arxiv: 2604.21184 · v1 · submitted 2026-04-23 · 🌌 astro-ph.SR

Recognition: unknown

Predicting the thermodynamics in the chromosphere from the translation of SDO data into the IRIS² inversion results using a visual transformer model

Alberto Sainz Dalda, Bart De Pontieu, Juno Kim, Kyuhyoun Cho, Paul S. Killam, Viggo Hansteen, Vishal Upendran

Pith reviewed 2026-05-09 21:12 UTC · model grok-4.3

classification 🌌 astro-ph.SR

keywords solar chromospherevisual transformerSDOIRISthermodynamic inversionelectron densitytemperature predictionmachine learning

0 comments

The pith

A visual transformer model translates SDO images and magnetograms into estimates of chromospheric temperature and electron density that match IRIS inversion targets for most test cases.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a visual transformer model that takes intensity images from the Atmospheric Imaging Assembly in the chromosphere and transition region plus line-of-sight magnetograms from the Helioseismic and Magnetic Imager and produces maps of temperature, line-of-sight velocity, turbulent velocity, and electron density. These outputs are trained against thermodynamic quantities derived from inversions of Mg II lines observed by the Interface Region Imaging Spectrograph. The model reaches a correlation of about 0.80 with the target temperature and electron density values across roughly 80 percent of the held-out test data and a correlation of 0.63 for turbulent velocity in 70 percent of cases, while line-of-sight velocity remains weakly correlated. The entire process completes in a few minutes on ordinary hardware, offering a fast route to chromospheric thermodynamic information whenever SDO observations are available.

Core claim

The SDO2IRIS² visual transformer learns a direct mapping from combined AIA intensity images of the chromosphere and transition region and HMI photospheric magnetograms to the temperature, line-of-sight velocity, turbulent velocity, and electron density that result from IRIS Mg II inversions. On independent test data the predicted temperature and electron density agree with the inversion results at a correlation of approximately 0.80 for about 80 percent of the samples, turbulent velocity reaches a moderate-to-strong correlation of 0.63 for 70 percent of the samples, and line-of-sight velocity shows only weak correlation. The model therefore supplies usable estimates of chromospheric state in

What carries the argument

The visual transformer architecture that ingests multi-channel AIA images and HMI magnetograms and regresses four thermodynamic quantities learned from IRIS inversion targets.

If this is right

The model outputs can serve as stand-alone estimates of chromospheric thermodynamics or as complementary information when other data are available.
The full set of thermodynamic maps is produced in a few minutes on CPU or GPU, far faster than direct inversion of spectral lines.
SDO data alone become sufficient for estimating chromospheric state in locations or times when simultaneous IRIS observations are absent.
The trained model can be applied repeatedly to the continuous, full-disk SDO archive to generate long time series of chromospheric conditions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the mapping remains stable across the solar cycle, the model could support near-real-time tracking of chromospheric heating and cooling using only SDO's routine observations.
The same translation strategy could be retrained on other instrument pairs to estimate additional atmospheric parameters such as magnetic field strength or ionization state.
The notably weaker performance on line-of-sight velocity indicates that purely spatial image inputs may miss essential dynamical information that spectral line profiles supply.

Load-bearing premise

The IRIS Mg II inversions used as training targets accurately represent true chromospheric conditions and the learned mapping applies to solar regions and times outside the training set.

What would settle it

Side-by-side comparison of model outputs with independent temperature and density measurements obtained from high-resolution ground-based spectroscopy or from forward-modeled synthetic spectra in solar regions never seen during training.

Figures

Figures reproduced from arXiv: 2604.21184 by Alberto Sainz Dalda, Bart De Pontieu, Juno Kim, Kyuhyoun Cho, Paul S. Killam, Viggo Hansteen, Vishal Upendran.

**Figure 1.** Figure 1: Left: spatial and temporal distribution of the selected IRIS data on the solar disk. Right: distribution of the exposure time of these data. 3. METHODOLOGY We use the thermodynamic maps obtained by IRIS2 , the SDO/AIA intensity images, and the SDO/HMI magnetograms to train a visual transformer (ViT A. Dosovitskiy et al. 2020) model to predict the thermodynamic values in the chromosphere from the SDO/AIA im… view at source ↗

**Figure 2.** Figure 2: Distributions of the Pearson correlation coefficients (r) from the linear correlation between each thermodynamic variable (rows) obtained by IRIS2 and the corresponding one obtained by SDO2IRIS2 at the selected optical depths (columns), for the original data (blue) and the Z-score filtered data (orange) of the experiment “1600 1700 304 HMI”. The mean and the mode of each distribution for the Z-score filter… view at source ↗

**Figure 3.** Figure 3: Shapley attribution values for the AIA and HMI passbands for the correlation r between the predicted and target IRIS2 thermodynamic variables. umbra. In this case, we suggest to the potential user of SDO2IRIS2 to keep a critical view of the results provided by SDO2IRIS2 while at the same time considering also other data available by other sources for the FoV studied (such as IRIS2 ). Another important arti… view at source ↗

**Figure 4.** Figure 4: Comparison between the thermodynamic values (in rows) obtained by IRIS2 for the IRIS data observed on 2022-12-29 11:29:33 UT (left panel) and the ones predicted by SDO2IRIS2 (right panel) at log(τ)=-5.2 (first column), log(τ)=-4.2 (second column), and log(τ)=-3.2 (third column), by using AIA 1600 Å, AIA 1700 Å, AIA 304 Å, and the HMI magnetogram as the input of SDO2IRIS2 . chosen this dataset because it co… view at source ↗

**Figure 5.** Figure 5: Similar to [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗

**Figure 6.** Figure 6: Similar to [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗

read the original abstract

We present SDO2IRIS$^2$: a visual transformer model that translates a combination of images of the chromosphere and transition region (TR), observed by AIA, and a line-of-sight magnetogram, provided by HMI, into temperature, line-of-sight velocity (v$_{los}$), velocity of the turbulent motions (v$_{turb}$), and electron density (n$_{e}$) in the chromosphere. Using the thermodynamic variables obtained from the inversion of the chromospheric lines Mg II h&k, observed by IRIS, as the target of the model, and the intensity images in the chromosphere and TR, and the photospheric magnetogram as the input, the predicted T and n$_{e}$ show a strong correlation ($\approx 0.80$) for $\approx$80% of the test inverted data, a moderate-to-strong correlation ($\approx0.63$) for 70% of the v$_{turb}$ of the target test inverted data, while for the $v_{los}$, the correlation is weak. Therefore, the predicted values by SDO2IRIS$^2$ may be used as an estimation of the thermodynamics in the chromosphere, either as a stand-alone result or as complementary information to other chromospheric data observed simultaneously. The execution time employed by SDO2IRIS$^2$ to obtain the thermodynamic values in the chromosphere is of the order of a few minutes, being $\le10$ minutes when using a CPU, and $\le5$ minutes when using a GPU. SDO2IRIS$^2$ opens a new avenue for the use of SDO data thanks to the inversions provided by IRIS observables.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The model translates SDO data to chromospheric T, ne, and v_turb with reported correlations of 0.8 and 0.63 on most test points, but unspecified train/test splits leave generalization claims unsupported.

read the letter

The core offering is a visual transformer that takes AIA chromosphere and transition-region images plus an HMI magnetogram and outputs estimates of chromospheric temperature, electron density, line-of-sight velocity, and turbulent velocity. Training targets come from IRIS Mg II inversions. On the test portion the paper reports Pearson correlations near 0.80 for T and ne across roughly 80 percent of points, 0.63 for v_turb across 70 percent, and weaker results for v_los. Runtime is a few minutes on CPU or GPU, which is genuinely useful for bulk processing of the SDO archive. That speed and the straightforward architecture choice are the practical strengths; the work shows that an off-the-shelf transformer can be pointed at this particular instrument-translation problem without obvious architectural invention. The abstract also positions the output as either stand-alone or complementary data, which is a reasonable framing if the numbers hold up. The main weakness is the missing description of the train/test split. Solar imaging data are strongly autocorrelated in space and time, so any split that re-uses the same active region, overlapping time windows, or random pixel sampling risks leakage. Without explicit confirmation that the test set is drawn from entirely separate epochs and regions, the quoted correlations cannot be read as evidence of robust generalization. The abstract supplies no error bars, no hyperparameter details, and no discussion of how the IRIS inversion uncertainties propagate. The weak v_los result is noted but left largely unexamined. This paper is aimed at solar physicists who need denser chromospheric thermodynamic coverage than IRIS alone can provide. A reader already working with SDO and IRIS data might extract a usable tool once the split protocol is clarified, but the current write-up does not yet supply enough methodological transparency for immediate adoption. I would send it to peer review; the idea is concrete and the runtime claim is testable, but referees will need to see proper independence checks and validation before the performance numbers can be trusted.

Referee Report

3 major / 1 minor

Summary. The manuscript introduces SDO2IRIS², a visual transformer model that maps combined AIA chromospheric/TR intensity images and HMI line-of-sight magnetograms to chromospheric thermodynamic quantities (temperature T, electron density n_e, line-of-sight velocity v_los, and turbulent velocity v_turb) derived from IRIS Mg II h&k inversions. The abstract reports Pearson correlations of ≈0.80 for T and n_e on ≈80% of the test data, ≈0.63 for v_turb on 70% of the test data, and weak correlation for v_los, concluding that the model outputs can serve as stand-alone or complementary estimates with inference times of a few minutes.

Significance. If the correlations demonstrate robust generalization beyond the training distribution, the work would offer a practical method to derive chromospheric thermodynamics from the extensive, high-cadence SDO dataset, extending coverage beyond IRIS's limited field of view and enabling broader statistical studies of chromospheric heating and dynamics. The visual-transformer architecture is a reasonable choice for this image-to-parameter regression task, and the reported execution times highlight potential for operational use.

major comments (3)

[Abstract] Abstract: The training, validation, and test split protocol is not described. Solar imaging data exhibit strong spatial and temporal autocorrelations; without explicit confirmation that the test set uses independent active regions, disjoint time intervals, or a strict temporal buffer, the reported correlations cannot be interpreted as evidence of generalization to new solar regions or epochs.
[Abstract] Abstract: The statement that correlations hold 'for ≈80% of the test inverted data' (and '70%' for v_turb) provides no selection criterion for the subset, no per-pixel or per-profile correlation distributions, and no uncertainty estimates or statistical tests, preventing assessment of whether the headline numbers are representative or driven by a small high-performing fraction.
[Abstract] Abstract and model description: No information is supplied on the number of training samples, hyperparameter tuning, regularization, or any baseline comparisons (e.g., linear regression or simpler CNN), making it impossible to judge whether the visual transformer architecture is necessary or whether the performance is robust.

minor comments (1)

[Abstract] The abstract states that the model 'opens a new avenue' but does not quantify how the predicted quantities would be validated against independent observations (e.g., other spectral lines or ground-based data) when IRIS inversions are unavailable.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their detailed and constructive comments on our manuscript. We address each of the major comments below and outline the revisions we plan to implement in the updated version of the paper.

read point-by-point responses

Referee: [Abstract] Abstract: The training, validation, and test split protocol is not described. Solar imaging data exhibit strong spatial and temporal autocorrelations; without explicit confirmation that the test set uses independent active regions, disjoint time intervals, or a strict temporal buffer, the reported correlations cannot be interpreted as evidence of generalization to new solar regions or epochs.

Authors: We agree that the abstract does not provide details on the data splitting protocol. In the revised manuscript, we will include a clear description of the training, validation, and test split in the Methods section, explicitly confirming that the test set uses independent active regions, disjoint time intervals, and a strict temporal buffer to minimize the effects of spatial and temporal autocorrelations in solar imaging data. revision: yes
Referee: [Abstract] Abstract: The statement that correlations hold 'for ≈80% of the test inverted data' (and '70%' for v_turb) provides no selection criterion for the subset, no per-pixel or per-profile correlation distributions, and no uncertainty estimates or statistical tests, preventing assessment of whether the headline numbers are representative or driven by a small high-performing fraction.

Authors: We acknowledge that the abstract lacks information on the selection criterion for the reported percentages and does not include supporting distributions or statistical analyses. In the revision, we will specify the selection criterion used for the ≈80% and ≈70% subsets, provide per-pixel and per-profile correlation distributions, and include uncertainty estimates along with relevant statistical tests to allow proper assessment of the results' representativeness. revision: yes
Referee: [Abstract] Abstract and model description: No information is supplied on the number of training samples, hyperparameter tuning, regularization, or any baseline comparisons (e.g., linear regression or simpler CNN), making it impossible to judge whether the visual transformer architecture is necessary or whether the performance is robust.

Authors: The referee correctly identifies the absence of these details in the abstract and model description. We will revise the manuscript to report the number of training samples, describe the hyperparameter tuning and regularization methods, and add baseline comparisons with simpler models such as linear regression and convolutional neural networks. This will help demonstrate the robustness of the performance and the appropriateness of the visual transformer architecture. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard supervised ML mapping with independent test evaluation

full rationale

The paper trains a visual transformer to map SDO/AIA+HMI inputs to thermodynamic targets obtained from separate IRIS inversions. Reported correlations are computed on test-set outputs versus held-out IRIS targets; these are statistical results of learned weights, not algebraic identities or definitional equivalences to the inputs. No self-definitional equations, fitted parameters renamed as predictions, load-bearing self-citations, or ansatz smuggling appear in the abstract or described chain. The derivation is self-contained empirical modeling whose validity hinges on data-split independence rather than internal reduction to the training inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities beyond the implicit assumption that IRIS inversions constitute reliable ground truth.

pith-pipeline@v0.9.0 · 5653 in / 991 out tokens · 30863 ms · 2026-05-09T21:12:46.811120+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

300 extracted references · 225 canonical work pages · 4 internal anchors

[1]

, year = 2020, month = feb, volume =

The SunPy Project: Open Source Development and Status of the Version 1.0 Core Package , journal =. 2020 , month = feb, volume =. doi:10.3847/1538-4357/ab4f7a , url =

work page doi:10.3847/1538-4357/ab4f7a 2020
[2]

P., Tollerud, E

Astropy: A community Python package for astronomy , journal =. 2013 , volume =. doi:10.1051/0004-6361/201322068 , url =

work page doi:10.1051/0004-6361/201322068 2013
[3]

The Astropy Project: Building an Open-science Project and Status of the v2.0 Core Package<sup>*</sup>

The Astropy Project: Building an Open-science Project and Status of the v2.0 Core Package , journal =. 2018 , volume =. doi:10.3847/1538-3881/aabc4f , url =

work page doi:10.3847/1538-3881/aabc4f 2018
[4]

The Astropy Project: Sustaining and Growing a Community-oriented Open-source Project and the Latest Major Release (v5.0) of the Core Package

The Astropy Project: Sustaining and Growing a Community-oriented Open-source Project and the Latest Major Release (v5.0) of the Core Package , journal =. 2022 , month = aug, volume =. doi:10.3847/1538-4357/ac7c74 , url =

work page internal anchor Pith review doi:10.3847/1538-4357/ac7c74 2022
[5]

, title =

Hunter, John D. , title =. Computing in Science & Engineering , year =
[6]

and Millman, K

Harris, Charles R. and Millman, K. Jarrod and van der Walt, Stéfan J and Gommers, Ralf and Virtanen, Pauli and Cournapeau, David and Wieser, Eric and Taylor, Julian and Berg, Sebastian and Smith, Nathaniel J. and Kern, Robert and Picus, Matti and Hoyer, Stephan and van Kerkwijk, Marten H. and Brett, Matthew and Haldane, Allan and Fernández del Río, Jaime ...
[7]

1995 , publisher=

Python reference manual , author=. 1995 , publisher=

1995
[8]

, title =

Van Rossum, Guido and Drake, Fred L. , title =. 2009 , isbn =

2009
[9]

2020 , pages =

Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python , journal =. 2020 , volume =. doi:10.1038/s41592-019-0686-2 , url =

work page doi:10.1038/s41592-019-0686-2 2020
[10]

Scikit-learn: Machine Learning in Python , journal =

Pedregosa, Fabian and Varoquaux, Ga. Scikit-learn: Machine Learning in Python , journal =. 2011 , volume =

2011
[11]

Bradski, Gary , title =. Dr. Dobb's Journal of Software Tools , year =
[12]

Keras , year =

Chollet, Fran. Keras , year =
[13]

PyTorch: An Imperative Style, High-Performance Deep Learning Library , booktitle =

Paszke, Adam and Gross, Sam and Massa, Francisco and Lerer, Adam and Bradbury, James and Chanan, Gregory and Killeen, Trevor and Lin, Zeming and Gimelshein, Natalia and Antiga, Luca and Desmaison, Alban and K. PyTorch: An Imperative Style, High-Performance Deep Learning Library , booktitle =. 2019 , url =

2019
[14]

2024 , eprint=

The Faiss library , author=. 2024 , eprint=

2024
[15]

and Hart, P

Cover, T. and Hart, P. , journal=. Nearest neighbor pattern classification , year=
[16]

Fifth Berkeley Sympos. Math. Statist. and Probability. I: Statistics , pages =
[17]

arXiv , author =:2009.07896 , primaryclass =

Captum: A unified and generic model interpretability library for PyTorch. arXiv e-prints , keywords =. doi:10.48550/arXiv.2009.07896 , archivePrefix =. 2009.07896 , primaryClass =

work page doi:10.48550/arxiv.2009.07896 2009
[18]

doi:10.5281/zenodo.3828935 , license =

Falcon, William and. doi:10.5281/zenodo.3828935 , license =

work page doi:10.5281/zenodo.3828935
[19]

and Schmude, Johannes and Lal, Rohit and Gaur, Vishal and Lin, Amy and Mandal, Kshitiz and Singh, Talwinder and Mu

Roy, Sujit and Hegde, Dinesha V. and Schmude, Johannes and Lal, Rohit and Gaur, Vishal and Lin, Amy and Mandal, Kshitiz and Singh, Talwinder and Mu. SuryaBench: Benchmark Dataset for Advancing Machine Learning in Heliophysics and Space Weather Prediction , journal =. 2026 , issn =. doi:10.1038/s41597-026-06552-5 , url =

work page doi:10.1038/s41597-026-06552-5 2026
[20]

2025, Surya: Foundation Model for Heliophysics, arXiv, doi: 10.48550/arXiv.2508.14112

Surya: Foundation Model for Heliophysics. arXiv e-prints , keywords =. doi:10.48550/arXiv.2508.14112 , archivePrefix =. 2508.14112 , primaryClass =

work page doi:10.48550/arxiv.2508.14112
[21]

A Unified Approach to Interpreting Model Predic- tions, November 2017

A Unified Approach to Interpreting Model Predictions. arXiv e-prints , keywords =. doi:10.48550/arXiv.1705.07874 , archivePrefix =. 1705.07874 , primaryClass =

work page doi:10.48550/arxiv.1705.07874
[22]

Contributions to the Theory of Games II , editor =

A Value for n-Person Games , author =. Contributions to the Theory of Games II , editor =. 1953 , publisher =

1953
[23]

Xinlei Chen, Hao Fang, Tsung-yi Lin, Ramakrishna Vedantam, Saurabh Gupta, Piotr Dollar, and C Lawrence Zitnick

End-to-End Object Detection with Transformers. arXiv e-prints , keywords =. doi:10.48550/arXiv.2005.12872 , archivePrefix =. 2005.12872 , primaryClass =

work page doi:10.48550/arxiv.2005.12872 2005
[24]

On layer normalization in the transformer architecture, 2020

On Layer Normalization in the Transformer Architecture. arXiv e-prints , keywords =. doi:10.48550/arXiv.2002.04745 , archivePrefix =. 2002.04745 , primaryClass =

work page doi:10.48550/arxiv.2002.04745 2002
[25]

Adam: A Method for Stochastic Optimization

Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[26]

Edouard Grave, Armand Joulin, and Nicolas Usunier

Convolutional Sequence to Sequence Learning. arXiv e-prints , keywords =. doi:10.48550/arXiv.1705.03122 , archivePrefix =. 1705.03122 , primaryClass =

work page doi:10.48550/arxiv.1705.03122
[27]

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. arXiv e-prints , keywords =. doi:10.48550/arXiv.2010.11929 , archivePrefix =. 2010.11929 , primaryClass =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2010.11929 2010
[28]

Attention Is All You Need

Attention Is All You Need. arXiv e-prints , keywords =. doi:10.48550/arXiv.1706.03762 , archivePrefix =. 1706.03762 , primaryClass =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1706.03762
[29]

Sur la division des corps mat\'eriels en parties. Bull. Acad. Polon. Sci. , volume = 4, pages =
[30]

doi:10.25495/7GXK-RD71 , url =

Zenodo , publisher =. doi:10.25495/7GXK-RD71 , url =

work page doi:10.25495/7gxk-rd71
[31]

and Orozco Su

Bellot Rubio, L. and Orozco Su. Quiet Sun magnetic fields: an observational view. Living Rev Sol Phys , year = 2019, volume = 16, pages =. doi:https://doi.org/10.1007/s41116-018-0017-1 , adsurl =

work page doi:10.1007/s41116-018-0017-1 2019
[32]

Welsch, B. T. and Longcope, D. W. , title = ". ApJ. doi:10.1086/368408 , adsurl =

work page doi:10.1086/368408
[33]

The solar internetwork

Go. The solar internetwork. I. Contribution to the network magnetic flux. , year = 2014, volume = 797, pages =. doi:https://doi.org/10.1088/0004-637X/797/1/49 , adsurl =

work page doi:10.1088/0004-637x/797/1/49 2014
[34]

The Solar Internetwork

Go. The Solar Internetwork. II. Flux Appearance and Disappearance Rates. , year = 2016, volume = 820, pages =. doi:https://doi.org/10.3847/0004-637X/820/1/35 , adsurl =

work page doi:10.3847/0004-637x/820/1/35 2016
[35]

, keywords =

On the nature of moving magnetic feature pairs around sunspots. , keywords =. doi:10.1051/0004-6361:20021807 , adsurl =

work page doi:10.1051/0004-6361:20021807
[36]

, keywords =

Signatures of ubiquitous magnetic reconnection in the deep atmosphere of sunspot penumbrae. , keywords =. doi:10.1051/0004-6361/202040171 , archivePrefix =. 2101.11321 , primaryClass =

work page doi:10.1051/0004-6361/202040171
[37]

2024 IRIS 2+: A Comprehensive Database of Stratified Thermodynamic Models in the Low Solar Atmosphere.ApJS271, 24

IRIS ^ 2+ : A Comprehensive Database of Stratified Thermodynamic Models in the Low Solar Atmosphere. , keywords =. doi:10.3847/1538-4365/ad1e55 , archivePrefix =. 2211.09103 , primaryClass =

work page doi:10.3847/1538-4365/ad1e55
[38]

, keywords =

Differential emission measures from the regularized inversion of Hinode and SDO data. , keywords =. doi:10.1051/0004-6361/201117576 , archivePrefix =. 1201.2642 , primaryClass =

work page doi:10.1051/0004-6361/201117576
[39]

, keywords =

Semi-empirical model atmospheres for the chromosphere of the sunspot penumbra and umbral flashes. , keywords =. doi:10.1051/0004-6361/201935289 , archivePrefix =. 1905.08264 , primaryClass =

work page doi:10.1051/0004-6361/201935289 1905
[40]

, keywords =

Force-free Magnetic Fields: The Magneto-frictional Method. , keywords =. doi:10.1086/164610 , adsurl =

work page doi:10.1086/164610
[41]

Mark C. M. Cheung and Marc L. DeRosa , title =. The Astrophysical Journal , abstract =. 2012 , month =. doi:10.1088/0004-637X/757/2/147 , url =

work page doi:10.1088/0004-637x/757/2/147 2012
[42]

V. M. J. Henriques and D. Kuridze and M. Mathioudakis and F. P. Keenan , title =. The Astrophysical Journal , abstract =. 2016 , month =. doi:10.3847/0004-637X/820/2/124 , url =

work page doi:10.3847/0004-637x/820/2/124 2016
[43]

Impact on solar atmospheric heating

Evidence of the multi-thermal nature of spicular downflows. Impact on solar atmospheric heating. , keywords =. doi:10.1051/0004-6361/202141404 , archivePrefix =. 2108.02153 , primaryClass =

work page doi:10.1051/0004-6361/202141404
[44]

Evelyn Fix and J. L. Hodges , journal =. Discriminatory Analysis. Nonparametric Discrimination: Consistency Properties , urldate =
[45]

Evelyn Fix and J. L. Hodges , journal =. Discriminatory Analysis. Nonparametric Discrimination: Consistency Properties , year =
[46]

Frontiers in Astronomy and Space Sciences , keywords =

Chromospheric thermodynamic conditions from inversions of complex Mg II h & k profiles observed in flares. Frontiers in Astronomy and Space Sciences , keywords =. doi:10.3389/fspas.2023.1133429 , archivePrefix =. 2211.05459 , primaryClass =

work page doi:10.3389/fspas.2023.1133429 2023
[47]

Hunter, J. D. , Title =. Computing in Science & Engineering , Volume =
[48]

Billion-scale similarity search with

Johnson, Jeff and Douze, Matthijs and J. Billion-scale similarity search with. IEEE Transactions on Big Data , volume=. 2019 , publisher=

2019
[49]

, keywords =

Evolution of the Ratio of Mg II Intensities during Solar Flares. , keywords =. doi:10.3847/1538-4357/ad2a46 , archivePrefix =. 2402.11189 , primaryClass =

work page doi:10.3847/1538-4357/ad2a46
[50]

and Varoquaux, G

Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V. and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P. and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and Cournapeau, D. and Brucher, M. and Perrot, M. and Duchesnay, E. , journal=. Scikit-learn: Machine Learning in
[51]

and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and

Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and. Nature Methods , year =
[52]

, keywords =

Mass and energy flow in the solar chromosphere and corona. , keywords =. doi:10.1146/annurev.aa.15.090177.002051 , adsurl =

work page doi:10.1146/annurev.aa.15.090177.002051
[53]

Information-based spectral sampling

Designing wavelength sampling for Fabry-P \'e rot observations. Information-based spectral sampling. , keywords =. doi:10.1051/0004-6361/202346230 , archivePrefix =. 2303.13875 , primaryClass =

work page doi:10.1051/0004-6361/202346230
[54]

2026 The IRIS2+ Inversion Tool: Recovering the Radiative Losses and the Thermodynamics in the Lower Solar Atmosphere.ApJ997, 229

Sainz Dalda, Alberto and de la Cruz Rodríguez, Jaime and Hansteen, Viggo and De Pontieu, Bart and Gošić, Milan , title =. The Astrophysical Journal , abstract =. 2026 , month =. doi:10.3847/1538-4357/ae274c , url =

work page doi:10.3847/1538-4357/ae274c 2026
[55]

, keywords =

Spectropolarimetric NLTE inversion code SNAPI. , keywords =. doi:10.1051/0004-6361/201833382 , archivePrefix =. 1806.08134 , primaryClass =

work page doi:10.1051/0004-6361/201833382
[56]

, keywords =

TIC: A Stokes Inversion Code for Scattering Polarization with Partial Frequency Redistribution and Arbitrary Magnetic Fields. , keywords =. doi:10.3847/1538-4357/ac745c , archivePrefix =. 2205.15666 , primaryClass =

work page doi:10.3847/1538-4357/ac745c
[57]

Recovering Thermodynamics from Spectral Profiles Observed by IRIS. (II). Improved Calculation of the Uncertainties Based on Monte Carlo Experiments. , keywords =. doi:10.3847/1538-4357/acb2c7 , archivePrefix =. 2211.01563 , primaryClass =

work page doi:10.3847/1538-4357/acb2c7
[58]

, year = 2022, volume = 932, number = 1, pages =

Non-LTE Inversion of Prominence Spectroscopic Observations in H and Mg II h & k lines. , year = 2022, volume = 932, number = 1, pages =

2022
[59]

, keywords =

Fast Inversion of Solar Ca II Spectra in Non-local Thermodynamic Equilibrium. , keywords =. doi:10.3847/1538-4357/ab1d4c , archivePrefix =. 1904.11843 , primaryClass =

work page doi:10.3847/1538-4357/ab1d4c 1904
[60]

, keywords =

A non-LTE inversion procedure for chromospheric cloud-like features. , keywords =
[61]

, keywords =

Non-LTE inversion of chromospheric \ Ii\ cloud-like features. , keywords =. doi:10.1051/0004-6361:20000257 , adsurl =

work page doi:10.1051/0004-6361:20000257
[62]

Estructura y din \'a mica de concentraciones magn \'e ticas en la superf \' cie solarEstructura y din \'a mica de concentraciones magn \'e ticas en la superf \' cie solarStructure and dynamics of magnetic regions on the solar surface
[63]

, keywords =

The Effects of Rotation on the Evolution of Rising Omega Loops in a Stratified Model Convection Zone. , keywords =. doi:10.1086/318320 , archivePrefix =. astro-ph/0008501 , primaryClass =

work page doi:10.1086/318320
[64]

, year = 1970, month = feb, volume =

The Structure of the Magnetic Field in a Sunspot with a Photospheric Light-Bridge at Two Levels in the Solar Atmosphere. , year = 1970, month = feb, volume =

1970
[65]

, keywords =

Large-scale structure of a sunspot and its surrounding photosphere. , keywords =. doi:10.1007/BF00156373 , adsurl =

work page doi:10.1007/bf00156373
[66]

Observations of the Two-Level Structure of Sunspot Magnetic Fields (presented by N. S. Soboleva). Solar Magnetic Fields , year = 1971, editor =

1971
[67]

Bulletin of the Astronomical Institutes of Czechoslovakia , year = 1973, month = jan, volume =

On the Physical Relation between the Magnetic Field and the Brightness in the Sunspot Umbrae. Bulletin of the Astronomical Institutes of Czechoslovakia , year = 1973, month = jan, volume =

1973
[68]

, keywords =

On the Magnetic Fields and Motions in Sunspots at Different Atmospheric Levels. , keywords =. doi:10.1007/BF00162480 , adsurl =

work page doi:10.1007/bf00162480
[69]

, keywords =

On fine structure of the magnetic field and brightness in the penumbrae of sunspots. , keywords =. doi:10.1007/BF00153337 , adsurl =

work page doi:10.1007/bf00153337
[70]

, keywords =

On Mass Flow in a Complex Sunspot. , keywords =. doi:10.1007/BF00151393 , adsurl =

work page doi:10.1007/bf00151393
[71]

, keywords =

Magnetic Fields of Sunspots Based on Combined Optical and Radio Observations. , keywords =. doi:10.1007/BF00146645 , adsurl =

work page doi:10.1007/bf00146645
[72]

, keywords =

Observed Redshifts in the Solar Transition Region above Active and Quiet Regions. , keywords =. doi:10.1086/176454 , adsurl =

work page doi:10.1086/176454
[73]

, keywords =

An Observational Examination of Models for Sunspot Magnetic Field Configurations. , keywords =. doi:10.1007/BF00154777 , adsurl =

work page doi:10.1007/bf00154777
[74]

7490 (July 5, 1979)

Velocity measures in the sunspot Rome no. 7490 (July 5, 1979). , keywords =. doi:10.1007/BF00147503 , adsurl =

work page doi:10.1007/bf00147503 1979
[75]

Magnetic field observations for the sunspot C. M. P. 1966 September 19. , year = 1969, month = jan, volume =. doi:10.1093/mnras/145.1.1 , adsurl =

work page doi:10.1093/mnras/145.1.1 1966
[76]

, year = 1967, month = jan, volume =

Line contours in sunspot regions. , year = 1967, month = jan, volume =. doi:10.1093/mnras/136.1.71 , adsurl =

work page doi:10.1093/mnras/136.1.71 1967
[77]

, keywords =

A Search for Sunspot Canopies Using a Vector Magnetograph. , keywords =. doi:10.1007/BF00645086 , adsurl =

work page doi:10.1007/bf00645086
[78]

, year = 1908, month = apr, volume =

Preliminary Note on the Rotation of the Sun as Determined from the Displacements of the Hydrogen Lines. , year = 1908, month = apr, volume =. doi:10.1086/141547 , adsurl =

work page doi:10.1086/141547 1908
[79]

III Spectrophotometry and preliminary model of a 2-component umbra

Photometric analysis of sunspot umbral dots. III Spectrophotometry and preliminary model of a 2-component umbra. , keywords =
[80]

I - Dynamical and structural behaviour

Photometric analysis of the sunspot umbral dots. I - Dynamical and structural behaviour. , keywords =

Showing first 80 references.