ARTICEPT

ARTICEPT is an automated Python-based module that identifies and characterizes solar active regions (ARs) using near real-time National Solar Observatory/Global Oscillation Network Group (NSO/GONG) magnetograms as raw inputs. The module is primarily developed to drive the SEPCaster model. It applies image segmentation techniques to detect regions of interest (ROIs) with both positive and negative polarities at 1-, 2-, and 3-sigma intensity thresholds. An agglomerative hierarchical clustering algorithm, refined with a consensus-based framework, is then employed to robustly identify potential AR clusters (ARCs) from the detected ROIs. Each ARC is subsequently characterized using a set of 24 parameters, including polarity inversion line (PIL) gradients, strong PIL (SPIL) gradients, boundary- and area-based AR complexity indices, and potential eruption site locations. The resulting AR parameters are then used to initialize SEPCaster for generating CME-driven shock simulations. Currently, the following modified versions of ARTICEPT are being developed/integrated: (1) Into the SOFIE Pipeline (collaboration with the CLEAR Team): As part of the post-eruption physics-based prediction, the ROI pair detection algorithm is integrated into the SOFIE pipeline to automatically extract positive-negative ROI pair that is closest to the observed flare location (from DONKI), and (2) Into the iPATH Pipeline (collaboration with Junxiang Hu): As part of the flare-triggered version of iPATH, the ROI pair detection and the AR characterization algorithms are employed to predict whether the observed flare (from DONKI) is associated with a CME or not.

The primary limitation of this module is tied to the availability of NSO/GONG MRZQS magnetograms (i.e., zeropoint-corrected synoptic magnetograms), which are produced approximately every hour. Although the module processes each magnetogram in less than an hour, it cannot operate faster than the cadence itself. The input format of this magnetogram is required to be a FITs file. In addition, if a magnetogram contains anomalous or atypical errors, the module may require human intervention to ensure robustness. Currently, this version of the module exclusively processes NSO/GONG magnetograms only. However, it is important to note that the module could be easily modified to process SDO/vector magnetograms. Due to the synoptic nature of the magnetograms, the input data cannot provide information about the solar backside, which limits our ability to perform far-side analysis. Furthermore, our ARC identification approach differs fundamentally from that of NOAA/SWPC. Rather than relying on semi-manually identified sunspot groups, our module automatically detects ROIs and identifies potential ARs from the pre-processed magnetograms. This makes our detection and identification procedures highly statistically guided and algorithm-driven. Keeping these in mind, the potential ARCs identified by our module may not correspond directly to the official NOAA/SWPC sunspot groups.

For each run, an NSO/GONG MRZQS magnetogram is acquired from the NSO/GONG server (or any pre-defined API) and used as an input file. This is a zeropoint-corrected synoptic magnetogram. The input format is required to be a FITS file.

1.​ ROI data table from the detection algorithm: Bounding box coordinates (in pixel and physical units), Flux-weighted centroid coordinates (in pixel and physical units).

2.​ ARC data table from the clustering algorithm: Bounding box coordinates (in pixel and physical units).

3.​ AR parameter data table from the characterization algorithm: Total unsigned flux (gauss–pixels², gauss–kilometers²), Maximum unsigned flux (gauss–pixels², gauss–kilometers²), Total number of unsigned flux-weighted centroids or peaks (unitless), Total active region area (pixels², kilometers²), Total number of polarity inversion line fragments (unitless), Sum of lengths of polarity inversion lines (pixels, kilometers), Total number of strong polarity inversion line fragments (unitless), Sum of lengths of strong polarity inversion lines (pixels, kilometers), Sum of unsigned gradients along SPILs (gauss/pixel, gauss/kilometer), Maximum unsigned gradient (gauss/pixel, gauss/kilometer), Maximum sum of unsigned gradients (gauss/pixel, gauss/kilometer), Maximum unsigned longitudinal gradient (gauss/pixel, gauss/kilometer), Sum of unsigned longitudinal gradients (gauss/pixel, gauss/kilometer), Maximum sum of unsigned longitudinal gradients (gauss/pixel, gauss/kilometer), Maximum unsigned latitudinal gradient (gauss/pixel, gauss/kilometer), Sum of unsigned latitudinal gradients (gauss/pixel, gauss/kilometer), Maximum sum of unsigned latitudinal gradients (gauss/pixel, gauss/kilometer), Fraction of SPIL lengths to PIL lengths (unitless), Boundary-based complexity index (unitless), Area-based complexity index (unitless), Effective active region area (pixels², kilometers²), Unsigned effective magnetic field (gauss), Unsigned effective flux (gauss–pixels², gauss–kilometers²), Total number of potential eruption sites (unitless).

4.​ Plots and Annotated Magnetograms: Several plots and annotated magnetograms are generated at each step of the run.