Application of an information system with artificial intelligence to detect geoacoustic emission anomalies prior to strong earthquakes in Kamchatka.

• download

Since 1999 the investigations of high-frequency geoacoustic emission in the frequency range 0.1 Hz ... 10
kHz has been carried out in Kamchatka by hydroacoustic sensors, installed in small water pools. The results of
the investigation showed that during the observation period the majority of the seismic events with energy class
Ks> 11, located generally within a radius of 250 km, were preceded by acoustic anomalies recorded within a day
before earthquakes. The effectiveness of acoustic response depends on energy of an earthquake, distance, depth
and spatial arrangement. Analysis of the recorded data is laborious apart from the anomalies of geoacoustic
emission before earthquakes; sensors record the similar signals determined by meteorological and anthropogenic
factors. Having examined the characteristics of the emission of different nature, information system using Koho-
nen neural networks was developed. It allowed us to identify and project the anomalies of geoacoustic emission
on a neural map, and clearly divide them into clusters of different nature. Testing of the system on test data
sets was carried out. At present its adaptation to the data from "Paratunka", "Mikizha" and "Karymshina"
sites is realized.