Solar Imaging Pipeline (SIMPL)

• The paper pioneers an automated, self‐calibration-based approach that addresses challenges from the Sun's extreme brightness, variability, and extended structure.
• It employs specialized uv-domain flagging and iterative calibrator selection to reliably manage RFI and ionospheric errors during daytime observations.
• SIMPL significantly enhances dynamic range and image fidelity, enabling scalable, uniform processing of extensive LOFAR solar data archives.

Searching arXiv for recent and foundational papers on LOFAR solar imaging pipelines and SIMPL. Solar Imaging Pipeline for LOFAR, often abbreviated as SIMPL, is an automated solar radio imaging pipeline developed specifically for LOFAR interferometric observations of the Sun. In the historical LOFAR literature it first appeared as the “LOFAR Solar Imaging Pipeline,” an extension of the LOFAR Standard Imaging Pipeline tailored to a target that is extremely bright, highly variable in time, extended, and observed under daytime ionospheric conditions (Breitling et al., 2016). In its current form, SIMPL is described as “an automated, self-calibration based pipeline for high-fidelity solar imaging with LOFAR,” designed to produce science-ready spectroscopic snapshot images with minimal human intervention and to enable scalable and uniform processing of large archival datasets (Dey et al., 14 Aug 2025). Across these formulations, the central idea is consistent: low-frequency solar images are not treated as ordinary radio maps, but as time- and frequency-resolved products whose calibration, imaging, quality control, and interpretation must be specialized for solar radio emission.

1. Origins and conceptual scope

The earliest pipeline formulation was introduced in “The LOFAR Solar Imaging Pipeline and the LOFAR Solar Data Center” (Breitling et al., 2016). That work defined the LOFAR Solar Imaging Pipeline as an extension of the LOFAR Standard Imaging Pipeline for solar observations and paired it with the LOFAR Solar Data Center as an archive and dissemination layer. Its rationale was that standard LOFAR imaging assumptions do not transfer well to solar data, because the Sun is about $10^4$ Jy at 100 MHz in quiet conditions, can rise up to $10^{12}$ Jy during intense solar radio bursts, is highly variable in time, is extended rather than point-like, and is observed during daytime when ionospheric scintillation is worse (Breitling et al., 2016).

The modern articulation in “An automated, self-calibration based pipeline for high-fidelity solar imaging with LOFAR: SIMPL” (Dey et al., 14 Aug 2025) preserves that motivation but redefines the operational center of the pipeline. The paper states that standard LOFAR calibration and imaging strategies, which are largely calibrator-based and inherited from non-solar radio astronomy, fail to deliver reliable solar images because of contamination of the simultaneously observed calibrator beam by the Sun, strong direction-dependent daytime ionospheric phase errors, extreme source variability and brightness range, poor suitability of conventional time-frequency RFI flagging for solar bursts, and the difficulty of automating all of this at scale (Dey et al., 14 Aug 2025). This yields a more specific calibration philosophy: use the calibrator mainly for amplitude or flux scaling, and recover the crucial phases directly from the Sun via self-calibration (Dey et al., 14 Aug 2025).

LOFAR’s frequency range is central to this scope. The older pipeline paper describes solar work in the Low Band Antenna range $10\text{--}90$ MHz and the High Band Antenna range $110\text{--}250$ MHz, with the $10\text{--}250$ MHz range corresponding roughly to radial distances of 1 to 3 solar radii in the corona (Breitling et al., 2016). The modern SIMPL paper likewise frames LOFAR as operating over $10\text{--}240$ MHz and positions the pipeline as enabling studies of quiet Sun coronal structure, active regions, noise storms, type III bursts, type II bursts, faint non-thermal features, dynamic event studies, and long-term archival analyses (Dey et al., 14 Aug 2025). This suggests a continuity of purpose but a substantial change in calibration architecture.

2. Why solar imaging requires a specialized pipeline

The need for SIMPL follows from properties of the radio Sun that are repeatedly emphasized across the LOFAR solar literature. The overview paper “Solar Imaging using Low Frequency Arrays” states explicitly that solar low-frequency imaging must be designed around high-fidelity, high-cadence, narrow-instant snapshot imaging of a bright, spectrally and temporally complex source, rather than around the long integrations, Earth-rotation synthesis, and multi-frequency synthesis commonly used in standard radio astronomy (Lonsdale et al., 2018). It further states that for the Sun one generally cannot rely on Earth rotation or multifrequency synthesis to improve Fourier-plane sampling, so pipelines must operate on short time and narrow frequency bins (Lonsdale et al., 2018).

The original LOFAR solar pipeline paper encodes the same distinction operationally. It contrasts standard imaging and solar imaging in terms of temporal resolution, flagging strategy, uv-range, and deconvolution choice: solar imaging uses $\le 1$ s cadence, does not time-flag the Sun data, restricts the uv-range to $0\text{--}1000$ wavelengths, and uses the CASA imager with multi-scale CLEAN because solar emission is extended (Breitling et al., 2016). The reason Sun data are not flagged in time is stated directly: solar bursts can be short-lived and bright, so a standard temporal flagger could classify the burst itself as interference and remove real science signal (Breitling et al., 2016).

The later SIMPL paper sharpens these requirements by adding specific failure modes. Solar radio emission can evolve on sub-second timescales and over very narrow fractional bandwidths $(\delta\nu/\nu_0 \ll 1)$, and the corona spans a brightness temperature range from quiet thermal bremsstrahlung, typically $T_B \sim 10^6$ K, to coherent bursts such as type III events with $10^{12}$0 K (Dey et al., 14 Aug 2025). The same dataset may therefore contain faint diffuse quiet-Sun structure and ultra-bright compact burst emission simultaneously (Dey et al., 14 Aug 2025). A plausible implication is that a usable solar pipeline must treat dynamic range, model complexity, and time-frequency segmentation as first-class design constraints rather than as optional refinements.

Several science papers reinforce that implication. “LOFAR observations of the quiet solar corona” shows that image products below 100 MHz are not simply deconvolved maps but frequency-resolved, propagation-aware observables from which one can infer coronal structure (Vocks et al., 2018). “LOFAR imaging of the solar corona during the 2015 March 20 solar eclipse” demonstrates that standard imaging can overestimate apparent source size in the quiet corona, implying that measured morphology mixes instrumental resolution with physical propagation broadening (Ryan et al., 2021). “Interferometric Imaging with LOFAR Remote Baselines of the Fine Structures of a Solar Type IIIb Radio Burst” further shows that source position, size, and apparent motion can be dominated by scattering, refraction, and possibly ducting, rather than by intrinsic source geometry alone (Zhang et al., 2020). In this literature, SIMPL is therefore not merely an automation layer; it is an attempt to formalize solar-specific calibration, imaging, and quality-control logic around those physical and instrumental constraints.

3. Historical architecture of the LOFAR Solar Imaging Pipeline

The 2016 pipeline architecture begins with correlated LOFAR observations and produces both image products and beam-formed dynamic spectra (Breitling et al., 2016). Imaging visibilities are stored as CASA Measurement Sets, beam-formed spectroscopic data are stored in HDF5, and the flowchart described in the paper includes online processing, temporary storage, preprocessing, calibration, imaging, optional self-calibration, simulation and diagnostics, and ingestion into the LOFAR Solar Data Center (Breitling et al., 2016).

Preprocessing is performed with NDPPP. The paper states that calibrator data are RFI-flagged with AOFlagger, Sun data are not flagged, channels are averaged to one channel per subband, and the uv-range is restricted to $10^{12}$1 wavelengths (Breitling et al., 2016). Calibration is then performed with BBS, deriving amplitude and phase corrections every 30 seconds on a bright calibrator and transferring those solutions to the solar data (Breitling et al., 2016). The paper names Cassiopeia A, Cygnus A, Taurus A, and Virgo A as the usable bright calibrators, and imposes explicit geometry constraints: the calibrator should be neither closer than $10^{12}$2 nor farther than $10^{12}$3 from the Sun, and both Sun and calibrator should be above $10^{12}$4 elevation (Breitling et al., 2016).

Imaging is performed with the CASA imager using multi-scale CLEAN, and each image is transformed from equatorial coordinates to solar coordinates with the Sun centered and the Sun’s north pole oriented upward (Breitling et al., 2016). The paper also reports that CASA imaging was extended to support moving sources via the movingsource option with keyword SUN, allowing aperture synthesis without smearing due to the Sun’s changing right ascension and declination (Breitling et al., 2016). Optional phase-only self-calibration is available for selected cases, using the cleaned image itself as the model, with usually one or two cycles sufficient (Breitling et al., 2016). The pipeline further generates diagnostics and simulations, including Gaussian Sun models, PSFs, and side-lobe contamination estimates from bright sources (Breitling et al., 2016).

The archive component, the LOFAR Solar Data Center, is presented as the repository and web-access layer for these products (Breitling et al., 2016). It exposes observation summaries, diagnostics, subband pages, dynamic spectrum pages, log files, configuration files, and links to external spectra and context images. In that formulation, the pipeline is as much an operational infrastructure for standardized processing and browsing as it is a calibration or imaging algorithm.

4. The self-calibration architecture of SIMPL

The 2025 paper redefines the core workflow around three major parts: calibrator processing, solar processing, and a self-calibration sub-module (Dey et al., 14 Aug 2025). The logic is explicit. SIMPL first identifies a usable quiet portion of contaminated calibrator data, derives calibrator gain solutions, keeps only the amplitude gains from the calibrator, processes the solar data in time-frequency chunks, performs solar self-calibration to derive phases and then amplitudes, optionally applies differential per-slice self-calibration for higher dynamic range, applies beam correction, and generates final FITS snapshot images (Dey et al., 14 Aug 2025).

The most distinctive shift relative to the earlier pipeline is the rejection of transferred calibrator phases. The paper states that the calibrator beam contains direction-independent instrumental phases plus direction-dependent ionospheric phase terms in the calibrator direction, and that these cannot correct the different ionospheric phase corruption in the solar direction (Dey et al., 14 Aug 2025). A quantitative example is given: a Virgo A image in its own direction had dynamic range $10^{12}$5, applying those full calibrator gains to the Sun produced a solar image with dynamic range only $10^{12}$6, and SIMPL’s self-cal approach on the same solar data raised this to $10^{12}$7 (Dey et al., 14 Aug 2025). This suggests that the historical transfer-calibration architecture was serviceable but not sufficient for high-fidelity automation.

Self-calibration is formulated through the usual least-squares visibility model. For a known sky model, the paper gives

$10^{12}$8

and for the unknown-model case it gives

$10^{12}$9

with $10\text{--}90$0 the observed visibility, $10\text{--}90$1 and $10\text{--}90$2 the complex antenna gains, $10\text{--}90$3 the model visibility, $10\text{--}90$4 the model image, and $10\text{--}90$5 the Fourier transform operator (Dey et al., 14 Aug 2025). The paper notes that the initial solar model is a simple Gaussian placed at the phase center and that the first iteration uses only short baselines where the Sun is unresolved (Dey et al., 14 Aug 2025). As the model improves, progressively longer baselines are admitted; phase-only self-calibration is performed first, and amplitude-plus-phase self-calibration is introduced only after the phase solutions are well constrained (Dey et al., 14 Aug 2025).

A representative example in the paper illustrates the iteration schedule. With only amplitudes from the calibrator and $10\text{--}90$6, the first phase-only self-calibration reaches dynamic range 106; after a second iteration using $10\text{--}90$7 the dynamic range is 211; by iteration 5 with $10\text{--}90$8 the dynamic range is 343; the final phase-only stage reaches 364; and the final amplitude-phase self-calibration yields 471 (Dey et al., 14 Aug 2025). The paper states that convergence is evaluated by image improvement such as dynamic range, though it does not provide a numerical stopping threshold (Dey et al., 14 Aug 2025).

A further complication is positional registration. Because self-calibration loses absolute phase information, SIMPL performs a special self-calibration step on a quiet solar interval to align the solar disk and preserve positional registration, after which images can be transformed properly into the helioprojective coordinate system (Dey et al., 14 Aug 2025). The quiet interval is at most 2 seconds, and the disk-visible condition is required so that the phase-center Gaussian aligns with the solar center rather than with an off-center compact burst (Dey et al., 14 Aug 2025). This is a specifically solar solution to a problem that ordinary radio self-calibration does not encounter in the same way.

5. RFI excision, quality control, and image production

RFI excision is a major point of divergence between ordinary LOFAR processing and SIMPL. For the calibrator branch, SIMPL uses CASA flagdata with tfcrop for initial RFI flagging, then calibrates iteratively with CASA bandpass, and after each iteration flags residual visibilities with rflag before recalibration (Dey et al., 14 Aug 2025). To reduce solar contamination from the extended quiet Sun, calibration uses only visibilities in a suitable uv range, typically $10\text{--}90$9 (Dey et al., 14 Aug 2025).

For solar data, however, the paper argues that conventional time-frequency flagging is unsafe because genuine solar bursts are statistical outliers in precisely the way that RFI detectors expect (Dey et al., 14 Aug 2025). SIMPL therefore implements a solar-specific uv-domain outlier detector inspired by aNKflag. For each time-frequency slice, visibilities are analyzed as a function of uv distance rather than in time-frequency space, using logarithmic uv-distance bins, separate treatment of the real and imaginary parts, and a robust MAD-based statistic

$110\text{--}250$0

where $110\text{--}250$1 for the data values $110\text{--}250$2 within a uv bin (Dey et al., 14 Aug 2025). Thresholds are intentionally lenient, typically $110\text{--}250$3 depending on the number of samples in the bin, increased for sparse bins, with a stated example of $110\text{--}250$4 if fewer than 20 points are present (Dey et al., 14 Aug 2025). If more than 40% of the points in a uv bin would be flagged, the threshold is relaxed to $110\text{--}250$5 the original, and if the bin still exceeds 40% flagged, nothing is flagged in that bin (Dey et al., 14 Aug 2025). For bins with more than 30 points, the median uses a trimmed estimator removing 5% from each tail (Dey et al., 14 Aug 2025).

This solar-specific flagging is applied to the CORRECTED_DATA column independently for each time, frequency, and correlation $110\text{--}250$6 and $110\text{--}250$7, after self-calibration, and the paper states that processing is parallelized across multiple timestamps (Dey et al., 14 Aug 2025). One example shows rms improving from 13.7 Jy/beam to 7.5 Jy/beam and dynamic range improving from 37 to 74 after uv-domain flagging followed by another self-calibration round (Dey et al., 14 Aug 2025). A plausible implication is that SIMPL treats quality control as an iterative interaction between flagging and calibration rather than as a one-time front-end cleanup.

Calibrator quiet-time selection is also fully algorithmic. SIMPL evaluates a sliding-window composite contamination score using standard deviation, MAD, skewness, kurtosis, and normalized Shannon entropy, each min-max normalized as

$110\text{--}250$8

and combined as

$110\text{--}250$9

with empirically chosen weights $10\text{--}250$0, $10\text{--}250$1, $10\text{--}250$2, $10\text{--}250$3, and $10\text{--}250$4 (Dey et al., 14 Aug 2025). The score is smoothed with a moving median filter of length $10\text{--}250$5, with default sliding window $10\text{--}250$6 s, and windows below the 25th percentile are called quiet; the longest contiguous quiet segment is chosen for calibrator solving (Dey et al., 14 Aug 2025). This suggests an explicit attempt to make calibrator contamination management reproducible and archive-scalable.

Final imaging is done with WSClean at user-defined time intervals, time integrations, frequency intervals, and frequency averaging, with default Briggs weighting and robustness parameter 0.5 (Dey et al., 14 Aug 2025). After calibration, visibilities are corrected for beam gain in the solar direction using everybeam integrated within DP3 (Dey et al., 14 Aug 2025). The principal outputs are science-ready FITS spectroscopic snapshot images, and in archive-production mode also video previews (Dey et al., 14 Aug 2025).

6. Relation to broader LOFAR solar-imaging practice

SIMPL sits within a broader methodological landscape rather than replacing it. Several LOFAR solar studies describe workflows that resemble either predecessors or complementary branches of SIMPL. The quiet-Sun study at low band uses an external calibrator, transfer calibration, moving-target imaging, brightness-temperature conversion, sector-based radial profile extraction, and ray-tracing fits including refraction and free-free transfer (Vocks et al., 2018). That work is effectively a prototype end-to-end solar imaging and profile-analysis workflow for low-band LOFAR data and makes explicit that low-frequency solar maps are inputs to a propagation-aware inversion rather than endpoints (Vocks et al., 2018). This suggests that SIMPL’s imaging products are most scientifically useful when coupled to downstream profile extraction and physically constrained fitting.

By contrast, the tied-array studies of type III bursts and S bursts use beam-formed modes rather than interferometric imaging. They produce dynamic spectra at many fixed spatial pointings, then reconstruct sparse spatial maps by interpolating beam intensities, typically at 1 s and 5 MHz for type III bursts or 50 ms and 1 MHz for S bursts [(Morosan et al., 2014); (Morosan et al., 2015)]. Those studies are explicit that tied-array mode is a beam-sampled spectro-imaging method rather than a replacement for interferometric imaging, and that its strengths are high temporal and spectral resolution, while its limitations are beam-size-dominated spatial uncertainty and lack of conventional image fidelity [(Morosan et al., 2014); (Morosan et al., 2015)]. In a SIMPL context, this suggests a division of labor between beam-formed burst detection and localization on one side, and visibility-based higher-fidelity imaging on the other.

Other interferometric burst-imaging papers illustrate the value and the current limits of SIMPL-like processing. The type II band-splitting study uses DP3 for calibrator-based gain solving and WSClean with w-stacking CLEAN and Briggs robust $10\text{--}250$7, then derives contours and 2D elliptical Gaussian centroids from images and combines those with full-Stokes beamformed spectra to measure the degree of circular polarisation (Normo et al., 15 May 2025). The jet-driven piston shock study uses DPPP and WSClean to image fundamental and harmonic sources at 1 s and 195.3 kHz cadence, then interprets the apparent source displacement through a scattering model (Maguire et al., 2021). The type IIIb fine-structure study uses DPPP transfer calibration from Taurus A and WSClean with multiscale deconvolution and Briggs robust 0.2 to image each 0.168 s visibility block separately, then derives brightness temperature, source position, source area, temporal evolution, and apparent source velocity (Zhang et al., 2020). These papers do not describe SIMPL itself, but they repeatedly expose the same problems that SIMPL is built to standardize: burst-aware cadence, external calibrator limitations, dynamic range pressure, and the need to interpret morphology through scattering and refraction.

The eclipse study in the HBA range adds a further caution. Its comparison between standard interferometric imaging and de-occultation-based super-resolution shows that the quiet-Sun source width at 80% maximum is $10\text{--}250$8 in standard imaging, $10\text{--}250$9 in image differencing, and $10\text{--}240$0 in visibility differencing, whereas the active-region source remains near $10\text{--}240$1 in all methods (Ryan et al., 2021). The paper concludes that ordinary imaging can overestimate scattering in the quiet Sun by 70% when using standard imaging and by 40% when using image differencing (Ryan et al., 2021). A plausible implication is that SIMPL must preserve beam, PSF, and calibration provenance carefully, because high-quality images are still not equivalent to direct maps of intrinsic source structure.

In its current documented form, SIMPL is explicitly limited to LOFAR core stations and Stokes $10\text{--}240$2 only (Dey et al., 14 Aug 2025). The paper states that current remote-baseline incorporation is ineffective with the present strategy due to sparse uv coverage and stronger long-baseline ionospheric decorrelation, and it identifies full-polarization calibration and remote-baseline high-resolution imaging as future work (Dey et al., 14 Aug 2025). Even so, the reported performance is substantial: in HBA observations dynamic range routinely exceeds 400 in quiescent conditions and can exceed 1000 during intense radio bursts, while a quiet-Sun LBA snapshot at 35.64 MHz achieves dynamic range about 100 with only 1 s and 1 MHz integration (Dey et al., 14 Aug 2025). The paper concludes that SIMPL is now sufficiently mature to process more than a decade of LOFAR solar observations and deliver science-ready FITS products for community use (Dey et al., 14 Aug 2025).

Taken together, these developments define SIMPL as both a software pipeline and a codification of solar-specific interferometric practice. Historically, it extends the LOFAR Solar Imaging Pipeline architecture of calibrator transfer, solar-coordinate imaging, diagnostics, and archival access (Breitling et al., 2016). In its current form, it recenters the system on solar self-calibration, solar-specific uv-domain flagging, automated calibrator contamination management, and spectroscopic snapshot imaging at archive scale (Dey et al., 14 Aug 2025). Its significance lies not only in automation but in the formal recognition that, at LOFAR frequencies, the Sun is a calibration problem, an imaging problem, and an interpretation problem simultaneously.