Neural network modeling of solar wind‐magnetosphere interaction

We study solar wind-magnetosphere interaction using partially recurrent neural networks to find the best coupling functions. Hourly data of the solar wind and Dst are selected, from the period 1963 to 1992, covering 10,554 hours. The best coupling functions are found from the out-of-sample performance of trained neural networks, which are then compared with those obtained from cross-correlation analyses. In most cases, good agreement is found. This indicates that cross-correlation analysis can be used to optimize input parameters for neural networks, albeit at a limitation due to the assumption of linearity. However, there exist exceptions where coupling functions strongly correlated with Dst give poor predictions. Based on neural network modeling, if we take the coupling function in the form of P λ V μ B ν s , then the conditions, 0 ≤ λ ≤ 1, 0.5 < μ ≤ 2 and 0.5 ≤ ν ≤ 2, will result in solar wind input functions well suited for predicting geomagnetic storms, e.g., P 1/4 VB S , P 1/6 VB s , P 1/5 VB S , P 1/3 VB S , P 1/2 VB S , V 2 B S , and VB S . In contrast, e and the polar cap potential are less good candidates for storm predictions. Therefore the best coupling functions do not necessarily have the dimension of power, energy or electric field. Furthermore, we predict geomagnetic storms up to five hours in advance with good accuracy from the best coupling functions. The confidence limits on the prediction accuracy are computed to show the statistical significance of the predictions and to confirm the best coupling functions. Finally, we discuss how the partially recurrent network models solar wind-magnetosphere interaction.

[1]  C. Clauer.  The technique of linear prediction filters applied to studies of solar wind-magnetosphere coupling , 1986 .

[2]  Hiroshi Maeda,et al.  Impulse response of geomagnetic indices to interplanetary magnetic field. , 1979 .

[3]  Toshiki Tajima,et al.  Neural net forecasting for geomagnetic activity , 1993 .

[4]  O. Troshichev.  The physics and meaning of the existing and proposed high-latitude geomagnetic indices , 1988 .

[5]  C. Russell,et al.  An empirical relationship between interplanetary conditions and Dst , 1975 .

[6]  Masahisa Sugiura,et al.  Hourly values of equatorial dst for the igy , 1963 .

[7]  Martin Casdagli,et al.  Nonlinear prediction of chaotic time series , 1989 .

[8]  Jeffrey L. Elman,et al.  Finding Structure in Time , 1990, Cogn. Sci..

[9]  Henrik Lundstedt,et al.  Geomagnetic storm predictions from solar wind data with the use of dynamic neural networks , 1997 .

[10]  D. Baker,et al.  IMF control of geomagnetic activity , 1988 .

[11]  Richard S. Bird,et al.  An introduction to the theory of lists , 1987 .

[12]  Henrik Lundstedt,et al.  Neural networks and predictions of solar-terrestrial effects , 1992 .

[13]  H. Lundstedt,et al.  Prediction of geomagnetic storms from solar wind data using Elman Recurrent Neural Networks , 1996 .

[14]  A. Weigend,et al.  Time Series Prediction: Forecasting the Future and Understanding the Past , 1994 .

[15]  Daniel N. Baker,et al.  The organized nonlinear dynamics of the magnetosphere , 1996 .

[16]  Michael I. Jordan.  Serial Order: A Parallel Distributed Processing Approach , 1997 .

[17]  H. W. Kroehl,et al.  What is a geomagnetic storm , 1994 .

[18]  Daniel N. Baker,et al.  Statistical analyses in the study of solar wind-magnetosphere coupling , 1985 .

[19]  Daniel N. Baker,et al.  A description of the solar wind-magnetosphere coupling based on nonlinear filters , 1995 .

[20]  Daniel N. Baker,et al.  A nonlinear dynamical analogue model of geomagnetic activity , 1992 .

[21]  Daniel N. Baker,et al.  The relation between the northern polar cap and auroral electrojet geomagnetic indices in the wintertime , 1996 .

[22]  Daniel N. Baker,et al.  Magnetospheric Impulse Response for Many Levels of Geomagnetic Activity , 1985 .

[23]  Daniel N. Baker,et al.  The evolution from weak to strong geomagnetic activity: an interpretation in terms of deterministic chaos , 1990 .

[24]  H. Gleisner,et al.  Predicting geomagnetic storms from solar-wind data using time-delay neural networks , 1996 .

[25]  P. Wintoft,et al.  Prediction of geomagnetic storms from solar wind data with the use of a neural network , 1994 .

[26]  D. N. Baker,et al.  Nonlinear ARMA models for the D(st) index and their physical interpretation , 1996 .

[27]  D. Prichard,et al.  The non‐linear response of the magnetosphere: 30 October 1978 , 1993, comp-gas/9305003.

[28]  R. A. Smith,et al.  Prediction of geomagnetic activity , 1993 .