On the applicability of available regression models for estimating Newmark displacements for low to moderate magnitude earthquakes. The case of the Betic Cordillera (S Spain)

Abstract

Newmark displacement estimation is generally computed using empirical models. These models are estimated from large datasets that mainly comprise moderate-to-high magnitude events (Mw > 6.0). In this work, we study the performance of several of these models to study moderate-to-low magnitude scenarios. For this purpose, data from the Betic Cordillera, S Spain, with magnitudes ranging from Mw 3.5 to 6.3, were used to compare with model predictions. The results show that errors in the estimates depend on the magnitude of events or on the yielding acceleration considered to estimate the displacement. The availability of an appropriate range of data (magnitude and yielding acceleration), when defining the regression model, may overcome the differences due to specific characteristics of the seismotectonic context of the area where data derives from. The results also show that performance of models including several ground motion predictors is better than those based on a single parameter, regardless of the combination these predictors. Furthermore, regression models with polynomial forms present a better performance than other functions based on the logarithm of these predictors. Finally, new specific models for the Betic Cordillera are proposed, especially suitable for low magnitude events (< 5.0) and low critical accelerations (< 0.1 g), based on simplified polynomial forms of models.