The Amazon grows on some of the most exhausted soil on Earth. Its rain strips phosphorus out of the ground and flushes it to the sea — and every year a plume of pulverised Saharan lakebed crosses the Atlantic and puts almost exactly that much back. The rainforest is fertilised by a desert 5,000 kilometres upwind.
27.7
million tonnes of African dust settle on the Amazon basin every year — the seven-year average measured by NASA’s CALIPSO lidar, 2007–2013. Inside it: about 22,000 tonnes of phosphorus, the nutrient the forest cannot make and cannot hold on to.
The dust bridge · North Africa to South AmericaFluxes are annual means, CALIPSO 2007–2013
Dust plume, faint → denseWinter–spring track to the AmazonSummer track to the CaribbeanAmazon basin
01 · The source
A dead lake in Chad, doing an outsized share of the work
The Bodélé Depression is the floor of Lake Mega-Chad, which dried out roughly 7,000 years ago and left behind a bed of diatomite — the chalky, feather-light skeletons of freshwater algae.
10,800
km² of diatomite
The erodible sheet of ancient lake sediment, endlessly re-milled by the Bodélé low-level jet that funnels between two mountain ranges.
~100
dust days a year
On roughly one day in three the Bodélé is actively lifting dust — the most persistent dust source measured anywhere on Earth.
0.2%
of the Sahara’s area
Koren et al. put about half of the Amazon’s dust supply on this single spot — a patch equal to half a percent of the forest it feeds.
The map’s sharpest argument is about which dot matters. A 2020 satellite-trajectory study found that most Bodélé dust rains out long before mid-ocean, and that El Djouf — the sand sea straddling Mauritania and Mali, already half-way to the coast — is the preferred source of the dust that actually reaches South America.
Both dots are on the map, and the argument between them is live. What is not in dispute is the direction of travel: east to west, along a corridor that has been open for as long as the Sahara has been dry.
02 · The crossing
A three-kilometre-thick slab of hot, dry air
Dust does not drift across the Atlantic. It rides — inside the Saharan Air Layer, a distinct mass of desert air that lifts off the coast and travels intact for thousands of kilometres.
Source: NOAA Atlantic Oceanographic & Meteorological Laboratory. Cross-section is schematic; altitudes and wind speeds are the published figures.
The layer rides on top of the moist marine air rather than mixing into it, which is why the dust survives the trip. That same dryness, and the wind shear along its edges, is what makes the Saharan Air Layer hostile to hurricanes: the plume that fertilises a rainforest also flattens the storms that would otherwise be spinning up beneath it.
03 · The ledger
What leaves Africa, and what arrives
CALIPSO’s lidar counted the plume at two meridians and at both ends. Most of the dust is gone before landfall — and what the Amazon gets is the tail of the distribution.
Annual means, 2007–2013 (Yu et al. 2015). Mt = million tonnes.
Roughly a quarter of the dust that crosses 15°W never reaches the western Atlantic; it falls into the ocean, where the same iron and phosphorus feed plankton instead. Of what is still flying at 35°W, the Amazon intercepts about one part in five.
The rest keeps going. Forty-three million tonnes a year continue past South America into the Caribbean — which is why a summer sky over San Juan or Miami can turn the colour of weak tea, and why the sunsets go orange.
Read the spectrum strip below as the plume itself: dense over the Sahel, thinning across the ocean, arriving as haze.
04 · The payload
Twenty-two thousand tonnes of phosphorus, and a forest that needs every gram
Amazonian soils are old, deeply weathered oxisols. Heavy rain leaches phosphorus out of them continuously. Dust deposition and hydrologic loss land, remarkably, in the same range.
Dust input from Yu et al. 2015: 23 g P per hectare per year, range 7–39. Loss range from Vitousek & Sanford (1986), as cited by Yu et al.
+19%
canopy productivity
What two years of added phosphorus — and only phosphorus — did in the Amazon Fertilization Experiment. Nitrogen and base cations did nothing.
+29%
fine-root productivity
The same experiment. The forest is measurably hungry for exactly the element the dust carries.
~60%
of the basin
The share of the Amazon whose low-phosphorus soils that experimental plot is taken to represent.
One complication belongs on the map. A 2019 PNAS study that fingerprinted the aerosol arriving over the Amazon found that smoke from African savanna fires supplies up to half of the phosphorus delivered by African aerosol. The dust bridge is also, for part of the year, a smoke bridge — and burning-season smoke carries phosphorus in a form plants take up more readily than mineral dust.
05 · The pulse
The bridge swings north and south with the sun
There is no single dust route. The whole plume pivots with the Intertropical Convergence Zone: south-west toward South America in boreal winter and spring, due west toward the Caribbean in summer and autumn.
Where the dust lands, by season
December–April. The plume tips south. Dust load over the Amazon basin reaches up to 50 mg/m², arriving as pulse-like intrusions through the forest’s wet season — when rain washes it straight into the soil.
June–August. Emission peaks, but the track runs north. Dust over the Amazon drops below 10 mg/m² while outbreaks cross to the Caribbean every three to five days.
Yu et al. 2015; Atmos. Chem. Phys. 23, 9993 (2023).
June 2020 · “Godzilla”
The most intense African dust intrusion in the two-decade MODIS record. Aerosol optical depth hit 3.5 off West Africa and 1.8 over the Caribbean; the plume covered an area comparable to the continental United States.
In Puerto Rico, daily PM₁₀ peaked at 453 µg/m³ on 23 June 2020 — hazardous air, delivered from 5,000 km upwind.
We know the pulse is real because someone has been counting it since before the satellite era. At Ragged Point on the windward tip of Barbados — the next land east is Africa, 4,500 km away — aerosol filters have been changed almost daily since August 1965. It is the longest continuous record of atmospheric dust anywhere on Earth.
06 · The forecast
The bridge is projected to narrow
Warm the north Atlantic faster than the south, and the trade winds that carry the plume weaken. Models converge on less dust, not more. Modelled
Yuan et al. (2020), Geophysical Research Letters, and NASA Goddard. A model projection under continued warming — not an observation. More severe scenarios reach −60% by 2100, which would be a 20,000-year low.
Less dust means clearer Caribbean skies, fewer hazardous-air days in San Juan — and possibly more Atlantic hurricanes, since the dry Saharan Air Layer is one of the things suppressing them. It also means a thinner phosphorus subsidy arriving over a forest already being cleared and burned at the other end. The two ends of this map are joined; nothing changes at one without the other feeling it.
What to notice on the map
Four things the geography makes obvious
The plume is not a line, it is a fan. It leaves Africa narrow and arrives wide and faint. The figures at 15°W and 35°W are the same air mass shedding a quarter of its cargo into the open ocean mid-crossing.
The source dots are tiny and the sink is enormous. A patch of dry lakebed supplies a continent-sized forest. Nutrient geography does not respect area.
The two dashed tracks are the same plume in different months. Whether Africa fertilises the Amazon or hazes the Caribbean is decided by where the ITCZ happens to be sitting.
The ocean underneath is not a gap. The ~105 Mt that never make landfall fall into the tropical Atlantic, where the iron in them fertilises plankton. The dust bridge feeds two ecosystems, and the sea takes the larger share.
Sources & notes
Yu, H. et al. (2015) “The fertilizing role of African dust in the Amazon rainforest: a first multiyear assessment based on CALIPSO lidar observations.” Geophysical Research Letters 42, 1984–1991. — 182 / 132 / 43 / 27.7 Mt fluxes, 22,000 t phosphorus (23 g P ha⁻¹ yr⁻¹), 86% interannual swing, 2007–2013 mean.
NASA Goddard (2015) “NASA satellite reveals how much Saharan dust feeds the Amazon’s plants.” Press release accompanying Yu et al. (2015).
Koren, I. et al. (2006) “The Bodélé depression: a single spot in the Sahara that provides most of the mineral dust to the Amazon forest.” Environmental Research Letters 1, 014005. — ~half of the Amazon’s dust supply from ~0.2% of the Sahara.
Washington, R. et al. (2009) “Dust as a tipping element: the Bodélé Depression, Chad.” PNAS 106, 20564–20571. — diatomite bed ≈10,800 km², dust on ~100 days a year, Bodélé low-level jet.
Yu, Y. et al. (2020) “Disproving the Bodélé depression as the primary source of dust fertilizing the Amazon rainforest.” Geophysical Research Letters 47, e2020GL088020. — El Djouf (Mauritania/Mali) as the preferred trans-Atlantic source.
Barkley, A. E. et al. (2019) “African biomass burning is a substantial source of phosphorus deposition to the Amazon, tropical Atlantic Ocean and Southern Ocean.” PNAS 116, 16216–16221. — smoke supplies up to half the phosphorus.
NOAA AOML Saharan Air Layer overview — layer 2–2.5 miles thick with a base ~1 mile up, ~50% drier than tropical air, 25–55 mph (10–25 m/s) mid-level easterly jet at 2,000–4,500 m, outbreaks every 3–5 days, peak late June–mid August.
Francis, D. et al. / Yu, H. et al. (2021) “Godzilla” June 2020 dust intrusion: AOD 3.5 off West Africa and 1.8 in the Caribbean; PM₁₀ 453 µg m⁻³ in Puerto Rico on 23 June 2020. Atmospheric Chemistry and Physics 21, 12359–12383; BAMS 106 (2025).
Prospero, J. M. / University of Miami Ragged Point, Barbados — near-daily aerosol sampling since August 1965, the longest continuous ground record of atmospheric dust; nearest land to the east is Africa, 4,500 km away.
Yuan, T. et al. (2020) “Anthropogenic decline of African dust.” Geophysical Research Letters — tropical-Atlantic dust loading projected to fall by more than 30% (modelled projection, not a measurement).
Cunha, H. F. V. et al. (2022) “Direct evidence for phosphorus limitation on Amazon forest productivity.” Nature 608, 558–562. — adding phosphorus alone raised fine-root productivity 29% and canopy productivity 19% in two years.
Note on the map: it is drawn schematically, not to survey accuracy — coastlines, the plume envelope and the seasonal tracks are simplified. Every number and every place name on it is real and cited above. The only figure here that is a projection rather than a measurement is the mid-century dust decline in section 06, flagged as modelled.