Cross-catalog matched-window evidence for pre-mainshock mid-depth b-value differences in the Hindu Kush/Jurm deep-intermediate zone

Deep and intermediate-depth earthquakes beneath the Hindu Kush/Jurm/Pamir region constitute one of the most active continental seismic systems in Asia. This study tests whether the 365-day Gutenberg–Richter b-value of 150–250 km seismicity, estimated before a decision date, can retrospectively separate event-prone windows from matched quiet windows at a 31–60-day lead. A USGS pilot identified this H2a feature. A locked USGS audit yielded AUC = 0.721, p = 0.0002, Holm p = 0.0006, event-positive fraction 23/33, and jackknife AUC range 0.705–0.757. An independent ISC repeat produced AUC = 0.782, p = 0.0002, event-positive 32/45, and jackknife range 0.778–0.792. A magnitude-policy audit showed the ISC result was not driven by maximum-magnitude labeling: under a body-wave magnitude policy, H2a strengthened to AUC = 0.876, p = 0.0003, event-positive 49/54, and jackknife range 0.871–0.893. These findings constitute a cross-catalog retrospective risk-state separation clue. They do not establish operational earthquake prediction, exact timing, or public-warning capability.