Daily-mean water levels in 3,411 US wells did not step more often on earthquake dates than on placebo dates at any seismic energy density from 1990 to 2025

We tested if public USGS daily-mean groundwater levels step more often after earthquakes than on placebo dates. We used 4,031 wells (3,411 with a valid score; 14.2 million well-days, 1990-2025) and 47,064 ComCat earthquakes. We gave each well-day the seismic energy density e of the strongest event on that UTC day (Wang 2007 relation). A step is a robust z of 4 or more for the change between 5-day medians before and after the event. The placebo moves each event date by 30-400 days, with the same offset for all wells. The plan was fixed before analysis. Result: no e bin from 1e-8 to above 1e-2 J/m3 had a step excess after Holm correction. For e >= 1e-4 J/m3 (47,337 well-dates), 3.90% of well-dates stepped against 3.44% on placebo dates (95% range 2.47-4.65%, p = 0.20). The excess is -0.75 to +1.42 percentage points. Confined and unconfined wells, the |z| >= 3 threshold and a 30-day persistence test gave the same null. Post hoc, no single M >= 7 event date of 361 passed a false-discovery test. The largest responses after great teleseisms (Sumatra 2005, Iquique 2014, Haida Gwaii 2012, Tohoku 2011) were regional clusters of rises in northeastern states; the Haida Gwaii window contains Hurricane Sandy. Only near events (e >= 1e-3) showed a timing sign: 41% (95% CI 32-49%) of 201 steps had their largest daily change on the event days, against 28% for placebo steps. Main limitation: daily means detect only large steps (median threshold 0.28 m), so the test cannot see the small coseismic steps and oscillations that high-rate wells record. Network-scale earthquake-groundwater studies with daily data need precipitation controls or sub-daily records.

groundwater · earthquake hydrology · seismic energy density · dynamic triggering · USGS wells · placebo test · teleseismic · null result · Hurricane Sandy · daily values