Astronomy & Astrophysics (A&A) is an international journal which publishes papers on all aspects of astronomy and astrophysics

The application of machine learning in solar physics has the potential to greatly enhance our understanding of the complex processes that take place in the atmosphere of the Sun. By using techniques such as deep learning, we are now in the position to analyze large amounts of data from solar observations and identify patterns and trends that may not have been apparent using traditional methods. This can help us improve our understanding of explosive events like solar flares, which can have a strong effect on the Earth environment. Predicting hazardous events on Earth becomes crucial for our technological society. Machine learning can also improve our understanding of the inner workings of the sun itself by allowing us to go deeper into the data and to propose more complex models to explain them. Additionally, the use of machine learning can help to automate the analysis of solar data, reducing the need for manual labor and increasing the efficiency of research in this field.

Context. Studies of long-term solar activity and variability require knowledge of the past evolution of the solar surface magnetism. The archives of full-disc Ca II K observations that have been performed more or less regularly at various sites since 1892 can serve as an important source of such information. Aims. We derive the plage area evolution over the last 12 solar cycles by employing data from all Ca II K archives that are publicly available in digital form, including several as-yet-unexplored Ca II K archives. Methods. We analysed more than 290 000 full-disc Ca II K observations from 43 datasets spanning the period between 1892–2019. All images were consistently processed with an automatic procedure that performs the photometric calibration (if needed) and the limb-darkening compensation. The processing also accounts for artefacts affecting many of the images, including some very specific artefacts, such as bright arcs found in Kyoto and Yerkes data. Our employed methods have previously been tested and evaluated on synthetic data and found to be more accurate than other methods used in the literature to treat a subset of the data analysed here. Results. We produced a plage area time-series from each analysed dataset. We found that the differences between the plage areas derived from individual archives are mainly due to the differences in the central wavelength and the bandpass used to acquire the data at the various sites. We empirically cross-calibrated and combined the results obtained from each dataset to produce a composite series of plage areas. The ’backbone’ approach was used to bridge the series together. We have also shown that the selection of the backbone series has little effect on the final composite of the plage area. We quantified the uncertainty of determining the plage areas with our processing due to shifts in the central wavelength and found it to be less than 0.01 in fraction of the solar disc for the average conditions found on historical data. We also found the variable seeing conditions during the observations to slightly increase the plage areas during the activity maxima. Conclusions. We provide the most complete so far time series of plage areas based on corrected and calibrated historical and modern Ca II K images. Consistent plage areas are now available on 88% of all days from 1892 onwards and on 98% from 1907 onwards.

Future ground-based telescopes, such as the 4-metre class facilities DKIST and EST, will dramatically improve on current capabilities for simultaneous multi-line polarimetric observations in a wide range of wavelength bands, from the near-ultraviolet to the near-infrared. As a result, there will be an increasing demand for fast diagnostic tools, i.e., inversion codes, that can infer the physical properties of the solar atmosphere from the vast amount of data these observatories will produce. The advent of substantially larger apertures, with the concomitant increase in polarimetric sensitivity, will drive an increased interest in observing chromospheric spectral lines. Accordingly, pertinent inversion codes will need to take account of line formation under general non-local thermodynamic equilibrium (NLTE) conditions. Several currently available codes can already accomplish this, but they have a common practical limitation that impairs the speed at which they can invert polarised spectra, namely that they employ numerical evaluation of the so-called response functions to changes in the atmospheric parameters, which makes them less suitable for the analysis of very large data volumes. Here we present DeSIRe (Departure coefficient aided Stokes Inversion based on Response functions), an inversion code that integrates the well-known inversion code SIR with the NLTE radiative transfer solver RH. The DeSIRe runtime benefits from employing analytical response functions computed in local thermodynamic equilibrium (through SIR), modified with fixed departure coefficients to incorporate NLTE effects in chromospheric spectral lines. This publication describes the operating fundamentals of DeSIRe and describes its behaviour, robustness, stability, and speed. The code is ready to be used by the solar community and is being made publicly available.

We present observations of transverse oscillations in slender Ca ii H fibrils (SCFs) in the lower solar chromosphere. We use a 1 hr long time series of high- (spatial and temporal-) resolution seeing-free observations in a 1.1 Å wide passband covering the line core of Ca ii H 3969 Å from the second flight of the Sunrise balloon-borne solar observatory. The entire field of view, spanning the polarity inversion line of an active region close to the solar disk center, is covered with bright, thin, and very dynamic fine structures. Our analysis reveals the prevalence of transverse waves in SCFs with median amplitudes and periods on the order of 2.4 ± 0.8 km s⁻¹ and 83 ± 29 s, respectively (with standard deviations given as uncertainties). We find that the transverse waves often propagate along (parts of) the SCFs with median phase speeds of 9 ± 14 km s⁻¹. While the propagation is only in one direction along the axis in some of the SCFs, propagating waves in both directions, as well as standing waves are also observed. The transverse oscillations are likely Alfvénic and are thought to be representative of magnetohydrodynamic kink waves. The wave propagation suggests that the rapid high-frequency transverse waves, often produced in the lower photosphere, can penetrate into the chromosphere with an estimated energy flux of ≈15 kW m⁻². Characteristics of these waves differ from those reported for other fibrillar structures, which, however, were observed mainly in the upper solar chromosphere.

Context. At present, an exponential growth in scientific data from current and upcoming solar observatories is expected. Most of the data consist of high spatial and temporal resolution cubes of Stokes profiles taken in both local thermodynamic equilibrium (LTE) and non-LTE spectral lines. The analysis of such solar observations requires complex inversion codes. Hence, it is necessary to develop new tools to boost the speed and efficiency of inversions and reduce computation times and costs. Aims. In this work we discuss the application of convolutional neural networks (CNNs) as a tool to advantageously initialize Stokes profile inversions. Methods. To demonstrate the usefulness of CNNs, we concentrate in this paper on the inversion of LTE Stokes profiles. We use observations taken with the spectropolarimeter on board the Hinode spacecraft as a test bench mark. First, we carefully analyse the data with the SIR inversion code using a given initial atmospheric model. The code provides a set of atmospheric models that reproduce the observations well. These models are then used to train a CNN. Afterwards, the same data are again inverted with SIR but using the trained CNN to provide the initial guess atmospheric models for SIR. Results. The CNNs allow us to significantly reduce the number of inversion cycles when used to compute initial guess model atmospheres (‘assisted inversions’), therefore decreasing the computational time for LTE inversions by a factor of two to four. CNNs alone are much faster than assisted inversions, but the latter are more robust and accurate. CNNs also help to automatically cluster pixels with similar physical properties, allowing the association with different solar features on the solar surface, which is useful when inverting huge datasets where completely different regimes are present. The advantages and limitations of machine learning techniques for estimating optimum initial atmospheric models for spectral line inversions are discussed. Finally, we describe a python wrapper for the SIR and DeSIRe codes that allows for the easy setup of parallel inversions. The tool implements the assisted inversion method described in this paper. The parallel wrapper can also be used to synthesize Stokes profiles with the RH code. Conclusions. The assisted inversions can speed up the inversion process, but the efficiency and accuracy of the inversion results depend strongly on the solar scene and the data used for the CNN training. This method (assisted inversions) will not obviate the need for analysing individual events with the utmost care but will provide solar scientists with a much better opportunity to sample large amounts of inverted data, which will undoubtedly broaden the physical discovery space.