Cascadia tremor in 2013-2026 followed the M2 tide about twice as strongly 48 h or more after local onset as in the first 24 h; northern California tremor showed almost no M2 response

Houston (2015) found that Cascadia tremor becomes more sensitive to tides during several days after it starts at a location, in six slow-slip events of 2007-2012. We tested this result in the PNSN tremor catalog (791,720 detections, 2009-2026). We counted detections per hour in 0.2 x 0.3 deg cells and split them into 4,484 local episodes. A conditional Poisson model removed each cell-episode-day level and the UTC hour of day. We fit one M2 peak phase per region with onsets in 2009-2012. Then we kept the phase fixed and measured the M2 log-rate amplitude a in 2013-2026 (3,772 episodes, 503,463 detections) as a function of time since local onset. In the test period, a was 0.077 (SE 0.013) in the first 24 h and 0.164 (SE 0.014) at 48 h or more. The difference D = 0.086 has a cluster-bootstrap 95% interval of 0.054 to 0.116. The same pipeline at 34 non-tidal periods gave D with SD 0.015 and a maximum of 0.030. The increase stayed inside daily-rate tertiles, without near-saturated hours, in long episodes only, before and after the 2017 catalog change, and for other cell sizes, gaps and onset rules (D = 0.073 to 0.120). It was clear in northern Washington and Oregon and uncertain in southern Washington. Northern California tremor, 47% of the test detections, showed almost no M2 response (a <= 0.045). Modulation also increased with the share of the local episode that had passed, and the data cannot separate elapsed time from accumulated tremor. Limitations: we used a fitted phase, not modelled tidal stress, so we give no sensitivity per kPa; the catalog is a 150-s detection grid.

tectonic tremor · episodic tremor and slip · tidal modulation · Cascadia · slow slip · M2 tide · out-of-sample test · placebo frequencies · conditional Poisson · PNSN