Statistical Analyses of Solar Active Region in SDO/HMI Magnetograms detected by Unsupervised Machine Learning Method DSARD
R. Chen, W. Lu, Q. Hao, Y. Meng, P. F. Chen, C. Shi · 2025
Read Statistical Analyses of Solar Active Region in SDO/HMI Magnetograms detected by Unsupervised Machine Learning Method DSARD by R. Chen, W. Lu, Q. Hao, Y. Meng, P. F. Chen, C. Shi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-17990.txt.
Introduction
Ruishuo Chen and five co-authors let an unsupervised algorithm find the Sun’s active regions across a full solar cycle, in a 2025 paper. Active regions are the magnetically tangled patches where flares and eruptions are born. Cataloging them by hand across years of magnetograms is slow and inconsistent.
Their method, DSARD, uses density-based clustering to pick out regions in magnetic images from the Solar Dynamics Observatory. It then measured region shapes and evolution over solar cycle 24 and the start of cycle 25. Automated, uniform catalogs sharpen the statistics linking region traits to eruption risk. Space weather forecasters, who warn satellite and grid operators, depend on exactly such statistics to judge when the Sun is dangerous.
Catalog record
| CREATOR | R. Chen, W. Lu, Q. Hao, Y. Meng, P. F. Chen, C. Shi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-17990
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Statistical Analyses of Solar Active Region in SDO/HMI Magnetograms detected by Unsupervised Machine Learning Method DSARD R. Chen, W. Lu, Q. Hao, Y. Meng, P. F. Chen, C. Shi Abstract Solar active regions (ARs) are the places hosting the majority of solar eruptions. Studying the evolution and morphological features of ARs is not only of great significance to the understanding of the physical mechanisms of solar eruptions, but also beneficial for the hazardous space weather forecast. An automated DBSCAN-based Solar Active Regions Detection (DSARD) method for solar ARs observed in magnetograms is developed in this work, which is based on an unsupervised machine learning algorithm called Density-Based Spatial Clustering of Applications with Noise (DBSCAN). The method is then employed to identify ARs on the magnetograms observed by the Helioseismic and Magnetic Imager (HMI) onboard Solar Dynamics Observatory (SDO) during solar cycle 24 and the rising phase of solar cycle 25. The distributions of the number, area, magnetic flux, and the tilt angle of bipolar of ARs in latitudes and time intervals during solar cycle 24, as well as the butterfly diagram and drift velocities are obtained. Most of these statistical results based on our method are in agreement with previous studies, which also guarantees the validity of the method. In particular, the dipole tilt angles in ARs in solar cycle 24 and the rising phase of solar cycle 25 are analyzed which reveal that 13% and 16% of ARs, respectively, violate Hale’s law. 1 1 1 R. Chen and W. Lu are co-first authors. I Introduction Active regions (ARs) are the patchy volumes in the solar atmosphere with strong magnetic fields, where a variety of activities occur, manifesting in complex spatial and temporal behaviors extending from the photosphere to the corona ( van Driel-Gesztelyi & Green 2015 ) . Sunspots, being the initial features discernible as tracers of ARs on the Sun, appear dark areas on the surface of the Sun in continuum, occasionally perceptible to the naked eyes ( Solanki 2003 ) . They mark the locations where magnetic field is sufficiently strong to suppress the convective flows that transport heat from the interior to the solar surface. Observations of sunspots in ARs have revealed several characteristics that are crucial to understanding the physical mechanisms of the solar cycles. One such example is the equatorward migration of sunspots known as the “Butterfly Diagrams” ( Maunder 1904 ) , which further led to the discovery of various magnetic properties of sunspots, including the famous Joy’s law and Hale’s polarity law ( Maunder 1904 ) . The most significant aspect of ARs and sunspots are their capacity to generate strong solar flares and, sometimes, coronal mass ejections (CMEs), which have the potential to significantly impact the space weather. Consequently, the study of the evolution of ARs and their relationship with solar eruptions is not only useful in predicting eruptive events ( Schrijver 2007 ; Chen 2011 ; Bobra & Couvidat 2015 ; Barnes et al. 2016 ; Sun et al. 2022 ; Zhang et al. 2022 ) , but also in understanding the long-term evolution of solar activities ( Hathaway 2010 ; Ravindra et al. 2022 ) . The National Oceanic and Atmospheric Administration (NOAA) Space Weather Prediction Center (SWPC) has developed a widely referenced catalog of active regions (ARs) based on white-light images. NOAA assigns a unique number to an AR if it contains at least one sunspot, with the numbering determined either manually or through computer-aided inspection assisted by human operators. This catalog offers detailed information, including the location of sunspot groups, their areas, and magnetic morphology. However, significant advancements in observational techniques have led to a substantial increase in both the quantity and quality of multiwavelength observational data. Traditional manual detection or labeling of ARs on images has become increasingly inadequate in meeting the demands for rapid and accurate processing of large datasets, particularly for long-term statistical studies and solar dynamo modeling. Recognizing the limitations of human inspection, the development of automated detection methods has gained momentum. These methods aim to provide objective and consistent extraction of ARs, addressing the challenges posed by manual processes and enabling more efficient analysis of solar data. An early class of automated detection methods for ARs was primarily developed using image processing techniques ( Zharkov et al. 2005 ; McAteer et al. 2005a ; McAteer et al. 2005b ; Benkhalil et al. 2006 ; Colak & Qahwaji 2008 ; Zhang et al. 2010 ; Higgins et al. 2011 ; Cui et al. 2021 ) . Verbeeck et al. 2013 conducted a comprehensive comparison of four typical automated detection methods for ARs or sunspots observed in images with different wavelengths. These methods typically involve threshold segmentation on magnetograms or white-light images, followed by region-growing and morphological operations to derive the areas of ARs. For instance, Zharkov et al. 2005 employed an edge detection method after standardizing images to identify sunspot candidates. This was followed by local thresholding for image segmentation, and subsequent watershed and morphological operations to smooth features and fill gaps. The temporal variations in the detected sunspot areas showed a high correlation with NOAA’s records, validating the effectiveness of the proposed method. This sunspot detection approach was applied to white-light images captured by the Michelson Doppler Imager (MDI; Scherrer et al. 1995 ) aboard the Solar and Heliospheric Observatory (SOHO) spacecraft from 1996 to 2004. The results revealed that the number of sunspots increases near-exponentially as the area decreases over the observed period. Additionally, the sunspot areas exhibited a periodicity of approximately 7–8 years in terms of north-south asymmetry. Zhang et al. 2010 developed an automated AR detection method incorporating three image processing techniques: intensity thresholding, region-growing, and morphological operations. This method was tested and applied to the Carrington synoptic magnetograms constructed from SOHO/MDI images spanning 1996 to 2008. The method identified 1730 ARs and yielded significant statistical insights, including morphological features, magnetic flux, and drift velocities during solar cycle 23. The true positive and false positive rates relative to the NOAA AR catalog were found to depend on the detection threshold. For example, the frequency distribution of ARs in terms of area size and magnetic flux follows a log-normal function, but shifts to a power-law distribution when the detection thresholds are lowered. Despite computational biases, the authors suggested that the distribution of solar magnetic features, from large to small, might consist of two components: a log-normal distribution for large ARs and a power-law distribution for smaller ARs and other minor features. However, these methods rely on empirically assigned thresholds and struggle to effectively separate ARs in complex magnetic field regions on high-resolution images, leading to incomplete identification. Their robustness depends on thresholds determined through extensive trial and error, making them susceptible to instability due to detector sensitivity degradation. The advancement of machine learning methods in computer vision technology has significantly expanded their application in the detection of solar ARs. One widely used approach is based on clustering, a form of unsupervised machine learning. Barra et al. 2008 introduced a fuzzy clustering technique to automatically segment EUV images into regions of coronal holes, quiet Sun, and active regions. Noticing some artifacts of the results, they later improved the method and applied