California Groundwater Crisis

The problem

Central Valley farms lean heavily on groundwater, and the water table has been falling for decades. The question I wanted answered was simple: if it keeps falling at the recent rate, how long does the stored water last?

A widely cited 2012 study (Scanlon et al., PNAS) put the answer at about 390 years. That number is now old, and the loss has not slowed. I wanted to update it with the years since, and show how the number is built so that anyone can check it.

The series goes on to show who uses the water (agriculture is roughly nine-tenths of Central Valley groundwater use, slide 5) and what happens to households when wells run dry (slides 8 to 10).

What I did

1. Built the curve of stored water. From the early 1960s to 2014 I used modelled storage change from the USGS Central Valley Hydrologic Model. From 2006 to 2023 I used NASA satellite measurements (GRACE and GRACE-FO) as processed by California's Department of Water Resources. The two come from different methods, so where they join is approximate.
2. Started from the 2012 study's numbers. It counted 860 km³ of storage remaining in 2000, drawn down at 2.2 km³ a year. That is where 390 years comes from (860 ÷ 2.2).
3. Brought the storage up to date. I estimate about 71.6 km³ was lost between 2000 and 2023, which leaves about 788 km³.
4. Measured the recent rate. A straight line through the satellite series from late 2006 to the end of 2023 falls by about 3.9 km³ a year, roughly 1.8 times the rate in the 2012 study.
5. Divided. 788 ÷ 3.9 is about 202 years, counted from 2023.
6. Followed the people. The dry-well slides use the State's dry-well reporting records, with household costs sourced to the State Water Board, DWR and others. Every number on those slides is traced on the methods page.

Tools, and how they fit together

Python for the analysis and the charts, Google Sheets for converting the satellite data's units, and Figma to turn the charts into the slides. The numbers flow one way: public datasets, then the analysis, then the slides, with every caption written to say where its number came from.

Result

About 202 years of storage left at the recent rate, counted from 2023, roughly half the 2012 study's 390. Measuring the recent rate over different windows gives anything from about 200 to 245 years, so the honest summary is "a bit over two centuries, not a precise year".

This is independent work. It was not commissioned.

Limits

This extends a published model. I did not independently re-verify the original volume estimates. The satellite data and the hydrological model do not reconcile cleanly, and the estimate assumes the recent rate continues, which it may not: droughts have sped up the loss, and the Sustainable Groundwater Management Act could slow it.

The figure counts stored water, not usable water. As the water table falls, what is left is deeper, costlier to pump and can be of poorer quality, so the practical limit would arrive sooner. Land subsidence has also permanently reduced storage capacity (about 15% of what has been drawn, per USGS), which the calculation does not model.

Technical detail