Development of ground motion models from seismic monitoring data for decision support in geo-resource risk management

This study develops a probabilistic, locally calibrated ground motion forecasting framework for microseismic monitoring in the Val d’Agri Basin, Southern Italy. Monitoring low-magnitude, shallow seismicity in active geo-resource areas is essential for ensuring structural safety and supporting operational control. Traditional ground motion models often fail to accurately represent local attenuation characteristics, particularly at short source-to-site distances, which may lead to overestimation of shaking. To address this limitation, a data-driven and uncertainty-aware modeling approach was implemented. The analysis is based on more than 30,000 high-resolution recordings (1.2 ≤ ML ≤ 3.7) collected between 2020 and 2025 by the CMS–INGV monitoring network, ensuring dense spatial coverage. A minimum signal-to-noise ratio criterion was applied to exclude low-quality observations. Six alternative functional forms were evaluated to determine the most suitable magnitude–distance relationship for predicting peak ground acceleration (PGA) and peak ground velocity (PGV). Two parameter estimation strategies were employed: classical regression using the EPOS TCS-AH platform and a Markov Chain Monte Carlo (MCMC) framework based on the Metropolis–Hastings algorithm. The MCMC approach enables explicit quantification of epistemic and aleatory uncertainties, distinguishing total variability (σ) from between-event variability (τ). Model selection was guided by statistical performance indicators (R² and AIC) and physical plausibility. A three-parameter model (Model 5), incorporating magnitude and logarithmic distance, was identified as optimal. Results confirm geometrical spreading (logR) as the dominant attenuation mechanism. The probabilistic framework allows estimation of exceedance probabilities for PGA and PGV, aligned with Italian Monitoring Guidelines. These outputs can be integrated into a Decision Support Tool to support real-time seismic risk management.