Next‑Generation Ground‑Motion Models for Georgia and the Surrounding Caucasus Region

Abstract

The manuscript presents updated ground‑motion models (GMMs) for peak ground acceleration (PGA) and 5% damped spectral acceleration (SA) for shallow crustal earthquakes in Georgia and the surrounding area. These models are developed exclusively from the strong‑motion network of Georgia and its neighboring countries. The final dataset consists of 2262 high‑quality three‑component accelerograms with magnitudes ranging from 3.5 to 7.0 and hypocentral distances of up to 300 km. The developed GMMs incorporate moment magnitude, hypocentral distance, faulting mechanism, and site conditions represented by Vs30. Such a selection of the data reduces the heterogeneity associated with converting broadband velocity records into acceleration records. It also avoids mixing them with direct acceleration records, as in previous hybrid models. Regression analysis combined with a magnitude‑based weighting scheme was used to improve the representation of moderate‑to‑large earthquakes in the dataset. Model performance was evaluated using residual analysis and goodnessof‑fit statistics. The resulting models predict slightly lower PGA and short‑period SA amplitudes than the previously published hybrid models, particularly for periods shorter than 0.3 s, and exhibit stronger attenuation at intermediate and large distances. Site conditions and faulting mechanism remain important controls on ground‑motion amplitudes, with thrust‑fault earthquakes and soft‑soil sites producing the highest motions. The updated models show reduced inter‑event variability, with inter‑event standard deviations approximately 15‑20% lower than those of the previous hybrid models, indicating more stable and reliable predictions. These results suggest that GMMs derived exclusively from strong‑motion data provide a more robust basis for seismic hazard assessment and engineering applications in the Caucasus region.