The development of an earthquake early warning system (EEWS) has been done by applying the
extended integrated particle filter (IPFx) method using a strong ground motion network in the western
part of Java, Indonesia. The method was applied to continuous waveforms including 95 earthquakes (M
> 4 and seismic intensity ≥ II MMI) and to the one-day waveforms including the 2022 Banten earthquake
sequences. We used 190 stations divided into 99 conventional force-balanced accelerometer sensors
(FBA) and 91 Micro-Electro-Mechanical System (MEMS) sensors. Early warning criteria were given
when the number of picks was more than five and the maximum seismic intensity was ≥ 3.5 (IV MMI)
based on the seismic attenuation equation. This system has successfully detected 95 earthquakes and
provided warnings to 46 from 49 events with observed seismic intensity ≥ 3.5 MMI. The system also
successfully detected five earthquakes of magnitude ≥ 4.0 as converged earthquakes and provided
warnings to two events with an observed seismic intensity of ≥ 3.5 for the 2022 Banten earthquake
sequence. The IPFx method shows good accuracy to estimate earthquake source locations with median
errors of 12 km, 22.7 km, 0.27, and 0.62 for the epicenter, depth, magnitude, and seismic intensity,
respectively, relative to the Indonesia Agency for Meteorology, climatology, and Geophysics (BMKG)
catalog. The system takes 3.4 seconds after the first P-wave was detected, and it is 6 seconds faster than
the previous method tested by BMKG. Based on the good accuracy and speed in estimating earthquake
sources, the IPFx method has the potential to be developed into an earthquake early warning system in
Indonesia in the future.
Keywords:
Earthquake early warning, earthquake source, IPFx method.