Old Faithful eruption intervals followed the previous year's river flow in 2001-2025, but same-year flow did not predict them and flow does not explain a 10-minute increase after 2015

Question: do Yellowstone geyser eruption intervals follow river discharge as reported by Hurwitz et al. (2008), who gave r = -0.98 between annual mean Old Faithful intervals and Madison River flow in 1967-2006? Approach: we built 105,015 cleaned Old Faithful intervals for 2001-2025 and July-September intervals for seven other Upper Geyser Basin geysers from the volunteer GeyserTimes database, and compared them with USGS daily discharge. We tested same-year, lagged and differenced relations, a 2002-2015 to 2016-2025 forecast, and the seasonal cycle. Results: same-year flow did not predict the annual Old Faithful interval out of sample (2007-2025, r = -0.35, 95% CI -0.76 to 0.26). Year-to-year changes of the interval followed the change in the previous year's flow (r = -0.80, n = 24, surrogate p = 0.0005; -1.74 min per 100 cfs; peak discharge lead 9-12 months). Electronic-logger and webcam data each gave this result, and a 2002-2015 fit predicted 2016-2025 changes with 58% lower error than no change. But the interval increased from 92.4 min (2015) to 101.6 min (2025), and flow explains only 1.7 of these 9.2 min; a source change explains about 1 min. The seven other geysers showed no flow relation and no common post-2015 change. The seasonal cycle (longest May-July) had a 1.3 min amplitude, half the 2008 value. Limitations: volunteer data with a 2016 source change, one geyser for the main result, and no test of the cause of the post-2015 increase.

geyser · Old Faithful · Yellowstone · eruption interval · streamflow · Madison River · hydrothermal · groundwater recharge · out-of-sample test · replication · GeyserTimes · snowmelt