Abstract
Observations of earthquake precursors via geomagnetic anomalies from ultra-low frequency (ULF) could provide an expectation for short-term earthquake prediction. However, there are still several obstacles in determining this precursor, one of which is the presence of a precursor bias if the earthquake occurs at a close time and location. To cope with this problem, we analysed six earthquakes with a magnitude >5 on Pagai Island, Mentawai, Sumatra, during 2020. These earthquakes have epicentres close to each other and occur within a short time (one month). This study used geomagnetic data recorded by the Magnetic Data Acquisition System (MAGDAS) magnetometer network at Sicincin station (SCN), West Sumatra, and Kepahiyang station (KPY), Bengkulu, with a span of one month before the earthquake. The anomaly of ULF emission was analysed using the power spectral density method at a frequency of 0.012 Hz for the H and Z components of the geomagnetic data. The onset time of the ULF emission anomaly was determined by the standard deviation value (pz + σ, pz – σ) of the SZ/SH power ratio polarization. The disturbance storm time index (Dst) was used to ensure that the anomaly occurred was not caused by geomagnetic storm. Also, the single station transfer function was used to calculate the azimuth, and the empirical formula between the anomaly amplitude and the magnitude was used to calculate the earthquake magnitude as the validation of the anomaly source. Several ULF anomalies as earthquake precursors were observed, especially at the KPY station. The estimated azimuth shows a good accuracy compared to the earthquake's azimuth, with an average value of 97.8%. Furthermore, the earthquake magnitude calculated from the ULF anomaly shows a good agreement with the actual earthquake magnitude, with an average accuracy of 98.13%.