Using deep learning to identify and classify sunspots in magnetograms

Some phenomena that occur in the Sun have consequences on Earth. Among these phenomena, solar flares release large amounts of radiation and energy that impact on Earth's life and technological systems. These flares usually come from sunspots, which derive from solar magnetic activities. One strategy to predict solar flares is to identify active regions, i. e., a group of sunspots with a high potential to cause solar flares. This paper reports the use of the deep learning technique to identify and classify active regions from magnetogram analysis. To achieve these tasks, we created a dataset with magnetograms and performed tests to choose the best deep learning models for the identification and classification of active regions. The results of the best models reached accuracies higher than 80% for both the identification and classification tasks. Based on these results, we implemented a system in Python to automate the complete identification and classification process, also reported in this paper.

Using the facilities at Heliophysics Science Division of NASA Goddard Space Flight Center, Greenbelt, MD, USA, we attempted to investigate the impact of solar magnetic activities on the climate of Wet Zone West Africa. The solar activity data namely, Sunspot Number (SSN) was obtained from the Royal Observatory of Belgium, Brussels; and Geomagnetic aa-index was obtained from World Data Center, Kyoto, Japan. Surface Air Temperature (SAT) and Rainfall data [for Port Harcourt in Nigeria and Abidjan in Cote D’Ivoire] were obtained from the HadCRUT-4 project of Climate Research Unit of University of East Anglia, United Kingdom. Firstly, we carried out Time Series Analysis of SSN and Geomagnetic aa-index spanning from 1950 to 2016. Secondly, we performed Regression Analysis on both solar activity data and climate variables to estimate the impact of solar magnetic activity on the Wet Zone West African climate. The Time Series Analysis showed that SSN variation was in-phase with Geomagnetic aa-index in all the solar cycles studied. Thus, Geomagnetic aa-index can be used as a proxy for studying solar magnetic activities. Performance of Regression Analysis showed that SSN regressed on SAT and Rainfall amounted to an average of 0.49 and 0.02% respectively throughout Solar Cycles 22–24. Furthermore, a regression of Geomagnetic aa-index on SAT and Rainfall yielded an average of 0.145 and 0.125% respectively. Our models showed that the variability of SAT and Rainfall in Wet Zone West Africa during Solar Cycles 22–24 are far less than 1%. Hence, the influence of SSN and Geomagnetic aa-index on SAT and Rainfall is less than 1%; and could cause ‘very small’ effect. These weak impacts are proofs that the variability of SAT and Rainfall were most probably not effected by SSN and Geomagnetic aa-index. Consequently, the variability of SAT and Rainfall in Wet Zone West Africa could not be attributed to SSN and Geomagnetic aa-index. We therefore, attempt to conclude that climate variability in Wet Zone West Africa is most probably not driven by solar magnetic activity, but could be attributed to anthropogenic activities.

A new daily composite of the solar flare index (SFI) and the hemispherically‐resolved versions (hSFI) are presented for 1937 to 2024. The data set confirms that the northern hemisphere (NH) dominated solar flare activity during Solar Cycles 17 to 21, but that the southern hemisphere has dominated from Solar Cycle 22 to present. That said, the highest SFI value occurred in the NH during the recent superstorm of May 2024. In sunspot activity, the “Gnevyshev‐Ohl rule” shows that the sum of sunspot numbers during even‐numbered cycles is related to those of adjacent odd‐numbered cycles. A similar rule appears to apply to SFI. The “Gnevyshev gap” phenomenon where solar maximum activity sometimes has two peaks separated by up to 1–2 years of a gap is confirmed for SFI. Although our data set represents the longest continuous daily data set for solar flare activity to‐date, it is known that stronger solar flare events occurred before 1937. Therefore, a brief discussion of earlier solar flare events in the historical record is also provided for context. The statistics of the SFI and hSFI series are compared to other solar and geomagnetic activity indices, including the May and October 2024 solar storms. Statistical analysis of past geomagnetic storms confirms they are more frequent during active cycles and less frequent during solar minima. Strong geomagnetic storms are also more likely to occur during the positive phase of a 1.7 year's quasi‐biennial oscillation in solar activity. The likelihood of low‐magnetic latitude aurorae seems to have a 30 year periodicity component.

The Sun’s variability is controlled by the progression and interaction of the magnetized systems that form the 22-year magnetic activity cycle (the “Hale Cycle”) as they march from their origin at ∼55° latitude to the equator, over ∼19 years. We will discuss the end point of that progression, dubbed “terminator” events, and our means of diagnosing them. In this paper we expand on the Extended Solar Cycle framework to construct a new solar activity “clock” which maps all solar magnetic activity onto a single normalized epoch based on the terminations of Hale Magnetic Cycles. Defining phase 0*2π on this clock as the Terminators, then solar polar field reversals occur at ∼ 0.2*2π, and the geomagnetically quiet intervals centered around solar minimum start at ∼ 0.6*2π and end at the terminator, thus lasting 40% of the cycle length. At this onset of quiescence, dubbed a “pre-terminator,” the Sun shows a radical reduction in active region complexity and, like the terminator events, is associated with the time when the solar radio flux crosses F10.7 = 90 sfu. We use the terminator-based clock to illustrate a range of phenomena that further emphasize the strong interaction of the global-scale magnetic systems of the Hale Cycle: the vast majority, 96%, of all X-flares happen between the Terminator and pre-Terminator. In addition to the X-rays from violent flares, rapid changes in the number of energetic photons—EUV spectral emission from a hot corona and the F10.7 solar radio flux—impinging on the atmosphere are predictable from the Terminator-normalized unit cycle, which has implications for improving the fidelity of atmospheric modelling.

A full-disk solar magnetogram has been measured each day since 1970 January 19, and the daily Magnetic Plage Strength Index (MPSI) and the daily Mount Wilson Sunspot Index (MWSI) were calculated for each magnetograrn. at the Mount Wilson Observatory. The MPSI and MWSI are used to investigate the periodicity of the solar full-disk magnetic activity through autocorrelation analyses. Just two periods, the solar cycle and the rotation cycle, are determined in both the MPSI (the solar full-disk weak magnetic field activity) and MWSI (the solar full-disk strong magnetic field activity) with no annual signal found. The solar cycle for MPSI (10.83 yr) is found to be obviously longer than that for MW SI (9.77 yr). The rotation cycle is determined to be 26.8 0.63 sidereal days for MPSI and 27.4 2.4 sidereal days for MWSI. The rotation cycle length for MPSI is found to fluctuate around 27 days within a very small amplitude, but for MWSI it obviously temporally varies with a rather large amplitude. The rotation cycle for MWSI seems longer near solar minimum than at solar maximum. Cross-correlation analyses of daily MPSI and MWSI are carried out, and it is inferred that the MPSI components partly come from relatively early MWSI measurements.

Solar activity could have significant impacts on various Earth and near-Earth space systems, such as satellite communication and power grid systems. The prediction of solar activity and active solar events plays a major role when preparing for these disturbances. Various satellite-based instruments constantly observe the Sun. However, only a few ground-based solar instruments could provide versatile enough information for the space weather prediction. Metsähovi Radio Observatory of Aalto University (Finland) has a unique collection of millimetre (8 mm) solar radio maps over the past 40 years, and even denser observational solar radio data catalogue since 2011. About 75–80 % days yearly are covered nowadays. This gives opportunity to make statistical estimation of solar flare occurrence based on solar radio maps. In this study, we had 2253 days when both solar radio map and GOES (Geostationary Operational Environmental Satellites) classified solar flare were observed. In this work, we used solar flare classification done by the Space Weather Centre (SWC) of the National Oceanic and Atmospheric Administration (NOAA). The data were observed between 1 January 2011 and 12 September 2022. Our study shows that the maximum intensity of radio brightenings is a good indicator to tell which kind of GOES classified solar flare could be expected to happen. The article presents that intense radio brightening is needed to produce a certain GOES classified solar flare.

Sunspot is known as the area with the highest solar magnetic activity and has the coolest region compared to its surrounding. It is also influenced by sunspot complexity and magnetic field strength. The huge explosion of solar energetic particles in the interplanetary space due to Coronal Mass Ejections (CMEs) and solar flares is believed due to instability of magnetic field and complexity magnetic field in the sunspot. Due to these events, the intense solar radio emission also known as solar radio bursts were released to the Earth. Delta (δ) sunspot is one of the largest magnetic classification of sunspot and seen mostly released solar flares and CMEs, however, Beta Gamma (βγ) sunspot has the highest formation from 2012 to 2016 compared to other type of bursts and has tendency to produce solar flares and CMEs. The data were collected using Space Weather Website (SWPC) and NOAA, e-CALLISTO network and LPDA. The data were collected within 5 years (2012-2016) and reveals βγ sunspot has (349 counts) (62.77%), followed by βγσ sunspot (160 counts) (28.78%), δ sunspot (39 counts) (7.01%) and βγ sunspot (8 counts) (1.44%), thus, highlights the highest type of magnetic classification occurred on Active Region (AR) was βγ sunspot. Highest formation of βγ sunspot illustrated the complexity of the βγ sunspot due instability of magnetic field inside AR and has tendency to produce flares and CMEs compared to other types of magnetic classifications. Furthermore, βγ sunspot demonstrated a tendency to generate Type II burst, Type III burst, Type IV burst and combination of bursts which we classified as selected events. This finding proved that Type IV burst has the highest formation of burst followed by Type III burst, Type II burst and selected events. Type IV burst has counted for 161 counts (56.29%), Type III burst (100 counts) (34.96%), Type II burst (14 counts) (4.9%) and selected events (11 days) (3.85%). The type of bursts was determined by the range frequency of burst and the duration of burst. The different types of bursts have different ranges of frequency and duration of bursts. Type IV burst displayed the highest formation due its associated phenomena mostly was sunspot where the location of events occurred, followed by Type III burst (solar flares), Type II burst (CMEs) and selected events. These findings were important in solar activity study as prediction and forecast for Solar Cycle 25. These significant findings manage the process of shutting down the system from breakdown due to high energetic particles being released from solar flares and CMES. This prediction can shut down the system in satellites and prevent the power grid and telecommunication system from breakdown.

The method of singular spectral analysis (SSA) is described and used to analyze the series of Wolf numbers that characterizes solar activity from 1748 until 1996. Since this method is relatively new, we detail its algorithm as applied to the data under study. We examine the advantages and disadvantages of the SSA method and the conditions for its applicability to an analysis of the solar-activity data. Certain regularities have been found in the dynamics of this series. Both short and long (80–100-year) periodicities have been revealed in the sunspot dynamics. We predict the solar activity until 2014.

This study investigated the effects of solar flare activity using a baseline during solar minimum from July 2019 through December 2019 to determine a threshold and then assessing the impacts during the solar maximum period from January 2022 through May 2022 on Global Positioning System (GPS) timing sources. GPS satellite data received at the National Institute of Standards and Technology (NIST) was collected and overlaid with the solar activity archive data from SpaceWeatherLive.com to determine whether any correlation existed to showcase effects of solar flare activity on GPS satellite timing sources. Focus of the resulting data was on M-class and X-class flares due to the severity based on prior research papers on the topic. The resulting data analysis of the datasets indicates that each flare class not only impacted the GPS satellite offset value in nanoseconds, but that after each M-class and X-class flare, it was determined that GPS offset value in nanoseconds remains high for several days after the event. Given that the sun is entering solar maximum, more research is needed to better understand the impacts of higher-class solar events on GPS satellite technology (i.e., timing, navigation, etc.) to ensure systems are better protected before events occur especially upon a direct impact from a solar flare.

Magnetic field structures on the solar atmosphere are not symmetric distribution in the northern and southern hemispheres, which is an important aspect of quasi-cyclical evolution of magnetic activity indicators that are related to solar dynamo theories. Three standard analysis techniques are applied to analyze the hemispheric coupling (north-south asymmetry and phase asynchrony) of monthly averaged values of solar Hα flare activity over the past 49 years (from 1966 January to 2014 December). The prominent results are as follows: (1) from a global point of view, solar Hα flare activity on both hemispheres are strongly correlated with each other, but the northern hemisphere precedes the southern one with a phase shift of 7 months; (2) the long-range persistence indeed exists in solar Hα flare activity, but the dynamical complexities in the two hemispheres are not identical; (3) the prominent periodicities of Hα flare activity are 17 years full-disk activity cycle and 11 years Schwabe solar cycle, but the short- and mid-term periodicities cannot determined by monthly time series; (4) by comparing the non-parametric rescaling behavior on a point-by-point basis, the hemispheric asynchrony of solar Hα flare activity are estimated to be ranging from several months to tens of months with an average value of 8.7 months. The analysis results could promote our knowledge on the long-range persistence, the quasi-periodic variation, and the hemispheric asynchrony of solar Hα flare activity on both hemispheres, and possibly provide valuable information for the hemispheric interrelation of solar magnetic activity.

Solar activity prediction studies and services in NAOC

The solar magnetic activity is governed by a complex dynamo mechanism and exhibits a nonlinear dissipation behavior in nature. The chaotic and fractal properties of solar time series are of great importance to understanding the solar dynamo actions, especially with regard to the nonlinear dynamo theories. In the present work, several nonlinear analysis approaches are proposed to investigate the nonlinear dynamical behavior of the polar faculae and sunspot activity for the time interval from 1951 August to 1998 December. The following prominent results are found: (1) both the high- and the low-latitude solar activity are governed by a three-dimensional chaotic attractor, and the chaotic behavior of polar faculae is the most complex, followed by that of the sunspot areas, and then the sunspot numbers; (2) both the high- and low-latitude solar activity exhibit a high degree of persistent behavior, and their fractal nature is due to such long-range correlation; (3) the solar magnetic activity cycle is predictable in nature, but the high-accuracy prediction should only be done for short- to mid-term due to its intrinsically dynamical complexity. With the help of the Babcock–Leighton dynamo model, we suggest that the nonlinear coupling of the polar magnetic fields with strong active-region fields exhibits a complex manner, causing the statistical similarities and differences between the polar faculae and the sunspot-related indicators.

The solar magnetic activity consists of two periodic components: the main cycle with a period of 11 yr and a shorter cycle with a period of ≈2 yr. The origin of this second periodicity is still not well understood. We use almost 15 years of long high‐quality resolved data provided by the Global Oscillation Network Group (GONG) to investigate the solar cycle changes in p‐mode oscillations with spherical degree Ł = 0–120 and in the range of 1600 µHz ≤ ν ≤ 3500 µHz. For both periodic components of solar magnetic activity our findings locate the origin of the frequency shift in the subsurface layers with a sudden enhancement in the amplitude of the shift in the last few hundred kilometers. We also show that the size of the shift increases towards equatorial latitudes and from minimum to maximum of solar activity. On the other hand, the signatures of the 2‐yr cycle differ from the one of the 11‐yr cycle in the magnitude of the shift, as the 2‐yr cycle causes a weaker shift in mode frequencies and a slower enhancement in the last few hundred kilometers. Based on these findings we speculate that a possible physical mechanism behind the quasi biennial periodicity (QBP) could be the beating between different dynamo modes (dipole and quadrupole mode) (© 2012 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)

The Earth’s geomagnetic field serves as a crucial protective barrier, shielding the planet from the potentially hazardous effects of solar and cosmic radiation. Despite its stabilizing presence, this magnetic shield is highly dynamic and susceptible to disturbances originating from solar and interplanetary phenomena. These disturbances often manifest as geomagnetic storms, which can have serious implications for technological infrastructure, including satellite integrity, communication and navigation systems, and ground-based power grids. This review provides a comprehensive synthesis of historical and contemporary research that explores the intricate pathways through which solar activity particularly Coronal Mass Ejections (CMEs), solar flares, and high-speed solar wind streams interacts with the interplanetary magnetic field (IMF) and subsequently with the Earth’s magnetosphere. Central to this interaction is the process of magnetic reconnection, especially when the IMF exhibits a southward Bz component, which enhances the coupling between solar and terrestrial magnetic fields and triggers energy transfer into the magnetosphere. These interactions are quantitatively assessed using geomagnetic indices such as the Dst (Disturbance Storm Time) index, the Kp index, and the AE (Auroral Electrojet) index, which offer measurable proxies for storm intensity and auroral activity. Furthermore, statistical analyses conducted across solar cycles 24 and 25 have provided critical insights into long term solar-terrestrial dynamics, revealing patterns and thresholds associated with storm onset and severity. The review also highlights the evolution of space weather forecasting methodologies, noting a growing shift towards machine learning and data-driven models that can integrate multi-parametric solar, interplanetary, and geomagnetic data to enhance prediction accuracy. As space weather events pose increasing risks in a technologically dependent world, understanding and forecasting geomagnetic disturbances remain pivotal scientific and operational priorities.

Hostile space weather is one of the principal threats to modern human technology because of the dependency of our lives on satellites. Solar coronal mass ejections, solar flares, and high‐speed solar wind streams often cause some sequences of damaging disturbances within the Earth's magnetosphere, in the atmosphere, and even on the Earth's surface. The time‐varying Sun, as the main source of space weather causes geomagnetic storms and substorms in many ways with coronal mass ejections (CME) being possibly the most harmful among them. CMEs cause transient magnetized plasma flows as interplanetary consequences, which drive the geomagnetic storms and substorms in the Earth's magnetosphere. It is shown that the cyclic solar activity has chaotic characteristics. It seems that the chaotic modeling of solar and geomagnetic activity indices is an important approach for analysis and prediction of the solar extreme events. This is particularly useful for engineers and designers interested in space weather prediction and its applications. One of the most important tools for eliciting the chaotic trends is the “Lyapunov exponents,” which is a useful measure of the stability of a dynamic system. This paper analyzes the variation of Lyapunov exponents (LE) for three solar and geomagnetic activity indices during CMEs: the sunspot number, disturbance storm time, and proton temperature. The analysis has been carried out using the technique of adaptive LE estimation adopted from previous works; the suitable window for which has been determined via trial and error. It is shown that the LE of these solar and geomagnetic activity indices varies rapidly during CMEs. The variation in LEs creates a pattern as a precursor for the forthcoming CME. This precursor, which is an oscillation in the values of Les, begins several steps sooner than the CME's occurrence. Then, during the CME, the LEs decrease to a small positive or a negative value, which demonstrates that during an anomaly such as a CME the chaotic characteristics of solar and geomagnetic activity indices decrease and solar and geomagnetic activity indices follow more regular dynamics.