INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR … · INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR...

*Captain, Boeing MD-80, Underhill, VT, Member AIAA

Copyright 2002 © by Steven D. Green. Published by American Institute of Aeronautics and Astronautics,Inc., with permission.

American Institute of Aeronautics and Astronautics1

INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR CARRIER PILOTS

Steven D. Green*

Abstract

During the early development of commercial airtransportation, inflight icing was approached witha policy of avoidance. This policy was appliedthrough a body of knowledge gained throughexperience. Pilots learned where and when toexpect icing encounters, and they developedoperational strategies to avoid it. However, thequality of forecasting was not sufficient to reliablypredict icing encounters. Thus, manufacturerstook the approach of equipping aircraft used inscheduled air transportation with ice protectionsystems. Currently, general aviation pilots stillapproach icing with a policy of avoidance.However, commercial transport pilots mustrationalize a dual approach. While avoidanceremains the best approach, the volume of airtraffic and the constraints upon routing imposedby hub operations place heavy restrictions onthe avoidance option. A study has been made of120 accidents involving air carrier aircraft in theUnited States and, to a limited extent, Canada.The data from these accidents has beencompiled and used to evaluate the paradigm isuse today, and to identify changes necessary toimprove the core knowledge possessed by theair carrier pilot.

Introduction

The mathematician Richard Hilson has said, "Agenuine education enables one to acquire, foroneself, the skills one happens, at a given stage ofone’s life, to need. A training, on its own, contributesalmost nothing to education and producesdistressingly ephemeral advantages." 1

Accurate situational awareness is dependent onan understanding of nuance that training cannotprovide. For much of its history, aviation has insteadrelied on experience to provide the necessaryunderstanding. Unfortunately, experience itselfoften falls short: one thousand hours of experiencemay turn out to be nothing more than one hour athousand times.

In contemporary times, the terms training andeducation have become convoluted. The process oftraining does not inherently lead to questions such

as "What do we know?" and "How do we know it?".The process of education is generally built aroundthese questions. In the absence of these questions,experience which has not been correctly interpretedcan be memorialized into axioms or premises whichstrongly, and often detrimentally, influence attemptsto understand nuance and therefore maintainsituational awareness.

The operational understanding of the effects ofice contamination on the aerodynamics of transportcategory aircraft is a case in point.

A review of 120 airframe icing accidentsinvolving aircraft commonly used in air carrier service†

dating from 1940 to the present day indicates thatthe average flight experience for the pilot incommand was 7356 hours. The median flightexperience was 6415 hours. Figure 1 illustrates thedistribution. This would suggest that a reliance onexperience to understand the nuance of icingdegradations is not sufficient.

A review of a much broader database containing312 synopses of accidents and incidents worldwide‡

yielded 136 cases in which enough informationexisted to determine both the flight crew awareness

† The 120 events that were evaluated for thiswork were drawn from a variety of sources. First,a search was conducted of the NTSB data usingthe Board's on-line Database Query Tool. Thisdata covers the period 1962 to present. Thesearch was limited to aircraft types in common aircarrier usage, and was limited to accidents only.Corporate aircraft and aircraft more typical togeneral aviation, as well as incidents to all types,were not considered. Second, a similar searchwas conducted of the digitized reports availablefrom the Department of Transportations's On-Line Special Collections website. This datacovers the period 1934 - 1965. Third, someevents were drawn from Flight Safety Digest (ref.16) . Finally, several reports were included fromCanada, although these do not represent theresults of a structured search as the U.S. reportsdo‡ This data was derived from the followingsources: NTSB, Transport Canada, the EURICEproject, ICAO, the Flight Safety Foundation, andthe NASA ASRS program. This data consideredboth accidents and incidents.

American Institute of Aeronautics and Astronautics

2

of the icing situation and whether the ice protectionsystem (IPS) had been operated. In 20 of thesecases, the flight crew was not aware of ice accretionon the aircraft. In 30 cases, they were aware butelected not to operate the IPS. In the remaining 86cases, the flight crew had operated the iceprotection system, although it is often not possibleto determine whether it had been operated correctly.Figure 2 shows this distribution. This makes a strongcase that the knowledge of structural icing hazardspossessed by flight crews is either not adequate or isnot adequately employed.

It is worth examining that knowledge andcomparing it to the accidents themselves. To do soon a case by case basis would be exhaustive. Tolook at this question from an air carrier perspective,there are four separate collections of events whichmight be useful. These are the DC-3 accidents, theICTS accidents, the takeoff accidents, and thecontemporary accidents. Before doing this, it isessential to examine the core knowledge basisemployed in the training of the air carrier pilot withregard to structural icing.

The Paradigm

In 1928, Carrol and MacAvoy of the NACA beganflight investigations of structural ice accretion. Theyimmediately identified two distinct types of iceformation, one forming in temperatures near 32F andthe other in a temperature range of 10 degrees lessthan 32F2. The former developed a large shape withconsiderable chordwise extent and occasionally amushroom shape at the leading edge. The latter wasmore confined to the leading edge and sharper.Indeed, the authors suggested that some of theselatter shapes might actually improve the airfoil'sperformance.

The following year they continued flightinvestigations, and in their paper of that year theyposited two fundamental concepts regarding themanagement of ice accretion by the pilot. First, theysuggested that avoidance of icing conditions shouldbe relatively easy. They stated that,

"This avoidance is not difficult since it [clearice] will not occur except in the presence ofmoisture in reasonable quantity, whichshould be clearly evident, visually, whetherthe moisture be in the form of rain, fog, orclouds."3

Second, they concluded that "all deposits of anicy nature are not dangerous", and they argued thatan educated discrimination made by the pilot wasnecessary. In their words,

"If pilots can be well and thoroughlyacquainted with the conditions controllingthe formation of ice and particularly if theycan learn that every deposit upon the wingsor parts of an airplane is not necessarilyhazardous, the problem will be in a largemeasure solved."3

In 1932, Samuels4 laid out more formalizeddefinitions for clear ice and rime ice. He stated thatclear ice would usually be smooth and glassy, andrime would build forward into a sharp nosed shape.He also introduced the argument that clear icingwhich occurred in the presence snow or sleet wouldproduce a rough ice shape, and that slow freezingwould produce ridges.

Despite this conclusion regarding the mixing ofsnow or sleet with clear icing, which is substantiatedby his data, Samuels classified 155 flightobservations as 108 cases of rime, 43 of clear and 4cases "where rime and clear ice formed during thesame flight". He did not break the clear iceclassification into the two distinct cases he haddefined.

Samuels drew the conclusion that clear icing hasa "more serious effect" than rime based on thepercentage of flights terminated due to the icing. Inhis data, 31% of the terminations were due to 1/4inch or more of clear ice, 12% due to less than 1/4inch of clear ice, 4% due to 1/4 inch or more of rimeice, and no terminations due to less than 1/4 inch ofrime.4 This data also strongly related the icing hazardto quantity (thickness). Samuels did not account forthe possible bias of the individuals making thedecisions to terminate (the pilots). Nor did he qualifythese decisions as based on metrics available to thepilot in the cockpit, i.e., airspeed degradations, rate-of-climb degradations, etc., which may notcomprehensively reflect the true degradations ofeach icing encounter.

Later in his paper he points out that

"one of the chief difficulties in a study of thiskind is the frequent impossibility for the pilotor observer to classify correctly the type ofice formation since it is usually melted by thetime the airplane reaches the ground...sincemany of these flights were made beforedaylight this difficulty was especiallypronounced..."4

Perhaps the most interesting point stated inSamuels' paper is that a third type of ice, frost, "doesnot adhere to the airplane very firmly and is neverdangerous (italics added) as it has very littleresistance to the vibration and wind forcesencountered in flight."4


3

German work on icing reported in 1935 by Nothand Polte5 (and translated by the NACA) set forththree types of ice, similar to Samuels' descriptions.They identified two types of warm temperature icing.The first is clear and creates little form change to theaerodynamic surface. The second is rough, granular,and has a "peculiar tendency to spread more towardthe sides than the front, thus building up a fairlybroad area". They attribute this second type of warmtemperature icing as probably due to supercooledwater droplets mixed with solid precipitation.

Noth and Polte then defined a cold temperatureice formation creating slight form changes andcharacterized by a "crystalline, snow-like coat".5

Findeisen6 in 1938 provides a detailed analysisof icing clouds and argues that knowledge of thefreezing level is easily predicted and is sufficient foravoidance flight planning. He states that

"According to the foregoing, the icinghazard can, in most cases, be avoided bycorrect execution of the flight according tometeorological viewpoints and bymeteorologically correct navigation...thezones of icing hazard are usually narrowlydefined. Their location can be ascertainedwith, in most cases, sufficient accuracybefore takeoff."

A paper by the French Committee for the Study ofIce Formation7, published in May 1938, makes fairlysimilar distinctions. In this case, two distinct iceformations were identified. The glaze formation issimilar to those identified in the American andGerman papers; the other formation, essentially rime,includes a double horned type of shape descriptor.This paper also advocates avoidance by reference tothe freezing level:

"Equipped with such information beforetaking off, the pilot can adopt some flighttactic, that is to say, modify his itinerary so asto avoid the dangerous zone..."

The regulatory requirements of 1938 whichapplied to scheduled airline operation reflect theavoidance approach. Civil Aeronautics Regulations61.7700 states that

"When an aircraft, not equipped withapproved propeller and wing deicingequipment...encounters, or, in theknowledge of the pilot, may encounter anyicing condition, the pilot shall immediately soalter the course or altitude of the flight so asto avoid or withdraw from the icingcondition."8

However, the development and introduction ofice protection equipment allowed a relaxation of thisapproach. The rule following the aforementionedrequirement, CAR 61.7701, addresses theadvantages of ice protection:

"When an aircraft, equipped with wing andpropeller deicing equipment...encountersan icing condition, the pilot shall so alter thecourse and altitude of the flight as towithdraw from the condition, if, in hisopinion, it appears that the icing conditionmay be of such duration or severity as tootherwise endanger the safety of theflight."8 (Italics added)

This brings us to the material contained withinCivil Aeronautics Bulletin No.25 of January 1943.9This document, entitled "Meteorology for Pilots",contains thorough descriptions of ice formations.These descriptions are largely based on thepreceding ideas, and include the notion that bothrime and clear ice, when mixed with snow or sleet,can produce very rough and hazardous shapes. TheBulletin provides a clearly written description of icingin cumuliform clouds, warm and cold fronts, and overmountainous terrain. Within these descriptions is therecommendation to avoid icing, but when that is notpossible, to fly in colder air so as to encounter onlyrime icing instead of clear.

The Bulletin also contains the definitions to beused in reporting icing in flight. These definitions willbe familiar to modern pilots:

Trace of ice - an accumulation of noconsequence, which does not affect theperformance of the aircraft but should bereported by air carrier pilots formeteorological purposes.

Light Ice - a condition which can be handledsafely by normal functioning of the aircraft'sdeicing equipment. On encountering lightice, it is assumed that the aircraft can beflown indefinitely provided the deicingequipment is used.

Moderate ice - an icing condition whichdeicing equipment will safely handle butwhich for practical flight purposes can beconsidered the signal to the pilot that it istime to alter the flight plan so as to avoidcruising in that condition.

Heavy ice - an icing condition which deicingequipment cannot handle. On encounteringheavy ice, the pilot will change altitude orreturn to a suitable airport and land,


4

inasmuch as to continue under thiscondition of icing would render the aircraftunairworthy. 9

These definitions, adopted by the Air TransportAssociation and approved by both the CAA and theU.S. Weather Bureau, reflect both the idea positedearly on by Carroll and MacAvoy that "all deposits ofan icy nature are not dangerous"3, and were writtenin a way to provide further guidance to the pilot indeveloping the opinion required by CAR 61.7701.

The Bulletin, actually a 244 page book, alsodiscusses the effects of structural icing. These arelisted as loss of lift, increase in drag and increase inweight. It points out that the airplane will likely stall atspeeds much higher than normal. This discussionalso contains warnings against steep turns, turns atlow altitude, and describes the phenomona whichwould become known as ice bridging with regard to apneumatic boot system. 9

The state of civil pilot knowledge, as reflected inpublished material, remains essentially unchangedfrom this point forward. In the December 1955edition of C.A.A Technical Manual 104, "Pilots'Weather Handbook"10, the discussion of icingformations is essentially the same as that of 1943.The section on weather planning points out that theavoidance of icing as a strategy may beoversimplified, "since it is not always possible toselect an altitude at which one would completelyescape these ice-forming conditions...". However,the emphasis is still that "in most cases, icing hazardscan be minimized by full use of meteorologicalinformation", a statement that sounds remarkablysimilar to Findeisen's phraseology.

C.A.A TM 104 introduces a diagram (Figure 3)and concept that had been suggested in much ofthe previous work but never quite formalized. This isthe notion that "the importance of icing is increasedby the fact that the effects of icing are cumulative".10

This concept and diagram were carried forward intothe FAA Advisory Circular 00-6A, "Aviation WeatherFor Pilots and Flight Operations Personnel", thelatest revision of which appeared in 1975.11 Thisdocument states that:

"Icing is a cumulative hazard. It reducesaircraft efficiency by increasing weight,reducing lift, decreasing thrust andincreasing drag."

An interesting concept which can be used as asort of marker in pilot knowledge of icing is that ofadded weight. This has always been mentioned ineducational and training material. However, from thebeginning, Carroll and MacAvoy3 in 1929 pointedout that the added weight of ice is unlikely to exceedthe weight of the fuel consumed during the icing

encounter, and thus is not particularly critical. All ofthe following researchers conclude that ice weight isof little or no consequence when considered in thecontext of the aerodynamic degradations associatedwith ice. Newton12 pointed out that a heavy icingencounter might yield at most 7 pounds of ice perfoot of span. Yet weight remains a concernexpressed in a number of the educational materialsproduced since that time. Indeed, it perseveres intothe 1996 version of AC 91-51A, "Effect Of Icing OnAircraft Control And Airplane Deice And Anti-IceSystems"13:

"Also, if the extra weight caused by iceaccumulation is too great, the aircraft may notbe able to become airborne and, if in flight,the aircraft may not be able to maintainaltitude."

Noth and Polte5 in 1935 discounted the effectsof weight and instead focused on changes inhandling qualities and stability due to icing, statingthat:

"The weight increase is of secondaryimportance...The effect on airplane stabilityis altogether different. Nose and tailheaviness have been observed, as well astorsion, about the longitudinal axis. Thereason lies perhaps less in the shifted centerof gravity than in the changed air flow on thetail, elevators, and their balance."

It turns out that these may have been propheticwords; however, training material in the UnitedStates has remained primarily focused on thecumulative effects concept.

Following World War II, a great deal of theresearch done on icing had to do with two initiatives:1) characterizing the icing environment for thepurpose of designing thermal ice protectionsystems, and 2) developing methods to use inforecasting inflight icing conditions. The principalissue in the design of a thermal ice protection systemis both quantitative and probabilistic: namely, howmuch heat is required to deal with the quantity ofwater that the wing will most likely encounter. Theanswers to these questions lay primarily with cloudliquid water content. It so happened that the toolsavailable for in-situ measurement were optimal forlooking at liquid water content. One of the outputs ofthis work was a more or less linear relationshipbetween liquid water content and icing intensity.

In 1947, Lewis14 related the following scale inuse by the United States Weather Bureau for themeasurement of icing intensity:


5

Trace of ice - 0 to 1.0 grams per squarecentimeter hour

Light ice -1.0 to 6.0 grams per squarecentimeter hour

Moderate ice - 6.0 to 12.0 grams per squarecentimeter hour

Heavy ice - 12.0 and over grams per squarecentimeter hour

This scale was specifically defined for use withicing reports from mountain stations. Lewisproposed an alternate scale, equating the sameterms, trace, light, etc., to cloud liquid water contentbased on assumed droplet diameter, anddemonstrated good agreement between the scales.

Lewis and other authors of the time specificallyrefer to these scales as measures of icing "intensity",which is not the same word as "severity". The readerwill no doubt have noticed the parallel between thedefinitions used by the CAA for aircraft icing and theweather bureau definitions. The CAA definitions arequalitative, while the Weather Bureau and Lewisdefinitions are quantitative. Nevertheless, bothaddress intensity. When looked at through themodel of McAvoy and Carroll, that "all deposits of anicy nature are not dangerous"3, and from theperspective of the energy requirements to removethe ice, it would appear that lesser intensities wouldequate to less hazard. Indeed, this is consistent withSamuels'4 data on flight terminations. Theconvolution of the argument was completed in thelate 1960s, when the term "heavy" was dropped infavor of the term "severe".

However, nothing in these definitions makes anyspecific reference to the severity of the aerodynamiceffects on an airfoil.

Thus, by the early 1950s, the paradigm was set.The icing hazard was considered to be cumulative,continued exposure could progressively render anairplane "unairworthy", clear ice was more hazardousthan rime ice, and small amounts of ice accretionwere not generally considered hazardous. The icinghazard could in large part be avoided by competentuse of all available meteorological information, andthat which could not be avoided by this approachwould rarely be of serious consequence.

Review of Accidents

The DC-3 Accidents

During the 1930s, 1940s and 1950s, structuralicing contributed to numerous air carrier accidents.

The accident reports provide a usefulcharacterization of icing knowledge in theoperational community, both from the standpoint ofthe flight crews involved and from the standpoint ofthe accident investigators. They also provide someuseful factual information with which to examine theparadigm described above.

Probably the two best documented accidents ofthis era in which icing played a role are those ofUnited 21 and Northwest 5. These accidents tookplace within 10 months of each other; United 21crashed at Chicago on December 4, 1940 andNorthwest 5 crashed at Fargo, North Dakota onOctober 30, 1941. Both accidents involved iceprotected DC-3 aircraft. Both were approach/landingaccidents. Between the two, there was one survivor,the Northwest captain.

United 21 crashed while circling the fieldfollowing an instrument approach at Chicago. Theconditions on the surface at the time were light snowand a temperature of 32F.15 Freezing drizzle hadprevailed until approximately one hour before at thesurface. During the five hour period surrounding thetime of the crash, there were thirteen landings andeighteen takeoffs involving seven different airlines.

The accident is interesting for a number ofreasons. The captain had intentionally held abovethe clouds until he was cleared for the approach,specifically in order to remain clear of the reportedicing conditions. The total time that the reportestimates the aircraft was in icing conditions was only8 minutes. The airplane appeared to be establishedon final approach when it simply rolled off on the leftwing, and, despite the application of power, crashed.The flight crew was aware of the icing.15

Because of the high volume of operations atChicago that afternoon, there were a number ofqualified witnesses who flew in these conditionsboth before and after the accident. They describedthe icing as anything from light to heavy rime andglaze. In all cases, the pneumatic boots weredescribed as effective. However, in one case thepilot evaluated the ice accretion on his airplane afterlanding. He found 3/8 of an inch to one inch of clearice on the leading edges, but also found 1/2 inch ofvery rough ice extending aft of the boots about 4and 1/2 inches. A similar formation was found on theempennage. The pilot believed that all of this ice hadaccumulated during the four minutes between whenhe shut off his pneumatic boots and his landing.15

The accident airplane exhibited 3/8 of an inch ofrough, granular ice on the leading edges extendingaft about 2 inches and a thin film of clear ice aft of thisto the boot seams. No ice was found aft of the boots.In this case it was also believed that the ice hadaccumulated after the boots had been shut down.This was a standard practice due to the effects ofinflated boots during the landing.15


6

Subsequent flight tests several days laterencountered 1/4 inch of clear ice on the boottogether with a substantial mass of rough iceextending about six inches aft of the boot. A secondflight test the same evening encountered 5/8 of aninch of very rough ice over the leading edge.15

While the stalling speeds of several flightslanding that afternoon were reported as higher thannormal (an informational advantage of an era in whichfull stall landings were the norm), the flight testsrevealed only an increase of approximately 6 milesper hour over that of a clean wing. The accidentreport noted that

"Although pilots do not agree as to theeffects of ice on the DC-3, it may beconcluded that the ice does raise the stallingspeed to some unpredictable extent andthat the effect cannot be stated in terms ofthe amount of ice alone, but is dependentupon both the amount of accumulation andits location on the wing."15

Horn or wedge shapes were not identified in thisreport. None of the ice shapes were particularlylarge. They all exhibited roughness of varyingdegrees.

A very similar accident took place on December21, 1947 at North Platte, Nebraska. A Seattle AirCharter DC-3 was making an approach in drizzle(probably freezing drizzle, since ice was forming)when it stalled and dropped off on the left wing. Inthis case, a recovery was made in time for a hardlanding which did substantial damage.16

The Northwest accident is a different case. In thiscase, the airplane stalled after leveling at a minimumdescent altitude during an approach to Fargo. Thisaccident is interesting because of the viewpoints ofthe surviving pilot and the investigation itself. Thepilot had been very alert for ice, and identified it assoon as it began accreting.17 He described the firstaccretion as very light ice, "but not any amount toeven be bothered about,". During the approachdescent, the ice accretion increased, and the captaindescribed it as rime, "the type that forms in irregularchunks and irregular chips, little extensions out onthe windshield,". He still did not consider it unusual.

Five hundred feet lower, the captain stated that"we did start to pick up quite a lot more ice,". Thereport notes that, "having on previous flightsexperienced what he considered to be heavier icing,he was still unduly concerned." In this case, uponlowering the landing gear, the captain ordered thefirst officer to operate the pneumatic boots.17

Upon leveling at the minimum descent altitude,the airplane stalled. Although he applied maximumpower and retracted the landing gear, the captainwas unable to avoid impacting the ground.17 An

aviation mechanic arriving at the scene withinminutes of the accident found a coating of rough icefrom 1/2 to 2 inches thick along the deicer boot onthe outboard right wing. The temperature at theground was approximately 32F.17

Subsequent flight testing revealed that anuncontaminated DC-3 could remain in an aggravatedstall with power on, continuously descending untilthe nose was pushed over and recovery effected.17

Propeller strike marks on the ground indicatedthat the airplane had impacted at an airspeed of 90miles per hour. The flaps up, power-off stallingspeed of the DC-3 is reported to be 80 MPH, andwould be lower with power on. The investigation wasperplexed about the normal flying characteristicsexhibited during the descent: "If the wing had beenstalled to the degree...required, it is hard to conceiveof the airplane having been kept continuously undercontrol during the descent..."17

The investigators considered the effects of ice.They stated that:

"A careful consideration of the evidence hassatisfied us that the partial loss of control wasnot caused solely by the ice which had beenaccumulated on the airplane. A collection ofice on the upper surfaces is not anuncommon experience and, while it is to beavoided to the fullest extent possible by theexercise of great caution, in the nature ofthings it cannot be eliminated entirely.Although the amount of ice which hadaccumulated on the airplane was substantial,experience has shown that aircraft maysafely be flown with a far greateraccumulation of ice than that which obtainedin this case. The testimony in the record ofthis accident, as well as general knowledgepreviously acquired, convincingly shows thatthe accident was not caused solely by ice. Itis equally clear, however, that the amount ofice which had been gathered by the airplanewas sufficient to affect materially the flightcharacteristics of the plane. The effect of iceis to reduce airspeed and increase thestalling speed."17

The effects of encountering the non-linearity inthe contaminated lift curve could not have beenmore dramatic. But it is important to consider theNorthwest report's analysis, because it succinctlydescribes the prevailing operational understandingof the effects of structural ice accretion.

On March 2, 1951, Mid-Continent 16 crashed atSioux City, Iowa while circling for landing. Theweather was characterized by a 500 foot ceiling, 1mile visibility in light snow showers, and atemperature of 29F.18 Ice was observed by both


7

passengers and rescue personnel. The airplaneclearly stalled; however the report does not detail aserious analysis into the effects of icing on the stallcharacteristics or performance, except to state that :

"Ice accumulation would not have beencritical for normal flight operations, but,under a condition of low air speed in a turn,might have been a factor in causing theaircraft to stall at a slightly higher than normalair speed."18

On January 20, 1954, a Zantop Airways DC-3crashed during an approach at Kansas City. Theweather was characterized by a 600 foot overcastceiling, light freezing drizzle, light snow, and a strongnorthwest wind.19 1/2 inch of clear ice was found onthe leading edges of wings and tail; the evidenceindicated that the deicing system had been in use.Indeed, the investigation criticized the crew formaking a turn during the approach with the boots inoperation, as it was well known that pneumatic boots,when inflated, would increase the stall speed.

On March 8, 1964, another DC-3 crashed atChicago. Although this accident was heavilyinfluenced by an encounter with wake turbulencecaused by a Boeing 707, the weather in this casewas a 700 foot overcast, light drizzle, surfacetemperature 34F, and a north wind at 10 knots.20

The ice accretions found on the right wing an rightstabilizer were characterized as "mixed rime and clearwhich was extremely rough textured", about 3/8 ofan inch thick with numerous projections extendingabout 1 inch from the airfoil leading edge. The flightcrew did not operate the pneumatic boots in thiscase, and the report criticizes them for this action.

Four days later, on March 12, 1964, another DC-3 crashed on approach to Miles City, Montana. Theweather was an indefinite ceiling of 500 feet, lightsnow showers, temperature 32F and wind from thenorthwest at 20 knots with gusts to 30. No iceaccretion was found after the crash, perhaps due tothe post-crash fire. Based on propeller slash marks,an indicated airspeed at impact of 134 knots wasestablished; the investigation believed that thisprecluded ice as a cause because, it said, thisairspeed should have been more than sufficient tocounteract the effects of severe airframe icing.21

None of these accidents involved largequantities of ice accretion. To the extent that anarrative account is available, none of them involvedobviously abnormal handling characteristics until acritical angle of attack was reached.

The Ice Contaminated Tailplane Stall (ICTS)Accidents

On April 6, 1958, during an approach to land atFreeland, Michigan, a Capital Airlines VickersViscount crashed in a steep nose-down attitude.The aircraft had just rolled wings level onto finalapproach following a fairly steep turn.22 The weatherin this case was a ceiling of 900 feet variable to 1100feet, visibility 3 to 4 miles, light snow and freezingdrizzle, with winds from the northeast at 18 to 27knots. No ice was recovered from the crash site,however, the impact was quite violent and asubstantial post-crash fire ensued.

Icing conditions had been forecast, and aConstellation which had landed 13 minutes earlierreported one inch of ice at landing. The originalinvestigation focused heavily on a wing stall. Theeffect of an inoperative stick shaker was explored,and icing was investigated in the wind tunnel.Propeller pitch malfunctions were looked at.22 But noclear explanation for this accident was forthcoming.

Then, on January 29, 1963, it happened again atKansas City. This time, a Continental AirlinesViscount crashed following an attempted landing.While the landing was not definitively aborted, it wasclear that the airplane was in trouble and that theflight crew was preoccupied with maintaining controlas the airplane flew down the full length of therunway. It nosed over steeply and crashed on the farend. The weather in this case was 3000 feetovercast, visibility 12 miles, no precipitation and atemperature of 17F.23 No ice was found after thecrash, but, again, there was a post-crash fire.

During this investigation, the Board discoveredthat this type of event had taken place before and asuccessful recovery had been achieved. The mostinteresting event took place on February 20, 1963 atColorado Springs.23 In this case, the Viscount creweffected recovery with both pilots pulling the noseup together. The aircraft returned to a normalbehavior, then encountered a second set of pitchexcursions with another recovery, and then handlednormally to the landing. Examination of the wingsfound a light rime accretion, but the empennagesurfaces exhibited a one inch thick double hornshape of what was described as "rough rime ice". Itturned out that, after a ten minute exposure to cloud,the flight crew had visually checked the aircraft andfound no ice. Between this inspection and theevent, the aircraft was exposed to cloud for onlyanother two minutes.

Another Viscount had survived a similar event atWillow Run in Michigan. In this case, an ice shapewas found on the empennage described as a "cove"buildup, presumably alluding to a double hornshape.23


8

The manufacturer was able to duplicate the iceshape in an icing tunnel; however it required 20minutes exposure to .72 grams per cubic meter ofliquid water with an MVD of 20 microns. TheContinental flight which crashed at Kansas City wasestimated to have been exposed to icing conditionsfor six to eight minutes. In none of the cases in whichthe crew survived was the thermal ice protectionsystem operated, because the flight crews did notrealize that ice was accreting. In the accident cases,the evidence extracted from the wreckagesupported, but did not prove, the idea that the IPShad not been operated.23

Of course, a subsequent Viscount accident atStockholm in 1977 led to further tunnel testing.Trunov and Ingelman-Sundberg were able to showthat the double horn shape is not required; a thinroughness will do the job as well. In that accident, itwas believed that the crew did operate the IPS, butthat descent power resulted in a less than adequateheat supply to the surfaces.24

On December 21, 1963, a corporate Convair580 experienced this phenomona during anapproach to Midland, Texas. The weather at the timewas 200 feet overcast, visibility less than 2 miles,light snow, and a temperature of 27F. When landingflaps were extended, the aircraft began a series ofdivergent pitch oscillations, culminating in groundimpact. Nine hours later, 1/2 inch of rime iceapproximately 2 inches wide was found on the leftwing leading edge; the flight crew had not operatedthe IPS.16

On March 10, 1964, the event happened again,this time involving a Slick Airways DC-4 landing atBoston. The weather was, surprisingly, 700 feetovercast, visibility 1 and 1/2 miles in moderate sleetand fog, temperature 32F, wind from the northeast at22 knots with gusts to 28. Again, due to the highenergy ground impact with subsequent fire, no iceaccretion was recovered. This DC-4 was notequipped with airfoil ice protection.25 However, inthat sense, it bears perfect resemblance to theViscounts with ice protection systems not operatedby the crew.

The Slick accident report is the first to considersome of the NACA work done in the late 1950s,particularly by Gray and von Glahn.26 The reportpoints out that "Rotation of an airfoil to angles ofattack higher than that at which icing occurred,generally creates an even greater loss of lift than ifthe airfoil iced when at higher angles of attack.".

The Continental accident report, released onJune 17, 1964, included a reasonably gooddiscussion of ice contaminated tailplane stall; theonly oddity in the report is a statement that nose upelevator may induce a tailplane stall, which is notconsistent with the findings of the NASA TailplaneIcing Project.27

On March 15, 1989, a Nihon YS-11 operated byMid-Pacific Airlines crashed at West Lafayette,Indiana. The available report indicates that theweather in the terminal area was good, with noreported ceiling and a visibility of 8 miles.; howeverthe aircraft's behavior was completely consistent withan ICTS event. Examination of the wreckagerevealed 1/2 to 3/4 inch of rime ice accreted on thehorizontal stabilizer, with none found anywhereelse.16 The airframe deicing system was notoperated. A year later, another Mid-Pacific YS-11experienced the same type of event on approach toWest Lafayette. In this case, the crew recovered.After landing, "substantial" ice, by some reports upto 2 inches, was found accreted on the empennageaft of the protected surfaces; in this case, the flightcrew had also not operated the IPS.28

Several other aircraft have experienced thisphenomona since this time, notably the BritishAerospace BaE-3100 Jetstream, the AerospatialeATR-42, and the Saab 340 (prior to themanufacturer's modification of the horizontalstabilizer). In January of 1998, a NASA ASRS reportwas filed which describes an ICTS event involving anMD-80 aircraft. In this case, the IPS was beingoperated, and the system had been cycled to heatthe tail. Nevertheless, when the flaps were extendedto 40 degrees, the nose pitched over and 500 feetof altitude was lost prior to recovering.29 This aircrafthad the singular advantage of configuring for landingat a significantly higher altitude than many of theaforementioned cases, which is typical for jettransport aircraft operated in contemporary times.

These accidents are interesting whenconsidered in light of the statements made by Nothand Polte during the 1930s, as cited above. In anyevent, the effects of an ice contaminated tailplanecan hardly be described as cumulative.

The Ground Deicing Accidents

Although the Air Florida 90 accident atWashington, D.C., on January 13, 1982, is perhapsthe most well-known accident involving a failure toadequately deice prior to takeoff, this type ofaccident has been occurring for many years. Twothat appear in earlier years are the accidents onJanuary 2, 1949 involving a Seattle Air Charter DC-3and on January 4, 1951 involving a Monarch AirlinesCurtiss C-46. In the DC-3 case, numerous attemptswere made to deice the airplane before the pilot wasadvised to be sure and get "plenty of speed" beforelifting off.30 The Monarch case is more interesting inthat the investigation refused to acknowledge a rolefor icing. The airplane had been deiced inside ahangar with an alcohol solution. Although the firstofficer reported observing frost on the wings forming


9

prior to takeoff, the investigation concluded that,since other airplanes with frost on the wings hadsuccessfully taken off, this should not have been aproblem. Rather, the investigation concluded thatthe captain's election to use a lower than standardtakeoff power setting caused the crash.31 This reportmakes interesting reading when read side by sidewith the Air Florida 90 report.

On February 12, 1979, a Mohawk/Frakes 298version of the Nord 262 crashed during takeoff atClarksburg, West Virginia. The aircraft had accreted a1/4 inch layer of wet snow prior to takeoff, and hadnot been deiced.32 In this case, the investigationrefers to Ralph Brumby's paper, "Wing SurfaceRoughness: Cause and Effect", published inJanuary 1979. The report states that,

"According to a recent review of the effectsof wing surface roughness, frost, snow orfreezing fog adhering to wing surfacescauses a reduction in maximum liftcoefficient, a reduction in the angle of attackat which stall occurs, and a rapid post stallincrease in drag. The above effects are mostpronounced when the roughness is on ornear the leading edge of the wing."32

In March of 1979, Trunov and Ingelman-Sundberg of the Swedish-Soviet Working Group onScientific-Technical Cooperation in the Field of FlightSafety published their paper entitled, "Wind TunnelInvestigation of the Hazardous Tail Stall Due toIcing".24 Although this report is more relevant to thepreceding discussion on ICTS, it details their resultswith distributed roughness on the leading edge of amodified 18 degree swept tailplane model used toinvestigate the Viscount accident at Stockholm.They concluded that a leading edge roughness ofonly 1/1300 of the chord resulted in degradationsalmost as significant as the large ice shapes they hadalso tested. This conclusion parallels that of Brumby,quoted above by the NTSB, at about the same time.

In the Air Florida case, the Safety Boardindicated an awareness of 22 events involving"aircraft pitchup or rolloff immediately after takeoff inweather conditions which were conducive to theformation of ice or frost on the wing leadingedges."33 Boeing had investigated these eventsand published several Operations Manual Bulletinsaddressing them. The particular condition had to dowith the use of reverse thrust while taxing on slipperysurfaces, which resulted in a melt/refreeze cycle ofblown snow. Following flight tests with a simulatedroughness applied to the leading edge slats, Boeingfound that "The stall characteristics with bothsymmetric and asymmetric leading edgecontaminations were characterized by a veryapparent pitchup, yaw rate and rolloff." The flight test

program concluded that "Wings should be kept clearof ice...and rotation rates should not exceed 3degrees per second." The NTSB also quotedBoeing as concluding that "additional speed marginswere advisable when operating in adverse weathersuch as snow, sleet, or rain at near freezingtemperatures." The subsequent Operations ManualBulletin emphasized care in controlling pitch rateduring rotation and the use of additional flaps, whenpossible, to increase stall margins. It stated that,

"If leading edge flap roughness is observedor suspected for any reason, care should beexercised to avoid fast rotation rates inexcess of 3 degrees per second and/or overrotation."33

On November 15, 1987, Continental 1713crashed during takeoff at Denver. During theinvestigation of this accident, the Board quotedMcDonnell Douglas as stating that a roughness ofonly 1/10,000 of the wing chord can "adverselyaffect the maximum lift coefficient and significantlyincrease the stall speed."34 Testimony was alsoheard during this investigation that "icecontamination may also produce roll oscillations andunexpected pitch-up tendencies during flight." (Thepitch up tendency is related to outboard wing stall,resulting in a significant change in the overallpitching moment.)

On March 22, 1992, USAir 405 crashed duringtakeoff from LaGuardia airport in New York. Onceagain, a discussion of wing leading edge roughnessensued. In this case, the airplane involved was aFokker F-28. The report details differences betweenthe conclusions drawn by McDonnell Douglas andFokker regarding whether or not slatted wings havean advantage. McDonnell Douglas had alreadyconcluded that non-slatted wings were moresusceptible to roughness effects; Fokker citedSwedish research indicating that there was nodifference in the effects of frost between a slatted ornon-slatted wing.35

Between these cases can be seen someuncertainty as to the full effects of leading edgeroughness. Although the Civil Air Regulations had,since 1938, always clearly prohibited air carrier aircraftfrom being taken off when the wings or tail surfaceshave a "coating of ice or snow", regulatory languageregarding frost did not appear until 1953.36 The rulesthen prohibited takeoff with "frost, snow or iceadhering to the wings, control surfaces, orpropellers...". (One may presume that the rules hadfinally caught up with the decline of fabric coveredwings, rendering Samuels' 1932 conclusion aboutfrost obsolete, i.e., frost now did adhere to the wingfirmly, had considerable resistance to the vibrationand wind forces in flight and was, in fact, dangerous.)


10

This language is identical to the basic languagecontained in the present Federal AviationRegulations. Even the present rule, however, allowsthat "Takeoffs with frost under the wing in the area ofthe fuel tanks may be authorized by theAdministrator. ". 37 This statement gets to the heartof the matter, which is to optimize the ability totolerate a certain amount of roughness with theability to eliminate it. The current guidance is quitespecific, but the evolution of understanding whatcan be tolerated, and where, has been quite painful.

On December 27, 1968, Ozark 982 crashedduring takeoff from Sioux City, Iowa. Theinvestigation found that the airplane, a non-slattedDC-9-14, had accreted ice during the precedingapproach and that this ice had not been removedprior to the takeoff.38 The report said that "The crewwas aware that a 'small' accumulation of ice waspresent on the aircraft, but the captain did notconsider it significant." This is once again illustrativeof the fundamental aspect of the paradigm that"every deposit upon the wings or parts of an airplaneis not necessarily hazardous, ". Actually, this ice wasanywhere from 1/16 to 1/2 inch thick and extended 6to 8 inches aft of the leading edge. The accidentreport in this case presents an excellent analysis ofwhy many instances of ice accretion result inunremarkable events or, indeed, go unnoticed. Thereport states that:

"The approach and landing at Sioux Citywere most probably completed withoutincident because they were flown at near thesame angle of attack as the angle of attack atwhich the ice was accumulated and [had] thebenefit of the increased lift as the aircraftdescended into ground effect during thelanding flare; whereas, during lift-off theaircraft was rotated to an angle of attack mostprobably 7 to 9 degrees greater than theangle of attack at which the ice wasaccumulated, coupled with the reduced liftas the aircraft departed ground effect."38

The same type of accident had occurred manytimes before. On February 16, 1950, an EasternAirlines DC-3 had landed at Lexington, Kentucky.During the approach, ice had accreted on theleading edges of the wings. Additional speed wasused on final approach, and a normal landing wasmade. The ice was not removed, probably by virtueof the same reasoning used by the Ozark crew, andthe aircraft crashed on the subsequent departure.16

The Ozark report is the second report to refer tothe NACA work done by Gray and von Glahn. It isworth noting that this report was adopted onSeptember 2, 1970. The Slick 12 report which also

referred to this work was released on November 5,1964.

On February 17, 1991, Ryan 590, a cargoversion of a DC-9-15 (non-slatted), crashed duringtakeoff at Cleveland. Once again, the aircraft had notbeen deiced prior to takeoff.39 It is interesting tonote that, although this aircraft had also flown asuccessful approach through icing conditions, thereis no reference to the work of Gray and von Glahn orto the effects of increasing the angle of attackbeyond that at which the ice was accreted. In thiscase, the Board made the assumption that the flightcrew had operated the IPS during this approach andthat it would have been effective, although thedeceased crew and overwritten voice recorderprevented any certainty of this. The investigationposited the idea that the still-hot wing had, afterlanding, melted some snow accumulation whichsubsequently refroze as the wing cooled.

The emphasis in this investigation was on theconsiderable work done by McDonnell Douglas tomake flight crews aware of the hazards of smallroughnesses. As can be seen in many of thepreceding accidents, this is an extremely importantaspect. However, unlike the accidents involvingContinental 1713 and USAir 405, this aircraft had notbeen deiced. Consequently, it was not possible todetermine when the ice accreted. A comprehensiveapproach to this accident would have been to positboth types of analysis, that of Continental1713/USAir 405 and that of Ozark 982.

The absence of an analysis in the report similar tothat of Ozark 982 reflects a preference for recentknowledge in isolation from a broader historicalperspective. This may lead to bolted-on additions tothe paradigm described earlier without seriouslyinvestigating the effects of the new knowledge onthe whole thing.

The data reviewed indicated that this type ofaccident continues to occur to aircraft operated fromremote facilities without ready access to deicingfluids.

The Contemporary Accidents

The investigation into the accident at Roselawn,Indiana, on October 31, 1994, stands as the mostcomprehensive work on a single icing accident todate. Simmons 4184, an ATR-72-212, crashed afteraccreting ice while descending in a holding pattern.For a variety of reasons, the flight crew had extended15 degrees of flap while holding.40 Upon retractionof the flaps as the descent was begun, flowseparation occurred, probably as a thin-wing type ofstall, at the outer wing panel and an aileron becameunbalanced. The aileron fully deflected, rolling theairplane completely and leading to a rapid loss of


11

control. The circumstances of this event weresurprising to the operating community, and thebroad response within that community was that,although "every deposit upon the wings or parts ofan airplane is not necessarily hazardous", this onewas and the flight crew should have realized it.

Yet within two hours of this accident, anotherATR-72 crew experienced a similar ice formation.This aircraft experienced a buffet at 170 to 180knots; although the leading edges appeared clean,close examination revealed a ridge of ice built up aftof the deicing boots, which had been in operation.41

The actual mechanics of the Roselawn accidentwere not too disimilar from the ICTS mechanicsdescribed earlier. The effects of ice in this case wereclearly not cumulative; there was little if any drag rise,and the airplane behaved quite normally until theangle of attack was driven up by the flap retraction.The loss of control at a high altitude led to a fairlyprolonged struggle during the descent beforeimpact, and the CVR and DFDR provided verydetailed information on the entire sequence. Theextended length of the accident sequence and thequantity and quality of data available led to anunderstanding of the event which far surpassedprevious icing accidents.

At the same time, the tools available forevaluating the weather conditions had evolvedtremendously since the 50s and 60s. It wasdetermined with reasonable certainty that the aircrafthad been exposed to large droplet conditions, inparticular, freezing drizzle. It became immediatelyapparent that this environment exceeded that ofFAR Part 25, Appendix C, which defined theconditions under which certification had beengranted. The effect of freezing drizzle in this casewas to accrete a ridge of ice aft of the protectedsurfaces, which was sufficient to induce a significantchange in the aileron hinge moment.

On January 9, 1997, Comair 3272 crashed whilebeing radar vectored for an approach at Detroit. Thisinvestigation was also a very comprehensive one, ifnot as visible as that of the Roselawn accident. In thiscase, it was discovered that a thin roughness of iceaccreted during a 4 to 5 minute period while indescent, perhaps accompanied by a small ridge ofice, may have been sufficient to induce a stall whenthe airplane was leveled off. The accident reportemphasizes the hazards of surface roughnesses toairfoils, and points out that the traditional method ofoperating a pneumatic deice boot system may leaddirectly into this type of situation.42

While these accidents were both fairly visible,there have been other significant accidents in recentyears. Rocky Mountain Airways 217 crashed east ofSteamboat Springs on December 4, 1978. Theaircraft had departed Steamboat Springs withfreezing rain falling and climbed into icing conditions

which were described as "heavy" by the same crewupon arrival earlier in the day. While the deicingequipment was able to remove much of the ice, theaircraft encountered a mountain wave conditionwhich yielded severe ice accretion. It was unable toclimb to an altitude sufficient to cross the mountains,and the flight crew attempted to return to SteamboatSprings. The airplane settled into the terrain undermaximum power. The investigation concluded thatthe ice accretion combined with downdraft activity tocause the accident.43

On February 16, 1980, Redcoat 103, a BristolBritannia, crashed several minutes after takeoff fromBoston. The aircraft was having considerabledifficulty climbing after takeoff in icing conditions andsignificant turbulence. The captain's decision toradically increase the angle of attack in response torepeated low altitude warnings from air traffic controllikely resulted in a stall. Ice accreted in flight veryprobably played a role, and ice accumulation prior totakeoff may also have been involved. Although theaircraft had been deiced, approximately one hourhad elapsed and some witnesses thought that somesnow had accumulated, although the surviving flightengineer stated that he had specifically examinedthe wing prior to takeoff and saw no ice.44

On March 28, 1989, Air Canada 571, aMcDonnell Douglas DC-8-73, experienced a hardlanding at Edmonton. The aircraft had flown anapproach in weather conditions characterized by a100 foot ceiling, 1/4 mile visibility in light freezingdrizzle, and a temperature of -4C. Rough, jagged iceshapes were found on the airfoils after the landing,ranging from 1/4 inch to 1 inch in thickness. Again,the IPS had not been operated during the approach.The accident was attributed to inadequate controlduring the landing procedure; an actual stall couldnot be verified. However, the manufacturer notedthat a 25 to 30 percent increase in stall speed wouldhave been likely with the reported ice shapes. In thisreport, the authorities pointed out that, had a go-around been required, the ice accretion would haveseriously compromised performance.45

Finally, performance degradations were notedby the Transportation Safety Board of Canada in twocontemporary accidents. On March 8, 1996,Canadian 48, a Boeing 767-300, experienced a tailstrike while landing at Halifax. The weatherconditions at the time were characterized by a 300foot ceiling, visibility 1 and 1/2 miles in light freezingdrizzle, and a temperature of -3.7C.46 Although iceaccretion could not be verified, the IPS was notoperated and a performance degradation was notedin the last 400 feet of the approach, The accidentwas attributed to a visual illusion and a failure torespond to glide path indications from the precisionapproach path indicator (PAPI). A second eventoccurred at Fredericton involving Air Canada 646, a


12

Bombardier CL-600. The flight crew initiated a go-around from an unstabilised, unspooled Category IIapproach in freezing fog.47 The aircraft stalled andcrashed. It was considered likely that some iceaccretion had influenced the aircraft's performanceduring the late stages of the approach andsubsequently, the stall characteristics. The accidentwas largely caused by the flight crew following theflight director go-around guidance before theengines had spooled up; however, the flightguidance system was also not certificated with anyicing effects in mind, and this accident raises thequestion of what relationship might exist betweenautomated flight guidance and contaminated wings.

Analysis

In order to consider improvements to theknowledge base that is necessary for commercialpilots to have, a spreadsheet was developedcontaining a variety of data extracted from thepreviously mentioned 120 accident reports,summaries or synopses. This data includes thefollowing fields:

Flight DesignationDateLocationManufacturerTypePilot Flight TimeCo-Pilot Flight TimePhase of Flight (Accretion)Phase of Flight (Accident)Stablised Approach DataSurface TemperatureCeilingSurface Precipitation IntensitySurface Precipitation Type 1Surface Precipitation Type 2VisibilityWindIce Shape DescriptorsIce ThicknessChordwise ExtentAoA ChangeIce Accretion in FlightExposure TimeIPS TypeIPS operated

The accidents reviewed are not a completeset. Additional accidents have taken place, andthe associated reports may not be a part of thedatabases searched for this paper. However, thisset is believed to be reasonably representitive ofthe complete set.

Exposure, Duration and Severity

The paradigm in use for operations in icing hasrelied heavily on an understanding of themeteorology conducive to icing in order to minimizethe exposure to, if not avoid entirely, thoseconditions. This is Findeisen's concept of"meteorologically correct navigation". It would beimpossible to evaluate this conceptcomprehensively, since no data is available for theflights which, apparently, made optimum navigationalchoices to avoid icing, or, for that matter, about whatwould have happened to them had they not doneso.

Of the 120 accidents and incidents evaluated forthis research, 82 involved ice accreted while in flight.Within this set, 60 took place during theapproach/landing phase (Figure 4a). A furthersubset of 21 presented reliable evidence that the icewas accreted during the approach phase (Figure 4b).

The approach phase brings with it geographicconstraints that are sufficient to reduce the conceptof meteorologically correct navigation to its simplestform: one either flies the approach or one doesn't. Inthis sense, the concept of icing as a cumulativehazard plays an influential role; consider again thelanguage used in the 1938 version of CAR 61.7701:

"encounters an icing condition, the pilotshall so alter the course and altitude of theflight as to withdraw from the condition, if, inhis opinion, it appears that the icingcondition may be of such duration or severityas to otherwise endanger the safety of theflight."8

The phrase, "icing condition may be of suchduration..." strongly suggests that if the exposurecan be minimized in time, the hazard can bemanaged. This is undoubtedly what the captain ofUnited 21 had in mind while holding before theapproach at Chicago. He insisted on remainingabove the clouds until receiving his approachclearance, a strategy that resulted in exposure to theicing conditions for only 8 minutes. Nonetheless,sufficient ice accreted to result in an unexpected stallduring the final stages of the approach.15

15 of the 82 events involving inflight accretionpresented adequate information to identify an icingexposure time prior to the event. This data isillustrated in Figure 5. Some of the reports provide arange of exposure times, such as "8 to 10 minutes",or "less than 20 minutes". Others are quite specific,for example, "9.5 minutes". In order to provide aweighted perspective on what the minimumexposures might have been, the median of therange cited in the report was used as a minimumexposure for those cases which inferred zero


13

minutes as the minimum. For example, 10 minuteswould be the minimum used for the "less than 20minutes" case.

It should also be borne in mind that 65 of theseinflight events do not provide adequate informationto determine exposure times. 2 cases provide dataindicating that the exposure was more or lesscontinuous. With this in mind, it is useful to considerthe minimum and maximum exposure data.

The average minimum exposure is 10.6 minutes;the median minimum exposure is 8 minutes. Theaverage maximum exposure is 12.8 minutes, with amedian maximum exposure of 10 minutes. Therange of the maximum data is 2 to 40 minutes. 12 ofthe15 events took place during theapproach/landing phase of flight. This data issignificant because a typical exposure time of 8 to10minutes is overlapped by the nominal time that ittakes a transport aircraft to fly an approach.

The above data speaks to the notion that onlythose aircraft which operate at lower altitudes, andwhich "spend more time in icing", are at risk. In fact,all aircraft are at risk from the standpoint of exposure.Regardless of whether the flight was one hour longor eight hours long, only a few minutes during theapproach is required to build sufficient ice for anevent to occur.

The strategy of minimizing exposure during theapproach can also significantly impact the executionof a stabilized approach. This was the case in theNorthwest Airlink 5719 accident at Hibbing,Minnesaota on December 1, 1993. In this case, thecaptain also elected to remain above the clouds tominimize exposure to icing during the approach.Unfortunately, he chose to delay his descent until apoint which required a very high rate of descent inorder to reach the minimum descent altitude prior toarriving over the missed approach point. Althoughthis descent rate was arrested, the airplane was notleveled at the minimum descent altitude (probablybecause the captain was looking outside for theapproach lights) and the airplane crashed short ofthe runway. The NTSB report is highly critical of thisstrategy of icing avoidance because of the impact ithas on a stabilized approach.48

The question then defaults from one of durationto the other conditional provided so many years agoin CAR 61.7701: severity. As has already beendiscussed, severity is strongly linked in concept withquantity and ice type. Current Part 121 regulatorylanguage requires the pilot to have an opinionregarding icing conditions that "might adverselyaffect the safety of flight". Again, the concept of icingas a cumulative hazard influences the argument. Inconjunction with the icing descriptors of trace, light,moderate and severe (heavy), and with theunderstanding that rime ice is not as threatening as

clear ice, the pilot's opinion is largely reliant onquantity.

In 20 of the 82 in flight events, enough data waspresent to estimate a thickness of the ice adhering tothe aircraft (Figure 6). Many of the ice thicknesseswere estimated after the crash. Often some time hadpassed; some of the reports are careful to point outthat local temperatures had remained below freezingduring the interval. However, in some cases, fire hadresulted, and the effects of this on what ice wasfound are not known. In no case is anything like thepossibility of sublimation discussed. However, thebest way to handle data from the reports is to take itas much at face value as possible, since there is noway to estimate any corrections.

In some cases, the reports cited a range ofthicknesses. For example, "1/4 to 1 inch of ice"might be stated. Since it is impossible to know whatthe distribution of thickness was in what was found, amedian thickness was taken in these cases. In theaforementioned example, that would be 0.625inches.

The average thickness was 0.80 inches; themedian was 0..5 inches. The smallest averagethickness derived from the reports was 0.19 inches;the largest was 2 inches. This highlights thedistinction between severity from the standpoint ofthe energy requirements to remove the ice andseverity from the standpoint of the aerodynamicrequirements for sustained flight.

It is important to consider these thicknesses withrespect to the operation of deicing systems. Theprocedures published for the operation of mostpneumatic and some thermal deicing systemstypically include a minimum accretion thickness priorto operation of the system (an "observable"thickness). This has been anywhere from 1/4 inch to1-1/2 inches. These operating procedures carry withthem the implication of safety up to the prescribedthickness. The data from the accidents, as well asconsiderable research data, suggest to the contrary.

For example, during the IRT tests in support ofthe investigation of Comair 3272, NASA reported amaximum overall thickness of 1/4 inch of ice,consisting of a "rough, 'sandpaper-like' icecoverage", with small ridges forming along thedeicing boot seams.42 Trunov reported significantdegradations when testing the 18 degree sweptairfoil used to represent the Viscount tailplane with aroughness height of only 1/1300 of the chordlength. 24 Lynch and Khodadoust have providedreferences and discussion of a wide range of dataaddressing roughnesses representitive of initial in-flight leading edge ice accretions.49 Not the least ofthis data is information from Abbott and vonDoenhoff, which has been around for over 50years.50


14

The reports were also surveyed for descriptorsof the ice accretion found after the event. In 71 ofthe 120 events, no descriptor is available. It must benoted that many of these reports are of the pre-1983format used by the NTSB, in which narrative text isextremely sparse. In the remainder, various wordsare used in the reports, such as white, milky, jagged,slushy, and granular. The term "rime" appears 11times in the reports of in flight events, and only oncein reports of ground events. The term "clear"appears 7 times for in flight reports, and also justonce for ground events. The term "mixed" appearsonly once in all of the reports; however, in theaforementioned rime and clear usages, these termsare used together 3 times.

Interestingly, the word "rough" is used on 8separate occasions to describe the ice found afterthe event; 5 of these cases were in flight events.Other descriptors of texture include "jagged","course", and "granular". Not too surprisingly, thereports of ground icing accidents use the term "frost"on 5 occasions and "snow" on 4.

Beginning with Samuels' work, then with theUnited 21 accident report, and continuing throughthe accident reports for Jetlink 2733, Simmons 4184and Comair 3272, the difficulty of seeing andcorrectly identifying ice formations has been welldocumented. Yet the paradigm emphasizes anevaluation of the ice shape as a means of estimatingrisk. The double ram's horn shape exhibited by formsof clear, or glaze, ice has always been described asthe most serious threat. But the ice formationsdescribed in the accident reports, as well as in flighttests or icing tunnel tests done to support accidentinvestigations, often do not bear much resemblanceto the classical training models. In fact, only oneaccident report actually described the observation ofa double horned shape (the Viscount ICTS events atColorado Springs and Willow Run). On the otherhand, a flight test conducted following the crash ofUnited 21 at Chicago produced:

"a clear glaze ice of approximately 1/4 inch inthickness, extending back 2-1/2 inches ontop and bottom of the center line of theleading edge of the de-icer boot, togetherwith a mass of rough ice extending from apoint behind the smooth ice over the capstrip and running back onto the wing properfor a distance of about 6 inches."15

Thus, while the concept of meteorologicallycorrect navigation can play a significant role in routeplanning and execution, its applicability to terminalarea operations is quite limited. More significantly,the concepts of duration and severity, while logical,are seriously biased by the notion that the ice shapemust attain a certain size before becoming a serious

threat. Finally, the expectation that a visualassessment of the ice formation can be madeaccurately predisposes the flight crew to drawquantitative and qualitative conclusions that aresimply not possible within the limits of humansensory perception.

But perhaps the most important concept calledinto question while evaluating these accidents andincidents is that of the "cumulative" threat. Nearly allof these events involve airplanes that were behavingmore or less normally up to a point in time, so far ascan be inferred from the reports. In 49 of the 120events , that point in time could be generallycorrelated with a change in the angle of attack ofeither the wing or the horizontal stabilizer.

Webster's defines the word "cumulative" as anadjective meaning "increasing with successiveadditions". The concept of icing as a cumulativethreat is consistent with early ideas such as "everydeposit upon the wings or parts of an airplane is notnecessarily hazardous", right on through the use ofprogressive descriptors of intensity based on liquidwater content. Indeed, relationships have beenestablished between drag and ice formation sizewhich show a relationship between ice shape sizeand drag change. Accidents such as RockyMountains 217 and the General Airways DC-3 atKerrville, Texas are strongly linked to these types ofeffects.

But there are other extremely importantrelationships. For example, while training materialnearly always emphasizes the reduction in stall angleor increase in stall speed, there is rarely anydiscussion about the shape of the lift curve at itspeak. The abruptness of the stall may beconsiderably different from that of the same wingwhen uncontaminated; conversely, it may be morebenign and thus less recognizable. The stall initiationand propagation may be wholly different from theclean wing, resulting in pitch characteristics notpreviously experienced by the flight crew. Theeffects of separated flow on control balance may besignificant.

Some significant relationships which aresupported by the accident history are discussedbelow.

Aerodynamic Relationships

A very significant relationship that is not partof the traditional model is that between theeffects of ice formations, the slope of the liftcurve and the shape of the lift curve after the liftpeak. The cumulative model suggests that liftloss is progressive, which would appear as achange in the shape and slope of the lift curve.In some cases, this may be true, but not always.


15

Broeren, et.al.,51 showed differences in thelift curves for a typical airfoil section, the NACA23012, when tested at high Reynolds numberswith a variety of intercycle ice shapes and astandard, distributed roughness. Figure 7ashows the results for the intercycle shapes;Figure 7b the results for the sandpaperroughness. Three things are important to thepilot. First, the lift degradation for a given iceshape is very dependent on angle of attack.Second, below a certain ice shape-dependentangle of attack, the contaminated lift curve andthe clean lift curve are virtually identical. Third,the behavior of the contaminated lift curve as itcrosses Cl max can vary significantly with the iceshape, and it can be quite dramatic.

The roughness results shown by Broerenagree well with the roughness results fromAbbott and von Doenhoff. In both cases, the liftcurve slopes do not depart significantly from theclean lift curve until within 2 to 3 degrees of thecontaminated Cl max, and then the departure isimmediate. This characteristic is not too disimilarto the clean stall. The lift curves with the zerodegree ice shapes, however, are quite a bitdifferent. While much greater lift loss is evident,the development of the lift loss is more benign.Were this to occur in a symmetrical manner to theentire wing, it might not be recognized in a timelymanner; more realistically, one or more lift curveswith the associated non-linearities will apply todifferent wing sections, depending on iceaccretion shape. Thus the entire stalled behavioris likely to be different from the crew's nominalexperience.

A second very important relationship is thatbetween the angle of attack at which the ice wasaccreted and the effects of the same ice formationwhen the angle of attack is increased. Several of theaccident reports in the 1960s cited work done byGray and von Glahn. In this work, the authors statedthat:

"A glaze ice formation on the leading edgesection for a simulated approach condition,during which the airfoil attitude is increasedfrom 2 to 8 [degrees] angle of attack, causeda severe increase in drag coefficient of over285 percent over the bare airfoil drag at 8[degrees] angle of attack and wasaccompanied by a shift in the position of themomentum wake that indicated incipientstalling of the airfoil." 26

The ice shapes tested by Broeren, et.al., in theNASA LTPT were all formed at a zero degree angleof attack in the BF Goodrich icing tunnel.51 The dragresults shown in Figure7 are dramatic. Although

these are section drag coefficients, and there arequestions regarding scale effects between the 36inch chord model used in these tests and a full scalewing, the effects of accreting ice at a low angle ofattack and then increasing that angle can be seenclearly.

Further, Broeren's results with the sandpaperroughnesses are consistent with the warning foundin Trunov and Ingelman-Sundberg:

"For thin but rough ice layers on the wingleading edge there can be a very largedisproportion between its effect on drag incruise and its effect on the maximum lift inthe landing case. An ice layer which has arelatively small effect on cruising speed canhave a large effect on the stall speed."52

The upshot of this is that there may be littlein the way of altered lift characteristics orsignificant, noticeable drag rises whencontinuing to operate at the angle of attack atwhich the ice shape was accreted. However,when the angle of attack is increased, significantdegradations can occur rapidly. If the angle ofattack is increased due to routine attitude and/orconfiguration changes, the cliff may beencountered before the magnitude of thedegradations can become clear. Consequently,the captain of Northwest 5 experienced normalbehavior during his descent only to be surprisedat the sudden stall when he leveled off, and theflight crew of Ozark 982 did not perceive anydegradation during their approach. This samecharacteristic has appeared many times, mostrecently with the Comair 3272 accident.

It follows, then, that the understanding of howangle of attack is controlled is critical to icingoperations. Conventional civilian pilot training placesnearly all of its emphasis on the control of angle ofattack through pitch attitude. Obviously, this is theprimary method, but in this simplification, the truenature of more subtle variations in angle of attack isoverlooked. Flight crews should first understand thatthey are constantly flying two wings, and two anglesof attack....that of the main wing, and that of thehorizontal stabilizer. They need to understand theeffects of all primary controls, vertical accelerations,ground effect, propeller slipstream, and, particularlyimportant, of all secondary controls, on the angle ofattack at each lifting surface. This point argues for amuch broader, more comprehensive education inthe function and operation of the airplane's liftingsurfaces than that produced by conventionaltraining, and that is what is needed. For example, theunderstanding of the effects of secondary flightcontrols is important because these controls aremost often motorized and usually preselected and


16

then automatically driven. This results in changes tothe angle of attack that are more or less forced. Whilethe scheduling of secondary flight controldeployment or retraction can be rescinded, this canrarely be done quickly in response to an undesirableaerodynamic effect. A perfect example of this can beseen in the Simmons 4184 accident at Roselawn, inwhich the flap retraction forced an increase in angleof attack, leading to flow separation. For that matter,all of the ICTS events fall into this category, as they alloccur following flap extension. In consideration ofthe sensitivity of a contaminated wing to angle ofattack changes, knowledge of how eachconfiguration change impacts the angle of attack atboth lifting surfaces is crucial.

A corrollary to the control of angle of attack is theunderstanding of flow separation and stall. All of theicing accidents evaluated involved separated flowand stall of some type. Conventional designpromotes a stall propagation which is intended toinsure that lateral control remains effective and thegeneral behavior of the wing is benign. Thecontaminated stall may develop very differently,leading to airplane behavior that is unusual in theexperience of the flight crew. The stall may initiate atthe leading edge instead of the more typical trailingedge. The stall may also initiate nearer the tip insteadof the root, leading to premature problems in lateralcontrol and substantial changes in pitching moment.

The effects of flow separation on lateral andlongitudinal stability are another part of this aspect.For example, the loss of longitudinal stability duringan instrument approach can lead to a disastrous pilotinduced oscillation. A good case in point is the UnionOil Convair 580 accident at Midland, Texas in 1963.In this case, the tail may never have actually stalled,but the flow separation may have led to a substantialreduction in longitudinal stability. Because the icecontaminated airfoil no longer retains the benigncharacteristics that were intended in the originaldesign, it is imperative that pilots understand howthe lifting surfaces may behave and what effect thismay have on the overall handling of the airplane. Thisis a subject that requires considerably more depththan the current discussion of whether or not stallwarning systems will function before the stall.

A final aspect of the flow separation discussion isthat of flight control balance. Without doubt, theflight crew of Simmons 4184 was not familiar withwhat could happen if the aileron hinge momentcoefficient was altered. The same is likely true forthose crews involved in ICTS events. A morecomprehensive understanding of how the flightcontrols are balanced and what aerodynamicconsiderations can change this balance is required.Because flight control designs vary, care must betaken to avoid inappropriate generalizations.

All of these relationships belie the myth that theeffects of icing are cumulative. Instead, they are verymuch a function of angle of attack, and,consequently, can present shifts and nonlinearitieswhich are sudden and catastrophic.

Ice Formation Descriptors

Beginning with the work done by Carrol andMacAvoy, the primary and virtually sole factor in theoperational description of inflight icing has beenwhether the shape is rime or clear. Around the timeof Samuels, the notion of what are today known as"mixed conditions" became known. Mixedconditions are defined as a combination ofsupercooled liquid precipitation and solidprecipitation. Sometime during the interveningperiod, the definition of mixed conditions appears tohave devolved into "mixed icing" which is generallyrecognized as a combination of rime and clear ice.

However, the accident record suggests that theutility of such terms in actual risk management isquestionable. Lynch and Khodadoust haveprovided what may be a much better organizationalhierarchy of icing shape descriptions. They discussfour categories of ice accretions: initial leading edgeaccretions, runback/ridge accretion, large (glaze)accretions, and ground frost.49

If, to this, three additions are made, then a muchbetter structure of descriptors for use in theeducation of the air carrier pilot will result. First, theterms "intercycle" and "residual" ice must be addedfor cyclic deice systems. Second, the runbackcategory must also include the ice developed by athermal system during the period that it runs wet.Finally, the effects of roughness, alone or incombination with any other shape, absolutely mustbe emphasized. These terms allow the pilot toevaluate the ice encountered with respect to boththe characteristics of the particular ice protectionsystem in use and, most importantly, with theaerodynamic aspects of ice shape and location.

Interpretation of Terminal SurfaceObservations

A final aspect of this analysis must bemeteorological. With respect to those events whichoccurred in close proximity to the surface (lowaltitude events), the reported surface weatherobservations were evaluated. For these events, theaverage surface temperature was -1.76C, with amedian of -0.28C (Figure 8). The range was from-9.0C to 2.8C. In the case of the ground icingaccidents, these numbers are not too different; theaverage surface temperature is -3.37C and themedian is -2.21C, with a range from -12.2C to 1.1C.


17

The cloud ceiling for the inflight events averaged1006 feet, with a median height of 650 feet and arange of 100 feet to 5000 feet. For the groundevents, it is hardly different. The average was 1123feet, the median 450 feet, and the range from 100 to5500 feet (Figure 9).

Most interesting was the precipitation data(Figure 10) In 27 of 69 inflight events for whichadequate data exists, snow was falling at the surface.In 7 of these snow cases, freezing drizzle was alsofalling concurrently. Freezing drizzle was falling alonein 10 additional cases - thus, freezing drizzle ispresent in 17 of the 69 inflight events. Freezing rainfell concurrently with snow in 2 cases, and fell alonein 1 case. In 2 other cases, sleet was reported to befalling alone. There was no precipitation reported in15 cases, and rain or drizzle prevailed in theremaining cases.

With regard to the ground icing accidents, snowwas falling in 21 of the 30 cases which presentedadequate data. Freezing drizzle was reported inisolation in 2 cases. The remainder involved noreported precipitation.

In 29 of the 47 in flight events which presentedenough data the reported precipitation intensity waslight; this was also true of 10 out of17 takeoff events.In 2 of the inflight events, moderate precipitation wasreported. This was the case with 2 of the 17 takeoffevents as well. Due to the format of a number ofthese reports, however, precipitation intensity wasnot recorded.

The in flight events yielded an average winddirection of 006 degrees and a median of 360degrees. The average wind velocity was 12.3 knots,and the median was 11.5 knots.

The inescapable conclusion from this data isthat, for air carrier operations, a surface observationof a ceiling below 1000 feet, a temperature of 0C or acouple of degrees less, a north wind of around 10 to15 knots and light snow or freezing precipitation is apretty good warning of serious icing potential duringthe approach and landing. Further, the sameconditions pose a serious threat with regard to theground deicing/takeoff situation.

It is interesting to consider that NACA publishedresearch papers and CAA documents of fifty yearsago, and FAA documents to this day, describe a sortof "special case" of icing when freezing or frozenprecipitation is mixed with liquid precipitation. Forexample, the current Advisory Circular 91-51a statesthat, "Ice particles become imbedded in clear ice,building a very rough accumulation."13 This isremarkably consistent with the frequency of freezingprecipitation and/or snow in the above accident andincident reports, and with the weak but noticableprevalence of the term "rough" and similar terms inthe icing descriptors used in the reports. Yet like the

large droplet environment, the mixed phaseenvironment is not part of FAR 25 Appendix C.

None of this is to say that other conditions arenot equally or perhaps more serious. However, thiswork has shown that the current paradigm in use bytransport aircraft pilots is essentially unchanged overthe course of 50 or 60 years. In that same timeperiod, presumably due to the influence of thatparadigm, these are the conditions in which air carrieraircraft seem to find the right combination ofexposure and susceptibility.

Application To Modern TransportOperations

The above relationships may form the coreknowledge of icing and its effects. However, the pilotmust also understand how to integrate thisknowledge into operations.

Aircraft Systems

It is important to first consider several aspects ofaircraft systems . These are:

1. Ice Protection System Design2. Flight Control Design3. Automated Flight Angle of AttackGuidance

It is important for the flight crew to understandthe design of the specific ice protection system theywill be using. In the case of pneumatic deicing bootsystems, this includes understanding the possibleeffects of pre-activation ice, intercycle ice andresidual ice. It involves understanding the limitationsof using the system throughout the landing,including what ice might be accreted after the systemis shut down prior to landing. In the case of thermalsystems, it includes understanding whether and/orwhen the system functions as a running wet systemand what the effects of runback ice might be. It alsoincludes understanding the power requirements foreffective operation of the thermal system, and whatchanges to configuration might be necessary tomaintain adequate power during the descent andapproach in icing conditions. Finally, it includesunderstanding the ramifications of a thermal systemwhich must be cycled, either automatically ormanually, between the wings and the tail and whatimplications that arrangement might have during theapproach. For example, Gray and von Glahn foundthat :

"the residual runback icing formed duringcycles at low angles of attack will determinehow closely the drag after an angle-of-attack


18

change to 8o will approach the bare wingdrag at 8o angle of attack following the heat-on period."26

It is also important for the flight crew tounderstand the design of the specific flight controlsused on their airplane, i.e., are they reversible orirreversible and, if reversible, how are they balanced?For example, besides aerodynamic balancing, hornbalances can respond poorly to ice accretion, andthe simple addition of hydraulic boosting does not initself make a control irreversible.

Finally, and perhaps the most subtle, is the issueof automated flight attitude guidance. This is bestseen in the Air Canada CL-600 accident atFredericton. Flight crews need to understand thatstall warning and recognition systems are not alwaysbiased for icing situations, and even when they are,they are biased on the basis of the effects of apresumed ice formation. But the Air Canada accidentserves to point out that there are other systems thatdo not recognize the ice condition. Most notably,this applies to automated go-around guidance. Theflight crew must consider ice formations when usingautomated pitch guidance, particularly when thatpitch guidance calls for significant increases in theangle of attack.

Operation of the IPS

All of this knowledge must then be integratedinto the planning of operations. First and foremost inthis planning is the vigilant operation of the airplane'sice protection system. In 55 of the 82 inflight iceaccretion events studied, an ice protection systemcould be identified. In 25 of these cases, the systemwas not operated. This is consistent with data fromthe larger database of 312 events worldwide. In thisdatabase, 143 events yielded information about theoperation of the IPS. In 30 of these cases, the pilotwas aware of icing conditions but the IPS was notoperated.

In many cases, the IPS probably was notoperated because it was a deice system, usuallypneumatic boots, and the manufacturer'sinstructions required delaying operation while an"observable thickness" developed. The discussionabove regarding how well transport pilots can see iceformations speaks to the fallacy of this. However, thepractice probably originated due to concerns aboutice bridging. The FAA Ice Bridging Workshop53

concluded that this phenomona is very unlikely tooccur with a rapid inflation/short duration type ofsystem typical today, and many manufacturers havechanged their procedures. It is very important forpilots operating pneumatic boots to understand therelationship between boot cycle times and icing

effects. For example, Figure 11 is an extract fromBowden's 1956 work.54 The sawtooth patternrepresents the cycle periods of a pneumatic system;the relationship between cycle period and the deltain aerodynamic coefficients can be seen. Theimportant aspect of this for the pilot is to understandthat delaying operation in order to obtain a clean,complete shedding of ice results in a higherpercentage of a given period flown with an iceaccretion, and the degradations themselves aregreater. A continuous cycle period, while perhapsnot as effective at achieving a complete shedding ofice, results in a lower percentage of a given periodflown with the ice accretion, and the degradationsare themselves less significant.

The notion of vigilant operation of the IPS mustbe integrated with the conclusion that most of theaccidents take place during the approach phase.This is a high workload phase, particularly during anapproach in icing (read: instrument) conditions. Manyairplanes do not have a well automated IPS, or haveone equipped with an automation schedule which ismay not dovetail well into the approach sequenceand workload. This is a situation which may lead tosome degree of complacency, particularly in theabsence of a reliable negative consequence.

The Takeoff

Enough has been said in many publicationsabout the function of deicing fluids and theprocedures to be used with them that little moreneed be added. However, curiously, nearly all of thisemphasis stops at brake release, so to speak. Verylittle is said in training material regarding ice accretionimmediately after takeoff.

It is important, however, for flight crews toconsider the icing situation immediately after liftoff.There will be a couple of forced changes in angle ofattack. The first change will come with departure fromground effect, and while that should not be an issueif adequate deicing has been accomplished, it isworth being aware of. The next change will comewith the first configuration change. This is usually aflap retraction. This has been associated with acouple of the Cessna 208 accidents. Shortlythereafter may come a slat retraction. Both of thesesubstantially change the wing characteristics and theangle of attack. In many cases, they come in closeproximity to the minimum altitude for IPS operation.Between the time of liftoff and that minimum altitude,it is possible for sufficient ice to accrete to causeproblems. It may be wise to delay reconfiguring theairplane, in certain cases, until the IPS has beenoperated. It is also important to consider the variousangle of attack changes that may come with a climbprofile.


19

The Stabilized Approach

A third aspect of this integration is based on theconclusion from the air carrier data that an icing eventrarely results from a stabilized approach. FAA Order8400.10 states that "Maintaining a stable speed,descent rate, vertical flight paths, and configurationis a procedure commonly referred to as the stabilizedapproach concept."55 In 60 of the 82 inflight icingcases evaluated, the event took place during theapproach phase. In only 6 of these cases could it beconcluded that the approach was truly stabilized. In33 cases, it was not stabilized; for example, in 9 ofthe cases, the aircraft was not configured in a timelymanner. Typically, this involves the ICTS events inwhich the flaps are lowered on very short final; thisremains a standard procedure for turboprop aircraft.In 10 of the cases, the aircraft was executing sometype of circling maneuver. In an additional 8, theaircraft was flying some type of non-precisionapproach which required intermediate level offs (thisset might be arguable as to whether the approachwas "stabilized" by conventional definition; however,even a well flown non-precision approach lacks someof the elements of stabilization considered essential,which is one reason why such approaches are notdesirable for transport operations). In 3 cases, theaircraft was below the glide slope; in 2 others, it wasabove the glide slope. Interestingly, in the oneknown case of an MD-80 aircraft experiencing anICTS event, the aircraft was recovered largelybecause it was configured early in preparation for astabilized approach profile.

Thus, the flight crew should seriously considerthe implications of icing on the type of approach thatthey are planning to fly. Configuration changes andcircling maneuvers close to the ground should beavoided. Attitude changes, with the consequentchanges in angle of attack, should be planned andmanaged carefully. It may well be appropriate toartificially increase the approach weather minimumsin the case of some non-precision approaches toaccommodate more conservative configuration andattitude changes. This notion is particularly importantwith regard to "fly-up" glide path corrections.

A case in point is TWA 924 at Gander in March of1949. Although the investigation did not citestructural icing, the airplane was flying a groundcontrolled approach (GCA) in icing conditions thatincluded freezing drizzle. The aircraft impactedground structures while correcting back up to theglide path.55 Other cases in which this couldconceivably have played a role are the HibbingJetstream accident and the EMB-110 accident atAlpena, Michigan in 1985. At Alpena, the aircraft wasalso flying an ILS in icing conditions that likelyincluded freezing drizzle. It crashed short of therunway with no apparent explanation except that the

flight crew was not monitoring the altitude.56

However, accidents such as Comair 3272 andNorthwest 5 have shown how rapidly degradationscan arise with increases in angle of attack.

Finally, the aforementioned FAA Order 8400.10states that a visual or circling approach should bestabilized by 500 feet above the surface, and anapproach in instrument conditions should bestabilized by 1000 feet above the surface. 57 Thedocument states that, "Operational experience hasshown that the stabilized approach concept isessential for safe operations with turbojet aircraft,and it is strongly recommended for all other aircraft."The icing event data suggests that, at the very least,an approach with ice accretion on the aircraft shouldbe stabilized by 1000 feet above the surface underany circumstances, visual or otherwise. Further, theicing event data seems to provide "operationalexperience" which shows that with ice accretion, astabilized approach is essential for any type aircraft,regardless of powerplant design.

The Go-Around

A fourth aspect is an extrapolation based on thenumber of approach events. The concentration ofevents within the approach phase suggests thatthere is a consequent exposure to the go-aroundphase. Continental 290 at Kansas City, Northwest324 at Sandspit, and Air Canada 646 at Frederictonall took place within the go-around phase. Theaccident report for Air Canada 571 at Edmondtonstated that "had the captain decided to overshootduring the final stages of the approach, the icecontamination would have been a detrimental factorin obtaining the lift performance required.".Consequently, operational planning for the go-around must take serious consideration of the icingsituation. Both Air Canada events involved ceilingsof 100 feet and visibilities of 1/4 mile or less. Suchconditions increase the liklihood of a go-aroundconsiderably. Planning for such a go-around mustinclude consideration of the ice contaminationeffects on required climb performance (includingengine-out performance), configuring so thatadequate bleed air is available during the final stagesof the approach to maintain as clean a wing aspossible, and considering the effects of go-aroundflight guidance, rotation rates, and configurationsequences on an ice contaminated wing and/orstabilizer.


20

Conclusion

The origins of the current operationalunderstanding of the structural icing hazard can betraced back to the years before World War II. Thecurrent paradigm is based on some generalizationswhich easily lead to the misinterpretation ofexperience. The major tenets of this paradigm, asdefined in the late 1930s, continue to be the areasof primary emphasis, while other aspects that havedeveloped during the intermediary period have notbeen integrated into the paradigm. Generally, theindustry has failed to capture many of the lessonslearned in the accident record. Only when a series ofcontemporary, high profile accidents occur is theparadigm modified, and then only by means ofbolted-on additional modules, such as those seen inthe last ten years regarding ground icing and SLDconditions.

In particular, the concept of icing as a cumulativehazard is misleading. The accident data shows thatthe icing hazard is strongly influenced by the angleof attack and characterized by the nature of lift curvebeyond the contaminated Cl max, evaluated acrossthe wing span and at the horizontal stabilizer. Thisbehavior is nonlinear and consequently not welldescribed by the cumulative concept. Further,accident data and research data, some of it datingback many years, indicate that the relationshipbetween the icing hazard and the quantity of the iceor the size of the formation is not linear, and thatsignificant hazard exists due to thin, rough iceaccretions.

Therefore, the paradigm must shift from onewhich is based on the concept of a cumulativehazard to one which recognizes the potential of verysmall, thin ice accretions and which emphasizes therole that angle of attack plays. While the currentemphasis on the avoidance of flight through icingconditions is an important part of the paradigm,greater emphasis on the evaluation of approachconditions is required. Finally, the air carrier pilot'sunderstanding of ice formations must integrateshape, roughness and location with the nature of iceformations expected before, during and followingoperation of the ice protection system, including pre-activation, intercycle, residual and runback iceformations.

Icing is widespread enough, and sufficientlydifficult to forecast or preemptively detect, that anunderstanding of its effects on the flightcharacteristics and handling qualities of the airplaneis critical. Perhaps no other hazard does more todisrupt the aerodynamic predictability of the flightvehicle while at the same time being so commonlyencountered. It simply will not tolerategeneralizations that are designed to simplify thediscussion or to reduce training expenditure. In

order for the pilot to operate competently in theseconditions, he must have a comprehensiveunderstanding which allows him to make wisedecisions, substantiate those decisions, andcorrectly interpret his operational experience as itdevelops.


21

References

1 Gullberg, Jan, "Mathematics From the Birth ofNumbers," W.W. Norton, New York, 1997.

2 Carroll, T., and McAvoy, W.H., "The Formationof Ice Upon Exposed Parts of an Airplane inFlight", NACA Report 293, 1928.

3 Carroll, T., and McAvoy, W.H., "The Formationof Ice Upon Airplanes in Flight", NACA Report313, 1929.

4 Samuels, L.T., "Meteorological ConsiderationsDuring the Formation of Ice on Aircraft", NACAReport 439, 1932.

5 Noth, E., and Polte, W. "The Formation of IceOn Airplanes", NACA Report 786, 1936.

6 Findeisen, W., "Meteorological-PhysicalLimitations of Icing in the Atmosphere", NACAReport 885, 1939.

7 French Committee for the Study of IceFormation, "Report on Ice Formation on Aircraft",NACA Report 919, 1939.

8 U.S. Department of Commerce, CivilAeronautics Regulations, Part 61, "ScheduledAirline Rules (Interstate)", as Amended May 31,1938.

9 U.S. Department of Commerce, CivilAeronautics Bulletin No. 25, "Meteorology forPilots", January, 1943.

10 U.S. Department of Commerce, CivilAeronautics Administration Technical ManualNo. 104, "Pilots' Weather Handbook",December 1955.

11 U.S. Department of Transportation, FederalAviation Administration Advisory Circular 00-6A,"Aviation Weather For Pilots and FlightOperations Personnel", 1975.

12 Newton, D., "Severe Weather Flying",Aviation Supplies and Academics, Inc., 1991.

13 U.S. Department of Transportation, FederalAviation Administration Advisory Circular 91-51A, "Effect Of Icing On Aircraft Control AndAirplane Deice And Anti-Ice Systems", 1996

14 Lewis, W., "A Flight Investigation of theMeteorological Conditions Conducive to the

Formation of Ice on Airplanes", NACA Report1393, 1947..

15 Report of the Civil Aeronautics Board - "Of theinvestigation of an accident involving civil aircraftof the United States NC25678 which occurrednear Chicago, Illinois on December 4, 1940" ,Release undated.

16 Flight Safety Digest, "Icing-related CommercialAviation Accidents, 1946–1996 (Summaries)",Vol. 16 No. 6/7/8/9 , June–September 1997.

17 Report of the Civil Aeronautics Board - "Of theinvestigation of an accident involving civil aircraftof the United States NC21712 which occurrednear Moorhead, Minnesota on October 30,1941", Released January 31, 1942.

18 Civil Aeronautics Board, AccidentInvestigation Report, "Mid-Continent Airlines,Inc., Sioux City, Iowa, March 2, 1951", ReleasedSeptember 27, 1951.

19 Civil Aeronautics Board, AccidentInvestigation Report, "Zantop Flying Service,near Kansas City, Missouri, January 20, 1954",Released June 29, 1954.

20Civil Aeronautics Board, Aircraft AccidentReport, "Hansen Air Activities, Douglas DC-3A,N410D, Near Chicago O'Hare InternationalAirport, Chicago, Illinois, March 8, 1964",Released February 24, 1965.

21 Civil Aeronautics Board, Aircraft AccidentReport, "Frontier Airlines, Inc.,Douglas DC-3C,N61442, Miles City, Montana, March 12, 1964",Released March 11, 1965.

22 Civil Aeronautics Board, Aircraft AccidentReport (revised), "Capital Airlines, Inc., Viscount,Tri-City Airport, Freeland, Michigan, April 6,1958", Released February 17, 1965.

23 Civil Aeronautics Board, Aircraft AccidentReport, "Continental Airlines, Inc., VickersViscount 812, N242V, Municipal Airport, KansasCity, Missouri, January 29, 1963", ReleasedJune 17, 1964.

24 Trunov, O.K., and Ingelman-Sundberg, M.,"Wind Tunnel Investigation of the HazaradousTail Stall Due to Icing", The Swediah-SovietWorking Group on the Scientiufic-TechnicalCooperation in the Field of Flight Safety, ReportJR-2, 1979.


22

25 Civil Aeronautics Board, Aircraft AccidentReport, "Slick Airways Division, The SlickCorporation, Douglas C-54B-DC, N384, CastleIsland, Boston, Massachusetts, March 10,1964", Released November 5, 1964.

26 Gray, V.H. and von Glahn, U.H., "Effect of Iceand Frost Formation on Drag of NACA 651-212Airfoil for the Various Modes of Thermal IceProtection", NACA Report 2962, June, 1953.

27 Ratvasky, T.P., van Zante, J., and Riley, J.,"NASA/FAA Tailplane Icing Program Overview",NASA/TM—1999-208901.28 National Transportation Safety BoardAccident/Incident Database Report NumberCHI90IA106

29 NASA Aviation Safety Reporting System,Report 392928.

30 Civil Aeronautics Board, AccidentInvestigation Report, "Seattle Air Charter,Seattle, Washington, January 2, 1949",Released May 4, 1949.

31 Civil Aeronautics Board, AccidentInvestigation Report, "Monarch Air Service,Chicago, Illinois, January 4, 1951", ReleasedOctober 22, 1951.

32 National Transportation Safety Board AircraftAccident Report, " NTSB AAR 79-12", ReleasedAugust 16, 1979.

33 National Transportation Safety Board AircraftAccident Report, "NTSB AAR 82-8", ReleasedAugust 10, 1982.

34 National Transportation Safety Board AircraftAccident Report, "NTSB AAR 88-09", ReleasedSeptember 27, 1988

35 National Transportation Safety Board AircraftAccident Report, "NTSB AAR 93-02", AdoptedFebruary 17, 1993.

36 Civil Aeronautics Board, Civil Air Regulations,Part 40, "Scheduled Interstate Air CarrierCertification and Operation Rules", EffectiveOctober 1, 1953.

37 Code of Federal Regulations, Chapter 14,Subchapter G, Part 121, "OperatiingRequirements: Domestic, Flag, andSupplemental Operations", January 24, 2002.

38 National Transportation Safety Board AircraftAccident Report, "File No. 1-0039", AdoptedSeptember 2, 1970.

39 National Transportation Safety Board AircraftAccident Report, "AAR 91-09", AdoptedNovember 16, 1991.

40 National Transportation Safety Board AircraftAccident Report, "AAR 96-01", Adopted July 9,1996.

41 NASA Aviation Safety Reporting System,Report 287615.

42 National Transportation Safety Board AircraftAccident Report, "AAR 98-04", AdoptedNovember 4, 1998.

43 National Transportation Safety Board AircraftAccident Report, "AAR 79-6", May 3, 1979.

44 National Transportation Safety Board AircraftAccident Report, "AAR 81-3", Adopted April 10,1981.

45 Transportation Safety Board of Canada,Aviation Occurrence Report, "Report No.89w0066".

46 Transportation Safety Board of Canada,Report Number A96A0035 .

47 Transportation Safety Board of Canada,Report Number A97H0011.

48 National Transportation Safety Board AircraftAccident Report, "AAR 94-05", Adopted May24, 1994.

49 Lynch, F.T., and Khodadoust, A., "Effects ofIce Accretions on Aircraft Aerodynamics",Progress in Aerospace Sciences, January 2002.

50 Abbott, I.H., and von Doenhoff, A.E., "Theoryof Wing Sections", Dover Publications, 1959.

51 Broeren, A., Addy, H.E., and Bragg, M.B.,"Effect Of Intercycle Ice Accretions On AirfoilPerformance", AIAA-2002-0240, 2002.

52 Trunov, O.K., and Ingelman-Sundberg, M,"Methods for the Prediction of the Influence ofIce on Aircraft Flying Characteristics", TheSwediah-Soviet Working Group on theScientiufic-Technical Cooperation in the Field ofFlight Safety, Report JR-1, 1977.


23

53 Proceedings, FAA Airplane Deicing Boot IceBridging Workshop, Ohio Aerospace Institute,November, 1997.

54 Bowden, D.T., "Effect of Pneumatic De-Icersand Ice Formations on AerodynamicCharacteristics of an Airfoil", NACA TN 3564,1956.

55 Civil Aeronautics Board, Aircraft AccidentReport, "Transcontinental and Western Air, inc.,

Gander, Newfoundland, March 2, 1949"Released February 1, 1950.

56 National Transportation Safety Board AircraftAccident Report, "AAR 87-02", AdoptedFebruary 18, 1987.

57 U.S. Department of Transportation, FederalAviation Administration, Order 8400.10, "AirCarrier Inspector's Handbook".

.

INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR … · INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR...

Documents

Transcript of INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR … · INFLIGHT ICING EDUCATIONAL OBJECTIVES FOR AIR...