Correlation between earthquake occurrence and the anomalous propagation of VHF radio waves indicated by the gain and the p-value of prediction maps produced by a simple objective algorithm at the Shimabara area, Kyushu, Japan

Electromagnetic precursors associated with the impending earthquake, such as variations of geoelectric current, total electron contents in the ionosphere, and anomalous transmission of radio waves in the VLF or VHF band, have been observed (ex. Hayakawa, 1996). Recently, some researchers have discussed how these precursory phenomena relate statistically to the impending earthquake (Le et al., 2010; Orihara et al., 2012, Hattori et al., 2013, Han et al. 2014). The observation of anomalous VHF-band radio-wave propagation beyond the line of sight prior to earthquakes is one of candidate method to predict an earthquake. It is considered that the anomalous propagation events were the result of scattering of VHF-band radio waves in preparatory process of immediately prior to earthquakes occurring around the observation area (Kushida and Kushida, 2002 Moriya et al.,2010). And, quantitative correlation between logarithm of the total duration of scattered wave transmission and the magnitude, or maximum seismic intensity has been proposed (Moriya et al., 2010). Hokkaido University monitors this anomalous VHF-band radio-wave propagation at some region which supposed to have the possibility of large earthquakes in Japan. Mw 7.0 earthquake occurred at Kumamoto, April 14th 2016. Before the event, anomalous wave propagation was observed at Shimabara receiving station (west part from the epicenter) which comes from Miyazaki, Shimanto, Kanoya and Niihama broadband station. Though that data is not enough quantity to consider, it is wonder whether these results have the significance statistically. In this study, we used data received at Shimabara station for 2011 to 2016 to make Earthquake prediction maps. If 6 min-average sampled data exceeds a threshold in 2 consecutive samples, we labeled that time as anomaly (Morita and Mogi, 2015, IUGG). After anomaly, we turn ON the alarm for a certain period of time L, and this divide all time into "Alarm ON", "Alarm OFF", and "Undecided (due to missing data)" periods. We use earthquakes M>4.5 that occurred within the epicentral distance at radius of 100km from the receiving and broadband station. After excluding the Undecided period, we measured "r" the fraction of the period of alarm ON and "s" the fraction of earthquakes that occurred during periods of Alarm ON.