Selection of anomalies in ionospheric parameters on the basis of combination of multiscale wavelet-decomposition and neural networks

• download

   The authors propose a method for the analysis of critical frequency
   parameters of the ionospheric layer F2, based on the combination of
   multiscale analysis and multilayer neural networks, which allows us to
   distinguish the abnormal features of ionosphere behavior. The complexity
   of solving the problems of ionospheric parameter processing and analysis
   is associated with their complex structure. They include a large number of
   components, contain local features of various shapes and duration,
   anomalous effects and noise factors. Traditional approaches and methods
   for ionospheric parameter analysis based on the smoothing procedure, lead
   to distortion and information loss. One of the major drawbacks of these
   methods is the lack of effective means for adaptation to the complex
   time-dependent data structure. The proposed method is based on the
   representation of the recorded time series of foF2 in the form of
   different scale components and their approximation by adaptive neural
   networks of variable structure. The method, algorithm and software,
   developed on its basis, allow us to perform a detailed analysis of each
   component and to distinguish anomalies that appear during increased
   seismic activity in Kamchatka. We used the recorded data of foF2 for the
   period 1969-2010. (''Paratunka'' station, Kamchatka). Comparison of the
   results of the ionospheric parameter processing with the Catalog of
   earthquakes and geomagnetic data showed the efficiency of the proposed
   method, and allowed us to allocate periods of anomalous behavior of the
   ionosphere.