arXiv subjects

A. Asensio Ramos

Publications and source records attributed to A. Asensio Ramos.

At least 37 records · Page 2Linked to original sources

arXiv2024

Solar multi-object multi-frame blind deconvolution with a spatially variant convolution neural emulator

The study of astronomical phenomena through ground-based observations is always challenged by the distorting effects of Earth's atmosphere. Traditional methods of post-facto image correction, essential for correcting these distortions, often rely on simplifying assumptions that limit their effectiveness, particularly in the presence of spatially variant atmospheric turbulence. Such cases are often solved by partitioning the field-of-view into small patches, deconvolving each patch independently, and merging all patches together. This approach is often inefficient and can produce artifacts. Recent advancements in computational techniques and the advent of deep learning offer new pathways to address these limitations. This paper introduces a novel framework leveraging a deep neural network to emulate spatially variant convolutions, offering a breakthrough in the efficiency and accuracy of astronomical image deconvolution. By training on a dataset of images convolved with spatially invariant point spread functions and validating its generalizability to spatially variant conditions, this approach presents a significant advancement over traditional methods. The convolution emulator is used as a forward model in a multi-object multi-frame blind deconvolution algorithm for solar images. The emulator enables the deconvolution of solar observations across large fields of view without resorting to patch-wise mosaicking, thus avoiding artifacts associated with such techniques. This method represents a significant computational advantage, reducing processing times by orders of magnitude.

astro-ph.IM↗

arXiv2024

NLTE modelling of water-rich exoplanet atmospheres. Cooling and heating rates

The hydrogen and water molecules respond very differently to the collisional-radiative processes taking place in planetary atmospheres. Naturally, the question arises whether H2O-rich atmospheres are more (or less) resilient to long-term mass loss than H2-dominated ones if they radiate away the incident stellar energy more (or less) efficiently. If confirmed, the finding would have implications on our understanding of the evolution of sub-Neptune exoplanets. As a key step towards answering this question, we present a non-local thermodynamic equilibrium (NLTE) model of H2O for the atmospheric region where the gas accelerates to escape the planet and conditions relevant to close-in sub-Neptunes. Our exploratory calculations for isothermal gas composed of H2, H2O and e- reveal that: 1) In the pressure region ~1e-2 - 1e-4 dyn cm-2 where the stellar extreme-ultraviolet (XUV) photons are typically deposited in the atmosphere, H2O is in rotational LTE but vibrational NLTE. Vibrational LTE is facilitated by high H2O abundances and fractional ionizations, and we report critical densities for the LTE-NLTE transition; 2) Vibrational cooling may locally dominate over rotational cooling, partly because of the comparatively small opacities of ro-vibrational lines; 3) Even low H2O abundances notably enhance the cooling, foreseeably offsetting some of the stellar heating; 4) Heating due to the deposition of stellar infrared (IR) photons is significant at pressures >=0.1 dyn cm-2. We estimate the contribution of H2O excitation to the internal energy of the gas and speculate on the photodissociation from the excited vibrational states. Ultimately, our findings motivate the consideration of NLTE in the mass loss rate calculations of H2O-rich atmospheres.

astro-ph.EP↗

arXiv2023

Magnetic field fluctuations in the shocked umbral chromosphere

Several studies have reported magnetic field fluctuations associated with umbral shock waves. We aim to study the properties and origin of magnetic field fluctuations in the umbral chromosphere. Temporal series of spectropolarimetric observations were acquired with the GREGOR telescope. The chromospheric and photospheric conditions were derived from simultaneous inversions of the He I 10830 Å triplet and the Si I 10827 Å line using HAZEL2. The oscillations are interpreted using wavelet analysis and context information from UV observations acquired with SDO/AIA and IRIS. The chromospheric magnetic field shows strong fluctuations in the sunspot umbra, with peak field strengths up to 2900 G. Magnetic field and velocity umbral oscillations exhibit a strong coherence, with the magnetic field lagging the shock fronts detected in the velocity fluctuations. This points to a common origin of the fluctuations in both parameters, whereas the analysis of the phase shift between photospheric and chromospheric velocity is consistent with upwards wave propagation. These results suggest that the strong inferred magnetic field fluctuations are caused by changes in the response height of the He I 10830 Å line to the magnetic field, which is sensitive to high photospheric layers after the shock fronts. The coronal activity seen in EUV data could possibly have some impact on the inferred fluctuations, but it is not the main driver of the magnetic field oscillations since they are found before EUV events take place. Chromospheric magnetic field fluctuations measured with the He I 10830 Å triplet arise due to variations in the opacity of the line. After shocks produced by slow magnetoacoustic waves, the response of the line to the magnetic field can be shifted down to the upper photosphere. This is seen as remarkably large fluctuations in the line of sight magnetic field strength.

astro-ph.SR↗

arXiv2023

Machine learning in solar physics

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.

astro-ph.SR↗

arXiv2023

Accelerating Multiframe Blind Deconvolution via Deep Learning

Ground-based solar image restoration is a computationally expensive procedure that involves nonlinear optimization techniques. The presence of atmospheric turbulence produces perturbations in individual images that make it necessary to apply blind deconvolution techniques. These techniques rely on the observation of many short exposure frames that are used to simultaneously infer the instantaneous state of the atmosphere and the unperturbed object. We have recently explored the use of machine learning to accelerate this process, with promising results. We build upon this previous work to propose several interesting improvements that lead to better models. As well, we propose a new method to accelerate the restoration based on algorithm unrolling. In this method, the image restoration problem is solved with a gradient descent method that is unrolled and accelerated aided by a few small neural networks. The role of the neural networks is to correct the estimation of the solution at each iterative step. The model is trained to perform the optimization in a small fixed number of steps with a curated dataset. Our findings demonstrate that both methods significantly reduce the restoration time compared to the standard optimization procedure. Furthermore, we showcase that these models can be trained in an unsupervised manner using observed images from three different instruments. Remarkably, they also exhibit robust generalization capabilities when applied to new datasets. To foster further research and collaboration, we openly provide the trained models, along with the corresponding training and evaluation code, as well as the training dataset, to the scientific community.

astro-ph.IM↗

arXiv2023

High-precision interpolation of stellar atmospheres with a deep neural network using a 1D convolutional auto encoder for feature extraction

Given the widespread availability of grids of models for stellar atmospheres, it is necessary to recover intermediate atmospheric models by means of accurate techniques that go beyond simple linear interpolation and capture the intricacies of the data. Our goal is to establish a reliable, precise, lightweight, and fast method for recovering stellar model atmospheres, that is to say the stratification of mass column, temperature, gas pressure, and electronic density with optical depth given any combination of the defining atmospheric specific parameters: metallicity, effective temperature, and surface gravity, as well as the abundances of other key chemical elements. We employed a fully connected deep neural network which in turn uses a 1D convolutional auto-encoder to extract the nonlinearities of a grid using the ATLAS9 and MARCS model atmospheres. This new method we call iNNterpol effectively takes into account the nonlinearities in the relationships of the data as opposed to traditional machine-learning methods, such as the light gradient boosting method (LightGBM), that are repeatedly used for their speed in well-known competitions with reduced datasets. We show a higher precision with a convolutional auto-encoder than using principal component analysis as a feature extractor.We believe it constitutes a useful tool for generating fast and precise stellar model atmospheres, mitigating convergence issues, as well as a framework for future developments. The code and data for both training and direct interpolation are available online at https://github.com/cwestend/iNNterpol for full reproducibility and to serve as a practical starting point for other continuous 1D data in the field and elsewhere.

astro-ph.IM↗

arXiv2023

Estimating the longitudinal magnetic field in the chromosphere of quiet-Sun magnetic concentrations

Details of the magnetic field in the quiet Sun chromosphere are key to our understanding of essential aspects of the solar atmosphere. We aim to determine the longitudinal magnetic field component (B_lon) of quiet Sun regions depending on their size. We estimated B_lon by applying the weak-field approximation (WFA) to high-spatial-resolution Ca II 854.2 nm data taken with the Swedish 1m Solar Telescope. Specifically, we analyzed the estimates inferred for different spectral ranges using the data at the original cadence and temporally integrated signals. The longitudinal magnetic field in each considered plasma structure correlates with its size. Using a spectral range restricted to the line core leads to chromospheric longitudinal fields varying from 50 G at the edges to 150-500 G at the center of the structure. These values increase as the spectral range widens due to the photospheric contribution. However, the difference between this contribution and the chromospheric one is not uniform for all structures. Small and medium-sized concentrations show a steeper height gradient in B_lon compared to their chromospheric values, so estimates for wider ranges are less trustworthy. Signal addition does not alleviate this situation as the height gradients in B_lon are consistent with time. Finally, despite the amplified noise levels that deconvolving processes may cause, data restored with the destretching technique show similar results, though are affected by smearing. We obtained B_lon estimates similar to those previously found, except for large concentrations and wide spectral ranges. In addition, we report a correlation between the height variation of B_lon compared to the chromospheric estimates and the concentration size. This correlation affects the difference between the photospheric and chromospheric magnetic flux values and the reliability of the estimates for wider spectral ranges.

astro-ph.SR↗

arXiv2022

FarNet-II: An improved solar far-side active region detection method

Context. Activity on the far side of the Sun is routinely studied through the analysis of the seismic oscillations detected on the near side using helioseismic techniques such as phase shift sensitive holography. Recently, the neural network FarNet was developed to improve these detections. Aims. We aim to create a new machine learning tool, FarNet II, which further increases the scope of FarNet, and to evaluate its performance in comparison to FarNet and the standard helioseismic method for detecting far side activity. Methods. We developed FarNet II, a neural network that retains some of the general characteristics of FarNet but improves the detections in general, as well as the temporal coherence among successive predictions. The main novelties are the implementation of attention and convolutional long short term memory (ConvLSTM) modules. A cross validation approach, training the network 37 times with a different validation set for each run, was employed to leverage the limited amount of data available. We evaluate the performance of FarNet II using three years of extreme ultraviolet observations of the far side of the Sun acquired with the Solar Terrestrial Relations Observatory (STEREO) as a proxy of activity. The results from FarNet II were compared with those obtained from FarNet and the standard helioseismic method using the Dice coefficient as a metric. Results. FarNet II achieves a Dice coefficient that improves that of FarNet by over 0.2 points for every output position on the sequences from the evaluation dates. Its improvement over FarNet is higher than that of FarNet over the standard method. Conclusions. The new network is a very promising tool for improving the detection of activity on the far side of the Sun given by pure helioseismic techniques. Space weather forecasts can potentially benefit from the higher sensitivity provided by this novel method.

astro-ph.SR↗

arXiv2022

Quiet Sun Center to Limb Variation of the Linear Polarization Observed by CLASP2 Across the Mg II h & k Lines

The CLASP2 (Chromospheric LAyer SpectroPolarimeter 2) sounding rocket mission was launched on 2019 April 11. CLASP2 measured the four Stokes parameters of the Mg II h & k spectral region around 2800 Angstroms along a 200 arcsecond slit at three locations on the solar disk, achieving the first spatially and spectrally resolved observations of the solar polarization in this near ultraviolet region. The focus of the work presented here is the center-to-limb variation of the linear polarization across these resonance lines, which is produced by the scattering of anisotropic radiation in the solar atmosphere. The linear polarization signals of the Mg II h & k lines are sensitive to the magnetic field from the low to the upper chromosphere through the Hanle and magneto-optical effects. We compare the observations to theoretical predictions from radiative transfer calculations in unmagnetized semi-empirical models, arguing that magnetic fields and horizontal inhomogeneities are needed to explain the observed polarization signals and spatial variations. This comparison is an important step in both validating and refining our understanding of the physical origin of these polarization signatures, and also in paving the way toward future space telescopes for probing the magnetic fields of the solar upper atmosphere via ultraviolet spectropolarimetry.

astro-ph.SR↗

arXiv2022

Accelerating non-LTE synthesis and inversions with graph networks

The computational cost of fast non-LTE synthesis is one of the challenges that limits the development of 2D and 3D inversion codes. It also makes the interpretation of observations of lines formed in the chromosphere and transition region a slow and computationally costly process, which limits the inference of the physical properties on rather small fields of view. Having access to a fast way of computing the deviation from the LTE regime through the departure coefficients could largely alleviate this problem. We propose to build and train a graph network that quickly predicts the atomic level populations without solving the non-LTE problem. We find an optimal architecture for the graph network for predicting the departure coefficients of the levels of an atom from the physical conditions of a model atmosphere. A suitable dataset with a representative sample of potential model atmospheres is used for training. This dataset has been computed using existing non-LTE synthesis codes. The graph network has been integrated into existing synthesis and inversion codes for the particular case of \caii. We demonstrate orders of magnitude gain in computing speed. We analyze the generalization capabilities of the graph network and demonstrate that it produces good predicted departure coefficients for unseen models. We implement this approach in \hazel\ and show how the inversions nicely compare with those obtained with standard non-LTE inversion codes. Our approximate method opens up the possibility of extracting physical information from the chromosphere on large fields-of-view with time evolution.

astro-ph.SR↗

arXiv2021

Bayesian Stokes inversion with Normalizing flows

Stokes inversion techniques are very powerful methods for obtaining information on the thermodynamic and magnetic properties of solar and stellar atmospheres. In recent years, very sophisticated inversion codes have been developed that are now routinely applied to spectro-polarimetric observations. Most of these inversion codes are designed for finding an optimum solution to the nonlinear inverse problem. However, to obtain the location of potentially multimodal cases (ambiguities), the degeneracies, and the uncertainties of each parameter inferred from the inversions, algorithms such as Markov chain Monte Carlo (MCMC), require to evaluate the likelihood of the model thousand of times and are computationally costly. Variational methods are a quick alternative to Monte Carlo methods by approximating the posterior distribution by a parametrized distribution. In this study, we introduce a highly flexible variational inference method to perform fast Bayesian inference, known as normalizing flows. Normalizing flows are a set of invertible, differentiable, and parametric transformations that convert a simple distribution into an approximation of any other complex distribution. If the transformations are conditioned on observations, the normalizing flows can be trained to return Bayesian posterior probability estimates for any observation. We illustrate the ability of the method using a simple Milne-Eddington model and a complex non-LTE inversion. However, the method is extremely general and other more complex forward models can be applied. The training procedure need only be performed once for a given prior parameter space and the resulting network can then generate samples describing the posterior distribution several orders of magnitude faster than existing techniques.

astro-ph.SR↗

arXiv2021

Approximate Bayesian Neural Doppler Imaging

The non-uniform surface temperature distribution of rotating active stars is routinely mapped with the Doppler Imaging technique. Inhomogeneities in the surface produce features in high-resolution spectroscopic observations that shift in wavelength depending on their position on the visible hemisphere. The inversion problem has been systematically solved using maximum a-posteriori regularized methods assuming smoothness or maximum entropy. Our aim in this work is to solve the full Bayesian inference problem, by providing access to the posterior distribution of the surface temperature in the star. We use amortized neural posterior estimation to produce a model that approximates the high-dimensional posterior distribution for spectroscopic observations of selected spectral ranges sampled at arbitrary rotation phases. The posterior distribution is approximated with conditional normalizing flows, which are flexible, tractable and easy to sample approximations to arbitrary distributions. When conditioned on the spectroscopic observations, they provide a very efficient way of obtaining samples from the posterior distribution. The conditioning on observations is obtained through the use of Transformer encoders, which can deal with arbitrary wavelength sampling and rotation phases. Our model can produce thousands of posterior samples per second. Our validation of the model for very high signal-to-noise observations shows that it correctly approximates the posterior, although with some overestimation of the broadening. We apply the model to the moderately fast rotator II Peg, producing the first Bayesian map of its temperature inhomogenities. We conclude that conditional normalizing flows are a very promising tool to carry out approximate Bayesian inference in more complex problems in stellar physics, like constraining the magnetic properties.

astro-ph.IM↗

arXiv2021

Performance of solar far-side active regions neural detection

Context. Far-side helioseismology is a technique used to infer the presence of active regions in the far hemisphere of the Sun based on the interpretation of oscillations measured in the near hemisphere. A neural network has been recently developed to improve the sensitivity of the seismic maps to the presence of far-side active regions. Aims. Our aim is to evaluate the performance of the new neural network approach and to thoroughly compare it with the standard method commonly applied to predict far-side active regions from seismic measurements. Methods. We have computed the predictions of active regions using the neural network and the standard approach from five years of far-side seismic maps as a function of the selected threshold in the signatures of the detections. The results have been compared with direct extreme ultraviolet observations of the far hemisphere acquired with the Solar Terrestrial Relations Observatory (STEREO). Results. We have confirmed the improved sensitivity of the neural network to the presence of far-side active regions. Approximately 96% of the active regions identified by the standard method with a strength above the threshold commonly employed by previous analyses are related to locations with enhanced extreme ultraviolet emission. For this threshold, the false positive ratio is 3.75%. For an equivalent false positive ratio, the neural network produces 47% more true detections. Weaker active regions can be detected by relaxing the threshold in their seismic signature. Conclusions. The neural network is a promising approach to improve the interpretation of the seismic maps provided by local helioseismic techniques.

astro-ph.SR↗

arXiv2020

Learning to do multiframe wavefront sensing unsupervisedly: applications to blind deconvolution

Observations from ground based telescopes are affected by the presence of the Earth atmosphere, which severely perturbs them. The use of adaptive optics techniques has allowed us to partly beat this limitation. However, image selection or post-facto image reconstruction methods applied to bursts of short-exposure images are routinely needed to reach the diffraction limit. Deep learning has been recently proposed as an efficient way to accelerate these image reconstructions. Currently, these deep neural networks are trained with supervision, so that either standard deconvolution algorithms need to be applied a-priori or complex simulations of the solar magneto-convection need to be carried out to generate the training sets. Our aim here is to propose a general unsupervised training scheme that allows multiframe blind deconvolution deep learning systems to be trained simply with observations. The approach can be applied for the correction of point-like as well as extended objects. Leveraging the linear image formation theory and a probabilistic approach to the blind deconvolution problem produces a physically-motivated loss function. The optimization of this loss function allows an end-to-end training of a machine learning model composed of three neural networks. As examples, we apply this procedure to the deconvolution of stellar data from the FastCam instrument and to solar extended data from the Swedish Solar Telescope. The analysis demonstrates that the proposed neural model can be successfully trained without supervision using observations only. It provides estimations of the instantaneous wavefronts, from which a corrected image can be found using standard deconvolution technniques. The network model is roughly three orders of magnitude faster than applying standard deconvolution based on optimization and shows potential to be used on real-time at the telescope.

astro-ph.IM↗

arXiv2020

Planet cartography with neural learned regularization

Finding potential life harboring exo-Earths is one of the aims of exoplanetary science. Detecting signatures of life in exoplanets will likely first be accomplished by determining the bulk composition of the planetary atmosphere via reflected/transmitted spectroscopy. However, a complete understanding of the habitability conditions will surely require mapping the presence of liquid water, continents and/or clouds. Spin-orbit tomography is a technique that allows us to obtain maps of the surface of exoplanets around other stars using the light scattered by the planetary surface. We leverage the potential of deep learning and propose a mapping technique for exo-Earths in which the regularization is learned from mock surfaces. The solution of the inverse mapping problem is posed as a deep neural network that can be trained end-to-end with suitable training data. We propose in this work to use methods based on the procedural generation of planets, inspired by what we found on Earth. We also consider mapping the recovery of surfaces and the presence of persistent cloud in cloudy planets. We show that the a reliable mapping can be carried out with our approach, producing very compact continents, even when using single passband observations. More importantly, if exoplanets are partially cloudy like the Earth is, we show that one can potentially map the distribution of persistent clouds that always occur on the same position on the surface (associated to orography and sea surface temperatures) together with non-persistent clouds that move across the surface. This will become the first test one can perform on an exoplanet for the detection of an active climate system. For small rocky planets in the habitable zone of their stars, this weather system will be driven by water, and the detection can be considered as a strong proxy for truly habitable conditions.

astro-ph.EP↗

arXiv2020

Determining the dynamics and magnetic fields in He I 10830 Å during a solar filament eruption

We investigate the dynamics and magnetic properties of the plasma, such as line-of-sight velocity (LOS), optical depth, vertical and horizontal magnetic fields, belonging to an erupted solar filament. The filament eruption was observed with the GREGOR Infrared Spectrograph (GRIS) at the 1.5-meter GREGOR telescope on 2016 July 3. Three consecutive full-Stokes slit-spectropolarimetric scans in the He I 10830 Å spectral range were acquired. The Stokes I profiles were classified using the machine learning k-means algorithm and then inverted with different initial conditions using the HAZEL code. The erupting-filament material presents the following physical conditions: (1) ubiquitous upward motions with peak LOS velocities of ~73 km/s; (2) predominant large horizontal components of the magnetic field, on average, in the range of 173-254 G, whereas the vertical components of the fields are much lower, on average between 39-58 G; (3) optical depths in the range of 0.7-1.1. The average azimuth orientation of the field lines between two consecutive raster scans (<2.5 minutes) remained constant. The analyzed filament eruption belonged to the fast rising phase, with total velocities of about 124 km/s. The orientation of the magnetic field lines does not change from one raster scan to the other, indicating that the untwisting phase has not started yet. The untwisting seems to start about 15 min after the beginning of the filament eruption.

astro-ph.SR↗

arXiv2020

Ubiquitous hundred-Gauss magnetic fields in solar spicules

Aims. We use high-resolution spectropolarimetric observations in the Ca ii 8542 A line obtained with the SST to study the magnetic field in solar spicules. Methods. The equations that result from the application of the Weak Field Approximation (WFA) to the radiative transfer equations are used to infer the LOS component of the magnetic field (BLOS). Two restrictive conditions are imposed on the Stokes I and V profiles at each pixel before they can be used in a Bayesian inversion to compute its BLOS. Results. The LOS magnetic field component has been inferred in six data sets totalling 448 spectral scans in the Ca ii 8542 A line and containing both active region and quiet Sun areas, with values of hundreds of G being abundantly inferred. There seems to be no difference, from the statistical point of view, between the magnetic field strength of spicules in the quiet Sun or near an active region. On the other hand, the BLOS distributions present smaller values on the disk than off-limb, a fact that can be explained by the effect of superposition on the chromosphere of on-disk structures. We find that spicules in the vicinity of a sunspot have a magnetic field polarity (i.e. north or south) equal to that of the sunspot. This paper also contains an analysis of the effect of off-limb overlapping structures on the observed Stokes I and V parameters and the BLOS obtained from the WFA. It is found that this value is equal to or smaller than the largest LOS magnetic field components of the two structures. In addition, using random BLOS, Doppler velocities and line intensities of these two structures leads in ~ 50% of the cases to Stokes I and V parameters unsuitable to be used with the WFA. Conclusions. Our results present a scarcity of LOS magnetic field components smaller than some 50 G, which must not be taken as evidence against the existence of these magnetic field strengths in spicules.

astro-ph.SR↗

arXiv2019

Improved detection of farside solar active regions using deep learning

The analysis of waves in the visible side of the Sun allows the detection of active regions in the farside through local helioseismology techniques. The knowledge of the magnetism in the whole Sun, including the non-visible hemisphere, is fundamental for several space weather forecasting applications. Seismic identification of farside active regions is challenged by the reduced signal-to-noise, and only large and strong active regions can be reliable detected. Here we develop a new methodology to improve the identification of active region signatures in farside seismic maps. We have constructed a deep neural network that associates the farside seismic maps obtained from helioseismic holography with the probability of presence of active regions in the farside. The network has been trained with pairs of helioseismic phase shift maps and Helioseismic and Magnetic Imager magnetograms acquired half a solar rotation later, which were used as a proxy for the presence of active regions in the farside. The method has been validated using a set of artificial data, and it has also been applied to actual solar observations during the period of minimum activity of the solar cycle 24. Our approach shows a higher sensitivity to the presence of farside active regions than standard methods applied up to date. The neural network can significantly increase the number of detected farside active regions, and will potentially improve the application of farside seismology to space weather forecasting.

astro-ph.SR↗