Real-time solar wind prediction based on SDO/AIA coronal hole data
We present an empirical model based on the visible area covered by coronal holes close to the central meridian in order to predict the solar wind speed at 1 AU with a lead time up to four days in advance with a 1hr time resolution. Linear prediction functions are used to relate coronal hole areas to solar wind speed. The function parameters are automatically adapted by using the information from the previous 3 Carrington Rotations. Thus the algorithm automatically reacts on the changes of the solar wind speed during different phases of the solar cycle. The adaptive algorithm has been applied to and tested on SDO/AIA-193A observations and ACE measurements during the years 2011-2013, covering 41 Carrington Rotations. The solar wind speed arrival time is delayed and needs on average 4.02 +/- 0.5 days to reach Earth. The algorithm produces good predictions for the 156 solar wind high speed streams peak amplitudes with correlation coefficients of cc~0.60. For 80% of the peaks, the predicted arrival matches within a time window of 0.5 days of the ACE in situ measurements. The same algorithm, using linear predictions, was also applied to predict the magnetic field strength from coronal hole areas but did not give reliable predictions (cc~0.2).