Estimation of an optimum velocity model in the Calabro-Peloritan mountains – Assessment of the variance of model parameters and variability of earthquake locations

Abstract

Accurate earthquake locations are of primary importance when studying the seismicity of a given area, they allow important inferences on the ongoing seismo-tectonics. Both, for standard, as well as for earthquake relative location techniques, the velocity parameters are kept fixed to a-priori values, that are assumed to be correct, and the observed traveltime residuals are minimised by adjusting the hypocentral parameters. However, the use of an unsuitable velocity model, can introduce systematic errors in the hypocentre location. Precise hypocentre locations and error estimate, therefore, require the simultaneous solution of both velocity and hypocentral parameters. We perform a simultaneous inversion of both the velocity structure and the hypocentre location in NE-Sicily and SW-Calabria (Italy). Since the density of the network is not sufficient for the identification of the 3D structure with a resolution of interest here, we restrict ourselves to a 1D inversion using the well-known code VELEST. A main goal of the paper is the analysis of the stability of the inverted model parameters. For this purpose we carry out a series of tests concerning the initial guesses of the velocity structure and locations used in the inversion. We further assess the uncertainties which originate from the finiteness of the available datasets carrying out resampling experiments. From these tests we conclude that the data catalogue is sufficient to constrain the inversion. We note that the uncertainties of the inverted velocities increases with depth. On the other hand the inverted velocity structure depends decisively on the initial guess as they tend to maintain the overall shape of the starting model. In order to derive an improved starting model we derive a guess for the probable depth of the MOHO. For this purpose we exploit considerations of the depth distribution of earthquake foci and of the shear strength of rock depending on its rheological behaviour at depth. In a second step we derived a smooth starting model and repeated the inversion. Strong discontinuities tend to attract hypocentre locations which may introduce biases to the earthquake location. Using the smooth starting model we obtained again a rather smooth model as final solution which gave the best travel-time residuals among all models discussed in this paper. This poses severe questions as to the significance of velocity discontinuities inferred from rather vague a-priori information. Besides this, the use of those smooth models widely avoids the problems of hypocentre locations being affected by sudden velocity jumps, an effect which can be extremely disturbing in relative location procedures. The differences of the velocity structure obtained with different starting models is larger than those encountered during the bootstrap test. This underscores the importance of the choice of the initial guess. Fortunately the effects of the uncertainties discussed here on the final locations turned out as limited, i. e., less than 1 km for the horizontal coordinates and less than 2 km for the depth.