Solar Flare Prediction Based on the Fusion of Multiple Deep-learning Models

Solar flare formation mechanisms and their corresponding predictions have commonly been difficult topics in solar physics for decades. The traditional forecasting method manually constructs a statistical relationship between the measured values of solar active regions and solar flares that cannot fully utilize the information related to solar flares contained in observational data. In this article, we first used neural-network methods driven by the measured magnetogram and magnetic characteristic parameters of the sunspot group to learn the prediction model and predict solar flares. The prediction fusion model is based on a deep neural network, convolutional neural network, and bidirectional long short-term memory neural network and can predict whether a sunspot group will have a flare event above class M or class C in the next 24 or 48 hr. The real skill statistics (TSS) and F1 scores were used to evaluate the performances of our fusion model. The test results clearly show that this fusion model can make full use of the information related to solar flares and combine the advantages of each independent model to capture the evolution characteristics of solar flares, which is a much better performance than traditional statistical prediction models or any single machine-learning method. We also proposed two frameworks, namely F1_FFM and TSS_FFM, which optimize the F1 score and TSS score, respectively. The cross validation results show that they have their respective advantages in the F1 score and TSS score.

Multiple CNN Variants and Ensemble Learning for Sunspot Group Classification by Magnetic Type

A solar active region is a source of disturbance for the Sun–terrestrial space environment and usually causes extreme space weather, such as geomagnetic storms. The main indicator of an active region is sunspots. Certain types of sunspots are related to extreme space weather caused by eruptive events such as coronal mass ejections or solar flares. Thus, the automatic classification of sunspot groups is helpful to predict solar activity quickly and accurately. This paper completed the automatic classification of a sunspot group data set based on the Mount Wilson classification scheme, which contains continuum and magnetogram images provided by the Solar Dynamics Observatory’s Helioseismic and Magnetic Imager SHARP data from 2010 May 1 to 2017 December 12. After applying some data preprocessing steps such as image cropping and data standardization, the features of magnetic type in the data are more obvious, and the amount of data is increased. The processed data are spliced into two frames of single-channel data for the neural network to perform 3D convolution operations. This paper constructs a variety of convolutional neural networks with different structures and numbers of layers, selects 10 models as representatives, and chooses XGBoost, which is commonly used in ensemble-learning algorithms, to fuse the results of independent classification models. We found that XGBoost is an effective way to fuse models, which is proved by the relatively balanced high scores in the three magnetic types. The accuracy of the ensemble model is above 92%. The F1 scores of the magnetic types of Alpha, Beta, and Beta-x reached 0.95, 0.91, and 0.82 respectively.

Analyses of Early Sunspot Records by Jean Tarde (1615 – 1617) and Jan Smogulecki (1621 – 1625)

Jean Tarde and Jan Smogulecki carried out sunspot observations in the 1610s and 1620s at the dawn of the telescopic era. We analysed their original observational records to revise their sunspot-group numbers in the existing database. In this study, we provide a new counting as a basis for future scientific discussions. Furthermore, we compared Smogulecki’s sunspot observations with those of Scheiner and Schönberger on the same observation days. We also detected a big sunspot group on 2 – 3 February 1622 in Smogulecki’s sunspot drawings and estimated its area to be approximately 1600 millionths of the solar disc. In addition, we measured the sunspot positions in Tarde’s and Smogulecki’s sunspot drawings to construct a butterfly diagram for this early period.

Number of Sunspot Groups and Individual Sunspots Recorded by Tevel for the Period 1816–1836 in the Dalton Minimum

Cornelis Tevel made sunspot observations during the period 1816–1836, including the Dalton Minimum. In this work, the first revision of these observations since Wolf incorporated them into his database is presented. On the one hand, the number of individual sunspots from Tevel’s drawings was counted. This is of special interest for the sunspot number reconstruction because this kind of information is not as common in historical sunspot records as the number of groups. Thus, Tevel could be considered for the future reconstruction of the sunspot number index. On the other hand, the number of groups counted according to modern sunspot group classifications finding significant misinterpretations with the number of groups assigned to Tevel in the existing databases. Tevel was a relevant sunspot observer in the Dalton Minimum. In fact, he was the observer with the highest number of groups observed in Solar Cycles 6 and 7 according to the existing sunspot group number databases. According to the raw group number recount in this work, the maximum amplitudes for Solar Cycles 6 and 7 are, respectively, 27% and 7% lower than those previously determined. Moreover, Solar Cycle 6 is the weakest solar cycle since the Maunder Minimum after applying these new counts. Group counts from Tevel’s observations were compared with those from relevant contemporary astronomers, demonstrating that Schwabe and Tevel systematically recorded a higher number of groups than Flaugergues and Derfflinger. In addition, sunspot areas and positions recorded by Tevel should be used with caution for scientific purposes.

Comparison of Geomagnetic Indices During Even and Odd Solar Cycles SC17 – SC24: Signatures of Gnevyshev Gap in Geomagnetic Activity

We show that the time series of sunspot group areas has a gap, the so-called Gnevyshev gap (GG), between ascending and descending phases of the cycle and especially so for the even-numbered cycles. For the odd cycles this gap is less obvious, and is only a small decline after the maximum of the cycle. We resample the cycles to have the same length of 3945 days (about 10.8 years), and show that the decline is between 1445 – 1567 days after the start of the cycle for the even cycles, and extending sometimes until 1725 days from the start of the cycle. For the odd cycles the gap is a little earlier, 1332 – 1445 days after the start of the cycles with no extension. We analyze geomagnetic disturbances for Solar Cycles 17 – 24 using the Dst-index, the related Dxt- and Dcx-indices, and the Ap-index. In all of these time series there is a decline at the time, or somewhat after, the GG in the solar indices, and it is at its deepest between 1567 – 1725 days for the even cycles and between 1445 – 1567 days for the odd cycles. The averages of these indices for even cycles in the interval 1445 – 1725 are 46%, 46%, 18%, and 29% smaller compared to surrounding intervals of similar length for Dst, Dxt, Dcx, and Ap, respectively. For odd cycles the averages of the Dst- and Dxt-indices between 1322 – 1567 days are 31% and 12% smaller than the surrounding intervals, but not smaller for the Dcx-index and only 4% smaller for the Ap-index. The declines are significant at the 99% level for both even and odd cycles of the Dst-index and for the Dxt-, Dcx- and Ap-indices for even cycles. For odd cycles of the Dxt-index the significance is 95%, but the decline is insignificant for odd cycles of the Dcx- and Ap-indices.

Reanalyses of the sunspot observations of Fogelius and Siverus: Two “Long-Term” observers during the Maunder Minimum

The solar activity during the Maunder Minimum (MM; 1645–1715) has been considered significantly different from the one captured in modern observations, in terms of sunspot group number and sunspot positions, whereas its actual amplitudes and distributions is still under active discussions. In its core period (1650/1660–1700), Martin Fogelius and Henrich Siverus have formed significant long-term series in the existing databases with numerous spotless days, as the 13th and 7th most active observers before the end of the MM. In this study, we have analysed their original archival records, revised their data, have removed significant contaminations of the apparent ‘spotless days’ in the existing databases, and cast caveats on the potential underestimation of the solar-cycle amplitude in the core MM. Still, they reported at best one sunspot group throughout their observational period and confirm the significant suppressed the solar cycles during the MM, which is also supported from the contemporary observations of Hook and Willoughby. Based on the revised data, we have also derived positions of notable sunspot groups, which Siverus recorded in 1671 (≈ N7.5° ± 2.5°), in comparison with those of Cassini's drawings (≈ N10° ± 1°). Their coincidence in position and chronology in corrected dates indicates these sunspot groups were probably the same recurrent active region (AR) and its significantly long lifespan (≥ 35 days) even during the MM.

Remeasurement of Solar Observing Optical Network sunspot areas

ABSTRACT The United States Air Force solar observing optical network (SOON) sunspot areas have been reported by several researchers over many years to be underestimated by as much as 50 per cent. Here, the areas of sunspots from scanned SOON disc drawings have been accurately remeasured for a period of two months from 2014 October and November – this being near the peak of Solar Cycle 24 and which includes the largest sunspot group of that cycle. The remeasured sunspot areas are now comparable with areas in sunspot catalogues.

Number of sunspot groups from the Galileo–Scheiner controversy revisited

ABSTRACT We revise the sunspot observations made by Galileo Galilei and Christoph Scheiner in the context of their controversy regarding the nature of sunspots. Those of their sunspot records not included in the current sunspot group database, used as a basis to calculate the sunspot group number, are analysed. Within the documentary sources consulted in this work, we can highlight the sunspot observations by Scheiner included in the letters sent under the pseudonym Apelles to Marcus Welser and the first sunspot observations made by Galileo, which can be consulted in Le opere di Galileo Galilei. These sunspot observations would extend the temporal coverage for these two observers and fill some gaps in the current group database in the earliest period, where the data available are sparse. Moreover, we have detected changes in the quality of the sunspot drawings made by Galileo and Scheiner in their observation series, affecting the number of groups recorded by the two observers. We also compare these records with sunspot observations made by other astronomers of that time. According to this comparison and regarding the same observation days, Scheiner was generally the astronomer who reported more sunspot groups, while Harriot, Cigoli and Galileo recorded a similar number of groups. We conclude that these differences are mainly because of the observational methods used by the observers.

Comparison of the shape and temporal evolution of even and odd solar cycles

Aims. We study the difference in the shape of solar cycles for even and odd cycles using the Wolf sunspot numbers and group sunspot numbers of solar cycles 1−23. We furthermore analyse the data of sunspot area sizes for even and odd cycles SC12−SC23 and sunspot group data for even and odd cycles SC8−SC23 to compare the temporal evolution of even and odd cycles. Methods. We applied the principal component analysis (PCA) to sunspot cycle data and studied the first two components, which describe the average cycle shape and cycle asymmetry. We used a distribution analysis to analyse the temporal evolution of the even and odd cycles and determined the skewness and kurtosis for even and odd cycles of sunspot group data. Results. The PCA confirms the existence of the Gnevyshev gap (GG) for solar cycles at about 40% from the start of the cycle. The temporal evolution of sunspot area data for even cycles shows that the GG exists at least at the 95% confidence level for all sizes of sunspots. On the other hand, the GG is shorter and statistically insignificant for the odd cycles of aerial sunspot data. Furthermore, the analysis of sunspot area sizes for even and odd cycles of SC12−SC23 shows that the greatest difference is at 4.2−4.6 years, where even cycles have a far smaller total area than odd cycles. The average area of the individual sunspots of even cycles is also smaller in this interval. The statistical analysis of the temporal evolution shows that northern sunspot groups maximise earlier than southern groups for even cycles, but are concurrent for odd cycles. Furthermore, the temporal distributions of odd cycles are slightly more leptokurtic than distributions of even cycles. The skewnesses are 0.37 and 0.49 and the kurtoses 2.79 and 2.94 for even and odd cycles, respectively. The correlation coefficient between skewness and kurtosis for even cycles is 0.69, and for odd cycles, it is 0.90. Conclusions. The separate PCAs for even and odd sunspot cycles show that odd cycles are more inhomogeneous than even cycles, especially in GSN data. Even cycles, however, have two anomalous cycles: SC4 and SC6. The variation in the shape of the early sunspot cycles suggests that there are too few and/or inaccurate measurements before SC8. According to the analysis of the sunspot area size data, the GG is more distinct in even than odd cycles. This may be partly due to sunspot groups maximizing earlier in the northern than in the southern hemisphere for even cycles. We also present another Waldmeier-type rule, that is, we find a correlation between skewness and kurtosis of the sunspot group cycles.