Hull-Loss Localisation Validation Corpus — 37 recovered cases / 185 seeded runs

Deep-ocean loss of control from cruise after unreliable airspeed. The real search burned two years across ~17,000 km²; the debris field sat ~6.5 NM from the last ACARS fix.

Solution

3.883, -30.779

Confirmed site

3.938, -30.500

90% search area

10,252 km²

Main debris field, located April 2011 by Remus AUVs after three failed campaigns. Source: BEA final report f-cp090601 (last ACARS position and wreckage site).

Cruise fire event with a long, poorly-constrained final descent — the hardest of the set and the one that stresses the heading-spread model.

Solution

33.834, 28.953

Confirmed site

33.600, 29.200

90% search area

9,970 km²

Recorder and main wreckage field, eastern Mediterranean, ~3,000 m depth. Source: Egyptian AAIC investigation and John Lethbridge survey positions.

Spiral dive from cruise into deep water. Recorders were pinpointed almost directly beneath the last radar plot.

Solution

-3.528, 118.183

Confirmed site

-3.683, 118.093

90% search area

8,104 km²

Recorders located on the Makassar Strait floor, ~2,000 m depth, August 2007. Source: KNKT final report KNKT/07.01/08.01.36.

Cargo fire, degraded tracking and the deepest recorder recovery ever attempted. Weak last-fix confidence is the point of this case.

Solution

-19.127, 59.715

Confirmed site

-19.170, 59.870

90% search area

1,619 km²

Main wreckage and CVR recovered from ~4,900 m north-east of Mauritius. Source: Margo Commission of Inquiry, South Africa (1990).

Blocked pitot, night departure over water, steep unrecovered upset from medium altitude into deep water.

Solution

19.988, -70.612

Confirmed site

19.980, -70.650

90% search area

738 km²

Wreckage located ~2,200 m deep off Puerto Plata, Dominican Republic. Source: Dirección General de Aeronáutica Civil (DR) final report, 1996.

Spatial-disorientation roll after a dark-night departure — short flight, high energy, deep water.

Solution

27.772, 34.306

Confirmed site

27.780, 34.310

90% search area

727 km²

Wreckage located in the Red Sea off Sharm el-Sheikh, ~1,000 m depth. Source: Egyptian Ministry of Civil Aviation final report (2006), BEA participation.

Go-around upset at night in rain. Short-baseline radar track, coastal current — the near-shore deep case.

Solution

43.520, 39.641

Confirmed site

43.480, 39.660

90% search area

460 km²

Debris field in the Black Sea ~6 km off Sochi, ~500 m depth. Source: MAK (Interstate Aviation Committee) final report, 2007.

Fuel exhaustion ditching that broke up on impact — low energy, and the lowest-altitude deep-water case in the corpus.

Solution

38.050, 13.271

Confirmed site

38.030, 13.280

90% search area

155 km²

Ditching site off Palermo; wreckage recovered from ~1,500 m. Source: ANSV Italy final report, 2005.

Upset and stall in convective weather over shallow water. Recovery site lay ~28 km south of the last radar return.

Solution

-3.277, 109.618

Confirmed site

-3.617, 109.717

90% search area

6,066 km²

Fuselage located on the Java Sea floor, January 2015. Source: KNKT final report KNKT.14.12.29.04.

High-energy near-vertical impact from low altitude — the tightest possible prior. A good engine must not over-disperse this one.

Solution

-5.793, 107.152

Confirmed site

-5.790, 107.148

90% search area

1,001 km²

Main impact point in the Java Sea, ~34 m depth. Source: KNKT final report KNKT.18.10.35.04.

Autothrottle asymmetry roll upset shortly after departure. Short, steep and shallow-water — the case where a tight prior directly saves search days.

Solution

-5.770, 106.589

Confirmed site

-5.909, 106.557

90% search area

1,487 km²

Main impact site off Pulau Laki, Java Sea, ~23 m depth. Source: KNKT report KNKT.21.01.01.04.

Deliberate dive from cruise with a partial recovery and second dive — a two-phase descent the single-phase descent sampler is not built for.

Solution

40.423, -69.583

Confirmed site

40.350, -69.770

90% search area

7,283 km²

Debris field ~100 km south of Nantucket, ~82 m depth. Source: NTSB aircraft accident brief DCA00MA006.

Jackscrew failure, inverted dive from a well-tracked position. Should be a near-bullseye; if it is not, the descent model is wrong.

Solution

34.026, -119.238

Confirmed site

34.060, -119.350

90% search area

2,299 km²

Impact site in the Pacific north of Anacapa Island, ~200 m depth. Source: NTSB AAR-02/01.

Cabin fire with radar and transponder lost before impact — degraded last fix, coastal current, shallow seabed.

Solution

44.378, -63.967

Confirmed site

44.400, -63.970

90% search area

1,122 km²

Impact site off Peggy's Cove, Nova Scotia, ~55 m depth. Source: TSB Canada report A98H0003.

Somatogravic-illusion push-over into very shallow tidal water. Tests the sink-phase model where depth is effectively zero.

Solution

29.760, -94.763

Confirmed site

29.720, -94.680

90% search area

955 km²

Impact in Trinity Bay, Texas, ~3 m of water and mud. Source: NTSB AAR-20/02.

Blocked static ports meant even the crew's own altitude and speed were false — the case where the last fix itself is corrupt.

Solution

-11.640, -77.519

Confirmed site

-11.650, -77.550

90% search area

1,311 km²

Impact in the Pacific off Pasamayo, Peru. Source: DGAC Peru final report, 1996.

Tail-fin separation and in-flight break-up over water — distributed debris on a shallow, current-swept seabed.

Solution

57.864, 9.047

Confirmed site

57.900, 9.000

90% search area

2,878 km²

Wreckage recovered from the Skagerrak, ~65 m depth. Source: AIBN / Norwegian commission of inquiry, 1993.

Near-supersonic dive from cruise onto land/river — the land-surface control case, where no drift term applies and dispersion must stay disciplined.

Solution

-2.311, 104.744

Confirmed site

-2.371, 104.579

90% search area

7,941 km²

Impact into the Musi River near Palembang; shallow riverbed recovery. Source: NTSC Indonesia investigation and NTSB comments (1997).

Very low altitude, low energy urban loss of control. The engine must not blow this into a multi-kilometre zone.

Solution

25.056, 121.518

Confirmed site

25.058, 121.526

90% search area

86 km²

Impact into the Keelung River, Taipei, after engine shutdown and stall. Source: ASC Taiwan final report ASC-AOR-16-06-001.

Vertical stabiliser separation just after departure. Urban land, short baseline, tight expected zone.

Solution

40.580, -73.849

Confirmed site

40.580, -73.850

90% search area

499 km²

Impact in Belle Harbor, Queens, New York. Source: NTSB AAR-04/04.

Missile strike shortly after departure with fire and progressive control loss — land impact with a partially disrupted airframe.

Solution

35.592, 51.085

Confirmed site

35.550, 51.100

90% search area

1,007 km²

Impact near Shahedshahr, south-west of Tehran. Source: CAO Iran final report, 2021.

The most extreme vertical dive on record. Almost no lateral travel — a hard test of the engine's ability to stay tight.

Solution

23.276, 111.061

Confirmed site

23.360, 111.130

90% search area

6,197 km²

Near-vertical impact into hillside near Wuzhou, Guangxi. Source: CAAC preliminary report, 2022.

Roll upset at night into dense mangrove. The wreckage was 5 km from the runway and took two days to find — the classic 'search failed on a tiny radius' case.

Solution

3.854, 9.735

Confirmed site

3.830, 9.720

90% search area

309 km²

Wreckage found in mangrove swamp ~5 km from Douala airport after a 2-day search. Source: Cameroon Civil Aviation Authority final report, 2010.

Both propellers feathered on final approach; stall into a gorge. Low energy, high terrain relief.

Solution

28.201, 83.976

Confirmed site

28.200, 83.980

90% search area

109 km²

Impact in the Seti River gorge, Pokhara, Nepal. Source: Government of Nepal accident investigation commission report, 2023.

Radio-altimeter failure, autothrottle retard and a stall short of the threshold. Almost no dispersion is acceptable here.

Solution

52.312, 4.739

Confirmed site

52.310, 4.750

90% search area

87 km²

Impact in a field short of runway 18R, Schiphol. Source: Dutch Safety Board report 2009026, 2010.

Deliberate descent into mountainous terrain. The truth is where terrain first intercepts the path — a case pure ballistics gets wrong without a terrain cut.

Solution

44.318, 6.420

Confirmed site

44.280, 6.440

90% search area

1,582 km²

Impact into the Estrop massif near Prads-Haute-Bléone, French Alps. Source: BEA final report BEA2015-0125.

Lost situational awareness in monsoon cloud during a circling approach — terrain intercept, not ground impact on the projected path.

Solution

33.737, 73.043

Confirmed site

33.750, 73.060

90% search area

364 km²

Impact into the Margalla Hills north of Islamabad. Source: Pakistan CAA Safety Investigation Board report, 2011.

Total fuel exhaustion in a hold over high terrain — an unpowered glide with no thrust and a terrain intercept.

Solution

6.360, -75.337

Confirmed site

6.110, -75.420

90% search area

1,510 km²

Impact on Cerro Gordo near La Unión, Colombia, after total fuel exhaustion. Source: Aerocivil Colombia final report, 2018.

Somatogravic push-over in severe wind shear. Very strong low-level wind — the layer term matters here more than anywhere else in the corpus.

Solution

47.248, 39.819

Confirmed site

47.250, 39.820

90% search area

374 km²

Impact on Rostov-on-Don airport grounds during a second go-around. Source: MAK final report, 2019.

Centre-tank explosion at altitude. Debris fell in three separate fields over several km — a single-point answer is structurally incomplete here, and the score should show that.

Solution

40.802, -72.374

Confirmed site

40.660, -72.630

90% search area

2,159 km²

Main debris field off East Moriches, New York, ~36 m depth. Source: NTSB AAR-00/03.

Bomb detonation at cruise over deep Atlantic with strong upper winds — the widest debris dispersion in the corpus.

Solution

51.538, -10.587

Confirmed site

51.050, -12.820

90% search area

22,740 km²

Debris field in the Atlantic ~190 km off the Irish coast, ~2,000 m depth. Source: Kirpal Commission report (India, 1986) and AAIU Ireland records.

Warhead detonation at cruise; wreckage spread across a 50 km² footprint on land. Scores the centroid of a field, not a point.

Solution

48.310, 40.299

Confirmed site

48.140, 38.650

90% search area

18,635 km²

Debris fields near Hrabove, eastern Ukraine, spread over ~50 km². Source: Dutch Safety Board final report, 2015.

Explosive device separating the tail at cruise. Desert terrain made recovery easy; localisation is still a dispersion problem.

Solution

30.369, 34.121

Confirmed site

30.170, 34.210

90% search area

5,990 km²

Debris field in the Hasana area of central Sinai, spread over ~15 km². Source: Egyptian Ministry of Civil Aviation preliminary report, 2015.

Wing separation at very low altitude over a shallow channel. Small aircraft, tiny scale — checks the engine does not have a floor error it cannot go below.

Solution

25.775, -80.132

Confirmed site

25.770, -80.130

90% search area

93 km²

Impact in the Government Cut shipping channel off Miami Beach, ~6 m depth. Source: NTSB AAR-07/04.

Night approach loss of control with an 18-hour delay before the organised search — the drift-window case.

Solution

-11.430, 43.309

Confirmed site

-11.467, 43.317

90% search area

324 km²

Debris field north-west of Grande Comore; recorders recovered from ~1,200 m. Source: Comoros / BEA-assisted investigation, published 2013.

Fuel-exhaustion ditching close inshore with a strong along-shore current — floating debris and structure separate quickly here.

Solution

-11.565, 43.297

Confirmed site

-11.600, 43.260

90% search area

128 km²

Ditching site in shallow water off Le Galawa Beach, Grande Comore. Source: Ethiopian Civil Aviation Authority report, 1998.

Successful ditching with a strong tidal set. The airframe floated and moved — the pure drift-dominant case, and the only survivable one in the corpus.

Solution

40.719, -74.013

Confirmed site

40.770, -74.000

90% search area

115 km²

Hudson River ditching; airframe drifted south with the ebb tide before recovery. Source: NTSB AAR-10/03.