≈   MH370 MODELING SYSTEM / OCEAN TRANSPORT
COMPLETED DIAGNOSTIC · TWO FULL 365-DAY RUNS · INDEPENDENT OF AIRCRAFT PATH INFERENCE

Where did the Indian Ocean take the debris?

A forward-transport experiment using actual 2014–15 daily GLORYS currents finds abundant looping, southward excursions and westward reversals—but few particles reached the western Indian Ocean within the first year. This is a circulation result, not an MH370 crash-location inference.

Modelled releases48140 geographic cells + G0
Days simulated3658 Mar 2014 – 7 Mar 2015
80°E westward13 / 116-hour / 3-hour runs
100°E eastward451 / 4526-hour / 3-hour runs
The corrected finding: In the 35.5–38.7°S release band, 57 of 240 particles (6 h) and 68 of 240 (3 h) went south and later made a substantial westward excursion. An earlier endpoint flag incorrectly suggested that this never happened. Yet only 4 / 6 particles in that band crossed 80°E westward, and none of those western crossers completed the stipulated significant southward-then-westward sequence. The motion exists; the proposed delivery route was not established.

1. A strong eastward tendency, with rare western access

The modeled releases cover 33–45°S and 85–98°E, each with 12 initial positions within about 5 km of its cell center. They are a deliberately exploratory circulation survey, not a distribution of aircraft crash locations. The same 481 seeds were used in both timestep runs.

Original eight-panel chart tracing one year of modeled GLORYS surface transport separately for each initial release latitude between 33 and 45 degrees south
Original figure A — One year by release latitude. Traces are from the six-hour simulation, grouped into eight latitude bands; black dots indicate hypothetical release positions. Most transport is eastward, and the lines at 119.92°E mark the data boundary, not the coast. A 37.5°S release did cross 80°E westward at six hours (about 2% of its band), despite the figure’s introductory statement about only the first two panels. Compare the separately computed 3 h / 6 h gate-access chart.

Across the full ensemble, 451 versus 452 particles crossed 100°E eastward. By contrast, 13 versus 11 reached 80°E westward, 4 versus 5 reached 70°E, and only 1 versus 2 reached 50°E. Crossing a longitude is not a beach arrival.

Gate first passages; 365 days
Gate6 h3 h
100°E east451452
80°E west1311
70°E west45
60°E west23
50°E west12

2. The western path mainly goes north before going west

Particles that did reach the western gates generally migrated equatorward, then entered westward transport along the subtropical gyre’s northern limb. This was not a persistent southern loop. The figure below shows the original simulated paths in both integrations, with the crossers highlighted. These are computed trajectories, not observed debris paths; their individual identities and timings are strongly sensitive to the timestep.

Original two-panel comparison of simulated western crossing trajectories at six-hour and three-hour resolution, with the four shared crossing particle IDs shown as solid traces
Original figure B — Western access, the rate is similar but the identities are not. Of 481 hypothetical releases, 13 crossed 80°E westward at six hours and 11 at three hours; only four of the same particle IDs crossed in both calculations. Solid lines mark these four; dashed lines identify crossers specific to one timestep. See the companion starting/ending-position plot.

The southern cartwheel idea was therefore partly right about circulation, but was not established as a westward-delivery mechanism within this first-year window. The analysis found prolonged westward episodes, reversals and returns to earlier positions; an eddy-residence diagnostic alone does not establish a coherent material barrier.

The ending matters: the original chronology figure below shows that the 57 qualifying south-then-westward objects in the six-hour run began their westward episodes at a median of day 284; approximately 74% were still moving west over the last 30 days. Their journeys were truncated by the 365-day observation window. This does not establish whether those episodes would ultimately deliver debris westward or reverse east again.

Original trajectory and longitude versus time panels for 57 south-then-westward objects, with the median westward-turn day 284 marked on the time series
Original figure C — Late westward turns. The two views show the 57 qualifying six-hour tracks geographically and as longitude histories. Westward runs begin late in the simulation; the shaded final interval ends while many are still heading west. This is right-censored transport behavior, not evidence that the particles inevitably reversed east or would eventually reach Africa.

3. The eastern boundary is the largest geographical limitation

The current dataset ends at 119.9167°E. At year-end, 235 particles in the 6 h run and 222 in the 3 h run had stopped because they left the data domain or entered masked/invalid velocity cells. Some masked cells lie near Australia's coast; they are not automatically beachings. For particles terminated at the artificial eastern edge, the model says nothing about their eventual return or destination.

Original two-panel position plot for all particles at termination or year-end showing active drifters, eastern data-boundary censoring, possible Australian shelf mask encounters, isolated masked cells, and one western boundary exit
Original figure D — Where the model loses particles. Positions and status codes at the end of each 365-day calculation. Nearly half the population was censored or encountered masked data, primarily toward the east. A single modeled western boundary exit appears in the three-hour calculation near 45°E, 11.7°S; it is not a confirmed landfall. See the companion group-by-latitude status chart.
Statuses at day 365
Status6 h3 h
Still active246259
Geographic domain exit91105
Mask or interpolation exit144117

The additional researcher-provided censoring map identifies one western-domain departure in the three-hour run, at approximately 45°E, 11.7°S. This is a modeled exit from the supplied current fields, not a verified landfall. The predominant termination pattern remains eastern, but the earlier blanket wording that every domain exit was eastward is too broad.

This is why zero recorded western crossings from a southern release group must not be read as a permanent exclusion: much of its population ceased being observed within the numerical domain before the end of the year.

4. What did the numerical refinement establish?

A full second year-long integration at a 3-hour step was completed using a verified checkpoint/restart procedure. The 80°E westward count changed from 13 to 11 and the 100°E eastward count from 451 to 452. This supports stability of the broad low-west/high-east population-level finding under the tested refinement, not demonstrated convergence to a unique limit.

Among 225 objects still active in both runs, corresponding year-end positions differed by a median of 846 km. Exact individual tracks and first-passage dates are therefore not numerically resolved. Path lengths and tortuosities also changed with timestep. It would be misleading to identify a physical debris route from any single simulated trajectory.

5. What this does—and does not—say about MH370

Debris eventually recovered on western Indian Ocean coasts is an observation-selected population. The actual recoveries depend on current transport, buoyancy, coastline encounters, detection and reporting; this simulation models none of those latter processes. It demonstrates how some starting locations can contribute to western pathways while many modeled objects travel elsewhere.

Interpretation boundaries: One historical GLORYS reanalysis, monthly-mean NCEP winds, an empirical 1.2% effective slip, no additional Stokes field, no diffusion or fragment-specific leeway, and no 508-day debris/beaching model. The geographic seeds are not posterior samples. G0 is a conditional sensitivity coordinate, not a verified crash contact. No seventh-arc likelihood, origin estimate, search recommendation or transport event horizon follows from these figures.

Reports, result tables & provenance

The following public copies come from the completed ocean-lab outputs, with both simulation resolutions preserved. The raw 1.6 GB GLORYS files and personal workspace information are intentionally not mirrored here.

Editorial note: The copied research reports are retained as written for provenance. A few statements in the original interpretation need qualification—for example, the 6 h run contains one 37.5°S particle crossing 80°E, despite language elsewhere saying no release at or south of 36.5°S did so. Read the publication errata alongside the source reports.

The four principal figures (A–D) are the original PNGs provided by the researcher, now reproduced here under their original filenames. Three supplementary SVG charts were independently derived from the archived CSV outputs and are linked in the figure captions. The original scientific code and raw GLORYS fields are not mirrored on this website. Published 14 October 2026.