Bangladesh is Trading

Bangladesh last told the United Nations what it trades in 2018. Every year since, the country's exports and imports have entered world statistics only as what other countries report buying from it and selling to it. The physical trade never stopped: trucks still queue at Benapole, ships still wait off Chattogram. This part of the series measures that physical trade directly, from orbit, and reads it against the mirror statistics that partners file in Dhaka's silence.

Three instruments, none of which require anything from Bangladesh's statistical system: a radar satellite that images the country every two to four days regardless of cloud, a European bilateral trade database built from partner customs records, and the UN Comtrade mirror, where a country's trade is reconstructed from the other side of every transaction.

The queue you can see from space

Benapole is Bangladesh's largest land port, the gateway for most overland trade with India. Its chronic truck queues are a news staple, but nobody publishes a queue series. Sentinel-1, the European Space Agency's radar satellite, passes over the border at 6 in the morning and 6 in the evening, and smooth asphalt scatters radar away from the sensor while metal trucks reflect it straight back. An empty road images dark. A queued road lights up.

We monitored 9.3 kilometers of the approach road from the Benapole gate east toward Navaron, using every Sentinel-1 scene from September 2025 through August 2026: 88 images across three orbital tracks.

The signal is the road's radar brightness relative to a control ring 150 to 300 meters away, which cancels out rain, soil moisture, and calibration differences between scenes. On the evening pass, the road averages 2.3 decibels brighter than its surroundings, and the value swings through a 3.8 decibel range across the year.

The swings are not noise. The two deepest troughs of the year, 0.5 dB on February 17 and 0.3 dB on March 1, bracket the start of Ramadan, when the 6 pm overpass coincides with iftar and the road empties. The reading jumps to 3.1 dB on March 25, immediately after Eid al-Fitr, consistent with the reopening backlog. The annual peak, 4.1 dB, falls on January 12, in the winter high season. These alignments are consistent with the border's known rhythm; verifying individual dates against port records is the next step, not a claim we make yet.

The control test is built into the orbits. The two dawn passes image the same road at 6 am, and there the road is indistinguishable from its surroundings all year, with half the variance. Trucks queue in the evening; at dawn the road is clear. A signal that appeared on all three orbits equally would be weather. This one is traffic.

Counting ships at anchor

Ships on water are the easy case for radar: bright metal targets on a dark, flat sea. We counted them in every Sentinel-1 scene over two areas: the outer anchorage off Chattogram, where vessels wait for berths at the port that handles over 90 percent of Bangladesh's seaborne trade, and the Pashur river reach at Mongla, the second seaport.

Across 52 usable scenes, the median Chattogram anchorage image held 23.5 large-vessel detections, with the middle half of scenes between 18 and 34. That matches the port's known operating range and is now observable every two to four days without a single AIS subscription. Four monsoon scenes were discarded honestly: wind-roughened seas scatter radar brightly enough to mimic thousands of false targets, and a September 2025 storm scene returned 1,149 detections, which is sea state, not shipping. The filter is disclosed, not hidden.

Mongla reads differently. Its reach is a working river: the raw detection count, median 85 objects per scene, mixes seagoing ships with barges, fishing craft, and moored vessels. The large-vessel filter, 3,200 square meters and above, cuts that to a median of 49, but even those are not 49 seagoing ships: in a narrow river, vessels moored side by side merge into single bright detections. Read the Mongla line as an activity index for the reach; the open-water Chattogram anchorage, where ships anchor apart, is the cleaner count.

The trade the exporter no longer reports

The satellite record answers what moved. What it was worth requires customs data, and here Bangladesh has been silent since 2018. Its trade with the world is reconstructed from mirrors: the import records of its partners.

Twelve major partners, which together cover the bulk of Bangladesh's export markets, reported buying 42.7 billion dollars of Bangladeshi goods in 2024, up from 33.7 billion in 2019, with the 2022 logistics-crisis spike to 50.0 billion and the 2023 correction to 41.3 billion clearly visible. Partial 2025 filings from 11 of the 12 already total 46.6 billion.

Germany reported 8.85 billion dollars of imports from Bangladesh in 2024 and the United States 8.78 billion, together nearly two fifths of the twelve-partner total. France, the United Kingdom, and Poland follow. The same twelve partners reported selling Bangladesh 40.4 billion dollars of goods.

The full geography comes from BACI, the CEPII database that harmonizes both sides of every bilateral flow. It puts Bangladesh's 2024 exports at 58.8 billion dollars against imports of 63.6 billion.

Drawn as flows, the same data shows how concentrated the export economy is: fifteen destinations, most of them in Europe and North America, absorb the bulk of what leaves Dhaka and Chattogram.

The import map is a different geography entirely. Where exports flow west to consumer markets, imports come from Asia: China supplied 22.8 billion dollars of Bangladesh's 2024 imports and India 11.3 billion, together more than half the total. The overland share of that Indian trade is what queues at Benapole.

For the Benapole story, one mirror matters most. India reports its trade with Bangladesh monthly, and the asymmetry is stark: in June 2026 India recorded 874.9 million dollars of exports to Bangladesh against 124.5 million of imports from it, a seven-to-one imbalance that rides the trucks through Benapole and Petrapole.

What this is for

Each instrument checks the others. If India's reported exports to Bangladesh fall while the Benapole queue index stays high, something other than demand is throttling the border. If partner-reported seaborne trade rises while the Chattogram anchorage count falls, the port got faster; if both rise, congestion is coming. None of this requires waiting for official statistics that stopped arriving seven years ago.

The method costs nothing to run. Sentinel-1 data is free and open under the Copernicus programme, the mirror statistics are public, and one year of monitoring one border crossing takes a few hours of compute on a laptop. The World Bank, whose satellite work on African border crossings prompted this exercise, reports the marginal cost of monitoring an additional crossing at under 100 dollars a year. Extending this to Burimari, Hili, Sonamasjid, and Akhaura is an afternoon of configuration, not a project.

Limits, stated plainly. The queue index is a brightness ratio, not a truck count: calibrating decibels to vehicle numbers needs ground truth from port records we do not yet have. The ship counts are detections, not identifications, and say nothing about vessel size beyond a pixel threshold or about cargo. Mirror statistics embed each partner's own valuation conventions, so partner-reported imports include freight and insurance that Bangladesh's own export figures would not. These are the instrument's error bars, and they are wide. But an instrument with wide error bars that reports every four days beats a perfect one that stopped reporting in 2018.

Sources

• Sentinel-1 SAR imagery: European Space Agency Copernicus programme, radiometrically terrain corrected scenes via the Microsoft Planetary Computer; scene counting and extraction code in scripts/satellite_trade/
• Ship detection water mask: JRC Global Surface Water v1.4 (Pekel et al.), via Google Earth Engine
• Approach road geometry: OpenStreetMap contributors, via the OSRM routing engine
• Mirror trade statistics: UN Comtrade, partner-reported flows with Bangladesh (partner2, transport mode, and customs dimensions held at totals); Bangladesh's own last annual submission is 2018, verified against the Comtrade data availability API
• Bilateral trade geography: BACI (CEPII), HS92, values converted from thousands of US dollars
• India monthly trade: UN Comtrade, India-reported monthly HS totals with partner Bangladesh
• World Bank border monitoring reference: "How Satellites and AI Can Transform Trade and Cut Congestion in Africa", World Bank, July 23, 2026