Neural Enhancement of the Traditional Wang–Sheeley–Arge Solar Wind Relation Pytorch ( A. Paszke et al. 2019 ) , pfsspy ( D. Stansby et al. 2020 ) , pysdtw ( M. Maghoumi et al. 2021 ) , timm ( R. Wightman 2019 ) , dtw ( T. Giorgino 2009 ) .

Prateek Mayank, Enrico Camporeale, Arpit K. Shrivastav, Thomas E. Berger, Charles N. Arge · 2025

Read Neural Enhancement of the Traditional Wang–Sheeley–Arge Solar Wind Relation Pytorch ( A. Paszke et al. 2019 ) , pfsspy ( D. Stansby et al. 2020 ) , pysdtw ( M. Maghoumi et al. 2021 ) , timm ( R. Wightman 2019 ) , dtw ( T. Giorgino 2009 ) . by Prateek Mayank, Enrico Camporeale, Arpit K. Shrivastav, Thomas E. Berger, Charles N. Arge (2025) 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-2509-06181.txt.

Introduction

Forecasting the solar wind from the Sun's magnetic map is improved by letting a neural network tune the physics, in a 2025 paper led by Prateek Mayank with four co-authors. The Wang-Sheeley-Arge model has anchored operational forecasts for nearly twenty years, but its empirical parameters are fixed or hand-tuned, limiting adaptation. Their WSA+ keeps the relation and learns its settings.

First, a neural optimizer adjusts parameters for each solar rotation to match spacecraft observations inside a differentiable, physics-constrained pipeline. Then a network learns to predict those tuned maps straight from magnetogram features, so new conditions need no manual work. WSA+ beats the classic relation across quiet and active periods by roughly forty percent on average, and drops into existing heliospheric pipelines.

Catalog record

CREATORPrateek Mayank, Enrico Camporeale, Arpit K. Shrivastav, Thomas E. Berger, Charles N. Arge
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-06181

Browse the full library · About the Tower of Babel

Full text

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

Neural Enhancement of the Traditional Wang–Sheeley–Arge Solar Wind Relation Pytorch ( A. Paszke et al. 2019 ) , pfsspy ( D. Stansby et al. 2020 ) , pysdtw ( M. Maghoumi et al. 2021 ) , timm ( R. Wightman 2019 ) , dtw ( T. Giorgino 2009 ) . Prateek Mayank, Enrico Camporeale, Arpit K. Shrivastav, Thomas E. Berger, Charles N. Arge Abstract The Wang–Sheeley–Arge (WSA) model has been the cornerstone of operational solar wind forecasting for nearly two decades, owing to its simplicity and physics-based formalism. However, its performance is strongly dependent on several empirical parameters that are typically fixed or tuned manually, limiting its adaptability across varying solar conditions. In this study, we present a neural enhancement to the WSA framework (referred to as WSA + ) that systematically optimizes the empirical parameters of the WSA solar wind speed relation using in-situ observations within a differentiable physics-constrained pipeline. The approach operates in two stages: first, a neural optimizer adjusts WSA parameters independently for each Carrington Rotation to better match the observed solar wind data. Then, a neural network learns to predict these optimized speed maps directly from magnetogram-derived features. This enables generalization of the optimization process and allows inference for new solar conditions without manual tuning. The developed WSA + preserves the interpretability of the original relation while significantly improving the match with OMNI in-situ data across multiple performance metrics, including correlation and error statistics. It consistently outperforms the traditional WSA relation across both low and high solar activity periods, with average improvements of approximately 40 percent. By integrating data-driven learning with physical constraints, WSA + offers a robust and adaptable enhancement, with immediate utility as a drop-in replacement in global heliospheric modeling pipelines. show][email protected] I Introduction Space weather events driven by solar activity can severely impact satellites, aviation, communication systems, and electrical infrastructure on Earth. Among the various drivers, the solar wind is a continuous structured outflow of plasma from the Sun. It serves as a persistent background condition that modulates heliospheric dynamics and preconditions the near-Earth environment. It also strongly influences coronal mass ejection (CME) propagation ( P. Mayank et al. 2023 ; C. Kay et al. 2024 ) and CME-CME interactions ( P. Mayank et al. 2024 ; S. V. Thampi et al. 2025 ) . Accurate modeling of the ambient solar wind is therefore essential for both research and operational space weather forecasting. A widely adopted tool for ambient solar wind modeling is the Wang–Sheeley–Arge ( C. N. Arge 2003 , WSA;) model, which connects synoptic photospheric magnetic field observations to solar wind speed near the Sun via a potential field extrapolation and empirical relations. The WSA model forms the backbone of many forecasting pipelines, including WSA-ENLIL, which is used operationally by NOAA’s Space Weather Prediction Center ( M. A. Sharpe & S. A. Murray 2017 ) and the UK Met Office ( S. Bingham et al. 2018 ) . Its efficiency, simplicity, and empirical grounding have contributed to its continued relevance across multiple solar cycles. However, the empirical nature of WSA also presents key limitations. The model’s wind speed prediction relies on a set of tunable parameters, which are often manually adjusted to optimize performance for individual Carrington Rotations (CRs). In practice, these parameters vary significantly across CRs and solar conditions ( O. Issan et al. 2023 ) . Operational pipelines often adopt fixed average values, which limits adaptability and accuracy. Consequently, while per-CR parameter fitting can improve forecast accuracy, that approach remains incompatible with real-time forecasting workflows. Several efforts have been made to improve the WSA model by refining its empirical formulation, adjusting boundary conditions, or addressing parameter uncertainty. Some studies have also focused on calibrating the speed relation. For instance, S. L. McGregor et al. 2011 proposed a revised formula based on Helios data. While, H. A. Elliott et al. 2022 introduced empirical corrections using residual error maps tied to magnetic topology. Additionally, P. Riley 2015 re-examined the expansion-factor term in WSA, showing that its influence can be limited under certain conditions, thereby motivating refinements to the empirical speed relation. S. Kumar et al. 2025 and S. Majumdar et al. 2025 showed that varying the source surface height and using higher-quality magnetograms improves forecast accuracy. M. A. Reiss et al. 2020 developed an adaptive system that integrates in-situ observations to better constrain near-Sun conditions. Together, these studies reveal the importance of flexibility and tuning. However, most approaches rely on manual calibration or static assumptions. This highlights the need for a generalizable, data-driven enhancement of WSA that can adapt across solar conditions in an automated and interpretable way. In this work, we introduce a hybrid machine learning (ML) framework that overcomes this constraint by first optimizing WSA parameters per CR using a neural optimizer, and then training a neural network to generalize across CRs. The surrogate model predicts full 2D solar wind speed maps at 21.5 solar radius (0.1 AU), capturing structural variability while preserving physical interpretability. This hybrid ML approach allows to address the limitation of tunable WSA parameters through learning the generalized pattern of their optimized set of values across the solar activity. The structure of this paper is as follows: in Section II , we describe the two-stage modeling pipeline, including the neural optimizer and the neural network. In Section III , we present performance results across 129 Carrington Rotations and analyze spatial and time series validation. Section IV discusses implications for operational forecasting and limitations of the current approach. We conclude in Section V with a summary and outlook. II Methodology The Wang–Sheeley–Arge (WSA) model computes solar wind speed by relating it to magnetic topology derived from observed synoptic magnetograms. This mapping relies on the Potential Field Source Surface (PFSS) model, which assumes a current-free inner-corona and solves the elliptic Laplace equation to compute the scalar potential of the magnetic field ( M. D. Altschuler & G. Newkirk 1969 ) . From this solution, the flux-tube expansion factor at the source surface ( f_{s} ) and the minimum angular distance ( D ) to the coronal hole boundary are derived, both of which serve as key inputs to the empirical WSA wind speed relation: V_{wsa}=V_{min}+\frac{V_{max}}{(1+f_{s})^{\alpha}}\times\,\bigg(1-a1\,exp\,\bigg(-(D/w)^{\beta}\,\bigg)\bigg)^{a2} (1) This empirical formula expresses the predicted WSA solar wind speed ( V_{wsa} ) as a function of f_{s} and D , modulated by a set of tunable parameters: V_{min} , V_{max} , \alpha , \beta , w , a1 and a2 . These parameters collectively determine the sensitivity of the model to changes in coronal magnetic topology, and effectively control how strongly the solar wind speed responds to changes in f_{s} and D . While fixed default parameter values are often used in operational settings, such static configurations do not adapt to varying solar conditions. In this study, we adopt a standard default set: V_{min}=250 , V_{max}=750 , \alpha=0.222 , \beta=1.25 , w=0.028 , a1=0.8 and a2=3 , consistent with values used in previous studies ( S. L. McGregor et al. 2011 ; M. A. Reiss et al. 2019 , e.g.,) and several contemporary solar wind modeling ( J. Pomoell & S. Poedts 2018 ; P. Mayank et al. 2022 , e.g.,) . Figure 1: Schematic diagram of the undertaken method

Read the complete text