VoL 4, No. Z, May 1979

UPDATE ON WW-LEVEL WIND SHEAR FORECASTING

Barry A. Richwien

Center Weather Service UnitBoston Air ~fic Control Center

Nashua, New Hampshire

Abstract

The National Weather Service (NWS) and Federal Aviation Administration (FAA) jointly sponsored a 6­month test of Northwest Orient Airlines' technique of forecasting low-level wind shear (LLWS) due toconventional fronts. The technique provided that any front moving> 30 kt and/or whose temperaturedifference was ~ lODF should produce LLWS sig~ificant to aviation operations. This technique wasverified by FAA reconnaissance, a Doppler acoustic-radar system, NWS radiosonde winds, and pilotreports.

A major conclusion was that successful and effective diagnosing and forecasting of mainly mesoscalefeatures that produce significant LLWS will require unprecedented dedication of manpower andresources.

1. INTRODUCnON

For this discussion we shall define low-level windshear (LLWS) to be the vertical shear of thehorizontal wind which persists over one location(such as an airport) for 10's of minutes (vs.turbulence-induced shear which lasts only 10's ofseconds). Low-level shall be meant to meanbelow about 600 m. The hazard of this steady­state type of shear to aviation is well docu­mented. Several major airliner crashes of thepast 5 years have been attributed, in large part,to this phenomenon. Most famous are the crashesof B727'5 at both Denver's Stapleton and NewYork's JFK airport in 1975. The LLWS contrib­uting to those accidents was due to thunderstorms- probably the thunderstorm's gust front.

Less spectactiJar and more subtle is LLWS causedby warm and cold fronts. Only one clear-cutexample of a crash due to the LLWS of a frontcould be found. It occurred at Boston's LoganInternational Airport when a DC-IO struck theapproach lights due to a LLW&-induced rate ofdescent. Northwest Orient Airlines has for manyyears been forecasting frontal LLWS (see Sowa,1974). Northwest's technique provides th~t anyfront whose temperature difference is 6 C orgreater, and/or is moving at 30 knots or greater,should produce LLWS significant to aviation oper­ations. Increasing concern ahout the hfl7.ardprompted the National Weather Service (NWS) andFederal Aviation Administration (FAA) to jointlysponsor a test of Northwest's technique. The testwas to determine whether or not LLWS wind shear

due to warm and cold fronts could be forecast upto 3 hours in advance with sufficient accuracy tobe of use to landing and departing pilots.

2. NW~FAA LLWS TEST

a. Test Particulars

The test was conducted for the six-month periodfrom November 1976 through March 1977 between0600 and 2200 hours each day of the week.Forecasts of LLWS (called LLWS Advisories) wereprepared for 7 east coast airports, namely: DullesInternational, Washington National, LaGuardia,JFK, Newark, Philadelphia and Atlantic City(home of the FAA National Aviation FlightExperiment Center, NAFEC). Advisories wereprepared by NWS Forecast Offices in Washington,D.C., Philadelphia and New York City. They werethen forwarded via direct telephone lines to theFAA Systems Command Center in downtownWashington, D.C. There, 1 of 2 meteorologistsspecially assigned for the test forwarded theAdvisories to appropriate FAA towers and AirRoute Traffic ("-ontrol Centers via direct tele­phone lines. FAA tower personnel then recordedthe information on Automatic Terminal Informa­tion Service (ATIS) or, if that was not available,the Advisory was verbally relayed to landing anddeparting pilots.

21

These observing problems aside, the next veri­fication question was, given a high-resolutionvertiCal wind profile, is the shear shown ''signifi­cant" to aviation operations? There are varyingopinions of what constitutes Significant shear.The International Civil Aviation Organization atits 5th Air Navigation Conference suggested thefollOWing categories be used:

universally accepted way of detecting or describ­ing LLWS from the cockpit. It seemed to us thatthe most logical way of doing this was for thepilot to observe the change in Indicated Airspeed(1\ lAS) of the aircraft. Accordingly we developedand sent out to FAA towers pilot reporting formswhich encouraged FAA personnel to ask for"'lAS's from pilots reporting LLWS.

National Weather Digest

The initial format of the Advisories was vervsimilar to that of Northwest Airlines and was asfollows:

LOW-LEVEL (abrupt/gradual)

(cold/warm) FRONTAL WIND SHEAR (WITH(moderate/severe) TURBULENCE (cross out ifnot applicable)) IS EXPECTED AT (airport)BETWEEN Z AND Z. WIND BELOWFRONTAL SURFACE FROM DEGREES AT

KNOTS AND WIND ABOVEFRONT FRO!VIDEGREES AT KNOTS.

"Abrupt" shear was defined to be that occurringthrough a layer 30 m or less in vertical thickness;"gradual", between 30 and 100 m. Midwaythrough the test this categorization of LLWS waseliminated because we had no real-time way ofdetermining the layer thickness. Likewise, theturbulence clause was withdrawn because turbu­lence was forecast by NWS AIRMETS and SIG­METS. Forecasts had a lead-time of 1-3 hoursand were valid for 3 hours or less unless cancelledsooner.

I l WS I n tens; ty

l i gnt'·lode rd teStrongSevere

~J ues (knots/3D m)

0-45-89-12> 12

Verification of Advisories was by 4 methods: l) adual (acoustic-radar) Doppler' sounder locatednear Dulles Airport (see Hardesty, et. al. (1977)for a description), 2) a meteorologically instru­mented FAA-NAFEC Aero Commander which wasdispatched by meteorologists at FAA SystemsCommand Center to expected LLWS locations, 3)pilot reports and 4) NWS radiosonde winds.

b. Verification Difficulties

Although developing a verification scheme formost any meteorological para meter is a chal­lenge, the phenomena to be verified are usuallyeasily measured and criteria are pre-set.. Thatwas not the case with LLWS. Even though wewere fortunate enough to have the most sophisti­cated of observing equipment available, the dual­Doppler system was only at one of the testairports, Dulles, and it used a 6-minute averagingsystem that masked some of the sudden windshifts. Also, the FAA Aero Commander was ableto fulfill only 25% of our reconnaissance requestsdue mainly to equipment malfunctions.

Problems with the other 2 observing methodswere as follows. NWS radiosonde balloons >Irereleased only every 12 hours at set times. There­fore, the probability of one ascending through afrontal surface - especially a fast-moving coldfront - is extremely small. Also, it is debatablewhether the vertical interVal usd for averagingwinds is small enough to measure significantLLWS. The remaining method of observing LLWS,using pilot reports, suffered from a high degree ofsubjectivity. This is because there is as yet no

22

It has been pointed out by various authorities thatone must also consider the integrated value ofLLWS over a layer greater than 30 m. Forexample, Sowa (personal communication) hasstated that a value of 5 knots/30 m over a 90 mlayer is sufficient to cause a significant MAS of15 knots in an aircraft traversing through thelayer. For this study we adopted flexible, consen­sus values of 3 to 5 knots/30 m as significantvalues.

3. RESULTS

a. Summary of Advisory-Events

During the 6-month test there were 22 days onwhich Advisories were issued. Twelve were forwarm or stationary fronts and 8 for cold fronts.There were 2 days on which Advisories wereissued for LLWS due to other causes. Windprofiles from the Dulles Doppler system (such asshown in Figure l) were available after-the-factfor 10 of the 12 events there. Aero Commanderprintouts (such as shown in Figure 2) were ob­tained for about 25% of Advisory-Events at vari­ous test airports. Because of the previouslyoutlined shortcomings of all other LLWS observingmethods, we decided to use pilot reports as theprimary method of verification. These wereavailable for every event. And, after all, thebottom line of any such verification attemptshould be the effect the vertical wind profile hason an aircraft traversing it.

VoL 4, No. Z, May 1979

.....? ....-. -.-.' ..- ..... ?..... ••• ' .. YO ..... ?..... • •• ' ..

'"it.. 4._'. .~1<.1<.I •••

I

I'

~ '''0.''.' ., cO',..00·... 'OC'"

J" ...,.....,'" '''",,...,,\.. ""0<'"

""_ ",,,,.,0<1 I...,,,.,, ,...,

(

)

t•\!

1... , o'" ,., l ,,,

figure 2. Example of vet'tica! profiles obtained byFAA-NAFEC Aero Commander. (Note slight low­level jet.)

,.

..-'"'.,

...'"

".

".

"'

".

"'

-

M' ' .. ",..........- ,.110 ,. 110'"

nO' ".,;:. .. .;:""".u

"'-'-'­"lO<'>' '10 .. ,,, ".

.."",

lie "''''''' .,_ '00"'" ,too ..._,

" ""

•

... "m

,-- " ,,,,., "-,. • '"~ • ""''',. .- Ill"''',.

"1"'"---- .,

-."-,.

"'jI;~~

- 'I" ~,'"'

{"~~-

~,

m

'"

.!-.--.,..""c.,...--,".-....,,~.-!.-....,_:--c,_=--~=!~c_=-...... _0 ,"'•• > ...... D' ........ 'OU>

Figure 1. Example of wind profile from DullesDoppler system after-the-fact. (Note low-leVeljet structure of profile.)

~ "" -"" .'"- "" ~

". ~

I- m. .",,~ .",

,. -,~ .",

'" ." -•• .",."0 .m

"e- .- ."

,.'. ' .... '...,,"', "',' (11' l' 00"1 ,!lr.

Table 2. Breakdown of the causes of LLWS pilotreports on Advisory-Event days.

Having all this information, the next step was tocort'elate the manifestation of frontal-LLWS onairct'aft (i.e. Ii lAS's) with out' forecast param­eters; fI T and speed of fronts. Of 79 pilot reportsjud~ed to have been due to fronts, 44 included6IAS1s. Tl1ese 44 are plotted vs. speed of frontsin Fig-ure 3 and vs. temperature gradient (flT/lOOn.mi.) in Fi~re 4.

Fig-ure 3 shows that only a few pilot reports werereceived with fronts moving 30 knots or greaterso no correlation was attemoted. The lat'genumber of reports near ft'onts moving at 10 knotsor less t'eflects a gt'eater number of LLWS pilotreports with warm frontal events. Since thescatter of those reoorts was so gt'eat and sincethe test criteria did nvt pt'edict significant LLWSwith fronts moving at less than 30 knots, arCfrression line was not computed for that region.

")f .' '\.' "1'-:

.. 0' "Ill Ii '10

" ""

"..........,.." , ..."f- ,.-: - :,;;;,_-

l_, .." n·

Pilot reports from each Advisory-Event werecollected in 3 ways: 1) by verbal relay from FAAtowers to FAA-Systems Command Center meteor­ologists, 2) mail forwarding of pilot reportingforms and 3) by normal service A teletype pilotreport messages.

Pilot reports were then listed in chronologicalorder for each event as shown in Table 1, Then,by using all available vertical wind profiles, NWSNational Meteorological Center (NlYIC) and ourlocal sut'face weather analyses and FAA runwaylogs, each of approximAtely 400 pilot reports wascarefully scrutinized and a judgment was madewhethet' ot' not the LLWS repot't was due to ttlefront for which the Advisory was issued. Thatdecision for each report is shown on the right­hand side of Table 1 followed by a "ProbableCause" if one was obvious. Table 2 shows ttlat anastounding numbet' of pilot reports on Advisory­Event days were attributed to causes other thanfronts.

Table I. ExAmnle of form used to 'icrutini7.e AnrJdetermine cause of f,LWS nilot reoort<; onAdvisot'y-Event days.

b. Interpretation of Pilot Reports

23

Figure 4. Plot of LL1\'S pilot reports tI AS's vs.temperature g-radients AS determined from localsurface analyses.

I...,. '"'"' I, LOLO 'RONT

• IUfEC

.'ReRoFT

~ 'llf-=~~--+--+-:--+--+--+--f---j

= !~ ~

<i. 10 f-+----+--+--+--+--+--f--'1~

~e ·\0 f-+----+--.j...-+--+--+--f---j~,

traversing- the shear layer - if any - at rightanf(les. To minimize this Dossible reason for thelac'k of LLWS pilot reports, we did two things.The first was to establish a "best-guess" differ­ence vector (A V) for each Advisory. "Best-guess"upper and lower winds were obtained by usingl"'iter, more accurate, observed winds. For this, ing-enerlll, Aero Com mander winds had the highestpriority followed bv the Dulles Doppler winds,then NWS radiosonde winds. A "best-guess" !'. VWIlS obtained bv vectorially subtracting the two"hest-guesstt winds. These values were not rou­tinelv divided bv a !'.z to derive the shear vector(!::, vi!'. z) hecause of the frequent uncertainty of aproper t> z. The t> z, though, shold always bethought of as ranging between 200 and 600 m.

The next step was to project these "best-guess"difference vectors onto active runways to givethe longitudinal (head- and tailwind) components.The sip;n convention chosen was such that a posi­tive projected t> V should have produced an lASp'"ain in an aircraft traversing the layer. Figure 5shows the results. The fact that the LLWS pilotreports are scattered throughout all 4 quadrantsrather than the 2 "proper" ones led us to thefollowin<r possible conclusions in our order ofpriority:

17.50 20.0012.510.0, .55.0

TEMPERATURE GRADIENT (OC/l00 n.m.)

2.5

• W"'AM'RONI .+ ""'''H~ COLD'RONI

. I ,NAHc AIRCRAFT

'" . .~--

e-;---- ---- ...-

, , , , ,

..

Figure 3. Plot of LLWS OIlot report AlAS's vs.speed of fronts. Pilot reports are those judp'"ed tohave been due to Advisorv-fronts af.ter carefulscrutinization. -

-------~

National Weather Digest

I- --.-- - . ~...-_ ..,".- -- n..

l• W!lRM FRO"T NAHC A1IRCRAFT

IX COLD FRO'" I..

l I. I"'AHC AIRCRAFT I

II>

" I~

Ii ' ,

, 5 "i ",~

.. .I,

§I ",! ~! •I • • .. .. " .. " "I SPEEO OF FRon (KNOTS)

.

To the eye, the scatter of points in Fig;ure 4 looksless random than in Fi!<Ure 3. Thus, a regressionline was computed and is shown. The correlationcoefficient for that line is only 0.26; statisticallysignificant at the 1% level. This small correla­tion between temperature gradient and MAS sug­gested to us that the problem was more complexthan we had hoped.

t I'f.-o'i---+--'L~-+--+-~+--f---j"

"f-+---+--+--+--+--+--f---j

"L-.J.,,__.J.,,__l.,,__'-__.,L.__,"'-__"L._-..1

RUNWAV 'ROJUlION OF '·IEST GUESS" OlFfERENCE VHTOR 0""':; ( >IOB)

One factor that the data in Figures 3 and 4 in noway took account of was the absence of pilotreports. During many events Advisories were ineffect for hours while aircraft were landing Rnddeparting Rt the rate of 1 every 5 minutes andabsolutely no pilot reports were received - evenwhile being: solicited hy FAA tower oersonnel. Itwas thou<rht that this could be due to aircraft

Figure 5. Plot of LLWS pilot reDort MAS's vs.lon~itudinlll components of "best-guess" differ­ence vectors, ",V. Pilot reports used lire thosejudged to have been (jue to Advisorv-fronts aftercareful scrutini7.lltion;

a. There were small-scale features (such as low­level jets 0r local inversions) which were em-

24

bedded or superimposed within the larg-er-scalefronts to which these pilot reports were originallyattributed. These features, then, were causingMAS's different from the conceptual winds of a

frontal model suggesting that significant LLWS isdue more to speed change rather than directionalchange of the wind.

b. Our assumption of straight-in approaches byaircraft below 600 m is not valid often enough forthis kind of correlation attempt.

c. There is more pilot (and possibly air trafficcontroller) misunderstanding of how LLWS affectsIAS and Ground Speed changes than we hadassumed.

4. CONCLUSIONS

a. IJmitatioTlS of Test

Before discussing conclusions and gwwg" recom­mendations based on this test, it is important tonote its limitations. First, all test airports werelocated on the coastal plain of the northeasternUnited States where the terrain is reilltively Oat.ThUS, terrain-induced LLWS was not investigated.Secondly, the test was conducted during- anunusual winter. During the first half of theseason the jetstream was unusually far southcausing many weak or moderate cold frontalpassages with few stroRlJ: ones and a small numberof warm front events. The latter half of thewinter was more normal in the eastern U.S. Wehave tried our best to take these limitations intoaccount. One of the wavs this was done was tomake a mental note or" the weather situationwhenever LLWS pilot ref'Orts appeared on thenational Service A teletype printer at the FAASystems Command Center. This ~ave us a roughre~onal and synoptic climatology of LLWS pilotreports.

b. Fronts

We have found it useful to divide the broad termof "front" into synoptic- and meso-scale categor­ies. Synoptic fronts are the everYday, TV weathermap types that extend vertically through much ofthe troposphere (i.e. have vertical height sCllleson the order of 10 km) and persist for a number ofdays. In contrast, meso-fronts hllve verticalscales on the order of I km and have life-timesranging from a few hours to, perhaps, a dRv.Examples of meso-fronts are "coastRI" fronts(pseUdo-warm or stationary fronts which fromroughly parallel to portions of the U.S. ellst COllstas described by Bosart et. al. (1972)), sea-brep7.efronts and thunderstorm gust fronts.

One key goal of our research was to establish atemperature gradient criterion rother than si·mplyusing some "temoerature difference" '11ethod of

VoL 4, No. Z, May 1979determining a front's LLWS potential. This wasdone in the following way. Consensus of whatvalue constitutes a significant LLWS-inducedMAS ranges from 15 to 20 knots (Joss or gain).

The regression line in Figure 5 crosses that lIlASat about a temperature gradient of 10oC/l00n.mi. Very few synoptic-fronts have gradientsthat large whereas most meso-fronts do. This isone indication that meso-fronts are the primarycause of significant LLWS. Another clue that thisis so is the observed rarity of the phenomenon.Since in meteorology time and space scales arewell correlated, it would seem that the short­lived nature of small-scale meso-fronts couldexplain why so few aircraft have experienced or"sampled" the effects of excessive LLWS. Ourconclusion, then, is that the strongest of synoptic­scale and most meso-scale fronts (i.e6 fronts withtemperature gradients exceeding 10 CI100 n.mi.)have the potential to oroduce signifi~ant LLWS.

c. LLWS Due to Other Phenomena

As shown in Table 2, the high percentage of pilotreports "weeded out" as not being due to frontsindicates that any advisory program should in­clude that LLWS due to phenomena such as inver­sions, low-level jets (LLJs), and frictional drag.We are concerned that there may be too muchemphasis ,>eing placed on LLWS caused by fronts ­especially synoptic-scale ones. If an advisorypro!'.!'am for only these fronts were to continuefor a period of time, a substantial number ofLLWS events woukl not be forecasted. For thisrellson the other 3 known causes of LLWS arediscussed here.

I\t this point the distinction between fronts, LLJs,and inversions is not clear in our minds. Thepresence of an inversion is the common denomi­nator for all causes of LLWS except frictionaldrill!. This is not surorising for it is well-knownthat static stability inhibits vertical mixing ofmomentum. Blackadar et. al. (1958) developed amethod of forecastinf( LLWS due to LLJs which isbased on nocturnal inversion strength. Our studyshowed at least two cases of LLJs associated withfronts. One of those is shown in Figure 1 and,incidentally, coincided with a pressure jumpsensor event at Dulles Airport as described byBedard et. Ill. (1977). We think that many of the"incorrect" signs of /lIAS's shown in Figure 5could be explained bv speed rather than direc­tional shear associated with frontal LLJs. It is,incidentallY, for this reason that we recommendnot including expected winds in frontal LLWSadvisories. A connection between LLJs and frontshas been noted by other researchers. Kreitzberg(I967) observed an 80-knot LLJ within 600 m ofthe ground neilr an occluded front in southern"ew Englanc1. f'e found that temperaturegradient alone could not account for the vertical

25

NatiOllaJ Weather Digest

shear (using the thermal wind relation) and con­cluded that much of the shear had to be Il.p.;eo­strophic. Brownin~ et. al. (1973) describes casestudies of LLJs ahead of certain 'Oid-Illtitude coldfronts in the British lsles. Thev found that tlleLLJs had maximum speeds of S'0-50 knots, wereembednen in a convective boundll.ry layer, ex­tended for WOOs of km and seemed to have littlediurnal or isallobaric relation. It seems that therelation between fronts and LLJs should be morefully investigated.

Table 2 shows that fUlly lI5 of all LLWS pilotreports on Advisory-Event days dl,rin" our Testwere due to frictional drag. Twice that fraction(40%) occurred with cold front events probablvbecause of the usual strong flow behind them.There were many more unsolicited pilot reportson non-Advisorv davs that I'll' attributed to tllefrictional drag- on sirong surface flow. Althougllsome LLWS is routine to pilots due to lag- orpower-law wind profiles, at some point thRt shearmust hecome excessive as shown by these reports.Our experience was that when sustained surfacewinds reached about 20 knots, pilots beg-an rc­porting [AS losses of 10-20 knots. [n many casesDoppler and Aero Commander wind profiles con­firmed LLWS as suggested by those oilot records.

5. GENERAL CONCLUSIONS

Based on the results of this test, I'll' feel com­pelled to s!lY that hefore any LLWS advisorvprop';ram is attempted, a soun<J <Jetermination ofits probable rate of success should first be made.That is, how much "cryinfT. wolf" too often will tht>credibilitv of LLWS 'ldvisories bear? \1any limesduring our test, Advisories were out for hours andno pilot reJlorts were received. Hopefullv, therefine",ents made on the test criteria will 'elimi­nate mapy of the f"lse "I"rms. 11'.It, toe ouest ionremains, how mueh is enouj!h?

Our final general conclusion is this. The small­scale nature of the LLWS flroble", has i",oortaotimplications to anv forec".t endeavors. First­with the exce;>tion of friction-induc€'d LIWS ­significant shear will only b€' d('tecled bv me""­analysis of fresh, clos€'ly sJl!lcerl <l"tq. It was ourexperience durin~ the test tha t N'''' C surf"cemaDs a~ fE'C!eived on faesirnile were only lIs€"fulfor gener,.1 self-briefin~ourJloses. This is becaliSI'thev were receiverl 2 hours aft('r dal" time an,idid' not include all availatlle "hs('rvations due tospa"" limitation. Frequently, the d"ta fllotterl onthese 'l1aDS WIlS too cluttered to be easily inter­preted. Therefore, hand-analyses YiNe <,onsis­tenth' used an<J are the kev to anI' suc<'ess thaImay have !wen achieved. ,\'150, the fioest resolu­tio~ numericfll !T!odel llS~rl hy N"I[C', thp. LimitedFine \"esh (LF'.I), <lid not or coulo not he exrc"tedto resolve features thAt Cfillse sig:nifie-Ant Lt\"'c..:..

26

In short, any success achieved in advising pilots ofCLWS will only occur if resources and personnelare dedicated to the problem and have the time todo meso-scale analyses of all available data.

ACKNOWLEDGEMENTS

'JIy colleague at the FAA Systems CommandCenter during the LLWS Test, Bob \'!cLeod, leadsthe list of those to be thanked for help with thisresearch. Vern Lindsay of the regular SystemsCommand Center staff displayed unusual enthusi­asm for the project and provided many helpfUlsuggestions. Frank Coons of the FAA and EdGross an<l Jerry LaRue of NWS arranged for timeto complete the report. Sanders Associates of!\lashua, New Hampshire graciously did some ofthe lITaohical work. And special thanks to mywife Debra for her typing skills and generalencouragement.

REFERENCES

Bedard, A. J., Jr., 1'1. H. Hooke and D. W. Beran,1977: The Dulles Airport Pressure JumpDetector Arrav for Gust Front Detection, Bul­letin of the American 'I1eteorological Society,Vol. 08, No.9, po. 920-926.

B1ackadar, Alfred K. and G. C. Reiter, 1958:Objective Forecasting of Low-level WindShear, Scientific Report #1 for Air ForceCambridge Research Center (No. AFI9(604j:2059), The Pennsylvania State University.

8osart, Lance F., Cosmo J. Vaudo and John H.fJelsdon, ,Jr. 1972: Coastal Frontogenesis.Journal of Appliecl ~leteorology, 11, pp. 1235­1258.

flrowning, K. A. lInd C. W. Pardue 1973: Struc­ture of Low-level Jet Streams Ahead of Mid­latitude Cold Fronts. Quarterly Journal of theRoval \1 eteorological Society, Vol. 99, No.422, ::>fl. 619-638.

'l"rdesty, R. \1., P. >\.. \landics, D. W. Beran andfl.. A. Strliusch, 1977: The Dulles AirportAcollstic-'\'icrowave Radar Wind and WindShear ¥.easuring System. Bulletin of theAmerican 'VIeteorological !'locietv, Vol. 08, No.9, op. 910-918.

T,"reil7.herf;, Carl W., 1968: The ~lesoscale WindFieln in lIn O"clusioo. Journal of Applied\1 eteorolo(rV, Vol. 7, op. 03-67.

~OW'"' n!iniel F., 1974: Low-level Wind Shear ...its "ffccts 00 ADflroaeh aod Climbout. DCFlir<h\ ,\floroach '" 9p:"zine #?O. Douglas Air­"reft ro., Lon~ l1ep."h, CA, pro 10-17.