Context. Large aperture ground-based solar telescopes allow the solar atmosphere to be resolved in unprecedented detail. However, ground-based observations are inherently limited due to Earth's turbulent atmosphere, requiring image correction techniques. Aims. Recent post-image reconstruction techniques are based on using information from bursts of short-exposure images. Shortcomings of such approaches are the limited success, in case of stronger atmospheric seeing conditions, and computational demand. Real-time post-image reconstruction is of high importance to enabling automatic processing pipelines and accelerating scientific research. In an attempt to overcome these limitations, we provide a deep learning approach to reconstruct an original image burst into a single high-resolution high-quality image in real time. Methods. We present a novel deep learning tool for image burst reconstruction based on image stacking methods. Here, an image burst of 100 short-exposure observations is reconstructed to obtain a single high-resolution image. Our approach builds on unpaired image-to-image translation. We trained our neural network with seeing degraded image bursts and used speckle reconstructed observations as a reference. With the unpaired image translation, we aim to achieve a better generalization and increased robustness in case of increased image degradations. Results. We demonstrate that our deep learning model has the ability to effectively reconstruct an image burst in real time with an average of 0.5 s of processing time while providing similar results to standard reconstruction methods. We evaluated the results on an independent test set consisting of high- and low-quality speckle reconstructions. Our method shows an improved robustness in terms of perceptual quality, especially when speckle reconstruction methods show artifacts. An evaluation with a varying number of images per burst demonstrates that our method makes efficient use of the combined image information and achieves the best reconstructions when provided with the full-image burst.

Physics-informed neural networks (PINNs) provide a novel approach for data-driven numerical simulations, tackling challenges of discretization and enabling seamless integration of noisy data and physical models (e.g., partial differential equations). In this presentation, we discuss the results of our recent studies where we apply PINNs for coronal magnetic field extrapolations of the solar atmosphere, which are essential to understand the genesis and initiation of solar eruptions and to predict the occurrence of high-energy events from our Sun.We utilize our PINN to estimate the 3D coronal magnetic fields based on photospheric vector magnetograms and the force-free physical model. This approach provides state-of-the-art coronal magnetic field extrapolations in quasi real-time. We simulate the evolution of Active Region NOAA 11158 over 5 continuous days, where the derived time profile of the free magnetic energy unambiguously relates to the observed flare activity.We extend this approach by utilizing multi-height magnetic field measurements and combine them in a single magnetic field model. Our evaluation shows that the additional chromospheric field information leads to a more realistic approximation of the solar coronal magnetic field. In addition, our method intrinsically provides an estimate of the height corrugation of the observed magnetograms.We provide an outlook on our ongoing work where we use PINNs for global force-free magnetic field extrapolations. This approach enables a novel understanding of the global magnetic topology with a realistic treatment of current carrying fields.In summary, PINNs have the potential to greatly advance the field of numerical simulations, accelerate scientific research, and enable advanced space weather monitoring.

Various instruments are used to study the Sun, including ground-based observatories and space telescopes. These data products are constantly changing due to technological improvements, different instrumentation, or atmospheric effects. However, for certain applications such as ground-based solar image reconstruction or solar cycle studies, enhanced and combined data products are necessary.We present a general AI tool called Instrument-to-Instrument (ITI; Jarolim et al. 2023) translation, which is capable of translating datasets between two different image domains. This approach enables instrument intercalibration, image enhancement, mitigation of quality degradations, and super-resolution across multiple wavelength bands. The tool is built on unpaired image-to-image translation, which enables a wide range of applications, where no spatial or temporal overlap is required between the considered datasets.In this presentation, we highlight ITI as a general tool for Heliospheric applications and demonstrate its capabilities by applying it to data from Solar Orbiter/EUI, PROBA2/SWAP, and the Solar Dynamics Observatory/AIA in order to achieve a homogenous, machine-learning ready dataset that combines three different EUV imagers. The direct comparison of aligned observations shows the close relation of ITI-enhanced and real high-quality observations. The evaluation of light-curves demonstrates an improved inter-calibration.ITI is provided open-source to the community and can be easily applied to novel datasets and various research applications. This research is funded through a NASA 22-MDRAIT22-0018 award (No 80NSSC23K1045) and managed by Trillium Technologies, Inc (trillium.tech)

Multiple satellites capture images of the Sun in Extreme Ultraviolet (EUV) light. However, only the Solar Dynamics Observatory (SDO) was equipped with instruments that measure the Sun's EUV spectral irradiance (i.e., MEGS-A and MEGS-B onboard the Extreme Ultraviolet Variability Experiment (EVE) suite). The MEGS-A instrument malfunctioned in 2014, making it impossible to measure the full irradiance spectrum ever since. Using AI, we explore the translation of a set of EUV images of the Sun into spectral irradiance, effectively replacing the malfunctioning MEGS-A instrument onboard SDO. In other words, we generate a virtual irradiance instrument, MEGS-AI, for SDO. Using an Image-to-Image translation tool (ITI), this virtual instrument can also be trained and added on other EUV-observing satellites like STEREO, GOES, SolO, and the upcoming VIGIL satellite, enabling unprecedented irradiance estimates from additional satellite missions. In the case of the STEREO twin-satellites and VIGIL, this enables estimates of spectral irradiance prior to the Sun rotating into Earth’s view, which directly enables the forecast of enhanced irradiance. Additionally, we explore different combinations of images in different EUV channels and evaluate their contributions in estimating different irradiance channels. Finally, when combined with a neural radiance field model of the Sun (SuNeRFs), MEGS-AI can estimate spectral irradiance from any viewpoint in the solar system, enabling for the first time a complete 4pi spectral irradiance map of the Sun. This can be directly used to estimate the Sun’s impact on other planets in the solar system and to determine the total solar irradiance output in multiple EUV spectral bands.

Spectropolarimetric inversions of solar observations are fundamental for the estimation of the magnetic field in the solar atmosphere. However, instrumental noise, computational requirements, and varying levels of physical realism make it challenging to derive reliable solar magnetic field estimates. In this study, we present a novel approach for spectropolarimetric inversions based on physics-informed neural networks to infer the photospheric magnetic field under the Milne–Eddington approximation (PINN ME). Our model acts as a representation of the parameter space, mapping input coordinates ( t , x , y ) to the respective spectropolarimetric parameters, which are used to synthesize the corresponding Stokes profiles. By iteratively sampling coordinate points, synthesizing profiles, and minimizing the deviation from the observed stokes profiles, our method can find the set of Milne–Eddington parameters that best fit the observations. In addition, we directly include the point-spread function to account for instrumental effects. We use a predefined parameter space as well as synthetic profiles from a radiative MHD simulation to evaluate the performance of our method and to estimate the impact of instrumental noise. Our results demonstrate that PINN ME achieves an intrinsic spatiotemporal coupling, which can largely mitigate observational noise and provides a memory-efficient inversion even for extended fields of view. Finally, we apply our method to observations and show that our method provides a high spatial coherence and can resolve small-scale features in both strong- and weak-field regions.

To understand its evolution and the effects of its eruptive events, the Sun is permanently monitored by multiple satellite missions. The optically thin emission of the solar plasma and the limited number of viewpoints make it challenging to reconstruct the geometry and structure of the solar atmosphere; however, this information is the missing link to understand the Sun as it is: a 3D evolving star. We present a method that enables a complete 3D representation of the uppermost solar layer (corona) observed in extreme ultraviolet (EUV) light. We use a deep-learning approach for 3D scene representation that accounts for radiative transfer to map the entire solar atmosphere from three simultaneous observations. We demonstrate that our approach provides unprecedented reconstructions of the solar poles and directly enables height estimates of coronal structures, solar filaments, coronal hole profiles, and coronal mass ejections. We validate the approach using model-generated synthetic EUV images, finding that our method accurately captures the 3D geometry of the Sun even from a limited number of 32 ecliptic viewpoints ( divide latitude divide <= 7 degrees). We quantify the uncertainties of our model using an ensemble approach that allows us to estimate the model performance in the absence of a ground truth. Our method enables a novel view of our closest star and is a breakthrough technology for the efficient use of multi-instrument data sets, which paves the way for future cluster missions.

Non-linear force-free extrapolations are a common approach to estimate the 3D topology of coronal magnetic fields based on photospheric vector magnetograms. The force-free assumption is a valid approximation at coronal heights, but for the dense plasma conditions in the lower atmosphere, this assumption is not satisfied. In this study, we utilize multi-height magnetic field measurements in combination with physics-informed neural networks to advance solar magnetic field extrapolations. We include a flexible height-mapping, which allows us to account for the different formation heights of the observed magnetic field measurements. The comparison to analytical and simulated magnetic fields demonstrates that including chromospheric magnetic field measurements leads to a significant improvement of our magnetic field extrapolations. We also apply our method to chromospheric line-of-sight magnetograms, from the Vector Spectromagnetograph (VSM) on the Synoptic Optical Long-term Investigations of the Sun (SOLIS) observatory, in combination with photospheric vector magnetograms, from the Helioseismic Magnetic Imager (HMI) onboard the Solar Dynamic Observatory (SDO). The comparison to observations in extreme ultraviolet wavelengths shows that the additional chromospheric information leads to a better agreement with the observed coronal structures. In addition, our method intrinsically provides an estimate of the corrugation of the observed magnetograms. With this new approach, we make efficient use of multi-height magnetic field measurements and advance the realism of coronal magnetic field simulations.

On 2024 May 10-11, the strongest geomagnetic storm since 2003 November occurred, with a peak Dst index of -412 nT. The storm was caused by NOAA active region (AR) 13664, which was the source of a large number of coronal mass ejections and flares, including 12 X-class flares. Starting from about May 7, AR 13664 showed a steep increase in its size and (free) magnetic energy, along with increased flare activity. In this study, we perform 3D magnetic field extrapolations with the NF2 nonlinear force-free code based on physics-informed neural networks (R. Jarolim et al.). In addition, we introduce the computation of the vector potential to achieve divergence-free solutions. We extrapolate vector magnetograms from the Solar Dynamics Observatory's Helioseismic and Magnetic Imager at the full 12 minute cadence from 2024 May 5 00:00 to 11 04:36 UT, in order to understand the AR's magnetic evolution and the large eruptions it produced. A decrease in the calculated relative free magnetic energy can be related to solar flares in similar to 90% of the cases, and all considered X-class flares are reflected by a decrease in the relative free magnetic energy. Regions of enhanced free magnetic energy and depleted magnetic energy between the start and end times of major X-class flares show spatial alignment with brightness increases in extreme-ultraviolet observations. We provide a detailed analysis of the X3.9-class flare on May 10, where we show that the interaction between separated magnetic domains is directly linked to major flaring events. With this study, we provide a comprehensive data set of the magnetic evolution of AR 13664 and make it publicly available for further analysis.

The three-dimensional (3D) coronal magnetic field has not yet been directly observed. However, for a better understanding and prediction of magnetically driven solar eruptions, 3D models of solar active regions are required. This work aims to provide insight into the significance of different extrapolation models for analyzing the preeruptive conditions of active regions with morphological parameters in 3D. Here, we employed potential field (PF), linear force-free field (LFFF), and nonlinear force-free field (NLFFF) models and a neural network-based method integrating observational data and NLFFF physics (NF2). The 3D coronal magnetic field structure of a "flaring" (AR11166) and "flare-quiet" (AR12645) active region, in terms of their flare productivity, is constructed via the four extrapolation methods. To analyze the evolution of the field, six prediction parameters were employed throughout, from the photosphere up to the base of the lower corona. First, we find that the evolution of the adopted morphological parameters exhibits similarity across the investigated time period when considering the four types of extrapolations. Second, all the parameters exhibited preeruptive conditions not only at the photosphere but also at higher altitudes in the case of active region (AR) 11166, while three out of the six proxies also exhibited preeruptive conditions in the case of AR12645. We conclude that: (i) the combined application of several different precursor parameters is important in the lower solar atmosphere to improve eruption predictions, and (ii) to gain a quick yet reliable insight into the preflare evolution of active regions in 3D, the PF and LFFF are acceptable; however, the NF2 method is likely the more suitable option.

Filaments are omnipresent features in the solar atmosphere. Their location, properties and time evolution can provide information about changes in solar activity and assist the operational space weather forecast. Therefore, filaments have to be identified in full disk images and their properties extracted from these images. Manual extraction is tedious and takes much time; extraction with morphological image processing tools produces a large number of false-positive detections. Automatic object detection, segmentation, and extraction in a reliable manner allows to process more data in a shorter time. The Chromospheric Telescope (ChroTel), Tenerife, Spain, the Global Oscillation Network Group (GONG), and the Kanzelh\"ohe Observatory (KSO), Austria, provide regular full-disk observations of the Sun in the core of the chromospheric H-alpha absorption line. We present a deep learning method that provides reliable extractions of filaments from H-alpha filtergrams. First, we train the object detection algorithm YOLOv5 with labeled filament data of ChroTel. We use the trained model to obtain bounding-boxes from the full GONG archive. In a second step, we apply a semi-supervised training approach, where we use the bounding boxes of filaments, to learn a pixel-wise classification of filaments with u-net. Here, we make use of the increased data set size to avoid overfitting of spurious artifacts from the generated training masks. Filaments are predicted with an accuracy of 92%. With the resulting filament segmentations, physical parameters such as the area or tilt angle can be easily determined and studied. This we demonstrate in one example, where we determine the rush-to-the pole for Solar Cycle 24 from the segmented GONG images. In a last step, we apply the filament detection to H-alpha observations from KSO which demonstrates the general applicability of our method to H-alpha filtergrams.

Context. The PLAnetary Transits and Oscillations of stars (PLATO) mission will observe the same area of the sky continuously for at least two years in an effort to detect transit signals of an Earth-like planet orbiting a solar-like star. Aims. We aim to study how short-term solar-like variability caused by oscillations and granulation would affect PLATO's ability to detect and size Earth if PLATO were to observe the Solar System itself. We also compare different approaches to mitigate noise caused by short-term solar-like variability and perform realistic transit fitting of transit signals in PLATO-like light curves. Methods. We injected Earth-like transit signals onto real solar data taken by the Helioseismic and Magnetic Imager (HMI) instrument on board the Solar Dynamics Observatory (SDO). We isolated short-term stellar variability in the HMI observations by removing any variability with characteristic timescales longer than five hours using a smooth Savitzky-Golay filter. We then added a noise model for a variety of different stellar magnitudes computed by PlatoSim assuming an observation by all 24 normal cameras. We first compared four different commonly used treatments of correlated noise in the time domain by employing them in a transit fitting scheme. We then tried to recover pairs of transit signals using an algorithm similar to the transit least squares algorithm. Finally, we performed transit fits using realistic priors on planetary and stellar parameters and assessed how accurately the pair of two injected transits was recovered. Results. We find that short-term solar-like variability affects the correct retrieval of Earth-like transit signals in PLATO data. Variability models accounting for variations with typical timescales at the order of one hour are sufficient to mitigate these effects. We find that when the limb-darkening coefficients of the host star are properly constrained, the impact parameter does not negatively affect the detectability of a transit signal or the retrieved transit parameters, except for high values (b > 0.8). For bright targets (8.5-10.5 mag), the transit signal of an Earth analogue can reliably be detected in PLATO data. For faint targets a detection is still likely, though the results of transit search algorithms have to be verified by transit-fitting algorithms to avoid false positive detections being flagged. For bright targets (V-mag <= 9.5), the radius of an Earth-like planet orbiting a solar-like star can be correctly determined at a precision of 3% or less, assuming that at least two transit events are observed and the characteristics of the host star are well understood.

Aims. This study focuses on the rapid evolution of the solar filament in active region 12975 during a confined C2 flare on 28 March 2022, which finally led to an eruptive M4 flare 1.5 h later. The event is characterized by the apparent breakup of the filament, the disappearance of its southern half, and the flow of the remaining filament plasma into a new, longer channel with a topology very similar to an extreme ultraviolet (EUV) hot channel observed during the flare. In addition, we outline the emergence of the original filament from a sheared arcade and discuss possible drivers for its rise and eruption. Methods. We took advantage of Solar Orbiter's favorable position, 0.33 AU from the Sun, and 83.5 degrees west of the Sun-Earth line, to perform a multi-point study using the Spectrometer Telescope for Imaging X-rays (STIX) and the Extreme Ultraviolet Imager (EUI) in combination with the Atmospheric Imaging Assembly (AIA) and the Helioseismic and Magnetic Imager (HMI) onboard the Solar Dynamics Observatory (SDO) and H alpha images from the Earth-based Kanzelh & ouml;he Observatory for Solar and Environmental Research (KSO) and the Global Oscillation Network Group (GONG). While STIX and EUI observed the flare and the filament's rise from close up and at the limb, AIA and HMI observations provided highly complementary on-disk observations from which we derived differential emission measure (DEM) maps and nonlinear force-free (NLFF) magnetic field extrapolations. Results. According to our pre-flare NLFF extrapolation, field lines corresponding to both filament channels existed in close proximity before the flare. We propose a loop-loop reconnection scenario based on field structures associated with the AIA 1600 & Aring; flare ribbons and kernels. It involves field lines surrounding and passing beneath the shorter filament channel, and field lines closely following the southern part of the longer channel. Reconnection occurs in an essentially vertical current sheet at a polarity inversion line (PIL) below the breakup region, which enables the formation of the flare loop arcade and EUV hot channel. This scenario is supported by concentrated currents and free magnetic energy built up by antiparallel flows along the PIL before the flare, and by non-thermal X-ray emission observed from the reconnection region. The reconnection probably propagated to involve the original filament itself, leading to its breakup and new geometry. This reconnection geometry also provides a general mechanism for the formation of the long filament channel and realizes the concept of tether cutting. It was probably active throughout the filament's continuous rise phase, which lasted from at least 30 min before the C2 flare until the filament eruption. The C2 flare represents a period of fast reconnection during this otherwise more steady period, during which most of the original filament was reconnected and joined the longer channel. Conclusions. These results demonstrate how rapid changes in solar filament topology can be driven by loop-loop reconnection with nearby field structures, and how this can be part of a long-lasting tether-cutting reconnection process. They also illustrate how a confined precursor flare due to loop-loop reconnection (Type I) can contribute to the evolution towards a full eruption, and that they can produce a flare loop arcade when the contact region between interacting loop systems has a sheet-like geometry similar to a flare current sheet.

The cause of excess spectral line broadening (non-thermal velocity) is not definitively known, but given its rise before and during flaring, the causal processes hold clues to understanding the triggers for the onset of reconnection and the release of free magnetic energy from the coronal magnetic field. A comparison of data during a 9-hour period from the extreme ultraviolet (EUV) Imaging Spectrometer (EIS) on the Hinode spacecraft - at a 3-minute cadence - and non-linear force-free field (NLFFF) extrapolations performed on Helioseismic and Magnetic Imager (HMI) magnetograms - at a 12-minute cadence - shows an inverse relationship between non-thermal velocity and free magnetic energy on short timescales during two X-class solar flares on 6 September 2017. Analysis of these results supports suggestions that unresolved Doppler flows do not solely cause non-thermal broadening and instead other mechanisms like Alfvén wave propagation and isotropic turbulence have a greater influence.

Automated detection schemes are nowadays the standard approach for locating coronal holes in extreme-UV images from the Solar Dynamics Observatory (SDO). However, factors such as the noisy nature of solar imagery, instrumental effects, and others make it challenging to identify coronal holes using these automated schemes. While discrepancies between detection schemes have been noted in the literature, a comprehensive assessment of these discrepancies is still lacking. The contribution of the Coronal Hole Boundary Working Team in the COSPAR ISWAT initiative to close this gap is threefold. First, we present the first community data set for comparing automated coronal hole detection schemes. This data set consists of 29 SDO images, all of which were selected by experienced observers to challenge automated schemes. Second, we use this community data set as input to 14 widely applied automated schemes to study coronal holes and collect their detection results. Third, we study three SDO images from the data set that exemplify the most important lessons learned from this effort. Our findings show that the choice of the automated detection scheme can have a significant effect on the physical properties of coronal holes, and we discuss the implications of these findings for open questions in solar and heliospheric physics. We envision that this community data set will serve the scientific community as a benchmark data set for future developments in the field.

To understand the solar evolution and effects of solar eruptive events, the Sun is permanently observed by multiple satellite missions. The optically-thin emission of the solar plasma and the limited number of viewpoints make it challenging to reconstruct the geometry and structure of the solar atmosphere; however, this information is the missing link to understand the Sun as it is: a three-dimensional, evolving star. We present a method that enables a complete 3D representation of the uppermost solar layer observed in extreme ultraviolet (EUV) light. We use a deep learning approach for 3D scene representation that accounts for radiative transfer, to map the entire solar atmosphere from three simultaneous observations. We demonstrate that our approach provides unprecedented reconstructions of the solar poles, and directly enables height estimates of coronal structures, solar flux ropes, coronal hole profiles, and coronal mass ejections. We validate the approach using model-generated synthetic EUV images, finding that our method accurately captures the 3D geometry even from a limited number of viewpoints. We quantify uncertainties of our model using an ensemble approach that allows us to estimate the model performance in absence of a ground-truth. Our method enables a novel view of our closest star, and is a breakthrough technology for the efficient use of multi-instrument datasets, which paves the way for future cluster missions.

The ISWAT clusters H1+H2 have a focus on interplanetary space and its characteristics, especially on the large-scale co-rotating and transient structures impacting Earth. SIRs, generated by the interaction between high-speed solar wind originating in large-scale open coronal magnetic fields and slower solar wind from closed magnetic fields, are regions of compressed plasma and magnetic field followed by high-speed streams that recur at the ca. 27 day solar rotation period. Short-term reconfigurations of the lower coronal magnetic field generate flare emissions and provide the energy to accelerate enormous amounts of magnetised plasma and particles in the form of CMEs into interplanetary space. The dynamic interplay between these phenomena changes the configuration of interplanetary space on various temporal and spatial scales which in turn influences the propagation of individual structures. While considerable efforts have been made to model the solar wind, we outline the limitations arising from the rather large uncertainties in parameters inferred from observations that make reliable predictions of the structures impacting Earth difficult. Moreover, the increased complexity of interplanetary space as solar activity rises in cycle 25 is likely to pose a challenge to these models. Combining observational and modeling expertise will extend our knowledge of the relationship between these different phenomena and the underlying physical processes, leading to improved models and scientific understanding and more-reliable space-weather forecasting. The current paper summarizes the efforts and progress achieved in recent years, identifies open questions, and gives an outlook for the next 5-10 years. It acts as basis for updating the existing COSPAR roadmap by Schrijver+ (2015), as well as providing a useful and practical guide for peer-users and the next generation of space weather scientists.

While the photospheric magnetic field of our Sun is routinely measured, its extent into the upper atmosphere is typically not accessible by direct observations. Here we present an approach for coronal magnetic-field extrapolation, using a neural network that integrates observational data and the physical force-free magnetic-field model. Our method flexibly finds a trade-off between the observation and force-free magnetic-field assumption, improving the understanding of the connection between the observation and the underlying physics. We utilize meta-learning concepts to simulate the evolution of active region NOAA 11158. Our simulation of 5 days of observations at full cadence (12 minutes) requires less than 12 hours of total computation time, allowing for real-time force-free magnetic-field extrapolations. The application to an analytical magnetic-field solution, a systematic analysis of the time evolution of free magnetic energy and magnetic helicity in the coronal volume, as well as comparison with extreme-ultraviolet observations, demonstrates the validity of our approach. The obtained temporal and spatial depletion of free magnetic energy unambiguously relates to the observed flare activity. The application of physics-informed neural networks enables an estimation of the solar coronal magnetic field in quasi real time. A comparison with extreme-ultraviolet observations reveals that the model provides a realistic approximation and the modelled coronal field has a clear relationship with flaring activity.

Context. The Spectrometer Telescope for Imaging X-rays (STIX) on board Solar Orbiter enables exciting multipoint studies of energy release and transport in solar flares by observing the Sun from many different distances and vantage points out of the Sun-Earth line. Aims. We present a case study of an M4-class flare that occurred on 28 March 2022, near Solar Orbiter’s first science perihelion (0.33 AU from the Sun). Solar Orbiter had a longitudinal separation of 83.5° west of the Sun-Earth line, making the event appear near the eastern limb from its perspective, while Earth-orbiting spacecraft observed it near the disk center. We follow the evolution of the X-ray, extreme-ultraviolet (EUV), UV sources, analyze their relation to plasma dynamics and heating, and relate our observations to magnetic field structures, including the erupting filament. Methods. The timing and location of the STIX X-ray sources were related to the plasma evolution observed in the EUV by the Extreme Ultraviolet Imager (EUI) on Solar Orbiter and the Atmospheric Imaging Assembly (AIA) on the Solar Dynamics Observatory, and to the chromospheric response observed in 1600 Å by AIA. We performed differential emission measure (DEM) analysis to further characterize the flaring plasma at different subvolumes. The pre-flare magnetic field configuration was analyzed using a nonlinear force-free (NLFF) extrapolation. Results. In addition to the two classical hard X-ray (HXR) footpoints at the ends of the flaring loops, later in the event we observe a nonthermal HXR source at one of the anchor points of the erupting filament. These results are supported by a robustness analysis of the STIX images and the co-temporal chromospheric brightenings observed by AIA. The full evolution of the AIA 1600 Å footpoints indicates that this change in footpoint location represents a discontinuity in an otherwise continuous westward motion of the footpoints throughout the flare. The NLFF extrapolation suggests that strongly sheared field lines close to, or possibly even part of, the erupting filament reconnected with a weakly sheared arcade during the first HXR peak. The remainder of these field lines reconnected later in the event, producing the HXR peak at the southern filament footpoint. Furthermore, we found several thermal X-ray sources during the onset of the impulsive phase and a very low-lying initial thermal loop top source that passes through a double structure during its rise. We were able to relate many of these observations to features of the complex flare geometry involving multiple interacting magnetic flux systems. Conclusions. The combined STIX and AIA observations, complemented by the NLFF extrapolation, allowed us to successfully constrain and verify the signatures of energy release and transport in the flare under study. Our results show that the reconnection between field lines with very different shear in the early phase of the flare plays a crucial role in understanding the motion of the HXR footpoint during later parts of the flare. This generalizes simpler models, such as whipping reconnection, which only consider reconnection propagating along uniformly sheared arcades.

Key Points• Constellations have an inherent tension between the scientific benefit of including more satellites and the additional cost that imposes.• Many white papers and Heliophysics Mission Concept Study reports are calling on the Decadal to recommend constellations, particularly to obtain 4π observations of the Sun.• Well-funded Artificial Intelligence (AI) methods can enhance the science return and forecasting capabilities of these constellations, for example to create virtual observatories capable of filling spatial gaps in the observations, and to inform/focus future mission designs.

Biomechanical and clinical gait research observes muscles and tendons in limbs to study their functions and behaviour. Therefore, movements of distinct anatomical landmarks, such as muscle-tendon junctions, are frequently measured. We propose a reliable and time efficient machine-learning approach to track these junctions in ultrasound videos and support clinical biomechanists in gait analysis. In order to facilitate this process, a method based on deep-learning was introduced. We gathered an extensive dataset, covering 3 functional movements, 2 muscles, collected on 123 healthy and 38 impaired subjects with 3 different ultrasound systems, and providing a total of 66864 annotated ultrasound images in our network training. Furthermore, we used data collected across independent laboratories and curated by researchers with varying levels of experience. For the evaluation of our method a diverse test-set was selected that is independently verified by four specialists. We show that our model achieves similar performance scores to the four human specialists in identifying the muscle-tendon junction position. Our method provides time-efficient tracking of muscle-tendon junctions, with prediction times of up to 0.078 seconds per frame (approx. 100 times faster than manual labeling). All our codes, trained models and test-set were made publicly available and our model is provided as a free-to-use online service on https://deepmtj.org/.