Researchers trace a third of Central Valley groundwater loss to 122 named producers

The denominator does most of the quiet work in that first number. One-third of *climate-related* groundwater loss is a share of a slice, not of the whole drawdown ^[1]. Central Valley aquifers have also been pulled down by cropping choices, surface allocations and decades of pumping, and none of that sits inside the fraction. What the study isolates is the depletion that follows from a warmer, drier, earlier-melting hydrology, then asks how much of that belongs to 122 named companies ^[2].

The mechanism is a subtraction. Build a simulated West in which those companies emitted nothing after 1950, run it forward, and difference it against the world we got ^[11]. That design rules out attributing the loss to natural variability shared by both worlds, and it rules out the linearity assumption the simpler method has to make ^[10]. It does not rule out the model itself being wrong about how snow and soil respond to the warming you removed. Every number downstream inherits that.

The snowpack accounting is also lopsided in a way worth holding. Of roughly 35 billion cubic meters of snow water lost each year, the Pacific Northwest accounts for 27.0 cubic kilometres and California 5.1 ^[6][7]. That is 32.1 together, about 92 percent of the annual total in two regions ^[15]. The 15 percent figure is a regional average that mostly describes those two places, which matters if your basin is somewhere else.

The 40 to 64 percent band is the line a plaintiff's expert will read twice ^[4]. If that share transferred cleanly to streamflow alone, the full human-caused reduction in warm-season flow would sit between about 9.4 and 15 percent, being 6 divided by 0.64 and by 0.40 ^[16]. The source does not present it that way and the share is reported for supply and demand changes taken together, so treat that as a scale check rather than a finding.

The thing the volume figures do not tell you is when the water shows up. Peak flows arriving 5.6 days early on average, and up to 30 days early on the western slopes of the Washington Cascades, leave rivers thinner in the hot months even where annual totals hold ^[8]. In the Upper Colorado Basin the drying shows up in winter as well as summer ^[9]. Those are two different problems for an operator working from a rule curve, and only one of them looks like a shortage in an annual ledger.

There is a gap between what this measures and what a complaint has to allege. The published quantities describe the 122 as a bloc; the account of the work says the methods can link climate change to individual companies, but reports no per-company figure ^[18]. Water litigation names parties and apportions shares among them. The bloc total is the input to that argument, not the argument.