2014 Volume 14 Issue 4 Pages 4_21-4_34
Current Japanese earthquake early warning system uses data from two seismic networks: Japan Meteorological Agency (JMA) accelerometer network and high sensitive velocity-type seismometer network (Hi-net). These two datasets are processed in the different scheme and the results are merged to issue a warning. Combining these two datasets and processing in the same framework should improve the accuracy and speed of the warning. In this study, we tried to develop a method to use these two dataset in the same framework. We applied time-domain recursive filters to correct instrumental response of Hi-net sensors and adjust them to the response of mechanical seismometers. We evaluated the saturation of the Hi-net data with the data in 2 month after the Tohoku earthquake, and found the variation due to the saturation was smaller than the variation of attenuation relationship. Therefore, the Hi-net data includes important information when other data are not available yet. Speed of the warning improved by 3.6 seconds in the average for the inland earthquake by combining these two networks.