Near real-time automatic moment magnitude estimation

Abstract

In this paper we describe a stable automatic method to estimate in real time the seismic moment, moment magnitude and corner frequency of events recorded by a network comprising broad-band and accelerometer sensors. The procedure produces reliable results even for small-magnitude events $\hbox {M}_{\mathrm{W}}\approx 3$. The real-time data arise from both the Transfrontier network at the Alps-Dinarides junction and from the Italian National Accelerometric Network (RAN). The data is pre-processed and the S-wave train identified through the application of an automatic method, which estimates the arrival times based on the hypocenter location, recording site and regional velocity model. The transverse component of motion is used to minimize conversion effects. The source spectrum is obtained by correcting the signals for geometrical spreading and intrinsic attenuation. Source spectra for both velocity and displacement are computed and, following Andrews (1986), the seismic moment and the first estimate of the corner frequency, $f_{0}$, derived. The procedure is validated using the recordings of some recent moderate earthquakes (Carnia 2002; Bovec 2004; Parma 2008; Aquila 2009; Macerata 2009; Emilia 2012) and the recordings of some minor events in the SE Alps area for which independent seismic moment and moment magnitude estimates are available. The results obtained with a dataset of 843 events recorded by the Transfrontier and RAN networks show that the procedure is reliable and robust for events with $\hbox {M}_{\mathrm{W}}\ge 3$. The estimates of $f_{0}$ are less reliable. The results show a scatter, principally for small events with $\hbox {M}_{\mathrm{W}}\le 3$, probably due to site effects and inaccurate locations.