Integrating b-Value and Background Seismicity Rate for Spatial Earthquake Forecasting in the Alborz Region, Northern Iran

Abstract

In this study, we evaluate the spatial forecasting skill of the $b$-value and background seismicity rate $\mu$ across the Alborz region using a homogenized catalog of 23,961 earthquakes ($M \geq 1.5$) recorded by the Iranian Seismological Center between 2006 and 2024. Forecast performance for $M \geq 4.0$ and $M \geq 4.5$ is assessed using Molchan error diagrams, probability gain, probability difference, and the modified area skill score. The results show that $\mu$ provides a consistently strong spatial signal, with Molchan curves well below the random baseline and probability gains of 5--6 at low alarm rates, reflecting the persistent clustering of seismicity along major Alborz faults. The $b$-value exhibits limited skill at lower magnitudes but improves steadily with increasing magnitude; its skill score becomes positive above $M \approx 5.3$, indicating that $b$-value anomalies begin to capture meaningful stress concentrations only for larger events. Spatial patterns reveal low $b$ zones along active reverse and strike-slip structures and high $\mu$ zones following long-term seismicity clusters, underscoring their complementary physical roles. Retrospective testing confirms this complementarity: the combined $b$--$\mu$ forecast achieves detection rates of 0.81--0.83 at spatial alarm rates of 0.43 and 0.36 for $M \geq 4.0$ and $M \geq 4.5$, respectively, representing the most efficient forecast configuration among all tested models. These findings demonstrate that integrating stress-state and tectonic-loading indicators yields a more efficient and physically grounded framework for operational earthquake forecasting in the Alborz region.