Our knowledge of Earth structure can be used to distinguish far-field seismic signals from near-field noise. In particular, earthquakes excite ground motion at the eigenfrequencies of the Earth's normal modes By observing the spectral levels in small bands centered on the known frequencies of the normal resonance peaks and comparing them to the spectral levels of ambient noise, it is possible to detect earthquakes at low frequenies without knowing anything about them at high frequencies. We have used this fact to develop an earthquake detection algorithm that compares the signal level in the mode bands, as measured by the modulus of zeroth-order moment of the Fourier spectrum, with that in the noise bands using an F test. The excitation of each mode over a global network of stations and of all the modes over the network is evaluated from a summed-score statistic using a binomial test and a Markov test, respectively, which are robust with respect to deviations from the Gaussian noise model.

We apply the technique to the Earth's fundamental spheroidal modes 0S8--0S43 as recorded by the IDA network for the 2-year interval 1978--1979. We calculate the level of excitation at 3-hour intervals using a time domain integration technique designed to optimize the signal-to-noise ratio for each mode. The algorithm detects 50% of all teleseismic earthquakes having seismic moments M0≥1¿1018 N m (Mw≥6.0), and it detects all events having M0≥3¿1018 N m (Mw≥6.3). Slow earthquakes, i.e., those having anomalously large characteristic durations, are observed to be enriched in low-frequency mode excitation relative to other events of comparable magnitude. Using this criterion, we have found a number of slow earthquakes that had not been previously identified as anomalous. Most newly identified slow earthquakes occurred on oceanic transform faults. In addition to the cataloged normal and slow earthquakes, we observe 27 episodes of substantial mode excitation that are reasonably well isolated from significant earthquakes. Such anomalies may represent ''silent earthquakes,'' events with propagation velocities sufficiently low that they do not generate globally detectable wave trains on high-frequency seismometers. ¿ American Geophysical Union 1990