“Flying rivers” (also called aerial rivers) above South America form organized drainage networks that structurally resemble terrestrial river systems, according to a 2026 study led by a National Taiwan University (NTU) interdisciplinary team and published in Nature Communications.

These are long- term preferential pathways of atmospheric moisture transport, not short- lived atmospheric rivers. Moisture enters from the ocean, is often supplemented by forest transpiration (evapotranspiration), and is carried by winds across the continent, eventually falling as rain far downwind. Deforestation or land-use changes in upwind areas can therefore reduce rainfall hundreds or thousands of kilometers away, sometimes across national borders.

Key findings

Researchers analyzed long- term atmospheric moisture transport over South America and developed a mathematical method based on “moisture drainage curves”. These curves track how moisture contributions from upwind areas accumulate.

They identified objective turning points on the curves: beyond these points, additional land area contributes progressively less efficiently to rainfall in a target region. This defines critical upwind basins more rigorously than earlier fixed-percentage approaches. The location of these turning points varies substantially across the continent.

The continental aerial-river system shows four distinct hydrological regimes analogous to land-based river networks:

• Headwater (G1): Entry points where ocean- enriched moisture first reaches the continent (not always strictly coastal; includes parts of the windward Amazon, Guiana Highlands, and areas of Chile/Patagonia).
• Drainage (G2): Areas where atmospheric drainage can increase and some water vapor is lost inland (e.g., bands from Colombia across the mid- Amazon and northeastern Brazil).
• Outfall (G3): Zones where aerial rivers slow abruptly, moisture content drops sharply, and regions rely more on continental evaporation; some lie upstream of major river basins and are particularly relevant for water management.
• Plain (G4): Farther downwind regions.

Two major continental systems progress from upwind to downwind through the G1- G4 sequence (one tropical system originates from the northeast and extends toward Paraguay and southern Brazil).

Implications

The four types differ in their potential for atmospheric moisture management and susceptibility to ecological tipping points.

This supports more targeted forest conservation and challenges traditional water-resource governance, which often focuses only on surface watersheds or administrative boundaries.

A region’s water security can depend on distant upwind forests. Lead author Prof. Wei Weng (NTU Department of Geography) and co-authors from atmospheric sciences and civil engineering emphasize that the turning- point method makes the concept of aerial rivers more practical for management.

What are “flying rivers”?

Not the short- lived atmospheric rivers you hear about in weather reports. The team defines aerial rivers as long- term, climatological pathways of moisture transport, the net, statistical flow that integrates all the daily breezes, low- level jets, and transient atmospheric rivers over years.

Like a surface river, they are:

• recharged by evapotranspiration in upwind regions
• routed by wind patterns
• discharged by precipitation downwind

Over South America, the Amazon acts like a giant pump, forest transpiration loads the trade winds, they hit the Andes like a wall and get diverted south, raining out over central and southern Brazil, Paraguay, La Plata.

The breakthrough: they have drainage networks like land rivers

Researchers analyzed moisture drainage curves for 724 grid cells across the continent, the relationship between how big an upwind area you include and how much of your rain comes from it.

Using a mathematical method to find the turning point where contribution intensity drops sharply, they found four major drainage types, G1 to G4, that behave like a fluvial system from headwaters to plains: 

G1 – Continental headwater regions
Where aerial rivers first enter the continent from the ocean, e.g. windward Amazon, Guiana Highlands, Chile/Patagonia. The critical upwind basin dominates: 90- 95% of terrestrial- sourced rain comes from a relatively compact ∼930,000 km² area. Only ∼12% of total rain is terrestrial here, so management potential is low.

G2 – Atmospheric drainage expansion
A west- east band from Colombia to mid- Amazon and NE Brazil. Drainage expands, dominance falls. Critical basin ∼410,000 km² supplies 60- 75% of terrestrial rain, about 20- 25% of total rainfall. This is where aerial rivers carry the largest amount of terrestrial moisture across the continent.

G3 – Aerial river outfall
Southern Amazon, Brazilian Highlands, Pampas, parts of Andes. This is where rivers slow abruptly, resembling river outfall areas. Delivery intensity plummets compared to G2, critical basin shrinks to ∼260,000 km² and 40- 55% of terrestrial rain. But these areas are often upstream of major surface river basins, so conserving forests here amplifies both sky and land water.

G4 – Atmospheric plain
Chaco, central-west Brazil, La Plata Basin. Furthest downwind, least efficient delivery, lowest per-area contribution. Turning point at only 30-40%. Yet because G4 depends most heavily on continental moisture, that small critical basin (∼140,000 km²) actually contributes the largest share of total rainfall, 15- 20%, making it the most effective target for intervention. Santa Cruz de la Sierra is the textbook case.

Two giant systems

The four types of sequence G1 → G2 → G3 → G4 in two super- basins:

1. Tropical system: from the northeast coast down to Paraguay and southern Brazil
2. Temperate system: from Patagonia up to La Plata Basin

They align with the austral Hadley and Ferrel cells, with a sharp “fall” between G2 and G3 at 5°S- 13°S where moisture delivery intensity collapses. They ultimately converge in La Plata.

Why it matters for water management

Current water law stops at watershed and political borders. Aerial rivers don’t. The authors show that deforestation in upwind Peruvian Ucayali, for example, can cut downwind rainfall 5-13% and runoff 19- 50% via aerial rivers.

Their framework says: get the drainage curve for your city, identify its type from the map, find the turning point, and you know your critical upwind basin, the upwind forest or land that functions like a headwater to protect. For G3 and G4 regions, targeted conservation or reforestation can secure up to a fifth of total rainfall.

Hydrological regimes and drainage systems of aerial rivers across South America

The team used 10 years of data (2006-2015), GPM rainfall, GLEAM evapotranspiration, ERA- Interim winds, in a moisture- tracking model (WAM- 2layers) to trace where the rain over 724 grid cells across South America actually came from.

Then they built moisture drainage curves, size of upwind area vs. % of terrestrial- sourced rain it provides and developed a math method to find the turning point where contribution intensity drops sharply. That’s how they define a critical upwind basin objectively, instead of using arbitrary 40% / 70% / 80% thresholds.

The 4 drainage types = a river system in the sky

They are geographic divisions shaped by long- term planetary transport, corresponding to fluvial positions: headwater, drainage, outfall, and plain. 

G1 – headwater regions: entry points of moisture from ocean. Windward Amazon, Guiana Highlands, much of Chile/Patagonia. Turning point 90- 95%, critical basin avg 930,000 km², but only 12% of rain is terrestrial, so management potential ∼11%.

G2 – atmospheric drainage: west- east band Colombia to mid- Amazon and NE Brazil. Expanding drainage. Turning point 60- 75%, critical basin ∼410,000 km², receives the largest terrestrial moisture load, 20- 25% of total rain manageable. 

G3 – outfall: southern Amazon, Brazilian Highlands, Pampas, Andes. “Occurs where aerial rivers slow abruptly, resembling river outfall areas”. Critical basin ∼260,000 km², 40- 55% contribution, but strategically in upstream areas of major surface basins, so interventions amplify through surface runoff.

G4 – atmospheric plain: Chaco, central- west Brazil, La Plata Basin. Furthest downwind, lowest delivery intensity across continent. Turning point 30- 40%, smallest critical basin ∼140,000 km², but highest dependence on continental moisture, so it actually contributes the largest share of total rainfall (15- 20%) for water- scarce cities like Santa Cruz de la Sierra.

From G1 to G4, basins decrease in both size and dominance, with a pronounced “fall” in delivery intensity at G2- G3 transition between 5°S and 13°S, interpreted as analogous to a river outfall.

Two continental systems

• Tropical system: NE coast → Paraguay and southern Brazil (much larger)
• Temperate system: SW origin Patagonia → La Plata Basin

They align with Hadley and Ferrel cells, plus low-level jets and terrain flows. They converge in La Plata.

Why “flying rivers” matter

Unlike surface rivers, aerial rivers are recharged by evapotranspiration. Deforestation in Peruvian Ucayali upwind can cut 5- 13% rainfall and 19- 50% runoff downwind via this link. 

The paper argues water management must move beyond catchment borders. Protecting a critical upwind basin in G3 and G4 is like protecting a headwater, and because in Amazon the downstream region lies upwind of aerial rivers, power relations flip: downstream forest users control upwind rain.

Journal information: Nature Communications  17, Article number: 9341 (2026)

DOI: 10.1038/s41467-026-76303-y

Provided: National Taiwan University

Authors: Wei Weng,
Ping Fu,
Ho Tin Hung,
Kai-Chih Tseng,
Li-Pen Wang,
Yun-Man Hsu,
Min-Hui Lo,
Luís Costa &
Vincent Dubreuil 

Abstract

Aerial rivers, long-term preferential pathways of atmospheric moisture flows, sustain ecosystems and regional hydrology. Maximizing their downwind discharge potential remains elusive without an objective way to identify critical upwind source regions. Here we analyze the drainage patterns of aerial rivers across South America and develop a mathematical method to identify turning points in moisture drainage curves, beyond which the intensity of upwind moisture contribution declines sharply. This provides a criterion for objectively delineating critical upwind basins and reveals pronounced continental variation in their spatial extent. These patterns reflect distinct regimes within the continental aerial river system, enabling its classification into four major drainage types: headwater, drainage, outfall, and plain areas. The four types differ in their potential for atmospheric moisture management and in their susceptibility to ecological tipping points, with implications for refining forest conservation priorities. Together, these findings support more effective land-water planning and moisture governance under a changing climate.