• A novel Gated Memory Network SSH Inversion Model (GMSIM) is proposed, integrating LSTM with GNSS-IR and meteorological data. • GMSIM significantly outperforms conventional GNSS-IR, reducing RMSE by 16.3% and increasing PCC by 30.0% on average. • Validated across four global coastal stations, GMSIM achieves a mean RMSE of 12.9 cm and a mean PCC of 0.93, demonstrating high reliability. Sea surface height (SSH) is a critical indicator of oceanic changes, and precise monitoring is essential for coastal, low-lying areas, ecosystems, and economic development. Global Navigation Satellite System Interferometry Reflectometry (GNSS-IR) can monitor SSH changes near GNSS station. However, traditional GNSS-IR sea surface height measurement is limited by multiple external error sources, making it difficult to meet the requirements of high-precision sea surface height altimetry. To address this, this paper proposes a Gated Memory Network SSH Inversion Model (GMSIM) based on Long Short-Term Memory (LSTM), which is driven by GNSS-IR parameters and meteorological inputs, and validated against tide gauge observations. To verify the performance of the model, four coastal GNSS stations (TAR0, PTLD, HKQT, and SC02) are analyzed. Compared with tide gauge value, GMSIM achieves a mean Root Mean Square Error ( RMSE ) of 12.9 cm (minimum 10.6 cm at PTLD) and a mean Pearson Correlation Coefficient ( PCC ) of 0.93 (maximum 0.96 at SC02), confirming its high reliability. SSH is retrieved using GMSIM and conventional GNSS-IR, and the results are compared with tide gauge data. GMSIM reduced Root Mean Square Error ( RMSE ) by 12.4% and increased Pearson Correlation Coefficient ( PCC ) by 2.2% on average. Therefore, GMSIM can improve the retrieve accuracy of traditional GNSS-IR SSH. In summary, this study provides a new retrieve idea for high-precision GNSS-IR altimetry, which has important theoretical and application value for marine environment monitoring and marine science research.
Shen et al. (2026) studied this question.