Телефон: 522-81-45

Geoinformatika 2014; 3(51) : 77-84 (in Ukrainian)


E.M. Kozlovskyy, D.V. Malytskyy, A.Yu. Pavlova

Carpathian Branch of Subbotin Institute of Geophysics NAS of Ukraine, Naukova st. 3b, Lviv 79060, Ukraine, susyinet@gmail.com

A neural network is a powerful data modeling tool that is able to capture and represent complex input/output relationships. The motivation for the development of neural network technology stemmed from the desire to develop an artificial system that could perform “intelligent” tasks. The purpose of this paper is to show the feasibility of using neural-network modeling for calculating and refining the depth distribution of earthquake sources and arrival times of the first seismic waves of local earthquakes in the Transcarpathians  seismic active region. In this paper, use was made of the most common neural network model (the multilayer perceptron (MLP)). This type of neural network is known as a supervised network because it requires a desired output in order to learn. The goal of this type of network is to create a model that correctly maps the input to the output using historical data so that the model can then be used to produce the output when the desired output is unknown. The authors have proved justified the introduction of azimuthal coefficient qі to be used in a teaching set for neural networks as a parameter responsible for the direction of the wave propagation in a real environment. Average values of the azimuthal coefficient qі for sectors with close values of qі have been calculated for one-, two- and three-layered medium, according to the depth distribution of the earthquake source in the corresponding layer. Neural-network modeling was used to calculate the depth of the earthquake source and arrival times of P-waves, and to specify the data from seismological bulletins. 370 examples for the period 2002–2012 to train the neural network were selected to refine the depth distribution of earthquake sources and arrival times of the first seismic waves. The interpretation of the results was carried out. Comparison has been made between the depths and arrival times of the first P-waves projected using neural-network modeling with the data from the seismological bulletins. The correlation coefficient between the arrival times obtained using neural-network modeling Tpr and the data from the seismological bulletins T is equal 0.98.

Keywords: The Transcarpathian seismic region, the azimuthal coefficient qі, the training set for the neural network, average velocity of wave propagation in a layer, depth of location of the earthquake source, direct P– and S– waves, a pair of ES (epicenter – seismic station), conventional velocity of wave propagation in combined layers.



  1. Kozlovskyy E., Malytskyy D., Pavlova A., Rozrahunok ta analiz azymutalnyh koefistieniv qі ta utochnennia shvydkisnoi modeli Zakarpatskoho seismoaktyvnoho rehionu [Determining and analysis of azimuthal coefficients qi for the seismically active region of Transcarpathians]. Visnyk KNU. Heologia [Visnyk KNU. Geology], 2014, issue 1(64), pp. 30-33.
  2. Malytskyy D. Pro dzherelo seismichnyh hvyl [The source of seismic waves]. Geofizicheskij zhurnal [Geophysical journal], 2005, vol. 27, no. 2, pp. 304-308.
  3. Malytskyy D., Kozlovskyy E., Khytriak O. Analis resultativ matematychnoho modeluvannia teoretychnyh seismogram [Analysis of the mathematical modelling results of theoretical seismograms]. Visnyk KNU. Heologia [Visnyk KNU. Geology], 2009, vol.47, pp. 51-54.
  4. Malytskyy D., Pavlova A., Kozlovskyy E., Hrytsai O. Zastosuvannya matrychnoho metodu v obernenyh zadachah z vykorystanniam realnyh zapysiv [Application of matrix methods in inverse problems using real records]. Geodynamika [Geodynamics], 2013, no. 2 (15), pp. 265-267.
  5. Seismolohicheskyy bulleten Ukrainy (2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012) [Seismological Bulletin of Ukraine (2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011)] / Pod red. B.H. Pustovitenko. Sevastopol, NPS “EKOSI-Hydrophysica”.
  6. Chekunov A., Livanova L., Geiko V. Hlubinnoe stroenie zemnoj kory i nekotorye osobenosti tektoniki Zakarpatskohk prohyba [Deep crustal structure and tectonics of some features of the Transcarpathian deflection]. Sov. Heologyia [Sov. Geology], 1969, no. 10, pp. 57-69.
  7. Abbott L.F. Learning in Neural Network Memories. Network: Comp. Neural Sys., 1990, vol. 1, pp. 105-122.
  8. Amit D.J., Tsodyks M.V. Effective neurons and attractor neural networks in cortical environment. Network, 1991, vol. 3, pp.121-138.
  9. Byrne J.H., Berry W.O. Neural Models of Plasticity. San Diego: Academic Press, 1989.
  10. Lin F., Ritzwoller M., Yang Y., Moschetti M., Fouch M. The stratification of seismic azimuthal anisotropy in the western US. American Geophysical Union. Fall Meeting. San Francisko, 2010.
  11. Perry H., Eaton D., Forte A. LITH5.0: a revised crustal model for Canada based on Lithoprobe results. Geophys. J. Int., 2002, vol. 150, pp. 285-294.