Sunspot drawings handwritten character recognition method based on deep learning

High accuracy scanned sunspot drawings handwritten characters recognition is an issue of critical importance to analyze sunspots movement and store them in the database. This paper presents a robust deep learning method for scanned sunspot drawings handwritten characters recognition. The convolution neural network (CNN) is one algorithm of deep learning which is truly successful in training of multi-layer network structure. CNN is used to train recognition model of handwritten character images which are extracted from the original sunspot drawings. We demonstrate the advantages of the proposed method on sunspot drawings provided by Chinese Academy Yunnan Observatory and obtain the daily full-disc sunspot numbers and sunspot areas from the sunspot drawings. The experimental results show that the proposed method achieves a high recognition accurate rate.     

A new interpretation of Christian Horrebow''s sunspot observations from 1761 to 1777

Christian Horrebow and his colleagues of Copenhagen, Denmark, actively observed sunspots from 1761 to 1777. These observations were examined by Thiele in 1859 and by d'Arrest in 1873 with markedly different conclusions. Thiele reported nearly twice as many sunspot groups as d'Arrest. To resolve this discrepancy, we have reexamined Horrebow's original notebooks. We find slightly more sunspot groups then did d'Arrest. Thiele apparently called individual sunspots sunspot groups, so he would call a bipolar group two groups. d'Arrest seems to have missed counting some of the smaller sunspot groups. A correct interpretation of Horrebow's observations is required in efforts to reconstruct solar activity. Wolf gave a sunspot number for 1769 of 106.1. On the basis of our re-examination of Horrebow's drawings and other observers, we deduce a sunspot number of about 80.5 for 1769.     

Digitization of Sunspot Drawings by Staudacher in 1749 – 1796

Original drawings by J.C. Staudacher made in the period of 1749 – 1796 were digitized. The drawings provide information about the size of the sunspots and are therefore useful for analyses sensitive to sunspot area rather than Wolf numbers. The total sunspot area as a function of time is shown for the observing period. The sunspot areas measured do not support the proposition of a weak, “lost” cycle between cycles 4 and 5. We also evaluate the usefulness of the drawings for the determination of sunspot positions for future studies.     

Digitization of sunspot drawings by Spörer made in 1861–1894

Most of our knowledge about the Sun's activity cycle arises from sunspot observations over the last centuries since telescopes have been used for astronomy. The German astronomer Gustav Spörer observed almost daily the Sun from 1861 until the beginning of 1894 and assembled a 33‐year collection of sunspot data covering a total of 445 solar rotation periods. These sunspot drawings were carefully placed on an equidistant grid of heliographic longitude and latitude for each rotation period, which were then copied to copper plates for a lithographic reproduction of the drawings in astronomical journals. In this article, we describe in detail the process of capturing these data as digital images, correcting for various effects of the aging print materials, and preparing the data for contemporary scientific analysis based on advanced image processing techniques. With the processed data we create a butterfly diagram aggregating sunspot areas, and we present methods to measure the size of sunspots (umbra and penumbra) and to determine tilt angles of active regions. A probability density function of the sunspot area is computed, which conforms to contemporary data after rescaling. (© 2015 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

On the Reliability of the de la rue Sunspot Area Measurements

In the 19th century De la Rue, Stewart, and Loewy carried out a compilation of drawings and photographs of the solar sunspots in the period 1832–1868. From these drawings and photographs, they determined fortnightly values of the sunspot areas. In this work, monthly values of the sunspot areas for the period 1832–1868 are calculated and the reliability of these data in terms of the solar activity indices is discussed.     

Sunspot Positions and Areas from Observations by Pierre Gassendi

Solar activity behaviour on the eve of the Maunder minimum may provide important information on the period of further suppression of sunspot population. We analyse sunspot positions and areas in the 1630s extracted from rare drawings published by Pierre Gassendi in Opera Omnia. This work was published in two different editions, the first in Lyon and the second almost 70 years later in Florence. The drawings published in Lyon are found to be slightly different from those published in Florence, which produces a discrepancy in the position of spots of a few degrees, while sunspot group areas may differ by a factor of two. We reveal that the orientation of the drawings in the book is not always the same as might be seen in the telescope. We conjecture that the time of Gassendi’s observations covers the beginning of a new Schwabe cycle in the southern hemisphere. The differential rotation rate in the 1630s is also assessed and discussed.     

Study of Multiple Coronal Mass Ejections at Solar Minimum Conditions

Solar activity alternates between active and quiet phases with an average period of 11?years, and this is known as the Schwabe cycle. Additionally, solar activity occasionally falls into a prolonged quiet phase (grand solar minimum), as represented by the Maunder Minimum in the 17th century, when sunspots were almost absent for 70?years and the length of the Schwabe cycle increased to 14?years. To examine the consistency of the cycle length characteristics during the grand solar minima, the carbon-14 contents in single-year tree rings were measured using an accelerator mass spectrometer as an index of the solar variability during the grand solar minimum of the 4th century BC. The signal of the Schwabe cycle was detected with a statistical confidence level of higher than 95?% by wavelet analysis. This is the oldest evidence for the Schwabe cycle at the present time, and the cycle length is considered to have increased to approximately 16?years during the grand solar minimum of the 4th century BC. This result confirms the association between the increase of the Schwabe cycle length and the weakening of solar activity, and indicates the possible prolonged absence of sunspots in the 4th century BC as during the Maunder Minimum. Theoretical implications from solar dynamo theory are discussed in order to identify the trigger of prolonged sunspot absence. A possible association between the long-term solar variation around the 4th century BC and terrestrial cooling in this period is also discussed.     

Solar activity around AD 775 from aurorae and radiocarbon

A large variation in ¹⁴C around AD 775 has been considered to be caused by one or more solar super‐flares within one year. We critically review all known aurora reports from Europe as well as the Near, Middle, and Far East from AD 731 to 825 and find 39 likely true aurorae plus four more potential aurorae and 24 other reports about halos, meteors, thunderstorms etc., which were previously misinterpreted as aurorae or misdated; we assign probabilities for all events according to five aurora criteria. We find very likely true aurorae in AD 743, 745, 762, 765, 772, 773, 793, 796, 807, and 817. There were two aurorae in the early 770s observed near Amida (now Diyarbakır in Turkey near the Turkish‐Syrian border), which were not only red, but also green‐yellow – being at a relatively low geomagnetic latitude, they indicate a relatively strong solar storm. However, it cannot be argued that those aurorae (geomagnetic latitude 43 to 50°, considering five different reconstructions of the geomagnetic pole) could be connected to one or more solar super‐flares causing the ¹⁴C increase around AD 775: There are several reports about low‐ to mid‐latitude aurorae at 32 to 44° geomagnetic latitude in China and Iraq; some of them were likely observed (quasi‐)simultaneously in two of three areas (Europe, Byzantium/Arabia, East Asia), one lasted several nights, and some indicate a particularly strong geomagnetic storm (red colour and dynamics), namely in AD 745, 762, 793, 807, and 817 – always without ¹⁴C peaks. We use 39 likely true aurorae as well as historic reports about sunspots together with the radiocarbon content from tree rings to reconstruct the solar activity: From AD ∼733 to ∼823, we see at least nine Schwabe cycles; instead of one of those cycles, there could be two short, weak cycles – reflecting the rapid increase to a high ¹⁴C level since AD 775, which lies at the end of a strong cycle. In order to show the end of the dearth of naked‐eye sunspots, we discuss two more Schwabe cycles until AD ∼844. The ¹⁴C record (from both Intcal and Miyake et al. 2013a) is anti‐correlated to auroral and sunspot activity, as expected from solar wind modulation of cosmic rays which produce the radiocarbon. (© 2015 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Statistical Properties Of Sunspots In 1996–2004: I. Detection, North–South Asymmetry And Area Distribution

The first statistical results in sunspot distributions in 1996–2004 obtained from the Solar Feature Catalogues (SFC) are presented. A novel robust technique is developed for automated identification of sunspots on SOHO/MDI white-light (WL) full-disk solar images. The technique applies image standardization procedures for elimination of the limb darkening and non-circular image shape, uses edge-detection methods to find the sunspot candidates and their edges and morphological operations to smooth the features and fill in gaps. The detected sunspots are verified with the SOHO/MDI magnetograms by strong magnetic fields being present in sunspots. A number of physical and geometrical parameters of the detected sunspot features are extracted and stored in the relational SFC database including umbra/penumbra masks in the form of run-length data encoding of sunspot bounding rectangles. The detection results are verified by comparison with the manual daily detection results in Meudon and Locarno Observatories in 2002 and by correlation (about 96%) with the 4 year sunspot areas produced manually at NOAA. Using the SFC data, sunspot area distributions are presented in different phases of the solar cycle and hemispheres which reveals a periodicity of the north–south asymmetry with a period of about 7–8 years. The number of sunspots increases exponentially with the area decrease with the index slightly increasing from −1.15 (1997) to −1.34 (2001).     

Wavelet Analysis of the Schwabe Cycle Properties in Solar Activity

Properties of the Schwabe cycles in solar activity are investigated by using wavelet transform. We study the main range of the Schwabe cycles of the solar activity recorded by relative sunspot numbers, and find that the main range of the Schwabe cycles is the periodic span from 8-year to 14-year. We make the comparison of 11-year‘s phase between relative sunspot numbers and sunspot group numbers. The results show that there is some difference between two phases for the interval from 1710 to 1810, while the two phases are almost the same for the interval from 1810 to 1990.     

Measurement of Kodaikanal White-Light Images – III. Rotation Rates and Activity Cycle Variations

The Kodaikanal sunspot data set covering the interval 1906–1987 is analyzed for differential rotation of sunspots of different sizes. As is known, smaller sunspots rotate faster than larger sunspots, and this result is verified in the analysis of this data set. These results agree well with the Mount Wilson sunspot results published earlier. The activity cycle dependence of sunspot rotation is studied. An increase in this rate at the minimum phase is seen, which has been reported earlier. It is demonstrated that this cycle variation is seen for sunspots in all size categories, which suggests that it is not a relative increase in the number of the faster-rotating small sunspots that causes the cycle dependence. These results are discussed as they may relate to subsurface dynamic properties of the Sun.     

10.7-cm Solar radio flux and the magnetic complexity of active regions

During sunspot cycles 20 and 21, the maximum in smoothed 10.7-cm solar radio flux occurred about 1.5 yr after the maximum smoothed sunspot number, whereas during cycles 18 and 19 no lag was observed. Thus, although 10.7-cm radio flux and Zürich suspot number are highly correlated, they are not interchangeable, especially near solar maximum. The 10.7-cm flux more closely follows the number of sunspots visible on the solar disk, while the Zürich sunspot number more closely follows the number of sunspot groups. The number of sunspots in an active region is one measure of the complexity of the magnetic structure of the region, and the coincidence in the maxima of radio flux and number of sunspots apparently reflects higher radio emission from active regions of greater magnetic complexity. The presence of a lag between sunspot-number maximum and radio-flux maximum in some cycles but not in others argues that some aspect of the average magnetic complexity near solar maximum must vary from cycle to cycle. A speculative possibility is that the radio-flux lag discriminates between long-period and short-period cycles, being another indicator that the solar cycle switches between long-period and short-period modes.Operated by the Association of Universities for Research in Astronomy, Inc. under contract with the National Science Foundation.     

What do the solar activity indices represent?

Sunspot number, sunspot area, and radio flux at 10.7 cm are the indices which are most frequently used to describe the long‐term solar activity. The data of the daily solar full‐disk magnetograms measured at Mount Wilson Observatory from 19 January 1970 to 31 December 2012 are utilized together with the daily observations of the three indices to probe the relationship of the full‐disk magnetic activity respectively with the indices. Cross correlation analyses of the daily magnetic field measurements at Mount Wilson observatory are taken with the daily observations of the three indices, and the statistical significance of the difference of the obtained correlation coefficients is investigated. The following results are obtained: (1) The sunspot number should be preferred to represent/reflect the full‐disk magnetic activity of the Sun to which the weak magnetic fields (outside of sunspots) mainly contribute, the sunspot area should be recommended to represent the strong magnetic activity of the Sun (in sunspots), and the 10.7 cm radio flux should be preferred to represent the full‐disk magnetic activity of the Sun (both the weak and strong magnetic fields) to which the weak magnetic fields mainly contribute. (2) On the other hand, the most recommendable index that could be used to represent/reflect the weak magnetic activity is the 10.7 cm radio flux, the most recommendable index that could be used to represent the strong magnetic activity is the sunspot area, and the most recommendable index that could be used to represent the full‐disk magnetic activity of the Sun is the 10.7cm radio flux. Additionally, the cycle characteristics of the magnetic field strengths on the solar disk are given. (© 2014 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Solar Rotation During the Period 1847?–?1849

Solar rotation rate has been measured using the sunspot positions recorded by W.C. Bond during the period 1847 – 1849 at the Harvard College Observatory. From the drawings carried out by Bond we have selected the sunspots and groups of sunspots with more reliable positions presented in three or more drawings on successive days. We have calculated from the positions of the selected sunspots (41 in total) a synodic rotation rate of ω=[(12.92±0.08)−(1.5±1.0)sin ² φ] degrees/day, where φ is the heliographic latitude. This rate, although slightly lower, is similar to the actual solar rotation rate, confirming no important changes in the solar rotation during the last 160 years.     

Periodicity of Total Solar Irradiance

We investigate the periodicity in the PMOD composite of the daily total solar irradiance (TSI) from 21 September 1978 to 9 June 2009. Besides the Schwabe cycle period (10.32 years), the quasi-rotation period is found to be statistically significant in TSI, whose value is about 32 days, longer than that in sunspot activity (27 days), and it intermittently appears around the sunspot maximum times. The quasi-rotation period in TSI is inferred to be mainly caused by sunspot activity, but to be modulated by bright features as well. It was previously found that variations of TSI over a Schwabe solar cycle mainly come from the combination of the sunspots’ blocking and the intensification due to bright faculae, plages, and network elements, with a slight dominance of the bright-feature effect during the maximum of the Schwabe cycle. For the sunspot-blocking and the bright-feature effect to contribute to TSI over a Schwabe solar cycle, the former is inferred to lead the latter by 29 days at least.     

Automatic Sunspots Detection on Full-Disk Solar Images using Mathematical Morphology

Sunspots are solar features located in active regions of the Sun, whose number is an indicator of the Sun’s magnetic activity. Therefore accurate detection and classification of sunspots are fundamental for the elaboration of solar activity indices such as the Wolf number. However, irregularities in the shape of the sunspots and their variable intensity and contrast with the surroundings, make their automated detection from digital images difficult. Here, we present a morphological tool that has allowed us to construct a simple and automatic procedure to treat digital photographs obtained from a solar telescope, and to extract the main features of sunspots. Comparing the solar indices computed with our algorithm against those obtained with the previous method exhibit an obvious improvement. A favorable comparison of the Wolf sunspot number time series obtained with our methodology and from other reference observatories is also presented. Finally, we compare our sunspot and group detection to that of other observatories.     

The Effect of Sunspot Weighting

Although W. Brunner began to weight sunspot counts (from 1926), using a method whereby larger spots were counted more than once, he compensated for the weighting by not counting enough smaller spots in order to maintain the same reduction factor (0.6) as was used by his predecessor A. Wolfer to reduce the count to R. Wolf’s original scale, so that the weighting did not have any effect on the scale of the sunspot number. In 1947, M. Waldmeier formalized the weighting (on a scale from 1 to 5) of the sunspot count made at Zurich and its auxiliary station Locarno. This explicit counting method, when followed, inflates the relative sunspot number over that which corresponds to the scale set by Wolfer (and matched by Brunner). Recounting some 60,000 sunspots on drawings from the reference station Locarno shows that the number of sunspots reported was “over counted” by \({\approx}\,44~\%\) on average, leading to an inflation (measured by an effective weight factor) in excess of 1.2 for high solar activity. In a double-blind parallel counting by the Locarno observer M. Cagnotti, we determined that Svalgaard’s count closely matches that of Cagnotti, allowing us to determine from direct observation the daily weight factor for spots since 2003 (and sporadically before). The effective total inflation turns out to have two sources: a major one (15?–?18 %) caused by weighting of spots, and a minor source (4?–?5 %) caused by the introduction of the Zürich classification of sunspot groups which increases the group count by 7?–?8 % and the relative sunspot number by about half that. We find that a simple empirical equation (depending on the activity level) fits the observed factors well, and use that fit to estimate the weighting inflation factor for each month back to the introduction of effective inflation in 1947 and thus to be able to correct for the over-counts and to reduce sunspot counting to the Wolfer method in use from 1894 onwards.     

Sunspot Plumes and Flow Channels

It is well known that sunspots are dark. This statement is not correct in the sunspot atmosphere between the chromosphere and the corona, where sunspots often are brighter than their surroundings. The brightest feature in the sunspot transition region is called a sunspot plume. Not all sunspots contain a plume. We find that 20 out of 21 sunspots show a plume when one magnetic polarity dominates the sunspot region out to a distance of 50 from the sunspot. Most sunspots show downflows that exceed 25 km s^–1 in the sunspot plumes at temperatures close to 250000 K. This downflow is not maintained by inflow from the corona, but by gas at transition region temperatures, streaming in flow channels from locations well outside the sunspot. We suggest that this inflow is a necessary requirement for the sunspot plume to occur and present a working hypothesis for the origin of sunspot plumes. This paper is the first thorough spectral analysis of sunspot plumes. It is based on simultaneous observations of ten or six EUV emission lines in 42 sunspot regions with the Coronal Diagnostic Spectrometer – CDS on the Solar and Heliospheric Observatory – SOHO. The line profiles are studied in detail with another SOHO instrument, the Solar Ultraviolet Measurements of Emitted Radiation – SUMER.     

Solar Rotation in the 17th century

Sunspot drawings made by Galileo Galilei in 1612 are used to derive the law of differential rotation at that time. The main interest of the work is during the time of observations, just at the beginning of telescopic observations and some decades before the Maunder Minimum (1645 – 1715), a period where the sunspots almost disappeared from the solar surface. For this purpose we have carried out careful corrections of the different sources of errors derived from the observing technique. By comparing with other results of the same century, a significant difference is only detected by comparing with data corresponding to the deep Maunder Minimum (Paris Observatory drawings). The characteristics of the solar differential rotation, and extrapolating the behavior of solar activity, did not differ before or after the Maunder Minimum. We also include an analysis of hitherto ignored sunspot drawings by N. Bion made in October and November 1672.     

太阳黑子面积数与黑子数指标在大气预报中的应用对比分析

分别应用太阳黑子视面积数和太阳黑子相对数代表太阳活动水平与天津夏季降水总量进行相关分析,结果表明黑子面积指标明显优于黑子数.