Scholarlink

<p>Increasing Global Navigation Satellite System-reflectometry (GNSS-R) satellites like Cyclone Global Navigation Satellite System (CYGNSS), Fengyun-3E (FY-3E), and Tianmu-1 (TM) boost data volume, but single systems lack short-term spatial coverage, limiting continuous surface monitoring. Conventional interpolation strategies may exhibit limited accuracy. This study proposes optimized multi-spaceborne GNSS-R data fusion to enhance temporal resolution, accuracy, and coverage. This study employs two categories of methods to fuse data from the CYGNSS, FY-3E, and Tianmu-1 satellites: statistical methods (linear regression and triple collocation, TC) and machine learning. The linear fitting of experimental data between CYGNSS and FY-3E yielded a coefficient of determination (R²) of 0.8761 and a Root Mean Square Error (RMSE) of 1.9827 dB. For CYGNSS and Tianmu-1 data, the R² value was 0.7962, with an RMSE of 2.0969 dB. CYGNSS data was used as the target output, while FY-3E and Tianmu-1 data served as input sources. Based on the linear relationship between reflectivity and Soil Moisture Active Passive (SMAP) soil moisture (SM), the feasibility of data fusion is evaluated by analyzing changes in the correlation coefficients between the fused data and SMAP data before and after fusion. Using data from the year of 2023 for model construction, daily and five-day fusion analyses were conducted on data from January to May, 2024. Results showed an increase of 43% in effective observation coverage and a significant improvement in correlation after daily fusion. This work shows that the fusion of multi-spaceborne GNSS-R data establishes a promising framework for future high spatiotemporal resolution observations, thereby enhancing both the timeliness and accuracy of related research endeavors.</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

Multi-spaceborne GNSS-R data fusion based on machine learning and statistical methods

  • 期刊论文

  • 发表日期：

    2025年12月22日

查看全文
  • Xiangyue Li,
  • Qingyun Yan,
  • Xudong Tong

摘要

Increasing Global Navigation Satellite System-reflectometry (GNSS-R) satellites like Cyclone Global Navigation Satellite System (CYGNSS), Fengyun-3E (FY-3E), and Tianmu-1 (TM) boost data volume, but single systems lack short-term spatial coverage, limiting continuous surface monitoring. Conventional interpolation strategies may exhibit limited accuracy. This study proposes optimized multi-spaceborne GNSS-R data fusion to enhance temporal resolution, accuracy, and coverage. This study employs two categories of methods to fuse data from the CYGNSS, FY-3E, and Tianmu-1 satellites: statistical methods (linear regression and triple collocation, TC) and machine learning. The linear fitting of experimental data between CYGNSS and FY-3E yielded a coefficient of determination (R²) of 0.8761 and a Root Mean Square Error (RMSE) of 1.9827 dB. For CYGNSS and Tianmu-1 data, the R² value was 0.7962, with an RMSE of 2.0969 dB. CYGNSS data was used as the target output, while FY-3E and Tianmu-1 data served as input sources. Based on the linear relationship between reflectivity and Soil Moisture Active Passive (SMAP) soil moisture (SM), the feasibility of data fusion is evaluated by analyzing changes in the correlation coefficients between the fused data and SMAP data before and after fusion. Using data from the year of 2023 for model construction, daily and five-day fusion analyses were conducted on data from January to May, 2024. Results showed an increase of 43% in effective observation coverage and a significant improvement in correlation after daily fusion. This work shows that the fusion of multi-spaceborne GNSS-R data establishes a promising framework for future high spatiotemporal resolution observations, thereby enhancing both the timeliness and accuracy of related research endeavors.