The Plasma-prescribed Active Region Static Extrapolation (PARSE) Dataset: A Machine-Learning-Ready Collection of Magnetohydrostatic Coronal Active Regions

TL;DR

This paper introduces the PARSE dataset, a large collection of magnetohydrostatic active region data cubes designed for training and validating physics-informed neural networks in solar physics research.

Contribution

It presents a novel, publicly available dataset of over five thousand simulated active regions constructed using a state-of-the-art extrapolation method, facilitating machine learning applications.

Findings

Over five thousand data cubes generated

High-resolution magnetic and plasma data near active regions

Accessible dataset for training physics-informed models

Abstract

As Physics-Informed Neural Networks and other methods for full-vector-field construction or analysis become more prominent, a need has developed for a large set of simulated active regions for training, validation and testing purposes. We use a state-of-the-art magnetohydrostatic extrapolation method to develop a public dataset of over five thousand data cubes based on the Spaceweather HMI Active Region Patch (SHARP) library of active region magnetogram images. Each cube resolves the magnetic field vector and plasma forcing at approximately 100,000 scattered points that are adaptively clustered near the high-flux regions of the domain. This paper describes the methodology of construction of the Plasma-prescribed Active Region Static Extrapolation (PARSE) dataset, as well as its structure and how to access it.