The present study deals with rapid, automatic, estimation of some earthquake parameters (location, focal depth, and magnitude) in a region of rather high seismic activity, in quasi-real time, through the analysis of incoming broadband records. The method can be applied, in particular, in poorly instrumented countries with high seismic-risk potential. It can also be applied when the analysis of a very important flow of data requires rapid, sophisticated, preferably automatic, data processing. The method requires, as a minimum, a three-component broadband seismographic station and a sufficiently populated database, that is, an instrument operating for a time long enough to have accumulated an appropriate data set, used to construct the knowledge base. The more extensive the knowledge base, the better the accuracy of the method.
We proceed in several steps. First, applying the spars algorithm to the only vertical component, available waveforms are classified according to the source location taken from National Earthquake Information Center (neic) catalog; it results in the sorting out of a subset of waveforms/events which will not be included in the knowledge base. Second, each element of the knowledge base is validated according to the epicentral distance with respect to the reference station (and eventually the azimuth of the corresponding source). Third, new input waveforms are analyzed and compared with one or more elements of the knowledge base to estimate their source location and size. The method can be used to search for doublets (or multiplets); if multiplets are found, their location and focal depth can be determined by using a fuzzy event relocation method.
We have tested the capability of the proposed algorithms, processing (broadband) waveforms collected during four and half years at the geoscope broadband station pvc, operated by Institut de Recherche pour le Développement, formerly ORSTOM (ird) at Port Vila, Vanuatu. Among 650 events recorded at this station, 254 ones, meeting a good criterion of quality, have been sorted. The results show that, in a range of distances up to 1000 km, the method is capable of yielding, in a very short time, the location of the input event, the accuracy depending on the local density of known events in the vicinity. We also obtain a reliable estimation of the energy by measuring the maximum surface wave (or S-wave) amplitude, related to the classical magnitude msz.