Depth gave no clear added size estimate gain beyond location for global M5.5+ events in 2013-2025

We tested whether depth improves later earthquake size estimates beyond coarse location. We fitted probabilities with 6,877 USGS ComCat events in 2000-2012 and tested 5,971 events in 2013-2025. The target was M6.5 or more among M5.5 or more events. Adding depth reduced Brier error by 0.509%. The conditional 95% three-year block interval was -0.003% to 0.969%. The primary test gave no clear added gain. Depth alone reduced error by 0.355% against a pooled probability. In a check for M7 or more among M6 or more events, added depth increased error by 0.327%. The results do not establish a general gain across size thresholds or identify a physical cause. They estimate size given occurrence and known location and depth, not event rates or hazard. Revised records, mixed magnitude types, coarse cells, clustered events and 13 test years limit inference. The package gives code, aggregate scores, a figure, an independent numerical check and external-data retrieval instructions. Source records remain external.

earthquake size · depth · chronological validation · Brier score · ComCat · null result