A channel that spins.

The textbook said “current.” The data said otherwise.

Between Mozambique and Madagascar lies a 400-km-wide channel that funnels water from the tropical Indian Ocean south toward the Agulhas Current. For decades it was drawn on maps as a single continuous flow — the “Mozambique Current.” But ships, fishers and spill responders kept seeing surface water behave in ways a steady current could not explain.

GWO set out to settle the question with measurement rather than assumption — by reading the channel through three independent observing systems and asking whether they told the same story.

Three observing systems, one question

Each platform constrains a different part of the water column, and all were sourced from the EU Copernicus Marine Service and processed in a single open-source Python pipeline. The test was deliberately simple: if the channel really moves as eddies, all three should independently show rotation — not a straight southward flow.

GNSS surface drifters 2002–2024

121 platforms, 70,175 velocity observations at 15 m drogue depth, wind-slip corrected — the direct trace of where surface water goes.

Argo profiling floats 2004–2024

53 floats on a park-and-profile cycle at ~1,000 m, reporting zonal & meridional velocity at depth and at each surfacing.

Satellite altimetry · DUACS L4 1993–2023

11,322 daily gridded sea-surface-height fields at 0.25°, with derived geostrophic velocities across the channel (30°–55°E, 25°–12°S).

Eddy detection · py-eddy-tracker method

Automated census from SSH — closed contours stepped at 0.002 m after a 400 km high-pass — giving each eddy a polarity, centre, radius and amplitude.

What the channel revealed

The three systems converged on the same answer — a channel filled with rotating cells, not a band of moving water. Six numbers carry the result.

Finding 1 — there is almost no steady current

Pooled across every near-surface observation, the basin-mean flow is vanishingly weak: a mean meridional velocity of just −0.03 m/s against a mean speed of 0.48 m/s. The kinetic energy of that mean flow is only 0.4% of the total for drifters and 0.6% for Argo — meaning more than 99% of the surface energy lives in the fluctuating eddy field, not in a through-current.

Finding 2 — trajectories loop, they don’t run south

Drifter and Argo floats curl and loop through the channel rather than tracking south. A straightness index — net displacement over path length, near 1 for a straight line and near 0 for tight loops — has a median of just 0.42 for drifters and 0.22 for the longer, deeper Argo paths. That looping is the Lagrangian signature of rotating cells; a uniform current cannot produce it.

Finding 3 — three platforms, one circulation

Interpolating the daily altimetry geostrophic velocity onto every in-situ observation and comparing component-by-component, the systems agree closely: Pearson correlations of 0.80–0.86 and root-mean-square differences near 0.23 m/s. The residual scatter is exactly what is expected from the wind-driven, ageostrophic motion that altimetry — being purely geostrophic — does not capture.

Table I — Inter-platform velocity agreement (1993–2023)

Platform pair	n	r (zonal)	r (merid.)	RMSD (m/s)
Drifter vs. altimetry	64,252	0.80	0.83	0.23 / 0.24
Argo vs. altimetry	1,719	0.82	0.86	0.24 / 0.24

Bootstrap 95% CI half-widths < 0.02 (drifter) and < 0.04 (Argo).

The altimetry view tells the same story directly: a single day of sea-level anomalies resolves a checkerboard of highs and lows — warm-core (anticyclonic) and cold-core (cyclonic) eddies — rather than a single banded current.

Finding 4 — a persistent population of eddies

An automated census — the open-source py-eddy-tracker — run over all 11,322 daily fields finds the channel is never empty. On an average day, 10.6 anticyclonic and 8.1 cyclonic eddies are present, with anticyclones systematically larger and stronger: a mean radius of 80 km (≈160 km across), with the largest cells reaching the ~300 km scale. Eddy-centre density and time-mean eddy kinetic energy both trace one continuous high-energy corridor down the channel axis, peaking near 0.16 m²/s².

Table II — Eddy census, Mozambique Channel (1993–2023)

Quantity	Anticyclonic	Cyclonic
Mean number present per day	10.6	8.1
Mean effective radius (km)	80 ± 49	73 ± 49
Mean amplitude (m)	0.088	0.062
Mean rotational speed (m/s)	0.35	0.33

Snapshot census from py-eddy-tracker over 11,322 daily fields.

Finding 5 — about six eddies shed south each year

Linking the daily detections into trajectories lets us measure the shedding rate directly rather than infer it. Over 1993–2023, 6.3 ± 1.7 long-lived (≥90-day) anticyclonic eddies migrate southward through the channel each year, with a mean trajectory lifetime of 144 days. That sits squarely within the four-to-seven per year established by earlier studies — now reproduced from a single 31-year record.

From velocities to structure

As a purely in-situ cross-check, the gridded drifter velocity field was differentiated — computing its spatial gradient to expose the boundaries between cells. The gradient map lights up the rims of individual eddies independently of the altimetry census: the very edges where rotation, upwelling and biological productivity concentrate.

The eddy field is intensifying

Aggregated into a monthly series, the channel’s eddy kinetic energy shows a statistically significant upward trend of about +1.2% per year over 1993–2023 (p < 10⁻¹⁵). We report this as an observed trend only — part of the rise may reflect the growing capability of the altimetry constellation across the record — but the direction is unambiguous.

Why a “spinning” channel matters

Knowing the channel is governed by discrete, trackable eddies — not an averaged current — is not an academic footnote. It changes the practical answer to almost every question stakeholders ask about these waters, because an eddy can be located, tracked and forecast in a way a smeared mean current cannot.

The recommendation: track the cells, not the average

The evidence points to one operating principle — the Mozambique Channel should be monitored and modelled as a population of eddies, each with a position, age and trajectory, rather than as a single mean flow. GWO’s recommendations build a living, eddy-resolving picture of the channel.

• 01

  Operational eddy tracking

  Stand up a near-real-time service that detects and follows each eddy from daily altimetry, validated against drifters and Argo.

• 02

  Decision layers for users

  Translate the eddy field into tailored products — fishing-ground guidance, spill-drift forecasts and routing advisories.

• 03

  Sustained in-situ network

  Maintain drifter and Argo coverage so the satellite picture stays anchored to measured velocities at and below the surface.

• 04

  Trans-boundary sharing

  Make the eddy record openly available to Mozambique, Madagascar and regional fisheries and disaster agencies.

The same method — cross-validating altimetry against in-situ velocities to resolve eddies — extends to any energetic boundary-current region, turning a one-off study into a transferable monitoring capability.