Determining hypocentral parameters for local earthquakes under ill conditions using genetic algorithm

We demonstrate that GA-MHYPO determines accurate hypocentral parameters for local earthquakes under ill conditions, such as limited number of stations (phase data), large azimuthal gap, and noisy data. The genetic algorithm (GA) in GA-MHYPO searches for the optimal 1-D velocity structure which provides the minimum traveltime differences between observed (true) and calculated P and S arrivals within prescribed ranges. GA-MHYPO is able to determine hypocentral parameters more accurately in many circumstances than conventional methods which rely on an a priori (and possibly incorrect) 1-D velocity model. In our synthetic tests, the accuracy of hypocentral parameters obtained by GA-MHYPO given ill conditions is improved by more than a factor of 20 for error-free data, and by a factor of five for data with errors, compared to that obtained by conventional methods such as HYPOINVERSE. In the case of error-free data, GA-MHYPO yields less than 0.1 km errors in focal depths and hypocenters without strong dependence on azimuthal coverage up to 45°. Errors are less than 1 km for data with errors of a 0.1-s standard deviation. To test the performance using real data, a well-recorded earthquake in the New Madrid seismic zone and earthquakes recorded under ill conditions in the High Himalaya are relocated by GA-MHYPO. The hypocentral parameters determined by GA-MHYPO under both good and ill conditions show similar computational results, which suggest that GA-MHYPO is robust and yields more reliable hypocentral parameters than standard methods under ill conditions for natural earthquakes.