Speaker
Description
Coronal mass ejections (CMEs) are among the most energetic solar eruptions and primary drivers of space weather. Despite decades of study, their internal velocity distributions remain poorly characterised. Coronagraph-based studies of CMEs have traditionally relied on leading-edge tracking or geometric fitting methods, which provide limited information about the internal velocity structure of eruptions. We present DOFCAT (Dense Optical Flow CME Analysis Tool), an open-source Python-based software package that applies dense optical flow algorithms to coronagraph image sequences to generate spatially resolved, pixel-level velocity maps of CMEs and their substructures. Unlike conventional methods, DOFCAT requires no prior assumptions about CME geometry or morphology, making it broadly applicable across instruments and events.
DOFCAT has been validated using high-resolution, high-cadence data from two next-generation space-based coronagraphs, ASPIICS onboard ESA's PROBA-3 and METIS onboard Solar Orbiter, as well as LASCO C2 onboard SOHO, providing complementary coverage of the middle corona (1.5–6.0 R☉), the critical region where CME impulsive acceleration and internal restructuring predominantly occur. The tool incorporates a dedicated pre-processing pipeline that includes background subtraction and edge-preserving noise smoothing while preserving feature edges to prepare coronagraph data for optical flow computation. We also developed a Gaussian-tapered Fourier (GTF) filter to suppress brightness flickering artefacts in ASPIICS running difference images, significantly improving velocity estimation stability. Validation across multiple structured CME events demonstrates that DOFCAT reliably captures internal velocity dispersion, front-core separation dynamics, and position-angle-dependent velocity gradients, which otherwise are inaccessible to traditional tracking approaches.
A key prospective capability of DOFCAT is physics-driven automated CME detection. Analysis of optical flow maps reveals that CME passage through the coronagraphic field-of-view produces a characteristic and reproducible statistical signature in the frame-by-frame velocity distribution: a sharp rise in high-velocity pixel fraction relative to background levels, broadening of the velocity histogram, and subsequent return to background levels. This signature is image-independent and physically motivated, providing a robust basis for automated detection without relying on brightness thresholds or morphological assumptions. Building on this, we are currently developing a machine learning and AI-based CME detection model trained on velocity distribution signatures extracted from a large sample of CME events across multiple coronagraphs. This approach moves beyond traditional intensity-based cataloguing, enabling scalable, systematic, and physically interpretable CME identification across large coronagraph archives with direct applicability to real-time space weather monitoring pipelines.
DOFCAT is publicly available as an open-source repository, designed for community use with data from existing and upcoming coronagraphs, including PUNCH and Vigil (at L5 vantage point). Standardised velocity map products from DOFCAT can serve as inputs for space weather forecasting pipelines, CME databases, and heliospheric propagation models. By providing a geometry-free, automation-ready framework for CME velocity analysis, DOFCAT addresses a significant gap in the current space weather instrumentation toolkit and contributes to the global effort to enable real-time, AI-enabled CME monitoring and forecasting.