Science1 distinct publisher3 min readUpdated
A machine-learning reconstruction finds severe, persistent groundwater losses across roughly 23% of Brazil, in the basins that supply much of the world's soy, sugarcane, coffee and oranges.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
A NASA-led team has reconstructed monthly groundwater changes across Brazil from 2002 to 2023 and found severe, persistent losses over about 1.97 million square kilometers, roughly 23% of the country's land area, spanning parts of the Sao Francisco, Parana, Southeast Atlantic and East Atlantic basins [17][14]. The work was published on June 3 in Science Advances, and it matters to procurement teams rather than to climatologists, because those basins hold the fields that produce most of the world's soybeans, sugarcane, oranges and coffee, plus a large share of its cotton and corn [16][4].
Groundwater supplies 55% of Brazil's water, and the study argues that many of its stores have shrunk enough over 25 years to face irreversible damage [1][2]. The drivers listed are drought, higher evaporation under a warming climate, and pumping for human use [3]. Some of the most critical aquifers went years without any recharge at all [18].
The method is worth understanding before the number is used in a model. Brazil holds the largest renewable freshwater reserves of any country, largely because of the Amazon Basin, but its monitoring network is thin next to the thousands of wells the United States and India operate [11][12]. The researchers trained AI models on records from 398 monitoring wells, satellite gravity measurements, and geology and land-cover data, then reconstructed groundwater at about 10-kilometer resolution [13][14]. Taking the paper's own figures, 1.97 million square kilometers at 23% implies a national area near 8.57 million square kilometers, or about one monitoring well per 21,500 square kilometers [2][3]. That is roughly 215 model grid cells per observed well [4]. The reconstruction is a defensible inference, not a measurement.
The authors are also careful about blame. The study did not attribute changes to specific pressures, though first author Augusto Getirana of NASA Goddard's Hydrological Sciences Laboratory said climate shifts clearly played a major role, with several of the sharpest declines following prolonged droughts and the 2015-2016 El Nino [21][10][22]. Patchy monitoring makes the agricultural, industrial and household shares hard to separate, but losses overlapped substantially with intense irrigation, urban and industrial demand, mining, land conversion and high well density [23][24]. Getirana's formulation: "climate creates the underlying water deficit, while extraction and land-use changes are probably amplifying its magnitude and duration in heavily used regions" [25]. Javier Tomasella of Brazil's National Institute for Space Research, who was not involved, told Live Science that climate change is the main cause and that droughts are lasting longer and intensifying across most of the country, particularly the central region and the Northeast [26][27].
The trade signal is not imminent. Getirana said declines probably will not reduce total agricultural output over the next 10 to 20 years, and that "the concern is what happens if these trends persist for several decades" [7][8]. He added that Brazil is one of the world's major breadbaskets, so its water security has consequences beyond its borders [9]. The second-order risk is electricity: experts told Live Science that depleted aquifers could collapse the river systems feeding Brazil's hydropower, pushing the country toward coal or biomass [6]. That would raise the energy cost of processing and crushing inside the same regions losing the water.
Watch whether the depleted basins post a recharge year, whether Brazil expands its well network so the next reconstruction rests on measurement, and whether the country's hydro output holds through the next El Nino. Also note that the study found no single national trend, with the Amazon and parts of southern Brazil flat or slightly wetter [20][19]. Sourcing shifts within Brazil are the cheap hedge, and they run out before the aquifers do.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
The groundwater depletion has been caused by drought, increased water evaporation due to climate change, and pumping for human consumption, the research suggests.
The study did not attribute groundwater changes to specific pressures, but Getirana said it is clear from the data that climate shifts played a major role in the observed patterns.
Compared with countries such as the U.S. and India, which have thousands of groundwater measuring sites, Brazil's network of monitoring wells is extremely sparse.
Getirana: "The most scientifically defensible interpretation is therefore that climate creates the underlying water deficit, while extraction and land-use changes are probably amplifying its magnitude and duration in heavily used regions."
Javier Tomasella, an environmental scientist and hydrologist at Brazil's National Institute for Space Research who was not involved in the study, agreed that the main cause of aquifer depletion in Brazil is climate change.
Due to patchy monitoring, it is difficult to untangle and quantify the impacts of agriculture, industry and households on Brazil's groundwater levels.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Peer-reviewed study, thin independent corroboration
The core findings rest on a named peer-reviewed paper in Science Advances with a NASA Goddard first author, quantified spatial extent (1.97 million km2, ~23% of Brazil), a stated method and period (2002-2023, ~10 km, 398 wells), and one outside hydrologist agreeing on climate causation. Evidence is capped by a single supplied publisher, no reported validation statistics or uncertainty bounds for the AI reconstruction, an extremely sparse in-situ base (about 215 grid cells per observed well), and the article's own acknowledgement that the study did not attribute losses to specific pressures.
No adoption signal in supplied sources
The single supplied source reports a research finding. It contains no releases, deployments, procurement, policy uptake, dataset or model distribution, pricing changes or disclosed usage of the reconstruction by water agencies, utilities or agribusiness, so no adoption level can be measured without guessing.
Framing runs ahead of the study's own caveats
The article's headline and lead ('very deep trouble', 'irreversible damage', 'Climate change is the culprit', 'most of the world's soybeans, sugarcane, oranges and coffee') sit above what the reported research establishes: the study did not attribute losses to specific pressures, the author expects no total production decline for 10-20 years and describes a gradual cost-of-access progression rather than aquifers drying up, and the crop-share superlative is unquantified. The framing window ('past 25 years') is also wider than the 2002-2023 reconstruction. The gap is moderate rather than severe because the spatial findings themselves are concrete and the piece does carry the author's hedges and a no-change/gain contrast for the Amazon and the south.
Mild institutional and traffic incentives, partly offset
Identifiable incentives are modest and disclosed in the source: the first author is a NASA Goddard researcher whose laboratory benefits from demonstrating the value of satellite hydrology and AI reconstruction, and the publisher is a consumer science outlet whose headline maximizes alarm. Offsetting factors are peer review in Science Advances, an explicitly uninvolved external hydrologist quoted, and no commercial vendor, funding round or product being promoted in the supplied text.
Moderate: solid primary study, one publisher, no adoption read
Confidence is limited by structural thinness rather than by contradiction: exactly one publisher and one article in the cluster, no access to the paper's own uncertainty figures, no adoption dimension at all, and a reconstruction that extrapolates from 398 wells. It is supported by a dated peer-reviewed venue, a named first author, explicit method parameters, quantified extent, and one independent expert corroborating the causal reading.
science
NASA finds hitchhiking microbes could persist at the lunar south pole, on a one-day clock3 distinct publishers
science
Feeds are triaging patients by engagement, and the proposed defence is literacy, not moderation1 distinct publisher
science
Ancient DNA keeps changing the sex of the richest graves. The backlog is the story1 distinct publisher
science
IBM's bottleneck was cold volume, and its answer is an 8-foot box you can bolt to another one1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.