Schorlemmer et al. (2005) found that the Gutenberg-Richter b-value is highest for normal faults, lowest for thrust faults and between for strike-slip faults. We tested this relation out of sample in the Global CMT catalog. We fitted b as a linear function of the folded rake for shallow (centroid depth 50 km or less) Mw 5.7+ earthquakes in 1977-2003 (4,726 events) and tested the frozen model on 2004 to mid-2026 (4,838 events). The plan was written before the b-rake statistics. The training slope was -0.138 per 90 degrees of rake (95% tile-bootstrap CI -0.203 to -0.072). The test slope was -0.073 (-0.129 to -0.022), half as large. The frozen rake model gave no test log-likelihood gain over a constant b (+0.0002 nats per event, CI -0.0037 to +0.0029), so the pre-set transfer rule failed. With one b intercept per Bird (2003) plate-boundary setting, the test slope was -0.005 (-0.048 to +0.034); in training it was -0.098. Spreading ridges and oceanic transforms have high b and few thrust events; subduction zones have b near 1.0 and most thrust events. Inside subduction zones there was no rake slope in either period. In an exploratory comparison, a frozen setting model gave +0.0078 nats per event (CI +0.0032 to +0.0125) in the test period, and rake added nothing. The within-setting slope was negative in all three pre-2004 windows and near zero in all three later windows. This step coincides with GCMT procedure changes in 2004, which we did not test as the cause. Limitations: one catalog, Mw 5.7+ only, a coarse nearest-boundary setting label, strong rake-setting correlation, and a low-power test of the corner-magnitude effect. Code, plan, aggregate tables and figures are included; raw catalogs are retrieved by script.
Kisslinger and Jones (1991) reported faster aftershock decay (higher Omori p) where surface heat flow is high in southern California. We tested this in three samples with the IHFC Global Heat Flow Database (release 2024, v2026.03): California-Nevada ComCat (M4.0+, 1986-2025, 501 sequences), Japan JMA (M4.5+, 2000-2024, 176 sequences) and global ComCat (M6.5+, 1995-2025, 158 sequences). A pooled likelihood fit lambda = mu + K t^-p with a fixed pre-event background, free productivity per sequence, and p linear in log10 of the local median heat flow, with magnitude, depth and year controls. Intervals are 95% cluster-bootstrap intervals over spatial cells. The change in p per tenfold heat flow was +0.14 (-0.23 to +0.43) in California-Nevada, -0.59 (-0.88 to -0.09) in Japan and -0.29 (-0.43 to -0.04) globally. Thus the pre-set hypothesis of a positive slope was not supported in any sample, and two samples showed the opposite sign. Productivity per tenfold heat flow changed by a factor of 1.9 (0.5 to 6.9), 4.1 (1.9 to 9.1) and 0.42 (0.21 to 0.84), so it had no common sign. The Japanese slope remained outside the 2011-2020 Tohoku zone and in a low-background subset, but it disappeared when the background was set to zero. The global slope weakened with a low-background subset or a 200 km heat-flow radius. In California, a foreshock-free subset gave a negative slope. Heat flow also marks tectonic setting, so this study cannot isolate a temperature effect.
Question: does the aftershock productivity of earlier nearby mainshocks forecast the productivity of a new mainshock? Approach: we used the USGS ComCat global M4.5+ catalog (1973 to November 2026) and Global CMT mechanisms. We selected 3,486 shallow (70 km or less) M6+ mainshocks from 1980 to 2026 and counted M4.5+ events within 10 days and max(50 km, 2 rupture lengths). We made rolling negative-binomial forecasts for each test year 2000-2026 with models fitted only on earlier sequences (1,992 test sequences). The neighbour predictor is a shrunken mean of magnitude-adjusted log residuals of earlier mainshocks within 200 km whose windows closed at least 1 year before. Results: the neighbour term increased the predictive log-likelihood by 0.061 nats per sequence (95% CI 0.041-0.079) relative to magnitude and year. Depth, mechanism class and background rate together gave 0.058. The neighbour term added 0.049 (0.032-0.066) beyond all three and was positive in 26 of 27 test years. The calibration slope was 0.87. Observed/expected counts went from 0.41 in the lowest forecast quintile to 1.32 in the highest. Residual correlation between mainshock pairs fell from 0.29 at less than 50 km to near zero beyond 3,200 km. Within 200 km, it did not decrease over lags of 1-47 years. Most of the gain came from a 1,000 km regional scale; a 100 km term added 0.013 (0.006-0.022). Limitations: regional catalog detection and magnitude-scale differences can cause part of the signal. A detection proxy decreased the gain to 0.034 (0.020-0.047). With M5+ counts, the gain was 0.014-0.022, and the local 100 km component was uncertain (0.0045, -0.0003 to 0.0093). We did not use ETAS declustering and did not identify a physical cause.