Prediction of recurrent geomagnetic disturbances by using adaptive filtering

Abstract

Recurrent geomagnetic disturbances are an important part of geomagnetic activities, which are associated with the neutral sheet structure in the heliosphere and the activities of long lived solar coronal holes. Another significant character is the periodic activities recorded by geomagnetic indices. In this paper an algorithm-Adaptive Filtering (AF), is introduced to forecast recurrent geomagnetic events based on the geomagnetic K index. Adaptive filtering can deal with nonstationary data and can adapt to changes in the data pattern. Therefore it is a very helpful method for forecasting the geomagnetic disturbances and the disturbances in the interplanetary space. By using AF technique a prediction for whole Bartels rotation can be obtained when output length is taken as 27-point. For recurrent periods the prediction efficiency is about 30%, the correlation coefficient is 0.55. For nonrecurrent periods the prediction efficiency and correlation coefficient decrease obviously, but the standard variance does not change very much.