Recent-hour counts gave no clear added Ridgecrest rate forecast skill on days 7 to 30 at magnitude 2.5 or more

We tested whether recent activity improved next-10-minute earthquake count forecasts in a fixed USGS Ridgecrest catalog. A rate based on the previous 24 hours formed the reference. We fit one recent-hour term on days 2 to 7 and froze its coefficient for tests on days 7 to 30 after the 2019 M7.1 event. The test had 3,312 bins and 387 events with preferred magnitude at least 2.5. The added term gave a total Poisson log score gain of 1.50 nats, or 0.00387 nats per event. A conditional 95% three-day block interval was -1.78 to 4.46 nats. This gives no clear added skill under the tested model. Other recent windows and penalties also had intervals that included zero. A later split gave a positive interval, while the magnitude 3.5 sample gave a negative interval. These checks share data and are not independent tests. The study links earlier clustering diagnostics to a new chronological forecast check. One known sequence, a short test, revised catalog data, simple rate models, and uncertainty conditional on fitted forecasts limit inference. The package contains exact source data, hashes, all forecasts and resamples, code, a figure, an independent count and score audit, and repeat instructions.

Ridgecrest · aftershocks · chronological validation · Poisson log score · recent event rate · null result