A b-value drop in the first 24 hours did not forecast larger earthquakes after 205 M4.5+ events in US, Italian and New Zealand catalogs (2000-2025)

The Foreshock Traffic Light System (FTLS) says that a b-value drop of more than 10 % after an earthquake, relative to the local background, signals that a larger earthquake will follow. We tested this rule as a pre-planned forward forecast. We used 1,955 first-largest shallow M4.5+ earthquakes (2000-2025) from ANSS ComCat regional networks, INGV (Italy) and GeoNet (New Zealand), with b-positive b-values from the first 24 h and from the prior 10 years in a rupture-scaled circle. 205 triggers had sufficient data, and 18 of them were followed by a larger event within 30 days. The b-value change gave an AUC of 0.56 (95 % CI 0.38-0.69; permutation p = 0.20). Larger-event rates were 8.0 % for red, 9.9 % for yellow and 7.8 % for green; 46 of 50 red alerts were false alarms. The b-value change did not improve out-of-sample logistic forecasts beyond productivity and magnitude. After events not followed by a larger one, b did not increase (median +1.5 %, CI -2.6 to +4.8 %). A simulation with the same sample sizes shows that an FTLS-size effect (20 % drop before larger events, 10 % increase otherwise) would give AUC 0.81 (5-95 %: 0.71-0.89); the data exclude it. A weak effect (about a 10 % drop, AUC near 0.6) remains possible. Sensitivity settings (6-72 h, 5 d-1 yr, Aki-Utsu estimator) mostly give AUC values from 0.5 to 0.7 with wide intervals. Limitations: few larger events, a fixed volume rule instead of expert choices per sequence, mixed magnitude types, and no Japanese data.

foreshock traffic light system · b-value · b-positive · foreshocks · earthquake forecasting · aftershocks · ComCat · INGV · GeoNet · forecast verification · null result