Synopsis of This Brief

The first section describes the events of the accident, as reported in the technical press, and includes some observations.

The postcript, added nearly six months later, contains a brief description of the USAF reporting procedures for this accident, and the conclusions that were published in the report.

The following information comes from the press, civilian and military sources. All sources are referenced except for one. I attempted to include only technical information about the equipment and the (possible) progress of the flight and the accident. Any political information is omitted.

Throughout, I offer some personal observations written on 26 April, 1996, before the final report was published. These should be taken as my opinion, not fact - maybe a prediction in April of what could later be concluded - and as appropriate subject for agreement or disagreement under reasoned discussion. Such opinions are offered in the context: [ A personal observation. .......] These observations have not (yet) been checked against the USAF final report. I refer readers to this report for detailed information.

The Events

A US Air Force T-43A aircraft, flight IFOR 21, carrying US Secretary of Commerce Ron Brown and other US Government personnel crashed on the approach to Dubrovnik airport, Croatia, on 3 April 1996. It disappeared from radar at 14.52 local time, about 5 mins before expected landing (FI.1).

The accident airplane, a `military 737-200' (FI.1) crashed 98 feet (20 meters) below the summit (NYTS) of a hill to the left of the NDB Rwy 12 approach to Dubrovnik airport.

Wreckage was found on top of a 2,300ft (700m)-high ridge 3km (1.5nm) north-west of the runway threshold. There was no evidence of hostile fire, the USAF says. (FI.1).

The aircraft was built in 1973, had accumulated 17,000h and 12,000 landings, and had undergone a major inspection in June 1995 (FI.1).

Minimum Descent Altitude (MDA) on the NDB approach for both straight-in Rwy 12 and circle-to-land is 2150ft. There are two NDBs: KLP at 11.8 miles (Final Approach Fix or FAF, 4000ft altitude) and CV at 1.9 miles (Missed-Approach Point or MAP, where one expects to be at MDA), both in line with the runway extended centerline.

The approach chart (JS-APP) indicates the accident hill clearly, close in to the CV NDB (the MAP). (IHT.1) reported that the aircraft was 1.8 miles off course on the approach; (FI.2) that the site is `1km (0.5nm) off the approach track'. The summit is in fact about 1.5 nautical miles off-course (JS-APP). Final approach course is 119 degrees magnetic. The crash site was reported to be at a bearing of 109 degrees from KLP (NYTS), but the chart shows the relative angle to be the inverse-tangent of 1/7, which is about 7.5 degrees.

[A personal observation. I have heard some lay speculation about potential aerodynamic problems, from people who are presumably thinking about the Colorado Springs and Pittsburgh B737 accidents, and are wondering about potential uncommanded rudder movements with B737s. Anybody who thinks the crash may have been caused by aerodynamic problems has to explain how to get this airplane to fly sideways 1.8 miles while at an altitude of somewhere between 4000ft and 2202ft. Doubtless with the right amount of fairy dust, everything will be possible to those who believe strongly enough!]

A brief look at a picture of the wreckage by an expert flight engineer elicited a comment that the planform of the wreckage was consistent with a CFIT accident, namely that the wreckage was not widely distributed and retained roughly the shape of the accident aircraft.

There's no timing given on the chart, thus flying the approach as a civilian, one must use CV as the MAP. Reception of both NDBs is thus required for the approach. In principle this may be accomplished with one ADF or RMI, but see the (personal) observation below. Anonymous USAF generals are quoted by (NYTS) as calling the navigation equipment of IFOR 21 `primitive' and `rudimentary'. (FI.2) denotes the navigation equipment aboard IFOR 21 as `primitive' and says:

The aircraft, however, had only one automatic direction-finder (ADF) receiver (the device for displaying the NDB's direction relative to the aircraft), so the crew could not use both beacons as a check. The USAF says that the beacons appeared to be serviceable. The aircraft was not fitted with any global-navigation-satellite system, says the USAF.

[A personal observation. A think-through suggests to me that, flying the course with one RMI tuned to KLP, it could be nearly impossible to tell one's position relative to CV when near to CV. Visually, in the cockpit, passing an NDB 100m away and passing it 1.5 nautical miles away have similar effects - the needle swings and one swing is merely quicker than the other. Good civilian technique would suggest that the switch to CV should occur somewhere in the mid-approach, because it's generally safer to fly to an NDB than from one: the ADF/RMI becomes less sensitive to course deviations the further away one is from the NDB; correspondingly more sensitive the closer one is. Flying the course with two ADFs/RMIs seems to me to be a lot easier. One could more easily notice course deviations. A deviation of 7.5 degrees from course would yield a difference of 15 degrees between the two needles at the course midpoint.]

No report I've seen has said which beacon the ADF was set to at the time of the crash.

Being 7.5 degrees, 1.5nm off-course would probably be enough to fail the NDB approach check on a civilian instrument flight check ride with an NDB at the MAP.

Some further points made by the press:

• The airplane didn't carry a Flight Data Recorder or a Cockpit Voice Recorder (FI.1, NYTS)
• Croatian authorities say the crash was caused by pilot error (NYTS)
• The USAF declines to comment (NYTS)
• Croatia Airlines will not let its crews land there if both pilot and copilot `have never done it before'. (NYTS)
• `The pilots were navigating with a compass and a radio receiver. Air Force generals called that equipment "primitive" and "rudimentary" (NYTS). According to (FI.2), the aircraft was navigating with `a single, relatively primitive non-precision navigation aid in weather which was significantly worse than forecast' (FI.2).
• `On Feb. 1, after an inspection, the Air Force's Air Mobility Command forbade all U.S.-based planes under its control, including VIP flights from Andrews Air Force Base in Washington, from landing at Dubrovnik except in clear weather and in daylight. (NYTS)
• The decision to fly there was taken by the USAF command in Europe, and it is free to do so under present procedures. (NYTS)
• `The rain, wind and fog were bad for April - but not as bad as early reports about the crash suggested. Five planes landed at Dubrovnik shortly before the accident.' (NYTS) The captain of a Croatian government Sabreliner who landed 99 minutes before the accident said that `We often see that kind of weather. [..] The wind was very steady - not gusty. The crosswind was almost nothing.' Croatian and Air Force officials way there was moderate rain, no lightning, broken clouds at 400 feet, an overcast sky at 2,000 feet, a 12-knot wind and five miles of visibility at the airport at the time of the crash' (NYTS). (No wind direction was given.) This coheres with the (FI.2) report of a `cloud base' of 2,000ft and visibility of 8km. The weather was significantly worse than forecast (13km and 25,000ft) (FI.2). These two reports conflict with the earlier one of (FI.1), which reports horizontal visibility as 1,000m and vertical visibility as 90m.
• The pilot of a Challenger carrying the Croatian Prime Minister and US Ambassador Galbraith spoke with the T-43 pilots from the cockpit 40 minutes after he'd landed and about 10 before they were due to land. He `was sure they would land' but advised them of lowering rain clouds near the `minimum ceiling' of 2,150 feet AGL (NYTS). `Those same conditions would have forbidden [sic] any attempt to land at Dubrovnik by U.S.-based Air Force jets transporting VIPs.' (NYTS). (FI.2) reports that the `cloudbase was 1,000ft lower than the minimums published for the approach'.

Finally, a procedural consequence of the accident is that US Secretary of Defence Perry has ordered `the military' to equip all its `passenger planes' with `much more precise navigation gear' (NYTS).

Additional Background Information

A glossary of terms is contained below.

Runways are designated according to the nearest 10-degree heading. Dubrovnik's runway ies at a magnetic heading of 119 degrees and is thus designated Runway 12.

The US military uses airframe type designations which are disjoint from the civilian designations of the same or similar aircraft. The engineering specifications of such aircraft may be significantly different (as in the B720 and the military EC-12) or similar. The T-43A is equivalent to a civilian B737-200 airframe (FI.1).

An Approach Plate or Approach Chart is a map which aviators use to guide them in an instrument approach procedure. The approach itself is developed and certified by the appropriate aviation authority (in the case of Dubrovnik, the Croatian, formerly the Jugoslavian, government). The idea of the approach plate came from a certain Captain Jeppesen, whose tradition is continued in the Jeppesen-Sandersen Company.

When one flies an approach, one starts at the Initial Approach Fix (IAF). In the US, one may either pass over an IAF or receive radar vectors (heading commands from a controller) to the approach. There is a Final Approach Fix which one must pass over, and for non-precision approaches a further Missed Approach Point (MAP). The MAP can either be reached by a time-out (i.e., dead reckoning) from the FAF, or can be a designated fix. For the Dubrovnik RWY 12 Approach, it's passage over the CV NDB. One passes over the FAF at or above a designated altitude (4000ft in this case) and must maintain at or above a designated Minimum Descent Altitude (2150ft here) until MAP. If the runway or its threshold or its immediate environment (lights, landing zone, etc) cannot be seen at this point by the pilots, the planned landing must be aborted and a Missed Approach Procedure followed. This Missed Approach Procedure is designated on the approach plate.

Postscript: The Investigation and Report

The USAF Reporting Process for the CT-43A accident

I am indebted for the information in this paragraph to Lt.-Col. Thomas Farrier, Chief of Flight Safety Issues in the Office of the US Air Force Chief of Safety.

The USAF does not normally release results of its safety investigations to the public. In cases in which an aircraft accident results in a fatality, the destruction of an aircraft, or more than one million dollars in damage, a separate investigation is conducted, in the USAF governed by Air Force Instruction 51-503. This report is always publically available, based on the same facts, but protects the rights of individuals who may have been responsible for the loss.

The AFI 51-503 investigation was the only one conducted in this case, with the assistance of experts from Boeing, the NTSB and the FAA. The public's right to know all of the information was considered of overriding importance, because of the death of the US Secretary of Commerce. Lt.-Col. Farrier notes that USAF safety personnel acted on that report the moment it was completed. The report is 22 volumes, contained in a box measuring roughly 30 x 30 x 50 cm (!!).

The investigator's `Statement of Opinion, limited in its possible uses by regulation 10 U.S.C. Section 2254, attributes the accident to

• a failure of command,
• aircrew error,
• an improperly designed instrument approach procedure

The Final Report's Conclusions

Lt.-Col. Farrier points out that the investigator's `Statement of Opinion, limited in its possible uses by regulation 10 U.S.C. Section 2254, attributes the accident to

• a failure of command,
• aircrew error,
• an improperly designed instrument approach procedure

A report by Ramon Lopez in Flight International (FI.3) summarised the conclusions of the USAF's Accident Investigation Board report.

• The flight took place in breach of orders. The airport's approach procedure had not been approved by the DoD. Court martials of commanders at 86th Airlift Wing (Ramstein) may follow.
• Flight planning is criticised because the crew `failed to appreciate' that the NDB approach `demanded the presence' of two ADFs, but the aircraft was fitted with only one (see my comments on this in the report I wrote on 29 April).
• The report says the crew rushed the approach, extended gear and flaps late, and were flying 80kt (150km/h) above final-approach speed some 9 degrees left of track.
• The report also blamed an `improperly designed' approach procedure (disputed by Croatian authories). The NTSB participated in the investigation and claims that if the approach had been designed with standard vertical obstacle clearance limits, the aircraft would not have hit the ground. [What is meant by `standard'?]

Glossary

ADF Automatic Direction Finder, a navigation instrument which indicates the bearing to the NDB generating the signal.
FAF Final Approach Fix
MAP Missed Approach Point
MDA Minimum Descent Altitude
NDB Non-Directional Beacon
RMI Radio Magnetic Indicator, a combined Directional Gyro and ADF

References

(IHT.1): Old technology Guided the Approach of Brown's Plane, by Don Phillips and Bradley Graham of the Washington Post Service. The International Herald Tribune, 6-7 April, 1996, p2.
Back to First Reference

(FI.1): Graham Warwick, USAF 737 was `off course' before crash, Flight International, 10-16 April, 1996, p8.
Back to First Reference

(FI.2): 737 crashed using `primitive' navigation aid, Flight International, 17-23 April, 1996, p6.
Back to First Reference

(JS-APP): NDB Rwy 12 Approach (16-1), Dubrovnik ( http://www.jeppesen.com/lddu161.html), approach chart published by Jeppesen-Sandersen Company.
Back to First Reference

(FI.3): Ramon Lopez, USAF `broke orders' on CT-43 disaster flight, Flight International , 19-25 June, 1996, p20.
Back to First Reference

Back to top

Back to `Computer-Related Incidents' Compendium.

Copyright © 1999 Peter B. Ladkin, 1999-02-08
	Last modification on 1999-06-15 by Michael Blume