Eighty-One Cubic Kilometers of Water the Cascades Never Reported

A Well You Never Have to Drill

Start with the obvious question. How do you find water a kilometer down, in rock nobody has ever poked a hole through looking for it?

The normal answer is a well. Sink a pipe, hit the water table, measure. That is expensive, slow, and gives you exactly one point, which is a poor way to inventory a mountain range. The 2025 study, published in PNAS by Leif Karlstrom and eleven coauthors, used something else entirely: heat [1]. Every rock on Earth has a background temperature that climbs steadily as you go down, because the planet's interior is hot and heat conducts outward. Drill a hole anywhere dry and boring, and the deeper thermometer reads warmer, in a straight, dull, entirely predictable line.

Water breaks that line. Cold rain and snowmelt sinking through a volcano carries its coldness with it, physically, faster than the surrounding rock can conduct heat into it from below. Where that is happening, a thermometer lowered down the hole finds temperature refusing to rise the way it should — sometimes staying nearly flat for hundreds of meters, occasionally even reading colder at depth than closer to the surface. The rock is not lying. It is telling you, honestly, that something colder than it is moving through it right now.

That is the entire method, in one sentence: find the boreholes where the temperature-depth curve goes wrong in that specific way, and you have found groundwater without ever asking a hydrologist's traditional question of where the water table sits.

What a Thermometer Actually Sees

Push the intuition one step further and it becomes a real physical model, the kind hydrogeologists have used since the 1960s to read vertical groundwater flow directly out of a temperature log, without a flow meter anywhere in the hole. Treat heat in a column of rock as moving two ways at once: conducting upward the ordinary way, and being carried, or advected, by whatever fluid is physically flowing through the pore space. The steady-state balance between the two gives a temperature profile shaped like

where $z^{*}$ is depth as a fraction of some reference depth, $\theta$ is temperature scaled the same way between 0 at the surface and 1 at the bottom, and $\beta$ is a dimensionless number built from the fluid's vertical velocity, its heat capacity, the rock's thermal conductivity, and that reference depth. Set $\beta=0$, no flow at all, and the fraction collapses to $\theta=z^{*}$: the boring straight line. Make $\beta$ large and positive, meaning water is being driven downward, and the curve does something specific — it stays pinned near zero for most of the profile, refusing to warm, and then bends sharply upward only near the bottom. That shape, flat on top and steep at the base, is a downward-advecting borehole's signature, and it is qualitatively exactly what the Cascades holes show: near-isothermal sections in the upper hundreds of meters that eventually rejoin a normal gradient at depth [1]. Our interactive companion lets you drag that $\beta$ yourself and watch the bend appear.

None of this is a new trick invented for this paper. USGS teams were running heat-flow surveys across this exact stretch of the Cascades back in the 1980s to trace hydrothermal circulation [9,11], and Michael Manga used precisely this advective-heat-transport reasoning in 1998 to explain why some Oregon Cascades springs discharge unusually cold, low-temperature water: the water is moving too fast through the rock to pick up its heat [8]. What the 2025 paper adds is scale — compiling that same style of reasoning across 72 holes instead of one or two, and turning a qualitative diagnosis into a volume.

The paper does not report fitting this specific type curve; its own diagnostic is simpler and more conservative. Any section of a hole, below 30 m to avoid near-surface seasonal noise, with a measured gradient under 20 °C per kilometer gets flagged as advective [1]. That threshold is a real number from the paper, not an invented one, and it is deliberately loose: 20 °C/km is well below what conduction alone would ever produce in this rock, so a flagged section is unambiguously telling you water is there, even before anyone tries to fit a curve to exactly how much.

Seventy-Two Holes, One Number: 6 Percent

A thermometer tells you where water is moving. It does not, by itself, tell you how much is standing there. For that the team needed a second number: how much of the rock's volume is actually pore space that water can occupy.

They got it the ordinary way, from a physical drill core, at the same Santiam Pass hole, SP 77-24, that anchors the temperature story. Porosity there averages 6%, with local patches running as high as 15 to 20% where the rock is more vesicular or fractured [1]. Six percent does not sound like much until you remember it is being multiplied against roughly a cubic kilometer of rock at a time. Combine that porosity with the depth extent of the near-isothermal sections, krige the pattern across all 72 compiled drill-hole profiles to interpolate between them, and the standing volume comes out to at least 81 cubic kilometers of actively circulating groundwater beneath the High Cascades crest [1].

The authors call this a lower bound and mean it literally, for two stated reasons: their drill holes do not map the full lateral extent of what they call the "rain curtain" — a term and a phenomenon first described at nearby Newberry Volcano in 1988, from core-hole drilling that found the same kind of high-elevation infiltration belt along a different Cascades volcano's flanks [10] — and their calculation makes no attempt to account for water flowing back upward, only downward advection [1]. Undercounting in both directions pushes the true number up, not down.

The rock they compiled those 72 holes from was mostly drilled for geothermal exploration, and a 1994 USGS survey of the range's hydrothermal systems had already mapped a great deal of it before anyone was asking about groundwater storage specifically [7]. Old data, asked a new question.

A second, independent estimate in the same paper approaches the question from the other end entirely: multiply the aquifer's recharge rate by how long water actually spends underground before it resurfaces. Regional recharge here runs 0.1 to 1 meter per year through an aquifer on the order of a kilometer thick, fed by precipitation reaching roughly 2 meters a year on the wet, west-facing side of the arc [1]. Tritium dating of spring water gives a mean transit time of 7 to 10 years [1]. Multiply flux by residence time and you get an independent estimate of about 35 cubic kilometers [1] — smaller than 81, built from a completely different set of measurements, and a useful reality check even where the two numbers disagree.

The State Shift: Where the Water Actually Lives

The paper's real argument is broader than one volume estimate. It says the Cascades' subsurface hydrology undergoes a wholesale change of state as the volcanic rock ages, and that the 81 cubic kilometers is specifically a measurement of the young end of that spectrum, the part of the range where the state shift has not yet happened. The bedrock ages doing the sorting come straight off the USGS's own geologic map of the range [6], and the basic link between bedrock age, hydrology, and topography here is not brand new either — a 2010 study had already traced how hydrology and topography coevolve on this same basalt landscape, watershed by watershed [5]. What the 2025 paper adds is the deep, kilometer-scale part of that story that a surface survey could never see.

At the topographic crest, where eruptions are geologically almost yesterday — Sand Mountain at 3,000 years old, Four in One Cone at 2,000, Collier Cone at roughly 1,300 to 1,500, Blue Lake Crater at 1,900 years or more [1] — the rock is young enough that its original volcanic plumbing, the lava-flow contacts, rubbly flow tops, and cooling fractures, has not yet been sealed shut by chemical weathering. Water sinks up to a kilometer deep through that open architecture before it does anything else. Electrical resistivity imaging at Olallie North Spring shows the groundwater there literally tracing the geometry of individual lava flows on its way to the surface, rather than moving in the tidy radial pattern a surface hydrologist would expect from the topography above it [1].

Push west, into bedrock more than about 7 million years old at the H.J. Andrews Experimental Forest, and the paper describes the same rock's plumbing as having shrunk to a critical-zone thickness "on the order of meters" [1]. Older still, past 20 million years in the Western Cascades, the transformation is essentially complete. Weathering has had geological ages to clog the pathways that once let water sink freely, and the transition between the two regimes plays out across bedrock ages of roughly 0.5 to 3 million years [1]. Figure 2 draws that collapse as the paper describes it: not a gentle taper, but a state shift, most of it compressed into a narrow slice of geologic time.

The word doing the heavy lifting in the paper's own title is "Critical Zone" — the label geoscientists use for the living skin of Earth from the top of the vegetation canopy down to unweathered bedrock, a research agenda formally proposed as a coordinated "weathering engine" initiative back in 2004 [2]. That community already knew basalt loses its chemical reactivity as it ages, consuming steadily less atmospheric carbon dioxide through weathering the longer it sits exposed [3,4]. This paper's contribution is showing that the same aging process is also quietly switching off a kilometer of groundwater storage, not just a chemistry problem.

Why Nobody Had Counted It Already

Here is the part that makes this more than a neat measurement trick: this water was not simply hard to find. It was actively invisible to the tools hydrologists normally use, for structural reasons the paper is explicit about.

Conventional watershed hydrology assumes water moves downhill along the surface topography, gathering into streams whose discharge you can gauge and whose catchment area you can draw on a map. That assumption is baked into the flow-routing algorithms behind almost every regional water survey. It fails here because the water is not following the topography at all. It is following buried lava-flow contacts and rubble zones that can run for kilometers in directions that have nothing to do with which way the mountainside slopes, exactly what the Olallie North Spring resistivity imaging shows directly [1]. The paper states plainly that this infiltration-dominated hydrology produces drainage-area-to-discharge relationships that violate what standard routing tools expect [1].

The mismatch shows up as a number, and it is a large one. Add up discharge at every mapped spring across this stretch of the High Cascades and you get a flux of roughly 1 cubic kilometer of water per year. Estimate the same region's total groundwater movement from a regional water balance instead, working from precipitation and recharge rather than from where water happens to daylight, and you get roughly 11 cubic kilometers a year [1]. Eleven times more water is moving through this rock than is showing up at any spring anyone has gauged. The authors note that this segment of the High Cascades ends up looking, in total groundwater flux, remarkably like the Hawaiian Islands — another young, porous, wet volcanic terrain where the same infiltration-dominated behavior is well documented [1]. A stream survey counts springs. It has no way to see the other ten-elevenths.

Club Arithmetic: How Big Is 81 Cubic Kilometers?

Cubic kilometers do not mean much until you attach them to something. None of the numbers in this section are the paper's; they are ours, built from its reported figures and one outside, independently verifiable public number.

We went looking for a reservoir comparison, and found the U.S. Bureau of Reclamation's own published figure for Lake Mead: 28,945,000 acre-feet of storage at full pool, elevation 1,221.4 feet [12]. Converting that at the standard 1,233.48 cubic meters per acre-foot gives 35.7 cubic kilometers. Divide 81 by that and you get about 2.3. Call it two and a quarter Lake Meads, at full pool, sitting under the Cascades crest as groundwater nobody had inventoried.

In more everyday units, 81 cubic kilometers is 81 billion cubic meters, and at 264.172 US gallons per cubic meter (a fixed unit conversion, not a measurement) that comes to roughly 21.4 trillion gallons. The alternate, flux-based estimate of 35 cubic kilometers converts the same way to about 9.2 trillion gallons, and to 0.98 Lake Meads — close enough to exactly one full reservoir that it is a genuinely useful second anchor for the range this water probably occupies.

A volume only tells you how much is there. It says nothing about how fast it replaces itself, which turns out to matter more for anyone actually planning around it. Divide 81 cubic kilometers by the springs-only flux of about 1 cubic kilometer a year and full turnover takes roughly 81 years. Divide it by the regional water-balance flux of about 11 cubic kilometers a year instead, and turnover drops to roughly 7 years. Both are the club's own division of the paper's own numbers; the paper does not compute either.

Table 1. The full ledger. Top block is quoted directly from the paper [1]; bottom block is the club's own arithmetic on those numbers plus one public reservoir figure [12].

A Reservoir With No Gate

Every number above describes a real resource, but it is a strange kind of reservoir: one with no dam, no gate, no allocation schedule, and until 2025, no official acknowledgment that it exists at the scale it does.

The paper's own language about this is understated but direct: "large amounts of stored subsurface water represent a critical resource for human populations," and "the total volume of the High Cascades aquifer and hydrologic response times — key metrics for sensitivity of water resources to perturbations such as climate change — depend on Critical Zone state" [1]. Translated: how much water is down there, and how fast it can respond to a dry year, are not separate questions from the geology. They are the same question.

The authors flag one specific vulnerability. Springs fed by the young, high-flux Critical Zone respond proportionally more to interannual climate variability than springs draining older, more buffered terrain [1]. That cuts two ways at once, and it is worth sitting with the tension rather than resolving it. The 81 cubic kilometers behind those springs is enormous compared to any single year's demand, which argues for resilience. But the springs delivering it are, by the paper's own finding, more climate-sensitive than the ones people are used to relying on, which argues against assuming that size alone guarantees steady supply through a run of dry years. A reservoir can be large and still behave unpredictably at the tap.

The authors' own recommendation follows directly from that tension, not from advocacy: "given increased human pressure on aquifer resources and decreasing snowpack in a warming climate," they write, the sensitivity of these young-terrain springs "implies a need for better characterization of these systems" [1]. That is a scientist's way of saying nobody currently knows this water well enough to manage it responsibly, whatever the eventual answers about water rights, allocation, or drought-year use turn out to be. Those are policy questions the paper does not attempt to answer, and neither do we; what changes after this study is that the questions now have a number attached to them where before they had none.

The Part That Made Us Slow Down

One implication of the paper sits outside the water-supply story entirely, and it caught the club off guard on a first read.

If water is sinking a kilometer into young Cascades rock, it is sinking into rock that also happens to be where magma is rising toward the surface. The paper states that magma-water interactions could plausibly occur over at least that same kilometer of depth during the final stages of magma ascent [1]. That matters for volcanic hazard in a very concrete way: magma meeting groundwater on its way up is the standard recipe for a phreatomagmatic eruption — explosive, rather than the comparatively gentle effusive lava flows the region is best known for. Several of the very young features that mark the wet end of this study's age spectrum, including Sand Mountain, Four in One Cone, Collier Cone, and Blue Lake Crater, already carry geological signatures of exactly that kind of explosive, water-influenced eruption [1].

Put plainly: the same deep aquifer that represents an uncounted water resource also represents an uncounted variable in how the next eruption in this range could behave. Both facts come from the same 81 cubic kilometers. Neither one was on anyone's map before this paper.

What We'd Want to See Next

A field note is a good place to say what would actually settle the open questions, rather than pretend the paper closes them.

Reconcile the two volume estimates directly. The heat-flow method (81 km³) and the flux-times-transit-time method (35 km³) measure genuinely different things — one is standing storage, the other an implied throughput volume — and the paper is candid that both carry real uncertainty. A combined inversion using both datasets together, rather than reporting them side by side, would narrow the range considerably.

Map the rain curtain's actual lateral extent. The 81 km³ figure is explicitly bounded by where existing drill holes happen to sit, not by the true edge of the young, high-infiltration terrain [1]. A geophysical survey designed to find that edge, rather than repurposing decades-old geothermal exploration holes, is the obvious next step, and it could only push the number up.

Instrument the young-terrain springs through an actual drought. The paper's climate-sensitivity finding is inferred from existing interannual variability, not observed through a multi-year drought event with dense monitoring in place beforehand. That is precisely the kind of dataset that would turn "implies a need for better characterization" into an answer.

What We Actually Believe

We believe the volume estimate, with the caveats the authors themselves attach to it. It rests on real drill-hole temperatures, a real measured porosity, and an explicit, stated lower-bound framing that the authors do not walk back or oversell. A second, independently derived estimate landing within a factor of 2.3 of the first, using entirely different data, is a genuinely reassuring cross-check rather than a coincidence to explain away.

We believe the mechanism, because it is not exotic physics; it is the same conduction-versus-advection competition that governs why a metal spoon left in hot soup warms your hand faster than a wooden one, applied at the scale of a mountain range instead of a kitchen. Water moving through rock steals the show from heat conducting through rock, and a sensitive-enough thermometer sees the theft directly.

We do not believe "roughly three Lake Meads" is a number we can stand behind, and we said so above rather than quietly using it. Our own arithmetic, from a public and checkable reservoir figure, gives 2.3.

What stays with us is not the volume at all, in the end. It is that the number sat there, physically present in Cascades rock, discoverable from data that in some cases had already been sitting in a filing cabinet for decades, until somebody thought to ask what a temperature log is actually for.

References

1. 1.

   Karlstrom, L., Klema, N., Grant, G. E., Finn, C., Sullivan, P. L., Cooley, S., Simpson, A., Fasth, B., Cashman, K., Ferrier, K., Ball, L. & McKay, D. (2025). State shifts in the deep Critical Zone drive landscape evolution in volcanic terrains. Proceedings of the National Academy of Sciences 122(3), e2415155122. doi:10.1073/pnas.2415155122

2. 2.

   Anderson, S. P., et al. (2004). Proposed initiative would study Earth's weathering engine. Eos, Transactions American Geophysical Union 85, 265–269.

3. 3.

   Dessert, C., et al. (2003). Basalt weathering laws and the impact of basalt weathering on the global carbon cycle. Chemical Geology 202, 257–273.

4. 4.

   Börker, J., et al. (2019). Aging of basalt volcanic systems and decreasing CO₂ consumption by weathering. Earth Surface Dynamics 7, 191–197.

5. 5.

   Jefferson, A., Grant, G., Lewis, S. & Lancaster, S. (2010). Coevolution of hydrology and topography on a basalt landscape in the Oregon Cascade Range, USA. Earth Surface Processes and Landforms 35, 803–816.

6. 6.

   Sherrod, D. R. & Smith, J. G. (2000). Geologic Map of Upper Eocene to Holocene Volcanic and Related Rocks of the Cascade Range, Oregon. U.S. Geological Survey.

7. 7.

   Ingebritsen, S. E., et al. (1994). Hydrothermal Systems of the Cascade Range, North-Central Oregon. U.S. Geological Survey Professional Paper 1044-L.

8. 8.

   Manga, M. (1998). Advective heat transport by low-temperature discharge in the Oregon Cascades. Geology 26, 799–802.

9. 9.

   Ingebritsen, S. E., Sherrod, D. R. & Mariner, R. H. (1989). Heat flow and hydrothermal circulation in the Cascade Range, north-central Oregon. Science 243, 1458–1462.

10. 10.

    Swanberg, C. A., et al. (1988). Core hole drilling and the "rain curtain" phenomenon at Newberry Volcano, Oregon. Journal of Geophysical Research: Solid Earth 93, 10163–10173.

11. 11.

    Blackwell, D. D., et al. (1990). Heat flow in the Oregon Cascade Range and its correlation with regional gravity, Curie point depths, and geology. Journal of Geophysical Research: Solid Earth 95, 19475–19493.

12. 12.

    U.S. Bureau of Reclamation. Lake Mead at Hoover Dam: FAQs. Total storage capacity at full pool, 28,945,000 acre-feet. usbr.gov/lc/hooverdam/faqs/lakefaqs.html