Sunspot Group Detection and Classification by Dual Stream Convolutional Neural Network Method

The automatic detection and analysis of sunspots play a crucial role in understanding solar dynamics and predicting space weather events.This paper proposes a novel method for sunspot group detection and classification called the dual stream Convolutional Neural Network with Attention Mechanism(DSCNN-AM).The network consists of two parallel streams each processing different input data allowing for joint processing of spatial and temporal information while classifying sunspots.It takes in the white light images as well as the corresponding magnetic images that reveal both the optical and magnetic features of sunspots.The extracted features are then fused and processed by fully connected layers to perform detection and classification.The attention mechanism is further integrated to address the“edge dimming”problem which improves the model’s ability to handle sunspots near the edge of the solar disk.The network is trained and tested on the SOLAR-STORM1 data set.The results demonstrate that the DSCNN-AM achieves superior performance compared to existing methods,with a total accuracy exceeding 90%.

Evolution of the Level of Sunspot Activity in Solar Cycles I. Evolution in the Descending Phase

Taking the 13-point smoothed monthly sunspot number,Ri, and the deviation of the 13 associated monthly sunspot numbers from the smoothed one, Di, as a number-pair describing the global level of sunspot activity, the evolution of the level is statistically studied for the period from the month which is just 48 months before the minimum to the minimum in the descending phase, using the observed data of Solar Cycles 10 to 22. Our results show （1） for 46 months （94%） of the studied 49 months it is found that for a given month, the distribution of the 13 pairs which come from the 13 solar cycles on a log Ri - Di plane may be fitted by a straight line with a correlation coefficient larger than the critical one at confidence level a = 5%, and for 36 months （73%） the fitting is even better, for a = 1%; （2） time variations of these two parameters and their correlations in the studied period can be described respectively by functions of time, whose main trends may be expressed by a linear or simple curvilinear function; （3） the evolutionary path of the level of sunspot activity may be represented by a logarithmic function as log/~ = 0.704 lnDi - 0.291.

The relationships of solar flares with both sunspot and geomagnetic activity

The relationships between solar flare parameters （total importance, time duration, flare index, and flux） and sunspot activity （R z ） as well as those between geomagnetic activity （aa index） and the flare parameters can be well described by an integral response model with the response time scales of about 8 and 13 months, respectively. Compared with linear relationships, the correlation coefficients of the flare parameters with R z , of aa with the flare parameters, and of aa with R z based on this model have increased about 6%, 17%, and 47% on average, respectively. The time delays between the flare parameters with respect to R z , aa to the flare parameters, and aa to R z at their peaks in a solar cycle can be predicted in part by this model （82%, 47%, and 78%, respectively）. These results may be further improved when using a cosine filter with a wider window. It implies that solar flares are related to the accumulation of solar magnetic energy in the past through a time decay factor. The above results may help us to understand the mechanism of solar flares and to improve the prediction of the solar flares.

The Formation and Evolution of the Sun and the Source of Star Energy as Well as the Sunspots and Flares of the Sun

Nebula theory is the most widely accepted hypothesis about the formation and evolution of the Solar System. This theoryholds that the Sun is formed from a collapsed gas cloud 4.57 billion years ago;when the core temperature of the gas cloud rises to 10million K, the thermonuclear reaction of hydrogen fusion into helium is ignited, then the Sun become a star;once the hydrogen in thecore is exhausted, the life of the star will end. But the limited hydrogen element obviously cannot satisfy such a long-termthermonuclear reaction, in order to sustain long-term thermonuclear reactions, a steady stream of fuel must be obtained from space.So the existing hypothesis about the formation and evolution of the Solar System has serious defects. Thus the author has studied theformation of the Moon, the Earth and the Sun, and discovered the formation of the Sun and the real source of star energy. The authorcould also explain many solar activity phenomena such as sunspots, flares, prominences, etc.

Magnetic cycles of Sun-like stars with different levels of coronal and chromospheric activity——comparison with the Sun

The atmospheric activity of the Sun and Sun-like stars is analyzed involving observations from the HK-project at the Mount Wilson Observatory, the California and Carnegie Planet Search Program at the Keck and Lick Observatories and the Magellan Planet Search Program at the Las Campanas Observatory. We show that for stars of E G and K spectral classes, the cyclic activity, similar to the 11-yr solar cycle, is different： it becomes more prominent in K-stars. Comparative study of Sun-like stars with different levels of chromospheric and coronal activity confirms that the Sun belongs to stars with a low level of chromospheric activity and stands apart among these stars by its minimum level of coronal radiation and minimum level of variations in photospheric flux.

Periodic variation and phase analysis of grouped solar flare with sunspot activity

Studies on the periodic variation and the phase relationship between different solar activity indicators are useful for understanding the long-term evolution of solar activity cycles.Here we report the statistical analysis of grouped solar flare(GSF) and sunspot number(SN) during the time interval from January 1965 to March 2009.We find that,(1) the significant periodicities of both GSF and SN are related to the differential rotation periodicity,the quasi-biennial oscillation(QBO),and the eleven-year Schwabe cycle(ESC),but the specific values are not absolutely identical;(2) the ESC signal of GSF lags behind that of SN with an average of 7.8 months during the considered time interval,which implies that the systematic phase delays between GSF and SN originate from the inter-solar-cycle signal.Our results may provide evidence about the storage of magnetic energy in the corona.

Multiple cycles of magnetic activity in the Sun and Sun-like stars and their evolution

The wavelet transform method for high-quality time-frequency analysis is applied to sets of observations of relative sunspot numbers and stellar chromosphere fluxes of 10 Sun-like stars. Wavelet analysis of solar data shows that in a certain interval of time there are several cycles of activity with pe- riods of duration which vary considerably from each other： from quasi-biennial cycles to lO0-yr cycles. Cyclic activity was detected in almost all Sun-like stars that we examined, even those that previously were not considered as stars with cyclic activity according to analysis using a Scargle periodogram. The durations of solar and stellar cycles significantly change during the observation period.

Low-dimensional chaos and fractal properties of long-term sunspot activity

Two primary solar-activity indicators- sunspot numbers （SNs） and sunspot areas （SAs） in the time interval from November 1874 to December 2012 - are used to determine the chaotic and fractal properties of solar activity. The results show that （1） the long-term solar activity is governed by a low-dimensional chaotic strange attractor, and its fractal motion shows a long-term persistence on large scales; （2） both the fractal dimension and maximal Lyapunov exponent of SAs are larger than those of SNs, implying that the dynamical system of SAs is more chaotic and complex than SNs; （3） the predictions of solar activity should only be done for short- to mid-term behaviors due to its intrinsic complexity; moreover, the predictability time of SAs is obviously smaller than that of SNs and previous results.

Latitudinal migration of sunspots based on the ESAI database

The latitudinal migration of sunspots toward the equator,which implies there is propagation of the toroidal magnetic flux wave at the base of the solar convection zone,is one of the crucial observational bases for the solar dynamo to generate a magnetic field by shearing of the pre-existing poloidal magnetic field through differential rotation.The Extended time series of Solar Activity Indices（ESAI）elongated the Greenwich observation record of sunspots by several decades in the past.In this study,ESAI＇s yearly mean latitude of sunspots in the northern and southern hemispheres during the years 1854 to 1985 is utilized to statistically test whether hemispherical latitudinal migration of sunspots in a solar cycle is linear or nonlinear.It is found that a quadratic function is statistically significantly better at describing hemispherical latitudinal migration of sunspots in a solar cycle than a linear function.In addition,the latitude migration velocity of sunspots in a solar cycle decreases as the cycle progresses,providing a particular constraint for solar dynamo models.Indeed,the butterfly wing pattern with a faster latitudinal migration rate should present stronger solar activity with a shorter cycle period,and it is located at higher latitudinal position,giving evidence to support the Babcock-Leighton dynamo mechanism.

On the Imminent Regional Seismic Activity Forecasting Using INTERMAGNET and Sun-Moon Tide Code Data

In this paper we present an approach for forecasting the imminent regional seismic activity by using geomagnetic data and Earth tide data. The time periods of seismic activity are the time periods around the Sun-Moon extreme of the diurnal average value of the tide vector module. For analyzing the geomagnetic data behaviour we use diurnal standard deviation of geomagnetic vector components F (North, East, Down) for calculating the time variance GeomagSignal. The Sun storm influence is avoided by using data for daily A-indexes (published by NOAA). The precursor signal for forecasting the incoming regional seismic activity is a simple function of the present and previous day GeomagSignal and A-indexes values. The reliability of the geomagnetic “when, regional” precursor is demonstrated by using statistical analysis of day difference between the times of “predicted” and occurred earthquakes. The base of the analysis is a natural hypothesis that the “predicted” earthquake is the one whose surface energy density in the monitoring point is bigger than the energy densities of all occurred earthquakes in the same period and region. The reliability of the approach was tested using the INTERMAGNET stations data located in Bulgaria, Panagurishte, PAG (Jan 1, 2008-Jan 29, 2014), Romania, Surlari, SUA (Jan 1, 2008-Jan 27, 2014), Italy, L’Aquila, AQU (Jan 1, 2008-May 30, 2013) in the time of EU IRSES BlackSeaHazNet (2011-2014) project. The steps of program for solving the “when, where and how” earthquake prediction problem are shortly described.

The causality between the rapid rotation of a sunspot and an X3.4 flare

Using multi-wavelength data of Hinode, the rapid rotation of a sunspot in active region NOAA 10930 is studied in detail. We found extraordinary counterclockwise rotation of the sunspot with positive polarity before an X3.4 flare. From a series of vector magnetograms, it is found that magnetic force lines are highly sheared along the neutral line accompanying the sunspot rotation. Furthermore, it is also found that sheared loops and an inverse S-shaped magnetic loop in the corona formed gradually after the sunspot rotation. The X3.4 flare can be reasonably regarded as a result of this movement. A detailed analysis provides evidence that sunspot rotation leads to magnetic field lines twisting in the photosphere. The twist is then transported into the corona and triggers flares.

Long-Term Sunspot Number Prediction based on EMD Analysis and AR Model

The Empirical Mode Decomposition （EMD） and Auto-Regressive model （AR） are applied to a long-term prediction of sunspot numbers. With the sample data of sunspot numbers from 1848 to 1992, the method is evaluated by examining the measured data of the solar cycle 23 with the prediction： different time scale components are obtained by the EMD method and multi-step predicted values are combined to reconstruct the sunspot number time series. The result is remarkably good in comparison to the predictions made by the solar dynamo and precursor approaches for cycle 23. Sunspot numbers of the coming solar cycle 24 are obtained with the data from 1848 to 2007, the maximum amplitude of the next solar cycle is predicted to be about 112 in 2011-2012.

Wavelet Analysis of Several Important Periodic Properties in the Relative Sunspot Numbers

We investigate the wavelet transform of yearly mean relative sunspot number series from 1700 to 2002. The curve of the global wavelet power spectrum peaks at 11-yr, 53-yr and 101-yr periods. The evolution of the amplitudes of the three periods is studied. The results show that around 1750 and 1800, the amplitude of the 53-yr period was much higher than that of the the 11-yr period, that the ca. 53-yr period was apparent only for the interval from 1725 to 1850, and was very low after 1850, that around 1750, 1800 and 1900, the amplitude of the 101-yr period was higher than that of the 11-yr period and that, from 1940 to 2000, the 11-yr period greatly dominates over the other two periods.

Low Dimensional Chaos from the Group Sunspot Numbers

We examine the nonlinear dynamical properties of the monthly smoothed group sunspot number Rg and find that the solar activity underlying the time series of Rg is globally governed by a low-dimensional chaotic attractor. This finding is consistent with the nonlinear study results of the monthly Wolf sunspot numbers. We estimate the maximal Lyaponuv exponent （MLE） for the Rg series to be positive and to equal approximately 0.0187 ± 0.0023 （month^- 1）. Thus, the Lyaponuv time or predictability time of the chaotic motion is obtained to be about 4.46 ± 0.5 years, which is slightly different with the predictability time obtained from Rz. However, they both indicate that solar activity forecast should be done only for a short to medium term due to the intrinsic complexity of the time behavior concerned.

How much has the Sun influenced Northern Hemisphere temperature trends?An ongoing debate

In order to evaluate how much Total Solar Irradiance(TSI)has influenced Northern Hemisphere surface air temperature trends,it is important to have reliable estimates of both quantities.Sixteen different estimates of the changes in TSI since at least the 19th century were compiled from the literature.Half of these estimates are"low variability"and half are"high variability".Meanwhile,five largely-independent methods for estimating Northern Hemisphere temperature trends were evaluated using:1)only rural weather stations;2)all available stations whether urban or rural(the standard approach);3)only sea surface temperatures;4)tree-ring widths as temperature proxies;5)glacier length records as temperature proxies.The standard estimates which use urban as well as rural stations were somewhat anomalous as they implied a much greater warming in recent decades than the other estimates,suggesting that urbanization bias might still be a problem in current global temperature datasets-despite the conclusions of some earlier studies.Nonetheless,all five estimates confirm that it is currently warmer than the late 19th century,i.e.,there has been some"global warming"since the 19th century.For each of the five estimates of Northern Hemisphere temperatures,the contribution from direct solar forcing for all sixteen estimates of TSI was evaluated using simple linear least-squares fitting.The role of human activity on recent warming was then calculated by fitting the residuals to the UN IPCC’s recommended"anthropogenic forcings"time series.For all five Northern Hemisphere temperature series,different TSI estimates suggest everything from no role for the Sun in recent decades(implying that recent global warming is mostly human-caused)to most of the recent global warming being due to changes in solar activity(that is,that recent global warming is mostly natural).It appears that previous studies(including the most recent IPCC reports)which had prematurely concluded the former,had done so because they failed to adequately consider all the relevant estimates of TSI and/or to satisfactorily address the uncertainties still associated with Northern Hemisphere temperature trend estimates.Therefore,several recommendations on how the scientific community can more satisfactorily resolve these issues are provided.

Predicting the 25th solar cycle using deep learning methods based on sunspot area data

It is a significant task to predict the solar activity for space weather and solar physics. All kinds of approaches have been used to forecast solar activities, and they have been applied to many areas such as the solar dynamo of simulation and space mission planning. In this paper, we employ the long-shortterm memory(LSTM) and neural network autoregression(NNAR) deep learning methods to predict the upcoming 25 th solar cycle using the sunspot area(SSA) data during the period of May 1874 to December2020. Our results show that the 25 th solar cycle will be 55% stronger than Solar Cycle 24 with a maximum sunspot area of 3115±401 and the cycle reaching its peak in October 2022 by using the LSTM method. It also shows that deep learning algorithms perform better than the other commonly used methods and have high application value.

Studies of grand minima in sunspot cycles by using a flux transport solar dynamo model

We propose that grand minima in solar activity are caused by simultane- ous fluctuations in the meridional circulation and the Babcock-Leighton mechanism for the poloidal field generation in the flux transport dynamo model. We present the following results： （a） fluctuations in the meridional circulation are more effective in producing grand minima; （b） both sudden and gradual initiations of grand minima are possible; （c） distributions of durations and waiting times between grand minima seem to be exponential; （d） the coherence time of the meridional circulation has an effect on the number and the average duration of grand minima, with a coherence time of about 30 yr being consistent with observational data. We also study the occurrence of grand maxima and find that the distributions of durations and waiting times between grand maxima are also exponential, like the grand minima. Finally we address the question of whether the Babcock-Leighton mechanism can be operative during grand minima when there are no sunspots. We show that an a-effect restricted to the upper portions of the convection zone can pull the dynamo out of the grand minima and can match various observational requirements if the amplitude of this a-effect is suitably fine-tuned.

What does the Sun tell and hint now?

Some historical records, which have held since the beginning of modern solar activity cycles, are being broken by the present Sun： cycle 23 records the longest cycle length and fall time; latitudes of high-latitude sunspots belonging to a new cycle around the minimum time of the cycle are statistically the lowest at present, compared with those of other cycles; there are only one or no sunspots in a month appearing at high latitudes for 58 months, which is the first time that such a long duration has been observed. The solar dynamo is believed to be slowing down due to： （1） the minimum smoothed monthly mean sunspot number is the smallest since cycle 16 onwards, and even probably among all modern solar cycles; and （2） once the time interval between the first observations of two neighboring sunspot groups is larger than 14 d, it should be approximately regarded as an observation of no sunspots on the visible solar disk, called a spotless event. Spotless events occur with the highest frequency around the minimum time of cycle 24, and the longest spotless event also appears around the minimum time for observations of the Sun since cycle 16. Cycle 24 is expected to have the lowest level of sunspot activity from cycle 16 onwards and even probably for all of the modern solar cycles.

Local sunspot oscillations and umbral dots

Data analysis of sunspot oscillations based on a 6-hr SDO run of an observation showed that low frequency （0.2 〈ω 〈 1 mHz） oscillations are locally similar to three and five minute oscillations. The oscillations in the sunspot are concentrated in cells of a few arcsec, each of which has its own oscillation spectrum. The analysis of two scenarios for sunspot oscillations leads to a conclusion that local sunspot oscillations occur due to a subphotospheric resonator for slow MHD waves. Empirical models of a sunspot atmosphere and the theory of slow waves in thin magnetic flux tubes are applied to modeling the subphotospheric resonator. The spectrum of local oscillations consists of a great number of lines. This kind of spectrum can occur only if the subphotospheric resonator is a magnetic tube with a rather weak magnetic field. Magnetic tubes of this sort are umbral dots that appear due to the convective tongues in monolithic sunspots. The interrelation of local oscillations with umbral dots and wavefronts of traveling waves in sunspots is discussed.

Solar flare forecasting based on sequential sunspot data

It is widely believed that the evolution of solar active regions leads to solar flares. However, information about the evolution of solar active regions is not employed in most existing solar flare forecasting models. In the current work, a short- term solar flare forecasting model is proposed, in which sequential sunspot data, in- cluding three days of information about evolution from active regions, are taken as one of the basic predictors. The sunspot area, the Mclntosh classification, the mag- netic classification and the radio flux are extracted and converted to a numerical for- mat that is suitable for the current forecasting model. Based on these parameters, the sliding-window method is used to form the sequential data by adding three days of information about evolution. Then, multi-layer perceptron and learning vector quanti- zation are employed to predict the flare level within 48 h. Experimental results indicate that the performance of the proposed flare forecasting model works better than previ- ous models.