, the outputvoltages even in the pass band are 50% of the input voltage. This conditioncan he corrected by simply doubling the Gutput of the filter.
Ujsinig the above values for the filter components results in coefficientst r thle d i i lerent i a I equa t ion of the form:
*Soe A Handbook on Electrical Filters, White Electromagnetics, InC.,Roc kv iIIe, Md ., 19 63.
203
2Wc3
4b c 2 (H-17)
c 4Wc
d =2
Thus, the differential Eq. (H-14) becomes:
2 I + . + i+ 21 = v. (t) (H-18)w 3 Z-2 W in
c c c
Vout is then obtained from the relation
vout(t) = I(t)-R 5-2 (H-19)
since the output voltage is developed across resistor R5 and must be doubledto compensate for the unavoidable attenuation. But because R5 = I,
Vout(5) = 21(t) (H-20)
where:
I(t) - the solution of the differential Eq. (H-18), and
Vout the low pass filtered signal using cut-off frequency wc .
From steady-state analysis of the filter, the steady-state frequencyresponse is determined to be
R5An
A = 5 in (H-21)out S
where:
Ain - amplitude of input voltage signal,
Ao ut - amplitude of output voltage,
S l IztlH = (x2 + y2)1/2,
Zt - Complex impedance - x + iy,
x = -bw 2 + d, and
y = -aw 3 + cw.
where:
, = the angular frequency of the input signal, and
a,b,c,d are defined in Eqs. (H-15) and (H-17).
204
--raw.
The phase angle of the filter circuit is given by:
= tan- (x) (i-22)
The steady-s Late response of this 3-stage, low-poi., B utterworth fit Ir ishown in Fig. H-3 for a cut -off period "c = 10 mun. T c = 2ti/we.) Noti,-,that the filter effectively attenuates the frequency -omponents greater thanWc, even though the rolloff is not extremely sharp.
Now, instead of using strictly voltage levels as the input signal to thefilter, use the wind speed and wind direction signals as if they were volt-ages. The output of the filter will then be the lew-pass filtered wind speedand wind direction that is suitable for use as a "moving mean," since it onlyincludes the low frequency components of the input signal. The advantages ofusing a filter to get a moving mean are:
1) the method ties in more closely with classicalturbulence theory; and
2) the mean follows the low frequency componentsof the signal and therefore does not introducearbitrarily large deviations into the eddyvelocities.
However, some important disadvantages are:
1) The insight into total pollutant dispersal poweris not as clear as with a fixed mean; and
2) The filter introduces a delay time into the outputsignal.
The last point is very important when the eddy velocities are to becomputed, since a matchup must be made between the total signal and thefiltered signal, if their difference is to represent only true deviations 'Ifthe wind speed and wind direction. If the delay time, td, introduced by thefilter is known, then the input signal Vin(t) can be matched with the outputsignal Vout(t + td) to get the desired moving average of Vi(t) at time t.
Recall that Eq. (H-22) defines the phase angle introduced by the filterc ircuitry:
6=tan 1 X
It t d is the delay time, then at angular frequency w, the phase angle (whichis a function of w) is:
O(w) = wtd. (1t-23)
Substituting the values for a, b, c , and d from Eq. (!t-17) into the values forx and y of Eq. (H-21), one obtains:
205
x = - C + 2, and
y = -2 + 4
I t ( << Wc then
x - 2,
y 4 WaC
e tndS-tan-I
(2 W, - )
xpanding this and neglecting higher order terms yields
(- 2- WC
and substituting this into Eq. (H-23) yields
t = 2 (H-24)d "twc
for the cases of w<<wc. Also note that td can be rewritten using the co-efficients of the differential equation:
t d = -C (H-2 5)d d
This equation gives a simple approximation for td, but it is strictlyv,-,lid only for the frequencies w(<wc. For frequencies greater than wc, thisis not a problem because these frequencies will have been filtered out of theoutput signal. For frequencies around w.C where attenuation is not complete,t j is underestimated by Eq. (H-24).
The phase angle 0 at w = wc is
0 2.356 radians,
so from Eq. (1-23), which gives the exact value of td, one obtains:
e 2.35bd --
200
Trhe approximate Eq. (H-24) would predict
2d w c
which amounts to an 18% underestimate in td at W~c-
For example, not ice in Fig. 11-4 hlow thle filtered signal of tic, Windspeeds (adjusted tor td) contains only thle low frequency comnponent s of thlewind speed trace arid serves as a moving average for thle wind speed. Alsonot ice that the approximation for td effectively compnrsatos for tire dL0a'.'ttine introduced by the f ilt er for almost every frequencY corijurwrelt of tilef iltered signal .
Filter ing tihe wind direct ion is more, diftficult because of tlrt, cuit at360'. So instead of directly filtering thle Oi values, sin~i and cosOi are,filtered separately and then recombined to obtain 6i via tihe expression:
(sine )0. tan 1
Hf(-26)i f
wh e re:
t denotes that the filtered values of sinDi
and cosDi are used.
Thiiis is not strictly a correct procedure, but it is a usettl and s imple te~hn ique 'hat gives reasonable results. In Fig. 1H-5, rnot ice that thris methondret urns a s ignal1 for ~t hat cor re sponids wellI to whrat wouild he expected 'ITa moving average of thle individual Di. Results of fil tering thle wirnd spei(dand wio idIrec tion data from a di fferenlt hiour are shown Ii Figs. 11-6 .ind Ii-1.
Values for 0i and Vi, then, can be obtained by filt ering thle r n'fiv i dna
t:ird Vi, irsing thiis low-pass But terworthi filter. T1he actual te tciq iitii
inv olvye; rnumeri cal11y solving the Differential Eq. (1i-l1 ) with the w i Ild spt.e4d
s igoal (Vi ) or tire sine and cosine of the wind direct ion signal ( sin0i andcosoi) as thle irnput forcing funct ion Vin(t ). The filtered outpUt V,,, is thenca-Ilulated frtsIn Eq. (H-20J). Tire Va lIe for Vi is obtained Ji rect ly in 11!
way , butL Di mutst be calcul a ted from Eq . (11-2b)
Toe so lve tire, third-order Di fferent ial Eq. (11-18), it mtist he, rewi itt'1 .rr athrroe f i rst-o rde r di i Iftrent ial eqirat ions that art- tierit solIve'd rirri, rii al v
- - thle Runge-Kit ta miethord. Thiis t idin iqirt is des irablet i or obta ilii ng t it, I I I t T'
Olt ptrt ecu tie' inlptt I rnct ion Vi nkt) need onrly be kno1wi art cen tinnI i meint ervalIs. ience , at seqntnce of data po ints carl he iie snpnitt, t ieiii rica!l me thodt w i thout knowledge of tilie continuonus I IncI I ioil tlc~ Ir w'Olilt s
represenlt .Moreove r, t it is iretired prodites a sequernce III po iiits a, aj I r I t -rednir f ptit thIa t , whren ad( jrrs ted for die lay L imr , can bet used -I,; ;I , )t I lllel tm~ i rik
mean for (I i sp1 -tv pturposes ( Figs . 11- 4 t irrergir Hi-7) or for -,Ii I it r iIn. .
Equat ions (I1-i12a anrd hb) ca niortw be xisuti to c a IcrrI at f t(1 I li arnd vjTir, mer sqtiare eddy ve'lilt it ies (JT77 and v 7 ) are t hen calciri at id ii,); Ion;(11-- a anti h). For a crit-onif period of 10 min and a 'qarn; ini I imo ,l iin-
207
hour, tie 7 and v 7 computed from tile 17 hr of rapid-scan meteorologicaldata appear in Tables t{-16 and H-17. As in previous sections, the computa-tions are done for both s = (6 sec) - 1 data and s = I in - l data.
From the large values of a(4), it appears that the § values are notn,'cessarilv the same as the value for 4. This factor indicates that averagingt ime has a noticeable effect on the calculation of u' and 77; although thefilter cut-off period Tc might also have an effect, since, an averaging timeof one minute means that after removal of the moving average, only fiticttia-tions with periods between I and 10 min are left. This band is not vet rywide, and it may not contain enough information to compute the R.M.S. edd%velocities.
The selection of the cut-off period for the low-pass filter Tc = 10 minwas almost arbitrary, being based on the desire to center Tc between the sam-pling time (T = I hr) and the sampling intervals (6 sec and 1 min). In thisway, the filter eliminates fluctuations with periods between 10 min and b sec.These periods can be detected with an averaging time of 6 sec or 1 min andwill contribute to the values for the mean square eddy velocities. Also, Tcis less than 1 hr so that the moving mean follows the fluctuations withperiods between 1 hr and 10 min and does not become much like a fixed mean
-.- i,:h would happen if Tc )> 1 hr).
For comparison, Tables H-18 and H-19 show the mean square along wind(u') and cross wind (vr2 ) eddy velocities computed, using a fixed linear meanfor ei and vi . The hourly average wind direction (WD) is used for Oi and thehourly average wind speed (S) is used for v i .
Notice that the values for W2 and VT7 computed with a fixed mean arehigher than those computed using a moving mean. However, the values of o(o)are just as large, indicating poorer agreement between the 0 and valuesin the case of u 7 and v- 'v than for other hourly quantities.
6. APPROXIMATE ErQUATIONS RELATING VARIOUS COMPUTED PARAMETEFS
Manipulating the equations used to define the hourly quantities WS, WD,VMWS, VMWD, uT'2, and ' leads to some new equations that interrelate some ofLhese quantities. Before starting, however, it is necessary to derive amethod for determining the correlation between a set of wind speed measure-ments V i and a set of wind direction measurements 8i . It can be shown thatthe agreement between WD and VMWD is an indication of how uncorrelated Viind ei are.
From Eq. (H-7) in Section 2:
NSVi sine i
-I IVMWD = tan N
Vi cose i
i2I
208
If Vi and ei are uncorrelated, then
VMW) anI N N ]Il Vi i e i
= tan-1 i. .I i CO se
as defined by Eq. (11-5) of Section 2.
Thus, if VMWD W , then Vi and 0 i are uncorrelated. This result isimportant, because the separation is used many times in the subsequent deriva-tions. From Table 1-1, notice the closeness between VMWD and WD; if ecompares the differences in VMWD and WD to o0, the largest difference as apercentage of 0 is found to be 13%, but most values fall below 10%.
A relationship between o0, WS, and VMWS can be derived as follows:
Given that
a [1~e - 2 / and
WS =-
N i '
an
VMWS = k Vi sine )2 + (1 V. cose)2]i12
Since V i and e i are uncorr lated
VMW V N s ,in i 211v x w s ~ X ~ i u ) 2 + ( I o s
- (sine) ( (oso) • il-ti)
209
I f
~e.= e - .(Note that -Z-6 0.)
then:
sine1 = sinO cos(AGi) + COST sin(O6i),
and so:
sin = -N! sinecos(he. ) + cos~sin(Ae)]
Since sin0 is an odd function, Eei = 0 implies that Xsin~ei 0 also.
Hence:
si1ne = sin o s(A 6) (H-28)
Similarly:
cOS = COSeT o(A) (H-29)
Substitution of Eqs. (H1-28) and (H-29) into Eq. (11-27) yields
2 =-S (SinWe)2 .(cos(AO))
2 + (cos0) (os(Ae))2
-(cosCLA0))2 .
If Ae is small, then use of the small angle approximation gives
(w)2 1 ()2=[~ (Ae.1) 2 +(AB e) 4 2
2T S 1 +~ 2-4~
rhis equation can be rewritten by using the definition of ce:
(IW) a 1 -02, (H-30)
whi, h is valid for small values of hei and, subsequently, ce.
210
Rewriting Eq. (H-30) one obtains:
[I _(VMWS 211/2(1- )
The validity of this equation is shown by Figs. H-8, H-9, and H-10 where G0
calculated from Eq. (H-31) is plotted against the co calculated directly from
the wind direction data. In Fig. 11-8, the hourly quantities used are those
computed from the six-second-scan interval data (tho quantities denoted by 0).
In Fig. H-9, the means of tile quantities computed from the one-minute-scan
interval data are used (the quantities denoted by i*). In Fig. Ii-10, the
individual quantities computed rrom the one-minute-scan interval data are used
(the quantities denoted by I)
Notice that for ao values below 40', Eq. (11-31) predicts a value for Go
that corresponds very well to what is actually observed. For co values
smaller than 20, Eq. (H-31) predicts a value that is within 10% of the
observed 06 value. For uo values of less than 400, the predictions are good
to within 20%. In almost all cases, however, Eq. (H-31) slightly underpre-
dicts the c0 value. Figure H-10 shows that even the individual 0 values
give good results, and more important, that the scatter of points is along
the one-to-one correspondence line. This shows that Eq. (H-31) is applicable
to any set of hourly averaged quantities computed from minute-interval data.
Manipulating Eqs. (H-4) and (H-7) of Sec. 2 and Eqs. (H-13a) and (H-13b)
of Sec. 5 in other ways leads to some different relationships between these
quantities. Two of the more useful ones are listed here without derivation:
0 - V2 2"l/2 ,and (H-32)
a= OV2 (1 - 20_ (1t-33)
wh ere :
V = average wind ;peed (WS), and
o V = standard deviation of the wind speed.
These equations use the small angle approximation to simplify expressions with
0 and assume that u7 and vT7 are computed using a fixed, linear mean, is was
done in Tables 11-15 and H-lb.
lin Fig. H-11, u0 calculated from Eq. (H-32) is plotted against the
calcuilated directly from the wind direction data. In lip. H-12, Eq. (11-33) is
tested in the s;ame manner. For both cases, tile € values are used in the cal-
cilat ions. Referring to Fig. It-Il, not ire that Eq. (1-32) is accurate to
within 10" tup to of) values of 20* and is accurate to within 30% up to 40'.
From Fi ,. 11-12, how,.ver, I is apparent that E . (11-33) is val id over only av i, \ narr ow range. rhc. calculate d value for i' is with in 10% of the. ,I t lial
vallo 11p to (1.2 m )/ s .and i5 accutate to within 40% up to 2 .0 m2/s 2 with u11.,.xcept i en -- a point cal ci 1 .t ed du ring -l hour wit i v ty low wi nd speed f a I I
we I I out o I t. li 4 OZ ,;i nd
211
ill~ i .....
.-N
Y
WV E_S x
Fig. Ul-1. Definition of Coordinates
-j1 s+5
I -S+~ZI S+ I 9 .
I + IIS 0
:C95f5 1 I% C2 95 -1 - C
1 -1, 0o
Fig. H1-2. Synthesized Three-Stage,
Butterworth, Low-Pass PrototypeValues of Components Chosen to Give
f 1.67x10-3 Hz or T =10 minC c
212
1.0 -0
S0.8 -2
U) -3
~0.6Io-
o;~ .4 I,
~~-10
- 20
f (Hz) 10-4 1-3 1 -T sec) 10 0 o
1 hr 10min 2 min$
Fi g. H- 3. Frequency Response of 3 -Stzige Low-Pass Buttur-worth Filte-r
213
anc
Li
iaC3
Fig H-4. Fitee Win Spe o a 4 , Hu 6015
I,214
LU
0.0-0 2;- 30-0 400 0 60.0
Fig. H--5. Filtered Wind Direction for Day 146, Hour 1650-1750
0'
200 0 io-O 50.0Q.TIME tMINj
Fig. H-6. Filtered Wind Speed for Day 145, Hour 1200-1300
216
0
I,
o
"SI,,r ~i"
4, * 4,
o ** 4,
-~ ~ ~: 4Li d' ' ~D i." .4,I", *1.4
* 1''' * ~~**~' *~* ,, . 4,,', 4.
1'' *0 r" .4.
* 4 . '.5, ,1*4,,'
Li I''.
/' ,, 4
o 4 44,, 4
z I' *'
4,''* 4'
40 ,*
'4
o ______ __________________________________
0~0 IOU aj.0 300 40.0 SO*0
TIME (MINi
Fig. H-7. Filtered Wind Direction for Day 145, Hour 1200-1300
217
I.
C!
('4(3
0. 100 2. 60 560 61 00 9.SIGM THT
Fi .H8 aCIatdE .( -1 eru bevd(o:C.oa al-
o218
0
0
0D
Io
I
0J
~Ii
o 0
U-
0.0 o0.0 20,0 30.0 40.0 50.0 60.0 70.0 000
SIGMA THETA
1"i .1 9. ('alculatted E'q. (tt-'3]) versw, O~h.s rvedl '7,-:
Average of Minute Values
219
-i g
crh
0
V)'
."
. Oo
~220
I- .
.'I
I0.0 0.0 20.0o 30.0 40.0o 50.0' 6o.0 70.0 80.0
SIGMA THETA
Fig. Ht-n. Calciilated Eq. (1-31) versus. Observed ',: MIinute Valuets
9
9-
00oCI) a
0 -,
0n 0
°!
oa
O.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 8.0SIG;MA THETR
Fig. Ht,11. Calculated Eq. (H-32) versus Observed :
Average of Minute Values
221
IK
IK
0 0,-..
I
El
UBUR2
Fig. tt-12. Calculated Fq. (H-33) versqus Observed i n:
Average of Minute Values
222
Table H-1. Average Wind Direction (t = WD)
Bsy Hour D () )* O(WD*)(dog) (dog) (deg) (deg)
145 1200 213.5 1.6 213.5 1.6
1300 208.9 1.2 208.9 1.2
1400 258.3 1.7 258.3 1.7
1500 268.6 1.6 268.6 1.6
1600 276.8 1.1 276.8 1.1
1700 297.2 0.9 297.2 0.9
1800 331.9 1.1 331.9 1.1
1900 348.6 0.9 348.6 0.9
2000 336.7 0.3 336.7 0.3
2100 337.6 0.2 337.6 0.2
2200 77.4 1.7 77.4 1.7
146 1650 250.7 1.0 250.7 1.0
1750 268.7 0.4 268.7 0.4
1850 302.2 0.3 302.2 0.3
147 0830 141.7 1.1 141.7 1.1
0930 168.9 2.4 168.9 2.4
1030 223.8 2.0 223.8 2.0
Table H-2. Vector Mean Wind Direction (4 = VMWD)
Day Hour VWD o(VMWD) VMWD* o,(VHWD*)(deg) (deg) (deg) (deg)
145 1200 212.4 1.5 212.4 1.5
1300 214.0 1.4 214.0 1.4
1400 262.4 1.7 262.4 1.7
1500 267.5 1.5 267.5 1.5
1600 275.3 1.1 275.3 1.1
1700 293.8 0.8 293.8 0.8
1800 330.6 1.2 330.6 1.2
1900 349.1 1.1 349.1 1.1
2000 336.3 0.3 336.3 0.3
2100 337.0 0.2 337.0 0.2
2200 71.1 2.2 71.1 2.2
146 1650 250.6 0.9 250.6 0.9
1750 266.8 0.6 266.8 0.6
1850 301.0 0.4 301.0 0.4
147 0830 140.9 1.1 140.9 1.1
0930 167.3 1.9 167.3 1.9
1030 229.5 1.6 229.5 1.6
223
Table H-3. Standard Deviation of the Wind
Direction ($ = a )
Day (our 00 0(0) e(o)(do&) (deg) (deg) (de)
145 1200 54.3 3.6 54.1 3.6
1300 50.3 1.8 50.3 1.8
1400 51.1 1.6 51.0 1.6
1500 25.3 2.0 25.2 2.0
1600 29.5 2.6 29.4 2.6
1700 30.9 1.2 30.9 1.2
1800 18.3 0.8 18.3 0.8
1900 11.4 1.4 11.2 1.4
2000 5.1 0.4 5.0 0.4
2100 6.4 0.2 6.4 0.2
2200 54.0 0.6 54.0 0.6
146 1650 19.5 1.2 19.4 1.2
1750 23.3 0.7 23.3 0.7
1850 15.0 0.4 14.9 0.4
147 0830 22.2 1.4 22.1 1.4
0930 69.7 2.3 69.6 2.3
1030 42.1 1.1 42.0 1.1
Table H-4. Average Wind Speed (0 = WS)
Day Hour VS o(wS) WS* (VS)(u/a) (w/a) (Wea) (a/a)
145 1200 2.67 0.09 2.67 0.09
1300 2.55 0.03 2.55 0.03
1400 3.52 0.06 3.52 0.06
1500 3.99 0.06 3.99 0.06
1600 3.57 0.04 3.57 0.04
1700 3.07 0.03 3.07 0.03
1800 3.54 0.04 3.54 0.04
1900 2.61 0.03 2.61 0.03
2000 2.59 0.01 2.59 0.01
2100 2.45 0.01 2.45 0.01
2200 1.41 0.01 1.41 0.01
146 1tI0 .4.4h t. OV 4.4b 0. 09
I 7 ,I1 4,A)7 II1 A.I 7 ]; 4,11
1850 3.37 0.03 3.37 0.03
147 0830 1.97 0.05 1.97 0.05
0930 1.82 0.03 1.82 0.03
1030 2.77 O.n4 2.?7 0.04
:, i'/
Table H-5. Vector Mean Wind Speed (P - VMWS)
Day Hour VmWS a(VMWS) VmWS* c (VWS*)(W/a) (ua) (m/8) (,-/.)
145 1200 2.09 0.07 2.09 0.07
1300 2.03 0.04 2.03 0.04
1400 2.95 0.05 2.95 0.05
1500 3.69 0.07 3.70 0.07
1600 3.23 0.05 3.23 0.05
1700 2.67 0.05 2.67 0.05
1800 3.36 0.05 3.36 0.05
1900 2.56 0.03 2.56 0.03
2000 2.58 0.01 2.58 0.01
2100 2.43 0.01 2.43 0.01
2200 0.80 0.01 0.80 0.01
146 1650 4.22 0.10 4.22 0.10
1750 3.76 0.04 3.76 0.04
1850 3.27 0.03 3.27 0.03
147 0830 1.85 0.04 1.85 0.04
0930 1.14 0.04 1.14 0.04
1030 2.17 0.06 2.17 0.06
Table H-6. Standard Deviation of the Table H-7. HourlyAverage No
Wind ,peed (1 = o ) 0 = NO)V
Day Hour J i(ov ) 0 * o(°*) Day Hour NO 0(NO) NO* '(NO*)V V V V
(m / 9, (r / ) (m / ) (r / ) (ppb) (ppb) (ppb ) (pp -)
145 1200 1.44 0.08 1.44 0.08145 1200 3.66 0.04 3.4 $1 .',4
1300 1.40 0.06 1.40 0.061300 1.53 r.08 3.53 j.P'.
1400 1.79 0.11 1.79 0.11
1500 1.20 0.05 1.20 0.05 1400 1. 0.05 3.68 0.0"
1500 1.62 0.08 3.62 0.081600 1.44 0.06 1.43 0.06
1600 4.13 0.07 4,13 0.071700 1.31 0.07 1.10 0.071700 3.40 0.06 3.40 0.06
1800 0.81 0.02 0.80 0.02 10 .0 00 .0 00
1800 3.52 0.07 3.52 0.0?1900 0.43 0.03 0.40 0.03
1900 3.11 0 .05 3.11 0.052000 0.37 0.02 0.37 0.02
2000 1. 5 o.03n 3.35 (.1)32100 0.3? 0.01 0.13 0.01 210'0 3.51 0.03 3.57 ((.03t2200 0.31 0.01 0.31 0.01
2200 3.6 0.04 3.68 r).04146 1650 1 16 0.OR 2.36 0.08 146 1650 I .4 0)(35 1 .48 3.05,
1750 ).84 0.04 0.84 0.041753 0.88 0.06f e,88 O001r
1850 0.5. 0.03 0.51 0.03
157 0830 0.72 0.04 0.71 0.04
147 0830 ((.75 ') 09 0.75 '.!0930 0.97 0.04 0.9; 0.040930 1.13 ). O8 1.11 )P
1030 1.20 0.04 1.20 0.041030 1.07 '(.07 1.O7 r.
223
Table H-8. Hourly Average NO (4= (NO)x x
Day Hour NOx o(NOx) NOx* (NOx*)
(ppb) (ppb) (ppb) (ppb)
145 1200 6.56 0.11 8.56 0.11
1300 8.81 0.07 8.81 0.07
1400 8.98 0.05 8.98 0.05
1500 9.46 0.10 9.46 0.10
1600 10.00 0.07 10.00 0.07
1700 8.54 0.06 8.54 0.06
1800 7.45 0.04 7.45 0.04
1900 8.42 0.06 8.42 0.06
2000 17.79 0.07 17.79 0.07
2100 28.84 0.03 28.84 0.03
2200 26.79 0.07 26.79 0.07
146 1650 5.71 0.13 5.71 0.13
1750 4.21 0.07 4.21 0.07
1850 6.98 0.10 6.98 0.10
147 0830 6.24 0.07 6.24 0.07
0930 6.48 0.06 6.48 0.06
1030 4.77 0.11 4.77 0.11
Table H-9. Standard Deviation of the VerticalWind Speed (D = ow )
Day Hour S W °(OW) (I W o(uw*)(sec) (m/s) (m/s) (m/s) (m/s)
144 1200 2.4 0.535 0.055 0.528 0.0551300 2.4 0.429 0.042 0.423 0.0411430 3.3 0.372 0.025 0.369 0.025
145 2305 3.0 0.096 0.010 0.095 0.010146 1130 3.0 0.722 0.052 0.717 0.052
Table H-10. Average Vertical Wind Speed (P = W)
Day Hour S W G(w) W*
(sec) (m/s) (m/s) (m/s) (m/s)
144 1200 2.4 0.166 0.064 0.166 0.0641300 2.4 0.184 0.058 0.184 0.0581430 3.3 0.209 0.044 0.209 0.044
145 2305 3.0 0.075 0.010 0.075 0.010146 1130 3.0 0.096 0.071 0.096 0.071
226
Table H-11. Calculation of p for Sampling
Times of 3 min and 60 min
Day Hour 0( (3 .llin) on (60 min) p
145 1200 26.903 5.3 •26b .'
1300 23. t90 50.309 0.",)
1400 24.422 51.092 0.?.6)
1500 1 S. 572 25.309 0. 162
1600 17.047 29.487 0. I2
1790 18.011 30.89 1
1800 9 .h,,I 18. 3014 o. 21,5
19001 5.491 1 l.3' 3 3.243
2-0(0 .8 7 5. C54 0. Ili
2100 2.121 6.407 ,1. 3(,Q
2200 8.343 54.046 0. 624
146 1650 11.063 19.454 O.188
1750 7.029 ?3.261 0. 359
1850 4.950 14,954 0.369
1"7 0830 16.623 22.184 0.01I
0Q i0 27.929 69. t.5 0.10)
1030 18.854 42.079 9. 2hS
Mean Value for p - 0.265
Standard Deviation - 0.116
Table H-12. Calzulation of p for Table H-13. Calculation of p fur
Sampling Times of Sampling Times of
10 min and 60 min 3 rain and 10 min
Day Hour 0 (0rin) oe (60 min) P Day l|our ml. ( In) , (I min)
145 120 40.723 54.266 0.160 145 121100 26.903 40.723 2. 1 4
1 300 35.202 50. 309 0.199 1300 .6)0 .5.12 .2
I4 00 39.580 51.092 0.142 1400 2.422 39 .580 -.
150(1 20.367 23.309 0.121 1500 15. 5) )..367 .1.222
1600 20.566 29.487 0.201 160(0 17 .1 7 20. ". 15 4
1700 26.337 30.891 0.089 1700 18.03I 26. 137 .3
1800 12.585 18.304 0.209 1800 9.061 12.585 0.27
1900 6.82S 11.359 0.284 10 5 .491 4.5 .180
2W0 1. 178 5.054 0.225 2000 2.837 1.378 ,145
2100 3.6.8 6.407 0.319 2100 2.121 3,618 0.444
2200 19.571 54.046 0.567 22)) S. 341 14 .)971 0.70R
14 I 16511 16.905 19.454 0.078 14, 6 11.0 1 'Of,' 1h,)" 0. ,2
1750 12. 1 7 23.261 0.353 1750 7.929 12,357 0. 169
1 MY) /.188 14 .954 0.394 18110 4. 1150 7. 3 8 . 1 13
1:3 3)1 1( I').768 22.184 0.064 147 O0 )0 1b.62 ] .1 7t,9 .144
0 1 ) 44.445 f)9.665 0.251 09 ()0 ,7 .';29 4.445 3. R..
1010 .'9 .40', 42.079 0.200 10i0 1 8,8-4 29 .40') <.36',
Mean Val,$r
for p - 0,.22) Mean Value for p - (1 ,2
standard )eviation - 0.17, ctandlrd Devilat Ion - I A4
227
Ibl. if-14. Comparison of the Actual Table H-i5. Compariso i ot thw IG0 LO the oU Value Calculated 10 to the II Value l', ,'uld.1tod
t rom Eq. (1-l1) using TS=3 min from Eq. (H-i usinl, 1."0 mill
oO des.)o (de0 g.)
Day Hour Actual Calculated Day Hour Actual Calculated
145 1200 54.3 55.3 145 1200 54.3 54.9
1300 50.3 51.8 1300 50.3 50.3
1400 51.1 52.6 1400 1.1 49.7
1500 25.3 25.4 1500 25.3 25.1
1600 29.5 29.5 1600 29.5 29.',
1700 30.9 31.1 1700 30.9 31.2
1800 18.3 18.5 1800 18.3 18.3
1900 11.4 11.5 1900 11.4 11.6
2000 5.1 5.1 2000 5.1 5.1
2100 6.4 6.4 2100 6,4 6,4
2200 54.0 54.2 2200 54.0 54,2
146 1650 19.5 19.5 146 1650 19.5 19,5
1750 23.3 23.3 1750 23,3 21.3
1850 15.0 15.0 1850 15.0 14.4
147 0830 22.2 22.2 147 0830 22.2 22.1
0930 69.7 77.9 0930 69.7 75.7
1030 42.1 42.2 1030 42.1 42,4
Table H-16. Mean Square Along Wind Eddy Velocity (0 -
Day Hour in o(- ) n * )(,2/@2) (,2,2 :) (a,,2/g2) (U /,, )
145 1200 0.959 0.125 0.898 0.109
1300 0.778 0.097 0.741 0.090
1400 1.440 0.286 1.367 0.276
1500 0.934 0.135 0.849 0.105
1600 0.729 0.102 0.667 0.081
1700 0.576 0,068 0.531 0.051
1800 0.401 0.046 0.377 0.040
1900 0.095 0.031 0.078 0.025
2000 0.026 0.007 0.021 0.004
2100 0.021 0.004 0.019 0.004
2200 0.017 0.003 0.015 0.003
146 1650 0.776 0.134 0.697 0.108
1750 0.356 0.066 0.305 0.043
1850 0.093 0.015 0.082 0.011
147 0830 0.342 0.065 0.302 0.051
0930 0.454 0,063 0.412 0.046
1030 0.700 0.067 0.640 0.063
228
r , _ _ ..... .. .. .. .. . ... ... .. . .. . ...... . ... .. _.
Table H-17. Mean Square Cross Wind Eddy Velocity (, - vr )
Day Hour v o() TT-' J ( v-T2i)
(2/2)
(u2/,2
(m I,,2 )
(22I31)
145 1200 0.764 0.263 0.823 0.256
1300 0.703 0.168 0.711 0.168
1400 1.757 0.208 1.830 0.195
1500 1.026 0.092 1.012 0.091
1600 0.899 0.102 0.853 0.090
1700 0.963 0.180 0.991 0.178
1800 0.308 0.025 0.298 0.023
1900 0.088 0.061 0.081 0.0b0
2000 0.016 0.005 0.013 0.004
2100 0.007 0.002 0.006 0.001
2200 0.032 0.007 0.035 0.007
146 1650 0.840 0.124 0.840 0.124
1750 0.310 0.042 0.286 0.035
1850 0.092 0.018 0.080 0.013
147 0830 0.256 0.036 0.242 0.033
0930 0.345 0.061 0.398 0.030
1030 0.604 0.071 0.602 0.071
Table 11-18. Mean Square Along Wind Eddy Velocity (4 u-)Computed Using a Fixed Mean
Day Hour 0 ,(u,7) t' 01( 2/ ) (In2/
' (M?/f4
2
) (w,/ G
145 1200 ,.122 0. 312 3.119 0. 12
1300 2.668 0.202 2.671 0.202
1400 4.6bi 0.387 4.626 0. A7
1500 1.79) 0.142 1.791 0.142
1600 2.262 0.207 2.256 0.207
1700 1.,318 0.136 1.816 0. 1h
1800 0.691 0.036 0.689 0.0 h
1900 0.15Q 0.037 0.158 0.037
200 0.117 0.012 0.13? 0.012
2100 0.109 0.007 0.1,)1 0.007
2200 0.626 0.022 0.6,5 0.022
146 1650 1.819 0.249 1.81)7 0.248
1750 0.910 0.058 0.919 0.058
1850 0.134 0.0.31 0.311 0.031
147 08)() 0.584 0.062 0.582 0.062
0930 2.000 0.149 1.997 0.14'
1030 2.223 0.196 2.217 0.196
229
Table 11-19. Mean Square Cross Wind Eddy Velocity ((P v7)Computed Using a Fixed Mean
Day Hour o(v'*)
(m2 /s2 ) (m2/s2 ) (m2 /s2 ) (n2/s 2)
145 1200 2.066 0.203 2.065 0.203
1300 2.003 0.223 1.999 0.223
1400 2.686 0.239 2.683 0.239
1500 1.197 0.159 1.965 0.159
1600 2.233 0.128 2.235 0.128
1700 2.393 0.326 2.392 0.326
1800 1.184 0.065 1.178 0.065
1900 0.275 0.057 0.273 0.057
2000 0.045 0.008 0.045 0.008
2i00 0.073 0.005 0.073 0.005
2200 1.225 0.048 1.224 0.048
146 1650 2.211 0.139 2.210 0.139
1750 2.315 0.149 2.31A 0.149
1850 0.633 0.041 0.633 0.041
147 0830 0.383 0.039 0.384 0.039
0930 1.428 0.080 1.430 0.080
1030 2.622 0.180 2.625 0.180
230
API't',NI)IX I
1110';IMINARY "'1ME ;;kES ANAI, YSI[,,
1. INTRODUCTION
It is often possible to describe complex physical phenomena and mech-anisms better in the frequency domain than in the time domain. The timedependent nature of air pollution and meteorological data and the likelihood
of serial correlation in it makes obvious the need for analyzing such data interms of its frequency composition.
Many statistical analysis methods that operate on the data in the
time domain make no use of the time dependency therein. For example, in theestimation of the frequency or cumulative frequency distributions, the dataare sorted into order, after which the time dependency information is nolonger available. An important property of time series analysis, on the other
hand, is that much of the time dependency in the data is preserved.
The use of time series in the anilysis of air pollution monitoringhas been suggested or carr ied out by several authors including: harlow
and Singpurwalla, I - I Saltzman - 2 Marcus,1 - 3 and Tiao and Hamming, 1- 4 to
mention only a few.
Data reduction, harmonic analysis, and filtering of Williams air qualitidata was performed previously on raw data from the HP9825. 1 - 8 Spectral
analysis was instrumental in identifying certain instrumentation problems
that were exhibited by a sawtooth pattern in the raw data. These problems,
which were diagnosed early in the experimental program, were subsequentlycorrected.
An extensive treatment of spectral analysis of time series has beengiven by Koopmans; I - 5 time series data analysis and theory are discussed
extensively by Brillinger;I - 6 and a bibliography of time series analysis
papers has been given by Wold.I-7
The principal goal of spectrum analysis is to decompose the power ofthe given time series into its harmonic components, that is, to estimate thepower spectrum from the available data. The estimated spectrum can then beused to gain information about the mechanism that generated the data.The stat istical theory of spectral anlaysis has been based on the hypo-theses that the underlying process is stationary and Gaussian, that the
process mean is zero, and that the spectrum is continuous. Some of thepreprocessing operations performed on the data before spectrum analysis isdone (for example, removal of the DC component of the series) are intended tobring the data into reasonable conformity with these hypotheses.
Two models of spectral analysis have been used hure, univariatt andbivariate. Univariate spectral analysis involves the computation of the powurspectrum or, as it is also called, the spectral density. The power spectrumis the Fourier transform of the autocorrelation function. The Fourier trans-
form of a sequence of autocorrelat ions contains precisely the same infor-
mation found in the original autocorrelations therr:-clves. The intensity otthe spectrum at any given frequency provides an evst imatt of the iowc.r or
2-11
energy at that frequency. Tb,: accumulated spectrum is essentially the firstintegral of the spectral dtasity and can be used to indicate the range offrequencies that contain a substantial proportion of the total power of the
time series.
Similarities between tile spectra of two series, such as peaks at similarfrequencies, may raise the possibility that tile series are related. 'oinvestigate such possibilities, the cross spectrum of two series miv becomputed. This computation is an extension of the definition of the spectrumand is usually estimated by smoothing the cross periodogram; this is performedin the bivariate spectral analysis of two series. Bivariate spectral analysisallows for the decomposition of two series into various spectral parameters,equivalent to descriptive statistics. These include: the estimated spectraldensity of each series; the phase function that indicates the relative shiftof harmonic components of the two time series at the same frequency; thetransfer function that indicates the linear relationship between the twoseries in the same way that regression coefficients relate tile linear rela-tionship between two variables; and the coherence function that provides anestimate of the frequency-dependent correlation between two series. It isinteresting to note that if the phase angle is a rapidly varying function offrequency, the estimated coherence can be biased downward to an extent that astrong coherence will be masked.
In summary, spectral analysis provides a means of analyzing serialcorrelations in the data, as well as a method of examining frequency dependentcorrelations. It is indispensable when searching for periodicity and grossvariations of power with frequency in the data.
Spectral analyses were performed on a continuous period of one-hour
averages, covering approximately one month, starting August 1, 1976. Missingdata and outliers identified by an interface program to the AQAM data tapewere assigned values via a simple linear interpolation of adjacent values.Univariate and bivariate spectral analyses were performed using methodsoutlined by Koopmans. I - 5 Before spectrum analysis was applied to the data,the mean (DC component) was removed from the series to bring the data intoconformity with stationarity.
Figure 1-I* shows the time series for observed CO at station 2. Whilethis series appears to exhibit a pronounced 24-hour period, one finds uponclose examination that the daily period has slight relative shifts. Where tileperiodicity is in fact a combination of several other periodic phenomena,closer investigation is required. It is felt that the roughly 24-hour periodsare actually beats, and unless the series is filtered, these will not exhibitthemselves in the spectrum as at first anticipated.
Figure 1-2 shows the spectrum for this time series; the peaks at 16-17hours, 8 hours, and 6 hours, are of interest. The largest peak at approxi-mately 16 hours is an artifact of the sunrise-sunset period. (Note thatthis data was drawn from summer months.) Periodicity reflected by peaks athighor frequencies appears to be an artifact of source emissions activityduring the (day. It is felt that the higher frequencies at 8 and 6 hours may
* igures appear consecut ively at the end of The app,.ndix.
232
t) r.lt 4 -l t, air raft ac.t iv it y . 'Ile x-aX is of tilt sip.ot r I has bt .n r.p lot -
td in terms of period but was original y coniputed a a ftquenfcy in trisoi c ycles in units of 1/(2*M*DTr), where DTI hoar, and M is th o tru'nat ii
parameter that has a value of LO0. ihe t runcat ion pa rameter, also referredto as the number of lags, controls the smoothing (it the periodoigram andhas been chosen approximately equal to 10,', of the dac-3 le::gth.
Figure 1-3 shows the accumulated spectrum tor the CO series, It isinteresting to note that the higher frequencies, above the frequencies relatedto diurnal fluctuation, appear to contribute significantly to the total power
of the observed CO series.
Figure 1-4 shows the time series for observed Nox at station 2 duringthis sane period. That the structure of this series is very similar to the CO
series is confirned by the locations of the peaks in the spectral density shownin Fig. 1-5. Figure 1-6 shows the coherence between the CO and NO. series.
It is interesting to note that while the peak coherence is associated with thediurnal cycle, the coherence remains strong at the higher frequencies, sug-gesting that these two pollutants may be associated with the same set of
sources.
Figure 1-7 shows the time series of aircraft CO emissions on ,roundlevel lines. This emission rate is computed in AQAM on the basis of the
hour-by-hour records of aircraft activity. The spectrum of this series showsthe strong diurnally induced peak at 16 hours along with small enhancements at
8 and 6 hours where significant peaks were observed in the observed CO
concentration spectrum (Fig. 1-2). The coherence between the spectra cfobserved CO concentrations and aircraft CO emission rate also shows peaks
corresponding to these periodicities; however, the wildly oscillating magni-tude of the coherence and accompanying phase (not shown) are suggestive ofcomputational instabilities possibly being driven by the strong low-frequenv
peaking in both spectra.
2. CONCLUSIONS AND RECOMMENDATIONS
While the background concentrations due to strong natural cycles ar,significant, harmonics at higher frequnenci es thought to be associated withemissions source activity have been identified. The presence of large peaksin the spectrum, have the effect of badly biasing estimates in the low endof the spectrum due to the inevitable side lobe distortion of the spectralestimates (filter leakage). A quasi differeuce filter can be used to balancethe spectrum of a time series with a large peak at low frequincif-s. 'The.
ba laine process is cal led prewhitenin, by Blackman and Tukey (1959).1-9
A prewhitening filter can be used to help smooth the larger natural peaks,improving the resolut ion of the higher frequency spectra.
Spectral analysis can be used to study the effect of filters used on
the air quality time series and also in the design of such filters. Filtersare used for purposefil modification of a discrete time series. It isimportant to understand the operation of the filter, and to know how to select
the "parameters" of the filter to achieve specific objectives of data
modification.
23
_41111111
The spectral analyses presented here illustrate the benefits of analyzingtime-dependent air quality data in the frequency domain. Estimation ofcorrelation vis a vis the coherence function between two series providesimportant insights into complex physical systems generating the data, andthe ability of the AQAM to reproduce features prevalent in the observatins.
234
Kl E' FE R K ;N(CE
1-1. Barlow, R.E., arid N.D. Singpurw;t la, Al ,tin& Time a-nd Maxc' 'Dependent Obeervations, Proc. of the Symposium on Statistical Aspectso0 Air Quality Data, EPA-650/4-74-038 (Cct. 1914).
1-2. Saltzman, B.E., Four-Lzer Analzlygis :,IT AbVr ;iLZ i, ?. ;,t(, 1'roc. of
the Symposium on Statistical Aspects of Air Quality Data, EPA-605/
74-038 (Oct. 1974).
1-3. Marcus, A.H., A S'tochastic Model for Fstimwtfin 01o74r anrt 4jo, e 'E,
Means of Air ,Qualit? Data, Proc. of Lhc SmposiuT. on StatisticalAspects of Air Quality Data, EPA-650/4-i4-038 (Oit. 1974).
1-4. Tiao, G.C., and W.J. Hamming, A StatisticaZ Anatlycis of" the Los AnyjelesAmbient Carbon Monoxide Data 1j9;.-1 72, J. of thf Air Pollution ControlAssociation, 2b:11 (Nov. 1975).
1-5. Koopmans, L.H., 7h(,;pectoal AnalysiG oj' Time Series, Academic Press,
New York (1974).
1-6. Brillinger, D.R., Time Series Data Anaysis and 7'feoiy, Holt, Rinehartand Winston, Inc. (1975).
1-7. Wold, H.O.A., Bibliographj of' Time Series and Stochastic troceseee,Oliver and Boyd, London (1965).
1-8. Dunphy, E.P., Diagnostic Investigation of Three Air Pollution Data Tapesfrom Williams AFB, final report for period I June 1976 to 30 July 197b,Letter report for C.E.R.F. Technical Directive 4.04/01.
1-9. Blackman, R.B., and J.W. Tukey, The Veasurement o:' :oetr Spcotra .(the Viewpoint of' Commezications Engineering, Dover, N.Y., reprintedfrom Bell System Tech. J 37 (1959).
235
0
CL0 ID
C)I
TIME (WEEKS)
Fig. 1-1. Time Series for Observed CO Concentrationsat Station 2 during August 1976
U')
C.)
C")
z
-)
Ic U
Li
C))
w 4 < T M Tr fr Tr
N c," (D nC)\w - . () U/) (Io U,) ,)
PERIODICITY
Fig. 1-2. Frequency Spectrum for tile Time Series of ObservedCO Concentrations at Station 2
237
Md
C)
CD60
0
~)
0-
0 A >
U r X izc
PERIODICITY
Fig. 1-3. Integrated Spectral Density for observed CO
Concentrations at Station 2
238
C-).
DI)
C)
C,r)
- F-
2 34D
TIE(WESF0.14 ieSre o bevdN~Cnttain
at Stto 2DrngA)st 17
L2 3
t.
F-
CD
C-V)w L ) U)Uw I n rcrI
24
~T-4
0)
CID
CD
(I - ) (n)lIJ 4 II X
PERIODICITY
Fig. 1-6. Coherence vs Periodicity of the Station 2 Observe'd
CO and NOX Time Series
241
CD
CD
0-0
z
Q0
CO C
ftnc
0
tnCD
Cc
C-D $
OJ_
0-
5)LILW .1 _LLT_ 1I~ t _1 1 1--
2 3 4 5TIME (WEEKS)
Fig. 1-7. Time Series for Aircraft CO Emission Rates on All
Ground-Level Lines Based on Actual Hourly Aircraft
Operations for August 1976
242
C:
z
C-~)LiJ
0
0-
W -
id)- < - c Cl) Cl)X
.N N 0C (D q ~ N
PERIODICITY
Fig. 1-8. Frequency Spectrum for the Time Series of AircraftCO Emission Rates
243
__ _ _ _ __ _ _ _ __ _ _ _ _
C-i
LOl
C
0$
W x , Cr. CO CO cO
N itM
PERIODICITY
Fig. 1-9. Coherence vs Periodicity of the Observed CO at Station 2and the Aircraft CO Emission Rate Time Series
244
INITIAL DISTRIBUTION
HQ AFSC/DLW IHQ USAF/LEEV 1HQ TAC/DEEV IHQ AFESC/DEV 1[IQ AFESC/TST 1HQ AFESC/RDVA 4
AUL/LSE 71-249 1AFArl,/DLODL IDTIC/DDA 2OSAF/MIQ 1OSAF /01 1AFI f/Library 1EPA/ORD IFAA (AEE-300) 1Argonne National Laboratory 1NAPC/Code PE71: AFK 1
245
(The reverse of this page is blank<)