Journal of Seismology 1: 205–218, 1997. 205 c 1997 Kluwer Academic Publishers. Printed in Belgium. Seismic monitoring of explosions: a method to extract information on the isotropic component of the seismic source P. Campus 1 & D. F¨ ah 2 1 Geophysical Institute, Academy of Sciences of the Czech Republic, Bocni II/1401, 141 31 Praha 4, Czech Republic (present address: Dept. of Earth Sciences, Parks Road, Oxford OX1 3PR, England) 2 Swiss Seismological Service, ETH-Hoenggerberg, CH-8093 Zurich, Switzerland Received 17 January 1997; accepted in revised form 1 September 1997 Key words: moment tensor, waveform inversion, modal summation, explosions, monitoring, lateral heterogeneity, point source, source time function Abstract The design of a monitoring system for detecting explosions is a very topical problem, both for routine data processing at seismological observatories as well as for the monitoring of a Comprehensive Test Ban Treaty. In this framework it is desirable to have the possibility to quantify the presence of the isotropic component in the seismic source. For this purpose a method is presented, which is based on waveform inversion for the full moment tensor retrieval. The method inverts either full waveforms or separate seismic phases and returns the mechanism and time history of a point source. Moreover, it allows to redefine the hypocentral depth of the event and, in a simplistic way, to optimize the structural model as well. In order to model strong laterally heterogeneous structures, different pairs of structural models can be used for each source-receiver path. The source is decomposed into a volumetric part (V), representing an explosive or implosive component, and into a deviatoric part, containing both the double couple (DC) and the compensated linear vector dipole (CLVD) components. The method is applied to an area in central Switzerland and to the network of the Swiss Seismological Service. The events of interest include both earthquakes and explosions. Despite some modelling inadequacies of the source-time function, the explosions can be well identified with the inverted isotropic component in the source, as long as the number of stations used for the inversion is larger than three. The results of the inversion are better for large epicenter-station distances of the order of 40-90 km. Introduction The problem of the retrieval of the source mechanism associated with a seismic event has been studied for many years and a considerable number of methods dealing with the inversion of seismic waveforms has been developed during the last ten years. Since the pioneering paper of Gilbert and Dziewonski (1975), who inverted free oscillation data to retrieve the source moment tensor of large earthquakes, many approach- es have been used to study the earthquake source by inverting waveforms, mainly associated to large teleseismic events (e.g., Langston, 1981; Dziewonski et al., 1981, Nabelek, 1984). When dealing with the problem of detecting explo- sions for the monitoring of a Comprehensive Test Ban Treaty it is highly desirable to quantify the presence of the isotropic component in the seismic source not only for events at teleseismic distances but also on a local scale. Therefore, a method able to identify non-double couple components in the source and using data record- ed by local networks is needed. It is often impossible, especially on local scale, to study the source mech- anism with the standard techniques based on P-wave polarities, unless the network is very dense. When the number of available stations is small, a suitable tool for source studies is the waveform inversion technique, using the dominant part (S- and surface waves) of the records. This allows to retrieve not only the fault-plane