A human factors analysis of general aviation accidents in Alaska versus the rest of the United...

8/12/2019 A human factors analysis of general aviation accidents in Alaska versus the rest of the United States

1/8

A human factors analysis of general aviation accidents in Alaska versus the rest of theUnited States

Cristy Detwiler, 1 Carla Hackworth, 1 Kali Holcomb, 1 Albert Boquet, 2 Elaine Pfleiderer, 1 DouglasWiegmann, 3

and Scott Shappell4

Federal Aviation Administration 1 , Embry-Riddle Aeronautical University 2 , Mayo Clinic 3 , Clem-son University 4

General aviation (GA) accidents that occurred in Alaska versus the rest of the UnitedStates were compared using the Human Factors Analysis and Classification System(HFACS). Overall, categorical differences among unsafe acts (decision errors, skill-basederrors, perceptual errors, and violations) committed by pilots involved in accidents inAlaska and those in the rest of the U.S. were minimal. However, a closer inspection ofthe data revealed notable variations in the specific forms of unsafe acts within the acci-dent record. Specifically, skill-based errors associated with loss of directional control

were more likely to occur in Alaska than the rest of the U.S. Likewise, the decision toutilize unsuitable terrain was more likely to occur in Alaska. Additionally, accidents inAlaska were associated with violations concerning VFR into IMC. These data providevaluable information for those government and civilian programs tasked with improvingGA safety in Alaska and the rest of the US.

INTRODUCTIONConsiderable effort has been expended

over the last several decades to improvesafety in both military and commercial avia-tion. Even though many people have died

and millions of dollars in assets have beenlost, the numbers pale in comparison tothose suffered every year within generalaviation (GA). For example, according tothe National Transportation Safety Board(NTSB), there were 1,741 GA accidents in2003 that resulted in 629 fatalities (NTSB,2005). While the numbers may not registerwith some, when considered within the con-text of commercial aviation, the losses suf-fered annually by GA are roughly equivalent

to the complete loss of three commercial passenger Boeing 727s.Why then has GA historically received

less attention? Perhaps it has something todo with the fact that flying has become rela-tively common as literally millions of trav-elers board commercial aircraft daily to getfrom place-to-place. Not surprisingly then,

when a commercial airliner crashes, it in-stantly becomes headline news, shaking theconfidence of the flying public.

In contrast, GA accidents happen virtuallyevery day yet they receive little attention

and seldom appear on the front page of USAToday . Perhaps this is because they happenin isolated places, involving only a couple ofunfortunate souls at a time. In fact, unlessthe plane crashed into a school, church, orsome other public venue, it is unlikely thatanyone outside the local media, government,or those intimately involved with the acci-dent even knew it happened.

Over the last couple of years, GA has de-servedly received increasing attention from

the FAA (FAA Flight Plan 2004-2008) andother safety professionals. Indeed, severalgroups from the government (e.g., theFAAs Civil Aerospace Medical Institute;

National Institute of Occupational Safetyand Health), private sector (e.g., the Medal-lion Foundation), and universities (e.g.,University of Illinois, Johns-Hopkins Uni-


2/8

versity) have conducted a number of studiesexamining GA accident causation.Alaskan Aviation

It is of note that many of these efforts havefocused on Alaska, where aviation is the

primary mode of transportation. Alaska isknown for its varied and often unique land-scape and when this is considered with tem-

peramental weather and seasonal lightingconditions, even the most experienced pilotwould have to agree that Alaskan aviationrepresents some of the most difficult flyingin the U.S., if not the world. The combina-tion of factors mentioned above, the numberof GA accidents that are occurring in Alaskaand the FAAs accident reduction goal

(FAA Flight Plan 2004-2008) were factorsin our decision to implement this study.Human Error and General Aviation

A variety of studies have been conductedin an attempt to understand the causes of GAaccidents. Most have focused on contextualfactors or pilot demographics, rather thanthe underlying causes of the accidents.When the leading cause of accidents, humanerror, has been addressed, it is often only toreport the percentage of accidents associatedwith aircrew error in general or to identifythose where alcohol or drug use occurred.What is needed is a thorough human erroranalysis. Previous attempts to do just thathave met with limited success (OHare,Wiggins, Batt, & Morrison, 1994; Wieg-mann & Shappell, 1997). This is primarily

because human error is influenced by a vari-ety of factors that are usually not addressed

by traditional classification schemes (Shap- pell & Wiegmann, 1997). Yet, with the de-velopment of the Human Factors Analysisand Classification System (HFACS) previ-ously unknown patterns of human error inaviation accidents have been uncovered(Shappell & Wiegmann, 2001; Wiegmann &Shappell, 2001a).

METHODGA accident data from calendar years

1990-2002 were obtained from databasesmaintained by the NTSB and the FAAs Na-tional Aviation Safety Data Analysis Center

(NASDAC). In total, 24,978 GA accidentswere extracted for analysis. Only accidentsoccurring during 14 CFR Part 91 operationswere included (22,987 cases). This analysiswas primarily concerned with powered air-craft and thus the data were further re-stricted to include only accidents involving

powered fixed-wing aircraft, helicopters,and gyrocopters. The remaining 22,248 ac-cidents were then examined for aircrew-related causal factors. In the end, 17,808 ac-

cidents were included in the database thatwere associated with some form of humanerror and submitted to further analyses usingthe HFACS framework.

RESULTSWhen using HFACS to examine the GA

accident data, the majority of the accidentsare coded with either a precondition for un-safe acts or an unsafe act. This is due pri-marily to the fact that there is typically notmuch of an organizational structure or su-

pervisory influence on the majority of GA pilots, as compared to their counterpartsconducting commercial or for hire opera-tions.

Indeed, with few exceptions (e.g., flightinstructors and flight training institutions),the top two tiers of HFACS (unsafe supervi-sion and organizational influences) remainedsparsely populated when examining the GAaccidents leaving the majority of causal fac-tors within the bottom two tiers of HFACS.Consequently, the balance of this report willfocus only on the unsafe acts of the operatorlevel of the HFACS framework.Unsafe Acts of Operators (Aircrew)

An overall review of the GA accidentdata yielded the following results (see Fig-


3/8

ure 1). The most prevalent error noted in theaccident data over the past decade was skill-

based errors (73%), followed by decisionerrors (28%), violations (13%), and percep-tual errors (7%). 1 The relatively flat lines in

the types of unsafe acts across the yearssuggest that past intervention strategies havehad little differential impact on any particu-lar category of error.

0

10

20

30

40

50

60

70

80

90

100

'90 '91 '92 '93 '94 '95 '96 '97 '98 '99 '00 '01 '02

Skill-basedErrors

PerceptualErrors

DecisionErrors

Violations P e r c e n

t a g e o

f A c c

i d e n

t s

Year Figure 1. Overall review of general aviationdata for HFACS unsafe acts.

To obtain a better sense of how humanerror differences between Alaska and therest of the United States (RoUS) are repre-sented in the data, the error types were bro-ken out accordingly (Figure 2). The analysisof the unsafe acts revealed that there wereslightly more decision errors, fewer skill-

based errors, perceptual errors and violationsin Alaska than there were in the RoUS.

Note, the following analyses did not dis-tinguish between those pilots who were na-tive to Alaska and were involved in an acci-dent versus those who were less familiarwith the state. That being said, the numbers

for Alaska reflect the accidents that occurredwithin the physical boundaries of the state.

1 These percentages do not add up to 100 because an accidentcould be assigned more than one HFACS code (i.e., DE, SBE, PE,etc..).

01020

304050607080

90100

Skill-basedError

DecisionError

PerceptualError

Violation

P e r c e n

t a g e o

f A c c

i d e n

t sAlaskaRest of US

Figure 2. Percentage of accidents associatedwith each of the unsafe acts of the operator.

Skill-based Errors . Differences that ex-isted between Alaska and the RoUS werefairly consistent across the years of study,with slightly more skill-based errors associ-ated with accidents in the RoUS (see Figure3). The only exception involved 1991, 1996,and again in 2002 where the percentageswere nearly equal.

0

10

20

30

40

50

60

70

80

90

100

'90 '91 '92 '93 '94 '95 '96 '97 '98 '99 '00 '01 '02

P e r c e n

t a g e o

f A c c

i d e n

t s

Year

AlaskaRest of US

Figure 3. Skill-based errors broken out byAlaska versus the RoUS.

Differences between Alaska and theRoUS were more distinct when the actualtype of skill-based error was compared (Ta-

ble 1). For instance, directional control wasthe most frequently cited skill-based errorfor both Alaska (19%) and for the rest of theU.S. (13%). Pilots in Alaska were morelikely to experience a loss of directional


4/8

control of their aircraft than those in the restof the U.S. (odds ratio = 1.593, 2 = 33.400,

p


5/8

Plan-ning/Decision

19(4.0%)

Reme-dial Ac-tion

339(6.3%)

Go-around

18(3.8%)

Go-around

336(6.3%)

Violations . In general, violations wereassociated with less than 20% of GA acci-dents (Figure 5). For the entire U.S. sample,nearly 50% of these accidents resulted in afatality. When examining accidents inAlaska separately from the RoUS, differ-ences were found. Accidents involving vio-lations in Alaska were 9 times more likely toresult in a fatality (odds ratio = 9.248, 2 =127.606, p


6/8

mann, 2003; Wiegmann & Shappell, 2001b;2003).

The accident data suggest that aircrafthandling should be taken into account whendetermining where interventions should be

applied. For instance, any training (both abinitio and recurrent) along these lines shouldinclude control of the aircraft on the ground(e.g., ground loops), crosswind landings,avoiding and recovering from stalls, andgeneral control of the aircraft in flight.Given the inherent risk associated with someof these maneuvers, it makes sense to utilizemodern simulators during this training. Un-fortunately, it is unclear whether therewould be adequate transfer of training for

these specific tasks to make simulation train-ing viable. Therefore, before utilizing simu-lation to address these issues, research needsto be conducted to determine the best rolesimulators might play. In the meantimehowever, it appears necessary to emphasizethese topics during actual in-flight training.

The only notable exception among theHFACS casual categories involved decisionerrors. Specifically, pilots in Alaska weremore likely to utilize unsuitable terrain for

landing, taxi, and takeoff. It would appearthat educating aviators on the hazards ofutilizing frozen rivers or gravel bars, for ex-ample, may reduce these types of errors.However, it may be that there are simplymore improved areas in the RoUS, provid-ing pilots with more options in case of anemergency (i.e., alternate airports, high-ways, roads, etc.) in which case education inand of itself may not prove successful. Addi-tionally, it is worth noting that unsuitable

terrain was a classification imposed by the NTSB investigators after the fact, and themoment-to-moment judgment of how suit-able terrain may be during a flight may beinfluenced by factors not considered fully in

post hoc analyses.Also of concern in both Alaska and the

rest of the U.S. was in-flight plan-

ning/decision making. After all, decisionsmade during flight are often more criticalthan those occurring on the ground. Thus,when confronted with important decisionsduring flight, pilots are often under pressure

to be right the first time while using limitedinformation. Scenario-based training alongthese lines like that provided within theFAA-Industry Training Standards (FITS)

program may improve decision-making inthe cockpit, particularly if examples aredrawn from the accident record.

Of the unsafe acts that aircrew commit,addressing violations may be the most diffi-cult and complex. Recall that violations arethe willful disregard for the rules and as

such are not necessarily something that can be easily deterred or mitigated. Neverthe-less, since nearly half of violations involvedfatalities, behaviors like VFR flight intoIMC are of great concern to the FAA andother aviation safety professionals.

Even though the percentage of accidentsassociated with violations did not differmarkedly between Alaska and the RoUS, thespecific types of violations did differ inmeaningful ways. In particular, when inten-

tional VFR flight into IMC and other ad-verse weather conditions were combined, analarming 47% of the violations occurring inAlaska were accounted for (27% for the restof the U.S.). Exactly why a larger proportionwas observed in Alaska remains unknown,

but one reason may be the rapid weatherchanges that often occur, especially aroundmountainous areas.

Current interventions like weather cam-eras in mountain passes and other locationshave proved useful by providing pilots withaccess to real-time weather information andtherefore allowing them to make informeddecisions. In addition, the Medallion Foun-dation has provided GA pilot training usinghigh-resolution flight simulators capable of

producing simulated weather and lightingconditions and terrain depictions which are


7/8

all appropriate to Alaska. With this technol-ogy, pilots are able to safely navigatethrough Alaska and see what flying through

places such as Merrill Pass in adverseweather conditions could entail, a difficult

task to successfully perform in clear condi-tions.Alaska, as perhaps the FAAs largest

aviation laboratory, has been the testbed foradvanced avionics like those associated withthe Capstone project. Enhanced weather ra-dar, global positioning sensors, AutomatedDependent Surveillance Broadcast (ADS-B), and other cutting-edge technologies pro-vide a more accurate picture of how theweather, terrain and traffic situation actually

look from inside the cockpit. These tech-nologies have proven useful with 14 CFRPart 135 (commuter) operations (Williams,Yost, Holland, & Tyler, 2002). However,their efficacy within GA remains to be seen.

CONCLUSIONSIn recent years, a growing concern has

been directed toward GA accident rates. TheFAA Administrator has set a goal of a 20%reduction in GA accidents by fiscal year2008. If this goal is to be realized, interven-tions that target the underlying humancauses as identified in this analysis need to

be developed. Only then can any greatstrides in improving the GA accident rate beachieved.

REFERENCESFederal Aviation Administration. (n.d.) FAA

Flight Plan 2004-2008. Retrieved Au-

gust 24, 2004, fromhttp://www.faa.gov/AboutFAA/FlightPlan.cfm.

Federal Aviation Administration. (n.d.)FAA-Industry Training Standards(FITS). Retrieved July 29, 2005, fromhttp://www.faa.gov/education_research/training/fits/

National Transportation Safety Board.(2005). Table 10. Accidents, Fatalities,and Rates, 1985 through 2004, U.S.General Aviation. Retrieved July 12,2005, from

http://www.ntsb.gov/aviation/Table10.htmOHare, D., Wiggins, M., Batt, R. & Morri-

son, D. (1994). Cognitive failure analy-sis for aircraft accident investigation.

Ergonomics, 37 (11), 1855-1869.Shappell, S. & Wiegmann, D. (1997). Why

would an experienced aviator fly a per-fectly good aircraft into the ground? In-ternational Symposium on Aviation Psy-chology , The Ohio State University, 26-

32.Shappell, S. & Wiegmann, D. (2001). Be-

yond Reason: Defining the holes in theSwiss Cheese. Human Factors in Avia-tion Safety, 1 (1), 59-86.

Shappell, S. & Wiegmann, D. (2003). A hu-man error analysis of general aviationcontrolled flight into terrain accidentsoccurring between 1990-1998. (DOT/FAA/AM-03/4). Washington, DC:Federal Aviation Administration, Officeof Aerospace Medicine.

Wiegmann, D.A. & Shappell, S.A. (1997).Human factors analysis of post-accidentdata: Applying theoretical taxonomies ofhuman error. The International Journalof Aviation Psychology, 7 (1), 67-81.

Wiegmann, D.A. & Shappell, S.A. (2001a).Human error perspectives in aviation.The International Journal of AviationPsychology, 11 , 341-357.

Wiegmann, D.A. & Shappell, S.A. (2001b).A human error analysis of commercialaviation accidents using the human fac-tors analysis and classification system(HFACS). (DOT/FAA/AM-01/3). Wash-ington, DC: Federal Aviation Admini-stration, Office of Aviation Medicine.


8/8

Wiegmann, D. & Shappell, S. (2003). A hu-man error approach to aviation accidentanalysis: The human factors analysisand classification system. Ashgate Pub-lishing Company: Aldershot, Great Brit-

ain.Williams, K., Yost, A., Holland, J. & Tyler,

R. (2002). Assessment of advancedcockpit displays for general aviationaircraft The Capstone Program. (DOT/FAA/AM-02/21). Washington,DC: Federal Aviation Administration,Office of Aerospace Medicine.

A human factors analysis of general aviation accidents in Alaska versus the rest of the United...

Documents

Transcript of A human factors analysis of general aviation accidents in Alaska versus the rest of the United...