Physics of auroral phenomena : proceedings of the 34th Annual seminar, Apatity, 01 - 04 March, 2011 / [ed.: A. G. Yahnin, A. A. Mochalov]. - Апатиты : Издательство Кольского научного центра РАН, 2011. - 231 с. : ил.

Classification o fspace weather complexes taking into account characteristics ofperturbing solar stream and its geomagnetic disturbance parameters IMF for 2000-2007. Is considered 50 isolated low-noises global geomagnetic disturbances of various intensities with duration more than six hours, established on Dst-index dynamic According to standard approach geomagnetic disturbances have been divided on intensity on weak (-50 <Dst <-20 nT), moderate (-100 <Dst <-50 nT), strong (-200 <Dst <-100 nT) and extreme storms (Dst <-200 nT) [Gonzalez et al., 1994]. For expansion of analyzed events types in research magnetic disturbance for which was observed only the increase of Dst-index caused by direct influence of solar stream on Earth magnetosphere (DCF) have been added. For the selected geomagnetic disturbances division under form variation Dst on sudden commencement phase on «classical» one-step storms, «поп-classical» two-step storms, disturbance only with increase of Dst-index and with increase and insignificant subsequent reduction (-20 <Dst <0 nT) has been carry out. It has been as result allocated eight types of global geomagnetic disturbances: increase of Dst-index (fig. la), increase of Dst-index with insignificant subsequent reduction (fig. lb), weak storms (fig. lc), moderate classical storms (fig. Id), moderate non-classical storms (fig. le), strong classical storms (fig. If), strong non-classical storms (fig. lg) and extreme storms (fig. lh). In the further neural network classification will be carry out also on eight classes, according to number allocated to types of global geomagnetic disturbances. 0 t -100 0 -100 0 -100 0 -100 \ _ Y . g -200 -200 -200 -200 -300 -300 -300 -300 8 15 8 15 9 17 9 17 8 15 8 15 8 15 8 15 Tim e, h e Time, h 1 Time, h s Time, h h 0 н -100 0 -100 Л . \ 0 -100 \ 0 -100 A TS -200 -200 -200 -200 \ \ -300 11 2 и -300 1 5 5 I -300 7 1 7 1 -300 i VS 6 1 V i 6 1 Time, h Time, h Time, h Time, h Fig. 1. Dynamics of Dst-index for allocated types of global geomagnetic disturbances (left parts of pictures) and results of processing of Dst-index dynamics spline (the right parts) Establishment of solar sources for selected geomagnetic disturbances was carried out by the analysis of data solar catalogues, the literature [Lynch et al., 2003; Cane et al., 2003; Leamon et al., 2004; Zhang et al., 2004] and dynamics PSW and IMF in near-Earth space. Following catalogues have been used: (1) Active prominences and filaments (ftp://ftp.ngdc.noaa.gov/STP/SOLAR_DATA/SOLAR_Website/FILAMENTS/) (2) LASCO CME Catalog (http://vso.nso.edu/cgi/catalogui) (3) The catalogue of the large-scale phenomena of solar wind for period 1976-2000. (ftp://ftp.iki.rssi.ru/pub/omni/catalog/) (4) A selected list of solar X-ray events observed by CELIAS/SEM (http://umtof.umd.edu/sem/sem_figs.html) (5) An incomplete list o f possible Interplanetary Shocks observed by the PM (http://umtof.umd.edu/pm/Shocks.html) (6) Ha solar flares (ftp://ftp.ngdc.noaa.gov/STP/SOLAR_DATA/SOLAR_Website/SOLAR_FLARES/FLARES_HALPHA/) (7) VSO Catalog Search Results GOES X-Ray Catalog (http://vso.nso.edu/cgi/catalogui) As result o f analysis were found solar sources for almost all considered geomagnetic disturbances. However unequivocally to establish type of disturbance solar stream previous some geomagnetic disturbances it was not possible. It is connected with the contradictions encountered in catalogues and literature, and also complicated PSW and IMF dynamics. 3. Neural network classification of space weather complexes Classification in the present work was carried out by means o f special software-computer complex by neural network «layer Kohonen» [Barkhatov and Revunov, 2010]. As input parameters were used: average value of magnetic field vector B, southern Bs magnetic field component, temperature T, hydrodynamic pressure P, electric field VBs of solar wind and Dst-index. As it was marked above, the basic difficulty of networks classification work is necessity to use of equal length input vectors. Besides, for improvement of network settings quality, the neurons number in input layer, i.e. variables number characterizing dynamics of each input parameter, it is desirable to choose the minimum. In this connection all input data have been presented in form of parametrical vectors. Preliminary dynamics of each parameter has been processed by a cubic spline, to determine its prominent features. For example in Fig.l in the right parts results of processing by spline of Dst-index dynamics are shown. Further processed parameter dynamics 107