Search

We demonstrate the use of machine learning algorithms in combination with segmentation techniques in order to distinguish coronal holes and filaments in SDO/AIA EUV images of the Sun. Based on two coronal hole detection techniques (intensity-based thresholding, SPoCA), we prepared datasets of manually labeled coronal hole and filament channel regions present on the Sun during the time range 2011–2013. By mapping the extracted regions from EUV observations onto HMI line-of-sight magnetograms we also include their magnetic characteristics. We computed shape measures from the segmented binary maps as well as first order and second order texture statistics from the segmented regions in the EUV images and magnetograms. These attributes were used for data mining investigations to identify the most performant rule to differentiate between coronal holes and filament channels. We applied several classifiers, namely Support Vector Machine (SVM), Linear Support Vector Machine, Decision Tree, and Random Forest, and found that all classification rules achieve good results in general, with linear SVM providing the best performances (with a true skill statistic of ≈ 0.90). Additional information from magnetic field data systematically improves the performance across all four classifiers for the SPoCA detection. Since the calculation is inexpensive in computing time, this approach is well suited for applications on real-time data. This study demonstrates how a machine learning approach may help improve upon an unsupervised feature extraction method.

We present a simple method of forecasting the geomagnetic storms caused by high-speed streams (HSSs) in the solar wind. The method is based on the empirical correlation between the coronal hole area/position and the value of the Dst index, which is established in a period of low interplanetary coronal mass ejection (ICME) activity. On average, the highest geomagnetic activity, i.e., the minimum in Dst, occurs four days after a low-latitude coronal hole (CH) crosses the central meridian. The amplitude of the Dst dip is correlated with the CH area and depends on the magnetic polarity of the CH due to the Russell – McPherron effect. The Dst variation may be predicted by employing the expression Dst( t)=(−65±25×cos λ)[ A( t *)]0.5, where A( t *) is the fractional CH area measured in the central-meridian slice [−10°,10°] of the solar disc, λ is the ecliptic longitude of the Earth, ± stands for positive/negative CH polarity, and t− t *=4 days. In periods of low ICME activity, the proposed expression provides forecasting of the amplitude of the HSS-associated Dst dip to an accuracy of ≈30%. However, the time of occurrence of the Dst minimum cannot be predicted to better than ±2 days, and consequently, the overall mean relative difference between the observed and calculated daily values of Dst ranges around 50%. [ABSTRACT FROM AUTHOR]

We analyze the relationship between the coronal hole (CH) area/position and physical characteristics of the associated corotating high-speed stream (HSS) in the solar wind at 1 AU. For the analysis we utilize the data in the period DOY 25 – 125 of 2005, characterized by a very low coronal mass ejection (CME) activity. Distinct correlations between the daily averaged CH parameters and the solar wind characteristics are found, which allows us to forecast the solar wind velocity v, proton temperature T, proton density n, and magnetic field strength B, several days in advance in periods of low CME activity. The forecast is based on monitoring fractional areas A, covered by CHs in the meridional slices embracing the central meridian distance ranges [−40°,−20°], [−10°,10°], and [20°,40°]. On average, the peaks in the daily values of n, B, T, and v appear delayed by 1, 2, 3, and 4 days, respectively, after the area A attains its maximum in the central-meridian slice. The peak values of the solar wind parameters are correlated to the peak values of A, which provides also forecasting of the peak values of n, B, T, and v. The most accurate prediction can be obtained for the solar wind velocity, for which the average relative difference between the calculated and the observed peak values amounts to $\overline{\vert\delta\vert}\approx10$ %. The forecast reliability is somewhat lower in the case of T, B, and n ( $\overline{\vert\delta\vert}\approx20$ , 30, and 40%, respectively). The space weather implications are discussed, including the perspectives for advancing the real-time calculation of the Sun – Earth transit times of coronal mass ejections and interplanetary shocks, by including more realistic real-time estimates of the solar wind characteristics. [ABSTRACT FROM AUTHOR]