Automated Classification of Plasma Regions at Mars Using Machine Learning

2026

Read Automated Classification of Plasma Regions at Mars Using Machine Learning (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-17131.txt.

Introduction

Qin and eight colleagues build a classifier for three plasma regions around Mars, solar wind, magnetosheath, and induced magnetosphere, using only ion energy spectra from the MAVEN Solar Wind Ion Analyzer. Two networks are tested: a multilayer perceptron and a convolutional network using short temporal sequences. That network cleanly separates all three regions, while the simpler one struggles to separate solar wind from the sheath.

Catalog record

YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30643
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.17131

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Automated Classification of Plasma Regions at Mars Using Machine Learning Abstract The plasma environment around Mars is highly variable because it is strongly influenced by the solar wind. Accurate identification of plasma regions around Mars is important for the community studying solar wind–Mars interactions, region-specific plasma processes, and atmospheric escape. In this study, we develop a machine-learning–based classifier to automatically identify three key plasma regions—solar wind, magnetosheath, and induced magnetosphere—using only ion omnidirectional energy spectra measured by the MAVEN Solar Wind Ion Analyzer (SWIA). Two neural network architectures are evaluated: a multilayer perceptron (MLP) and a convolutional neural network (CNN) that incorporates short temporal sequences. Our results show that the CNN can reliably distinguish the three plasma regions, whereas the MLP struggles to separate the solar wind and magnetosheath. Therefore, the CNN-based approach provides an efficient and accurate framework for large-scale plasma region identification at Mars and can be readily applied to future planetary missions. † † journal: AGU Journals † † authors: Yilan Qin, Chuanfei Dong, Hongyang Zhou, Chi Zhang, Kaichun Xu, Jiawei Gao, Simin Shekarpaz, Xinmin Li, Liang Wang Department of Electrical and Computer Engineering, Boston University, Boston, MA 02215, USA Center for Space Physics, Boston University, Boston, MA 02215, USA School of Natural Sciences, Institute for Advanced Study, Princeton, NJ 08540, USA Department of Mathematics and Statistics, Boston University, Boston, MA 02215, USA keypoints Machine learning models using MAVEN ion energy spectra are developed to automatically classify plasma regions at Mars. A CNN exploiting short-timescale spectral continuity achieves \sim 95% accuracy and outperforms an MLP. The CNN approach enables efficient identification of plasma regions under varying solar wind conditions for MAVEN and Mars missions. Plain Language Summary Mars does not have a global intrinsic magnetic field like Earth. As a result, its upper atmosphere interacts directly with the solar wind—a stream of fast-moving plasma continuously emitted from the Sun—along with the magnetic field carried within it. This interaction forms a global induced magnetosphere around Mars. Because the solar wind is highly variable, the electromagnetic fields and plasma environment around Mars are also highly variable and typically consist of several distinct plasma regions. Identifying these regions is important for scientists studying how the solar wind interacts with Mars and the associated magnetospheric processes. To help address this problem, we develop a machine-learning tool that can automatically identify plasma regions using observations from the MAVEN spacecraft. By comparing with ground-truth results, we show that the developed machine learning tool is accurate and robust, and it can also be readily applied to future planetary missions such as ESCAPADE. 1 Introduction Unlike Earth, Mars lacks a global intrinsic dipole magnetic field, and therefore the solar wind can directly interact with the upper atmosphere and ionosphere. This direct interaction allows the solar wind to efficiently remove atmospheric ions and is considered an important driver of atmospheric escape at Mars Dong et al. (2014) ; Dong et al. (2015) ; Dong et al. (2018) ; Dong et al. (2018) ; Jakosky et al. (2018) ; Lillis et al. (2015) ; Ramstad et al. (2018) ; Zhang et al. (2025) . The solar wind interaction with Mars generates an induced magnetosphere through the draping and pileup of the interplanetary magnetic field (IMF) around the planet Luhmann et al. (2004) ; Halekas et al. (2021) ; Ramstad et al. (2020) ; Zhang et al. (2022) . This induced magnetosphere includes major plasma boundaries and regions such as the bow shock, magnetosheath, and induced magnetosphere, forming a space environment that in some respects resembles the intrinsic magnetospheres of magnetized planets Nagy et al. (2004) . However, unlike intrinsic magnetospheres, the Martian induced magnetosphere is highly sensitive to upstream conditions, such as the IMF and solar wind dynamic pressure Cheng et al. (2025) . Variations in these external drivers can significantly modify the electromagnetic environment around Mars, thereby influencing planetary ion acceleration, transport, and escape Dubinin et al. (2017) ; Inui et al. (2018) ; Zhang et al. (2024) , as well as electron dynamics Zhang et al. (2025) . Therefore, studies of solar wind interaction with Mars, atmospheric escape, and related plasma processes must explicitly consider upstream solar wind conditions. However, due to the limited spatial and temporal coverage of spacecraft observations, continuous measurements of the upstream solar wind at Mars are generally unavailable. Consequently, researchers must manually identify time intervals when the spacecraft is located in the upstream region in order to determine the background solar wind conditions. This manual procedure significantly reduces the efficiency of data analysis. To address this limitation, several automatic methods have been developed to distinguish plasma regions around Mars. Halekas et al. (2017) proposed an approach based on ion bulk parameters and magnetic field measurements. Nemec et al. (2020) developed an empirical model using ion bulk parameters, which requires upstream solar wind parameters for normalization. Building on the work of Nemec et al. (2020) , Linzmayer et al. (2024) developed a machine-learning approach for automatic region identification. However, these methods generally rely on multiple input parameters, such as magnetic field measurements and ion bulk properties, and involve relatively complex processing procedures. Recent advances in machine learning have provided a powerful new approach for automatic plasma-region classification. In intrinsic magnetospheres, convolutional neural networks have been used to identify plasma regions in near-Earth space using MMS 3D particle energy distributions Olshevsky et al. (2021) , and random forest classifiers have been applied to classify magnetospheric regions and identify boundary crossings at Mercury using MESSENGER magnetometer and ephemeris data Hollman et al. (2026) . At Saturn, convolutional neural networks have been applied to detect major plasma boundaries, such as the bow shock and magnetopause, using Cassini magnetic field data and electron energy–time spectrograms Cheng et al. (2022) . These studies demonstrate that plasma regions and boundaries can be distinguished using relatively simple inputs, such as particle energy distributions, magnetic field measurements, and spacecraft position information, without requiring derived plasma parameters. This capability makes machine learning approaches particularly suitable for plasma-region classification. Motivated by these developments, we apply a machine learning approach to plasma-region classification at Mars. The Mars Atmosphere and Volatile EvolutioN (MAVEN) mission Jakosky et al. (2015) has provided more than a decade of observations spanning both upstream and downstream regions, together with high time-resolution plasma measurements, forming a comprehensive dataset for machine learning applications. Based on this dataset, we develop an automatic plasma-region classifier using a convolutional neural network (CNN) that incorporates temporal variations by using consecutive ion omnidirectional energy spectra. Unlike previous methods at Mars that rely on multiple parameters, the classifier uses only ion energy spectra as input. The model achieves an accuracy of approximately 95% and outperforms a multilayer perceptron (MLP) baseline. It enables automatic identification of the three major plasma regions—the upstream solar wind, magnetosheath, and induced magnetosphere—and provides an efficient tool for investigating solar wind interactions and plasma dynamics under varying conditions at Mars. 2 Methods 2.1 Experiment Design and Data Partitioning MAVEN was inserted into an elliptical orbit around Mars with a periapsis of approximately 150 km, an apoapsis of about 6,200 km, and an orbital period of 4.5 hours, enabling comprehensive sampling of the Martian space environment. During a typical orbit, MAVEN traverses multiple

Read the complete text