Scholarlink

<p>Thermospheric density models are essential for satellite operations, yet they exhibit significant discrepancies during geomagnetic storms. This study evaluates the performance of the Naval Research Laboratory Mass Spectrometer and Incoherent Scatter Radar Exosphere (NRLMSIS) 2.1 model during geomagnetic disturbances in Solar Cycle 25’s ascending phase (2021–2024) using Swarm-A, -B, -C, and GRACE-FO satellite measurements. Tree-based machine learning algorithms were employed to analyze model residuals and identify key factors influencing prediction accuracy. Random Forest Regression provided comparatively better performance among tested methods (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\hbox {R}^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mtext>R</mtext> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>&#xa0;= 0.52), explaining about half of the residual variance. Feature importance analysis revealed solar flux (F10.7, importance = 0.384), altitude (0.221), and Dst index (0.127) as the top three factors in this test. These results suggest that refined empirical formulations may improve the representation of storm-time density variations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Tree-Based Analysis of Geophysical and Orbital Influences on NRLMSIS 2.1 Residuals during Geomagnetic Storms

  • 期刊论文

  • 发表日期：

    2025年11月20日

查看全文
  • Patapong Panpiboon

摘要

Thermospheric density models are essential for satellite operations, yet they exhibit significant discrepancies during geomagnetic storms. This study evaluates the performance of the Naval Research Laboratory Mass Spectrometer and Incoherent Scatter Radar Exosphere (NRLMSIS) 2.1 model during geomagnetic disturbances in Solar Cycle 25’s ascending phase (2021–2024) using Swarm-A, -B, -C, and GRACE-FO satellite measurements. Tree-based machine learning algorithms were employed to analyze model residuals and identify key factors influencing prediction accuracy. Random Forest Regression provided comparatively better performance among tested methods ( \(\hbox {R}^2\) R 2  = 0.52), explaining about half of the residual variance. Feature importance analysis revealed solar flux (F10.7, importance = 0.384), altitude (0.221), and Dst index (0.127) as the top three factors in this test. These results suggest that refined empirical formulations may improve the representation of storm-time density variations.