NASA and France’s CNES space agency have delivered something hydrologists have wanted for decades: a direct, month-by-month accounting of how much water the world’s rivers gain and lose over a full year. During its first observation cycle, running from October 2023 to September 2024, the SWOT satellite logged a swing of roughly 313 cubic kilometers across the river reaches it monitors — 28% below the lowest estimate produced by existing models. The findings appeared in Nature on March 4, 2026.
That gap does not mean rivers hold less water than scientists thought. It means that for the first time there is a measured yardstick against which simulations can be checked — and the simulations, it turns out, drift a long way.
- First global benchmark: SWOT tracked storage changes in nearly every river on Earth using a single instrument, replacing a patchwork of separate altimeters and optical estimates.
- Models overshoot: Measured variability landed 28% under the lowest hydrological model estimate, prompting calls for substantial revisions to how global models represent surface water.
- The Amazon dominates: It gained and lost more than 172 cubic kilometers of water, far more than any other river, while the Nile barely budged.
What KaRIn Actually Measures
SWOT does not tally the total volume of a river; it tracks how that volume shifts. Its primary instrument, the Ka-band radar interferometer (KaRIn), captures a river’s width and the height of its surface simultaneously, bouncing microwaves off the water across two 50-kilometer swaths on either side of the spacecraft. From those readings, researchers derive how much volume each stretch of river gains or loses month to month.
The satellite launched in December 2022 from Vandenberg, California, with contributions from the Canadian and UK space agencies. The analysis drew on nearly 1.6 million observations spanning some 127,000 river reaches; the Nature note specifies 126,674.
The Amazon Leads, the Nile Surprises
No river came close to the Amazon, which gained and lost more than 172 cubic kilometers during the study period — enough, NASA notes, to fill 68 million Olympic swimming pools.
Across the Amazon, Mississippi, Orinoco, Yangtze, Ganges, Mekong and Yenisei, water levels varied by more than 10 meters between their annual highs and lows. At the opposite extreme sits the Nile, which oscillated by just 8.5 cubic kilometers, less than expected. NASA points to three possible explanations: upstream dams, a severe drought, and the learning curve of a brand-new instrument.
Why the Measured Signal Fell Short
Part of the discrepancy is tied to the extreme drought that gripped the Amazon basin during the observation window: with less water flowing in, there is less to gain and lose. But the authors’ conclusion extends well beyond a single dry year. Ohio State University, whose team took part in the work, says the comparison shows global models need major improvements in how they represent river and surface-water dynamics. The authors themselves acknowledge the data come from the mission’s early phase, carry uncertainty, and leave gaps over ice and floodplains.
Mapping the Shape of Riverbanks
A quieter second contribution of the study: near-global maps of active channel geometry, reflected in the paper’s title, “Wide-swath altimetry maps bank shapes and storage changes in global rivers.” One takeaway is that rivers carrying similar discharge can have very different shapes — a nuance that models tend to smooth over.
A Separate Estimate of Total River Storage
SWOT does not answer how much water rivers hold in total, but an earlier study from the same NASA group offers an estimate. In April 2024, research published in Nature Geoscience, drawing on data from 1980 to 2009 and roughly 3 million river reaches, calculated that the planet’s rivers store an average of about 2,246 cubic kilometers, with the Amazon accounting for around 38%, according to JPL.
That work combined models with about 1,000 gauging stations, and its authors anticipated SWOT would eventually put their numbers to the test. The two figures are not directly comparable: one is a modeled average for a different era and river network, the other a measured annual oscillation from orbit. What they share is the team behind them and the ambition — moving past simulations alone.
“We’re starting to answer some of the really difficult questions SWOT was built for,” said Cédric David, who leads the SWOT research team at JPL. “This is just the beginning.”
For Arnaud Cerbelaud, who co-led the study, the implications “go far beyond hydrology.”
The real significance here is not the headline figure but the fact that a measured figure exists at all. Water management, flood forecasting and drought planning all lean on models that SWOT shows can diverge substantially from reality — a point the authors make explicitly. One year of data, and an unusually dry one in the Amazon at that, is not enough to rebuild those models, but it is enough to stop treating them as ground truth. Waiting for several annual cycles before drawing conclusions about trends remains the prudent course. What can already be said is that satellite observation is shifting from supplementing hydrological models to correcting them.