The freshwater that defines life on this planet — every river system, every Great Lake, every reservoir, every glacial meltwater pond — accounts for a vanishingly small fraction of the water Earth actually contains. The largest reservoir is not at the surface. It is not even liquid in any conventional sense. It is locked inside the crystal structure of a blue-tinted high-pressure mineral called ringwoodite, sitting between 410 and 660 kilometres beneath the crust, and the best current estimates suggest it holds more H2O than all the planet’s surface freshwater put together, possibly by a wide margin.
The conventional picture of Earth’s water budget puts oceans at roughly 97 percent of the total and freshwater at about 3 percent, most of that frozen in ice caps. That accounting describes only the hydrosphere — the thin film of water riding on top of the silicate planet. It leaves out the mantle entirely.
What changed the picture was a single diamond, dredged up from a riverbed in Juína, Brazil, and the careful spectroscopic work done on the tiny inclusion trapped inside it.
A diamond from 660 kilometres down
The diamond was unremarkable to look at — small, brown, commercially worthless. What made it scientifically extraordinary was a microscopic green-blue inclusion sealed inside its lattice. The inclusion was identified as ringwoodite, a high-pressure polymorph of olivine that had been synthesised in laboratory diamond anvil cells since the 1960s but never before found in a sample originating from inside the Earth.
Ringwoodite is stable only under the pressures and temperatures of the mantle transition zone, the layer that sits between the upper and lower mantle from roughly 410 to 660 kilometres depth. The Brazilian inclusion was, in effect, a piece of that zone that had been brought to the surface by a kimberlite eruption and preserved in diamond like an insect in amber.
When infrared spectroscopy was run on the inclusion, it was found to contain water by weight, bound into the crystal structure as hydroxyl groups. Extrapolated across the volume of the transition zone, that single measurement implied something staggering: if ringwoodite throughout the zone is hydrated at anything close to that level, the transition zone alone could hold the equivalent of several times the volume of all surface oceans, and many times the volume of all surface freshwater on Earth.
What the mineral actually does with water
Calling it “water” requires a careful definition. Ringwoodite does not host liquid H2O in pores or cavities. The hydrogen is incorporated into the crystal lattice itself, substituting for cations at defect sites and bonding to oxygen as hydroxyl. Bring a hydrated ringwoodite crystal to surface pressure and temperature and it would not pour out a glass of water — it would destabilise, recrystallise into lower-pressure phases, and release that hydrogen, which would combine with available oxygen to form actual H2O.
The mineral physics is well-characterised. As work catalogued in the Nature Index on water influence on mantle mineralogy describes, even trace amounts of hydroxyl dissolved in nominally anhydrous minerals like olivine, wadsleyite, and ringwoodite profoundly alter the mantle’s viscosity, electrical conductivity, melting behaviour, and seismic velocity. Wadsleyite, the phase that dominates between 410 and 520 kilometres, can hold up to about 3 weight percent water. Ringwoodite, dominant from 520 to 660 kilometres, can hold roughly 2.5 to 3 weight percent at saturation. The lower mantle minerals beneath — bridgmanite and ferropericlase — hold far less.
That contrast is critical. The transition zone behaves as a hydrogeological trap. Water carried down by subducting slabs gets locked into wadsleyite and ringwoodite. When those minerals are pushed across the 660-kilometre boundary into the lower mantle, where the dominant phases cannot accommodate the hydrogen, the water is expelled. It causes melting.
The seismic evidence for the wet zone
The diamond inclusion was one data point. The argument that the transition zone is broadly hydrated rests on a second, independent line of evidence: seismology. Seismic data from the USArray network captured wave behaviour at the base of the transition zone beneath North America, revealing a signature consistent with partial melt sitting on top of the 660-kilometre boundary — exactly what the model predicts if hydrated ringwoodite is being forced into the lower mantle and dehydrating as it crosses.
The detected melt zone is large and extensive. It is the seismic fingerprint of water being squeezed out of a mineral that can no longer hold it. The mechanism only works if the transition zone above is, in fact, hydrated.
How widely that hydration extends is the question that current research is trying to resolve. The Brazilian diamond and the USArray observations both sample regions associated with present or past subduction. Work from the Geodynamics Research Center at Ehime University on hydrous magnesium silicates has continued to map the mineral physics of how subducting slabs ferry water downward, including the role of aluminum-enriched dense hydrous phases in transporting H2O deeper than previously thought possible.
How the number compares to surface freshwater
The comparison the title makes depends on what counts as the surface reservoir. According to a UN Chronicle assessment of lakes in the global hydrological cycle, lakes are by far the dominant component of liquid surface freshwater, vastly outpacing rivers in standing volume. The total liquid freshwater held in lakes, reservoirs, and rivers worldwide represents a vanishingly small fraction of Earth’s total water inventory — a figure dwarfed by the water locked in ice sheets and glaciers, and the water estimated to sit in groundwater.
Even using only the liquid surface number — rivers, lakes, and reservoirs — the ringwoodite estimate runs orders of magnitude higher. If the transition zone is hydrated at the level inferred from the Brazilian diamond, the water mass contained there is on the order of one to three times the volume of the surface ocean — five to six orders of magnitude above the liquid surface freshwater total.
The framing matters because it reorders what we mean by Earth’s hydrology. A 2025 freshwater valuation effort led by the University of Nevada, Reno emphasises how poorly surface freshwater is accounted for even within human economic systems. The deeper reservoirs are not accounted for at all, because they are not, in any usable sense, available.
Why it is not a resource
Mantle water cannot be drilled. The deepest borehole humans have ever made — the Kola Superdeep in northwestern Russia — reached 12.262 kilometres, less than 3 percent of the way to the top of the transition zone. The pressures involved at 410 kilometres exceed 13 gigapascals; at 660 kilometres they approach 23 gigapascals. Temperatures run between roughly 1,500 and 1,900 degrees Celsius. No drilling technology that currently exists or is plausibly on the horizon can engage that environment.
What the mantle water does instead is regulate the planet over geological time. The cycling of hydrogen between the surface and the deep interior, mediated by subduction and volcanism, has likely buffered ocean volume on billion-year timescales. Some models suggest the modern ocean represents a near-steady-state balance: water descending in slabs roughly equals water ascending in arc volcanism and mid-ocean ridge degassing. The transition zone is the warehouse that makes the cycle possible.
Where the water came from
The origin of mantle water is itself unresolved. One school holds that it was delivered late, after the Moon-forming impact, by hydrated carbonaceous chondrite material — essentially the same source proposed for the surface oceans. Another holds that significant water was incorporated during accretion and survived the magma ocean stage by partitioning into the silicate melt and then into hydrous phases as the mantle crystallised.
Both mechanisms probably contributed. Nature Index summaries of water transport mechanisms in Earth’s mantle describe how subducting oceanic plates carry hydrous minerals and sediments downward, with the dense hydrous magnesium silicates and the nominally anhydrous transition zone phases together forming a relay system that can move water from the trench to depths exceeding 1,000 kilometres under the right thermal conditions.
The Brazilian diamond inclusion remains the single most direct sample. A handful of additional ringwoodite-bearing inclusions have been identified since 2014, most also from Juína. Each one is a physical fragment of the transition zone, preserved by the only geological process capable of capturing it intact and bringing it to the surface.
What the freshwater claim actually means
Calling it “freshwater” is a category that deserves a footnote. The hydrogen in ringwoodite is not part of a saline ocean — there is no dissolved chloride at depth, no marine chemistry. If that hydrogen were extracted and combined with oxygen at the surface, it would form H2O without salt. In that strict chemical sense the description holds. It is water that, were it returned to the surface, would not be seawater.
The qualifier that matters more is locked. The reservoir is not interchangeable with surface freshwater. It does not buffer drought. It does not refill aquifers on any human timescale. It does, however, suggest that the planet’s habitability owes something to a hidden hydrological architecture that was not even suspected fifty years ago and was not directly confirmed until 2014.
Earlier dispatches on this site have tracked other places where the planet’s water budget is being revised — a new global atlas finding glaciers hold less ice than previously thought, the anatomy of glacial ice loss, and the warnings from Central Asia where glaciers have disappeared completely. The mantle reservoir does not change any of those numbers. What it changes is the denominator. The surface system is the small part. The hidden ocean inside the rocks is the rest of the inventory, and it is the part that has been there, in equilibrium with everything else, for most of the age of the Earth.
