Simulated seismicity highlights the <i>b</i> -value as a potential earthquake precursor

SUMMARY Understanding when and where strong earthquakes occur is crucial for assessing seismic hazard. Changes in the b-value, which describes how frequently earthquakes of different sizes happen, have been investigated as possible indicators in the space–time vicinity of upcoming large earthquakes. In this study, we investigate short-term b-value variations in the central-northern Apennines (Italy) by comparing an observed earthquake catalogue spanning the period 1987–2025 with a 10 000-yr synthetic catalogue generated by a physics-based earthquake simulator. The synthetic seismicity is produced using a 3-D seismotectonic model derived from the DISS data base and an elastic-rebound framework for earthquake nucleation. We apply a stacking procedure to compute average b-values within symmetrical time windows of ± 15 d and 30 km distance from selected pivot events of moderate to large magnitude. The same methodology is consistently applied to both observed and simulated data sets, enabling a direct comparison of their temporal behaviour. The observed catalogue shows a statistically significant decrease in the b-value in the days preceding earthquakes with Mw ≥ 4.0, followed by a post-event recovery. A comparable pattern is reproduced by the synthetic catalogue, where a pronounced b-value drop precedes pivot events of Mw ≥ 4.5 and is systematically followed by an increase after rupture. The persistence of this behaviour across different magnitude thresholds in the simulated data supports its robustness. These results indicate that physics-based simulations can reproduce short-term b-value variations associated with earthquake nucleation, supporting the relevance of this parameter for investigating the physical processes governing seismicity.