GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 1.1 GNGTS 2023 ● to calculate the event's M L with two different statistical methods (Huber Weighted Mean and outliers’ removal) Amplitudes In the present work, we only analyze waveforms where at least a P-pick is present. This approach allows a more precise setting of spectral analysis and search window. Automated signal-to-noise ratio analysis on each channel defines, when successful, lower and upper corner frequencies for adaptive band-pass pre-filter, before conversion to WA. We use this adaptive approach to reduce as much as possible the bias introduced by static filtering on differential frequencies related to different source-receiver distance, source magnitude and complexity, and the kind of interfering noise. Noise and signal windows are defined using P-onset and S-onset (or S-traveltime when a revised S-onset is not available), as reported in Figure 1a. The maximum elongation is searched for within a signal search interval that is set to: −−−− | | + 40(1 − 10 − 40 ) Where S tt is the S theoretical travel time. Within this interval, two alternative search methods are used. The first method, shortened as lmag (Fig. 1b), is the same approach used in Earthworm’s localmag module ( USGS, 2001 ) : the difference between the downward and upward maximum peaks are found in moving and overlapping windows of, typically, 0.8s, building a set of measurements among which the largest is taken as the maximum amplitude for the channel; in this approach the result strictly depends on the moving window size. The second method, shortened as swing , used for example in Di Bona 2016 , only looks for consecutive (i.e. on the same swing) downward-to-upward (or viceversa) peaks. The differences between these two methods on the resulting peak-to-peak amplitude is investigated on a 1-year subset (Fig. 1c). The results show very small differences with amplitudes systematically lmag ≥ swing. Magnitudes Amplitudes contributing to the event’s M L estimation are quality selected based on the results of the signal-to-noise analysis. The hypocenters used for the presented results are those from CLASS catalog of the Italian seismicity ( Latorre et al., 2023 ) in the time lapse 2009-2018 when a direct link with INGV magnitude estimation is always present. The station's M L is calculated by using the geometric mean of the horizontal channels’ amplitudes within two different attenuations laws. The first is the Hutton and Boore 1987 (HB), adapted to the Italian region with no station's corrections, while the second is Di Bona 2016 (DB), specifically calculated for the Italian region with stations corrections at 400 stations (extension by Mele and Quintiliani, 2020). The hypocentral distances validity range is 10-600 km for both. HB was chosen because it was (and still