Physics of auroral phenomena : proceedings of the 33rd Annual seminar, Apatity, 02 - 05 March, 2010 / [ed.: A.G. Yahnin, A. A. Mochalov]. - Апатиты : Издательство Кольского научного центра РАН, 2011. - 206 с. : ил.

Disturbances ofheliogeophysical parameters displaying in the maximum observedfrequencies on ionospheric oblique sounding traces this purpose the Elman neural network with internal nonlinear memory has been chosen. The received results are used for development o f a neural network restoration technique witch allows to receive MOF data on one trace by other trace data. On the first stage o f neural network experiments characteristic time delays of DMOF reaction on changes in heliogeophysical parameters are established. For each trace DMOF restoration by values of concrete heliogeophysical parameter is carried out with time delays from 0 till 300 minutes DMOF relatively this parameter. Calculated time delays contain a transfer time of Solar wind disturbance from spacecraft up to magnetospheric boundary and also reaction time of magnetosphere-ionosphere system on approached disturbance. Results o f the first stage experiments are collected in Table 1 where established characteristic time delays for each of used additional parameters are submitted. Table 1. Interplanetary space parameters and time delays o f DMOF reaction on changes in Interplanetary space parameter Time delay, min В 100-140 Bx 80 By 120-200 Bz 60-80 N 220-280 V not received XL 20-30 XS 0-10 given parameters. On the second stage of neural network experiments the influence of heliogeophysical parameters on DMOF on each trace is established taking into account received time delays. For this purpose it was carried out a neural network restoration o f DMOF on Cyprus - Rostov-on-Don, Norilsk - Rostov-on-Don and Irkutsk - Rostov-on-Don traces by DMOF data on Inskip - Rostov-on-Don trace and by values of chosen heliogeophysical parameter. The influence degree each o f parameters was determined by increase o f DMOF restoration efficiency at considered trace under its addition. Results o f executed neural network experiments show that DMOF data restoration o f Cyprus - Rostov-on-Don trace by DMOF data o f Inskip - Rostov-on-Don trace in December 2006 is possible. Attracting of IMF module В as additional parameter improved restoration quality on 4-6 %, parameter Bx - on 4-7%, parameter By - on 8%. Influence o f other parameters appeared less evident. Thus for traces Inskip - Rostov-on-Don and Cyprus - Rostov- on-Don the most considerable is influence of IMF module and components Bx and By in comparison with Solar wind parameters. Probably it is connected with location o f reflection points of examined traces - at middle and low latitudes. Therefore changes in Solar wind density and velocity are not shown in DMOF. Restoration o f Norilsk - Rostov-on-Don DMOF data by Inskip - Rostov-on-Don DMOF data appeared difficult due to long night intervals on Norilsk - Rostov-on-Don trace. For the given pair of traces it was not possible to reveal a nonlinear correlation connections between DMOF. Entering of additional parameters also has no influence on level of nonlinear correlation connections. It can be explained by the specified location of trace Norilsk - Rostov-on-Don in subauroral region unlike three other traces. It leads to significant difference of DMOF behaviour in this region from DMOF at middle and low latitudes. Research o f nonlinear connections between DMOF variations on Inskip - Rostov-on-Don and Irkutsk -Rostov-on- Don traces shows that restoration of Irkutsk - Rostov-on-Don DMOF data does not occur without attraction of additional parameters. However entering of additional parameters allows to receive such restoration. The main influence has such parameters as IMF module В (restoration quality increase by 40 %), components By and Bz (restoration quality increase by 30 and 40 %), Solar wind density (restoration quality increase by 40 %). On the third stage o f neural network experiments MOF data restoration of traces Cyprus - Rostov-on-Don and Irkutsk - Rostov-on-Don by MOF data of trace Inskip - Rostov-on-Don is executed. Thus the results received at the first and second stages were taken into account.. Restoration was carried out by initial MOF data of trace Inskip - Rostov-on-Don, i.e. a daily course was taken into account. Input additional parameters which have the greatest influence on a level o f nonlinear connection are used. For additional parameters also established earlier time delays are considered. As additional input parameter also we added a solar zenith angle which provides restoration of MOF daily course. For example at restoration o f Irkutsk - Rostov-on-Don MOF data by Inskip - Rostov-on-Don MOF data neural network input feed: Inskip - Rostov-on-Don MOF data, IMF module В with time delay 120 minutes, components By and Bz with time delays 120 and 60 minutes, accordingly, Solar wind density N with time delay 240 minutes and solar zenith angle. On Fig. 3 results o f such restoration are submitted. The solid line corresponds to real MOF values of Irkutsk - Rostov-on-Don trace, a dashed line - to the restored values by neural network. The high value o f correlation coefficient (0.8) between real and restored sequences is marked. I l l