Deep events had a larger M6.5+ share than intermediate events after fixed weights in nine common ComCat cells in 2000-2025

We tested whether broad geographic mix explains a larger large-event fraction at greater depth in a global ComCat snapshot. In nine common 20-degree cells in 2000-2025, the fixed-weight M6.5+ share among preferred M5.5+ events was 17.36% at depths of 300 km or more and 9.02% at 70 to below 300 km. The difference was 8.34 percentage points, with a conditional 95% three-year block interval of 3.54 to 13.52 points. Common support retained 1,140 intermediate and 687 deep events. Moment-family magnitudes and two other cell widths gave positive contrasts. The M7+ share among M6+ events gave a less certain result: 5.15 points, with an interval of -0.26 to 10.80. Thus the tested geographic adjustment did not remove the main contrast. This extends a prior campaign size-forecast test with direct standardized shares; the data overlap. Mixed magnitude methods, within-cell geography, depth errors, event sequences, data-selected support and conditional intervals prevent a causal depth or hazard claim. Code, aggregate evidence, uncertainty draws, an arithmetic audit, a figure and exact external-source provenance are included. External event records remain outside the package.

earthquake depth · magnitude distribution · ComCat · geographic adjustment · deep earthquakes · conditional uncertainty