12,626 Publications · 75,762 Citations

Mason et al. (2004) reported that more volcanic eruptions start from November to April than from May to October in the Smithsonian GVP catalog for 1700-1999, and connected this to seasonal loading by water and sea level. We did an out-of-sample test with the current GVP catalog. We used confirmed eruptions with a known start month and a day uncertainty of 31 days or less. Volcano-cluster bootstrap intervals and a per-volcano random-rotation null control for repeated eruptions. In 1700-1999 (3,826 onsets, 431 volcanoes) the November-April share is 0.531 (95% CI 0.514-0.547; 0.496 expected), R = 0.054, the peak is in mid-February, and p < 0.001. This agrees with the published size. In 2000-2025 (938 onsets, 219 volcanoes) the share is 0.486 (0.455-0.519, p = 0.74) and R = 0.027. The change is -0.045 (-0.082 to -0.010). The regional mix predicts 0.534, and the 183 volcanoes with onsets in both periods changed from 0.547 to 0.490. Global and regional seasonal models fitted to 1700-1999 give no forecast skill for 2000-2025. The local seasonal cycles of rain (GPCP) and coastal sea level (PSMSL) give small effects that do not agree between periods. A constant physical loading effect of the published size is not supported. The cause of the pre-2000 excess (detection and reporting changes or a time-varying process) is not identified, and effects smaller than a share of about 0.51 cannot be excluded.