Prompt identification of tsunamigenic earthquakes from 3-component seismic data

• Prompt identification of regional tsunamigenic earthquake is proposed.
• The technique utilizes single 3-component seismic record for identification.
• Artificial neural network is used for nonlinear inverse mapping.
• The technique is validated by estimating focal parameters of the earthquake.
• The technique is expected to be very effective in poorly instrumented regions.

An Artificial Neural Network (ANN) based algorithm for prompt identification of shallow focus (depth < 70 km) tsunamigenic earthquakes at a regional distance is proposed in the paper. The promptness here refers to decision making as fast as 5 min after the arrival of LR phase in the seismogram. The root mean square amplitudes of seismic phases recorded by a single 3-component station have been considered as inputs besides location and magnitude. The trained ANN has been found to categorize 100% of the new earthquakes successfully as tsunamigenic or non-tsunamigenic. The proposed method has been corroborated by an alternate mapping technique of earthquake category estimation. The second method involves computation of focal parameters, estimation of water volume displaced at the source and eventually deciding category of the earthquake. The method has been found to identify 95% of the new earthquakes successfully. Both the methods have been tested using three component broad band seismic data recorded at PALK (Pallekele, Sri Lanka) station provided by IRIS for earthquakes originating from Sumatra region of magnitude 6 and above. The fair agreement between the methods ensures that a prompt alert system could be developed based on proposed method. The method would prove to be extremely useful for the regions that are not adequately instrumented for azimuthal coverage.