Predictions of the solar wind speed by the probability distribution function model

Abstract

Abstract The near‐Earth space environment is strongly driven by the solar wind and interplanetary magnetic field. This study presents a model for predicting the solar wind speed up to 5 days in advance. Probability distribution functions (PDFs) were created that relate the current solar wind speed and slope to the future solar wind speed, as well as the solar wind speed to the solar wind speed one solar rotation in the future. It was found that a major limitation of this type of technique is that the solar wind periodicity is close to 27 days but can be from about 22 to 32 days. Further, the optimum lag between two solar rotations can change from day to day, making a prediction of the future solar wind speed based solely on the solar wind speed approximately 27 days ago quite difficult. It was found that using a linear combination of the solar wind speed one solar rotation ago and a prediction of the solar wind speed based on the current speed and slope is optimal. The linear weights change as a function of the prediction horizon, with shorter prediction times putting more weight on the prediction based on the current solar wind speed and the longer prediction times based on an even spread between the two. For all prediction horizons from 8 h up to 120 h, the PDF Model is shown to be better than using the current solar wind speed (i.e., persistence), and better than the Wang‐Sheeley‐Arge Model for prediction horizons of 24 h.