How many accidents is a collision? Hans de Jong Eurocontrol Safety R&D Seminar, Southampton, 24.10.2008
Introduction Interesting about moving is to experience people have different views Even more interesting is that local views are often perceived as obvious The number of accidents that occur when aircraft collide is an example a quick look through easily available documents confirmed this picture Interpretation 1 ( occurrence perspective ) a collision between aircraft, where the accident definition applies to at least one of the involved aircraft, is counted as one accident Interpretation 2 ( aircraft perspective ) a collision between aircraft constitutes as many accidents as there are aircraft involved for which the accident definition applies could be two accidents Most documents do not make their interpretation explicit Hans de Jong, Page 2
Contents Introduction Consulting various ICAO sources Annex 13 ADREP Annual Statistics Annex 11 and Doc 4444 Manuals on Airspace Planning, RVSM and A-SMGCS What interpretation does ESARR 4 use? ESARR 4 SRC POLICY DOC 1 Other SRC Publications Conclusions Hans de Jong, Page 3
Consulting various ICAO sources Annex 13 Annex 13 defines an aircraft accident as: An occurrence associated with the operation of an aircraft which takes place between the time any person boards the aircraft with the intention of flight until such time as all such persons have disembarked, in which: a) a person is fatally or seriously injured as a result of: being in the aircraft, or direct contact with any part of the aircraft, including parts which have become detached from the aircraft, or direct exposure to jet blast, except when the injuries are from natural causes, self-inflicted or inflicted by other persons, or when the injuries are to stowaways hiding outside the areas normally available to the passengers and crew; or b) the aircraft sustains damage or structural failure which: adversely affects the structural strength, performance or flight characteristics of the aircraft, and would normally require major repair or replacement of the affected component, except for engine failure or damage, when the damage is limited to the engine, its cowlings or accessories; or for damage limited to propellers, wing tips, antennas, tires, brakes, fairings, small dents or puncture holes in the aircraft skin; or c) the aircraft is missing or is completely inaccessible. Note 1. For statistical uniformity only, an injury resulting in death within thirty days of the date of the accident is classified as a fatal injury by ICAO. Note 2. An aircraft is considered to be missing when the official search has been terminated and the wreckage has not been located. Hans de Jong, Page 4
Consulting various ICAO sources Annex 13 The definition seems to have been written with a single aircraft in mind possibility of more than one aircraft associated with occurrence not indicated Points towards Interpretation 2, but leaves room for Interpretation 1 the occurrence of a collision could be associated with both aircraft involved In favour of Interpretation 2: definition is in terms of the end condition of the aircraft or associated persons artificial to consider e.g. a midair as one accident, whereas e.g. two CFITs resulting in the same end conditions would be counted as two accidents Nevertheless, Annex 13 provides no decisive answer to the question Hans de Jong, Page 5
Consulting various ICAO sources ADREP Statistics Überlingen is stated as follows in Part VIII List of accidents involving passenger fatalities in scheduled and non-scheduled operations in 2002 : Date State-Location Registration Aircraft Passenger fatalities 02-Jul-01 GERMANY- UEBERLINGEN RA-85816 TUPOLEV-TU- 154M/TU-164 Crew fatalities Remarks 57 12 Collision with scheduled cargo/ no passenger fatalities on cargo aircraft The cargo aircraft involved is not listed as there were no passenger fatalities Regarding Linate, ADREP Annual Statistics 2001 state Date State- Location Registration Aircraft Passenger fatalities Crew fatalities Remarks 01-Oct-08 ITALY- MILAN SE-DAM MC-DONNELL-DOUGLAS- MD80 SERIES 104 6 The business jet involved apparently did not satisfy the selection criteria, too Hans de Jong, Page 6
Consulting various ICAO sources ADREP Statistics The 911 occurrences did not involve collisions but are stated in a way analogous to Interpretation 2 the four involved flights are all separately listed as accidents note legal battle between leaser and insurer whether the destruction of the WTC were one or two occurrences On 12 November 1996, a midair collision occurred over India when Saudi Arabian Airlines Flight 763 collided with Air Kazakhstan Flight 1907 ADREP 1996 however does not yet list individual accidents The ADREP Annual Statistics from 1997 to 2001 (in which accidents are listed individually) have been scanned for comparable occurrences no examples have been found of collision accidents involving two aircraft in the considered categories Hans de Jong, Page 7
Consulting various ICAO sources Annex 11 and Doc 4444 Annex 11 Air Traffic Services provides no answer to the title question, although a TLS from Doc 9689 is mentioned Doc 4444 ATM Procedures for ANS also does not provide an answer Hans de Jong, Page 8
Consulting various ICAO sources Manuals on Manual on Airspace Planning Methodology for the determination of Separation Minima (Doc 9689) states in Methods of Evaluating Safety : 6.13 In order to evaluate the estimate of collision risk, this should be compared to a maximum tolerable collision risk for the system. Determining this level of risk is an independent process involving decision makers who represent State authorities, regional authorities or ICAO technical panels. The maximum tolerable risk is normally expressed in terms of a TLS. In the past, when applied to en-route collision risk, the TLS had been expressed in terms of the number of fatal accidents per flight hour, which could result from collisions between aircraft (where a collision between two aircraft represents two fatal accidents). Although ICAO has agreed that the concept of a global TLS is valid, the metric of fatal accidents per flying hour (as applied, for example, in the NAT Region) may not be appropriate for other regions. ICAO has agreed to the development and use of different metrics, provided that it can be demonstrated that any change in separation minima or other system parameter is subject to the overriding consideration that the risk of collision as a consequence of a loss of separation, from any cause, should be lower than that of the agreed level of system safety. The RVSM Manual (Doc 9574) relates collisions and accidents similarly: Collision risk. The expected number of mid-air aircraft accidents in a prescribed volume of airspace for a specific number of flight hours due to loss of planned separation. Note. One collision is considered to produce two accidents. The A-SMGCS Manual (Doc 9830) does not explicitly distinguish collisions, accidents and fatal accidents Hans de Jong, Page 9
What interpretation does ESARR 4 use? ESARR 4 itself ESARR 4 is an important regulatory requirement when making changes to ATM, applicable for the ANSPs in the Eurocontrol/ECAC region risks associated with changes need to be assessed and mitigated in accordance to a certain safety minimum before implementation ESARR 4 expresses this ECAC Safety Minimum as a maximum tolerable probability of ATM directly contributing to an accident of a Commercial Air Transport aircraft of 1,55x10-8 accidents per Flight Hour, where accident has been defined as in Annex 13 As preventing collisions between aircraft is a primary purpose for ATM, the way to count accidents in collisions is crucial counting collisions as two accidents, yields a minimum that is twice as strict for such accidents as when counting collisions as single accidents Hans de Jong, Page 10
What interpretation does ESARR 4 use? ESARR 4 itself The risk respectively severity classification scheme stated in ESARR 4 provide some information regarding the interpretation to be used: The maximum tolerable probability (of ATM directly contributing) to an occurrence of Severity Class 1 is 1,55x10-8 per Flight Hour The effect of Severity Class 1 on operations is Accidents (note the s ), examples of effects on operations include: one or more catastrophic accidents one or more mid-air collisions one or more collisions on the ground between two aircraft one or more Controlled Flights Into Terrain total loss of flight control Hence, the risk and severity classification scheme are logically insufficient to achieve the ECAC Safety Minimum Hans de Jong, Page 11
What interpretation does ESARR 4 use? ESARR 4 itself A second issue concerns the severity classification: the severity classification shall rely on a specific argument demonstrating the most probable effect of hazards, under the worst case scenario (p. 13) The worst credible effect in the environment of operations determines the severity class (p. 15) Two different notions, both disregarding accident frequencies for occurrences for which the classified severity is not 1 Difficult to deduce what interpretation has been used in the context of these logical issues Hans de Jong, Page 12
What interpretation does ESARR 4 use? SRC Policy Doc 1 Also no direct answer to interpretation question The first steps in deriving the ECAC Safety Minimum are 1. Review of 1988 1999 ADREP data, yielding 374 accidents in ECAC region, for aircraft with a maximum take-off weight (MTOW) of at least 2250 kg, which amounts to 31.17 accidents per year 2. A UK CAA study is mentioned for the estimate that there is at least one primary ATC cause for 1.1% of the fatal accidents for public transport aircraft with MTOW of at least 5700 kg No explicit reference, but CAP 681 Global Fatal Accident Review 1980-1996 yields that ATC causal factors are primary for 10 of 589 = 1.7 % of the accidents CAP 681 does not answer the interpretation question CAP 776 Global Fatal Accident Review 1997-2006 does: The sum, by individual aircraft type, of the number of fatal accidents was 284, one more than the total number of fatal accidents stated earlier in this document. This was due to the inclusion of both jet aircraft involved in the Uberlingen mid-air collision that occurred on 1 July 2002 (a Boeing 757 and a Tupolev TU-154). This mid-air collision was counted as one fatal accident in the overall statistics Hans de Jong, Page 13
What interpretation does ESARR 4 use? Other SRC Docs EAM 4/GUI 1, EAM 4/GUI 2, EAM 4/AMC, EAM 4/ICAO, SRC DOC 12, SRC DOC 20 and SRC POLICY DOC 2 provide no interpretations SRC DOC 1 (not referenced in ESARR 4): ATM is intended to prevent the following basic accident types leading to the loss of one or more aircraft and/or multiple (fatal) injury to occupants: a) collision between aircraft in flight or moving on the ground b) EAM 4/ GUI 4 (not referenced in ESARR 4): For ATM, one accident can involve 2 or more aircraft. See A2 of ESARR 2 ESARR 2 indeed mentions collisions as examples of types of accidents number of aircraft involved is contextual information regarding an occurrence Hans de Jong, Page 14
Conclusions Source Annex 13 ADREP Annual Statistics Annex 11 Doc 4444 Airspace Planning Manual RVSM Manual A-SMGCS Manual ESARR 4 SRC Policy Doc 1 SRC Doc 1, ESARR 2, EAM 4/ GUI 4 The other doc s related to ESARR 4 Interpretation 1 1???? 2? 2 Hans de Jong, Page 15
Conclusions The title question is answered differently by documents from different organizations such as Boeing, CAA UK, Eurocontrol, ICAO and NTSB some look at the collision from the occurrence perspective - one accident others adopt the aircraft perspective and count two accidents most references do not address the question Although there are ambiguities and exceptions, the sample of ICAO documents consulted predominantly use Interpretation 2 ESARR 4 and SRC Policy Doc 1 are unclear, but associated SRC documents point to Interpretation 1 it could not be verified whether this interpretation has been consistently used why has implicitly another interpretation been introduced? Hans de Jong, Page 16
Conclusions Even more important than the interpretation used, is clear indication of it, since misinterpretations may lead to Target Levels of Safety not being achieved misleading comparisons between accident statistics from different sources The variety and lack of answers to the title question illustrate how easily implicit and different interpretations are made that the actual existence of interpretations different from intentions is likely Hans de Jong, Page 17
Questions or comments are also welcome at hans.de.jong@dfs.de Hans de Jong, Page 18