Machine Learning Methods for Spaceborne GNSS-R Sea Surface Height Measurement From TDS-1

Abstract

Sea surface height (SSH) retrieval based on spaceborne global navigation satellite system reflectometry (GNSS-R) usually uses the GNSS-R geometric principle and delay-Doppler map (DDM). The traditional method condenses the DDM information into a single scalar measure and requires error model correction. In this paper, the idea of using machine learning methods to retrieve SSH is proposed. Specifically, two widely-used methods, Principal Component Analysis combined with Support Vector Regression (PCA-SVR) and Convolution Neural Network (CNN), are used for verification and comparative analysis based on the observation data provided by Techdemosat-1 (TDS-1). According to the DDM inversion method, ten features from TDS-1 Level 1 data are selected as inputs; The SSH verification model based on the Danmarks Tekniske Universitet (DTU) 15 ocean wide mean SSH model and the DTU global ocean tide model is used as output verification of SSH. For the hyperparameters in the machine learning model, a grid search strategy is used to find the optimal values. By analyzing the TDS-1 data from 31 GPS satellites, the mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R 2 ) of the PCA-SVR inversion model are 0.61 m, 1.72 m and 99.56%, respectively; and the MAE, RMSE and R 2 of the CNN inversion model is 0.71 m, 1.27 m and 99.76%, respectively. In addition, the time required to train the PCA-SVR and CNN inversion models is also analyzed. Overall, the technique proposed in this paper can be confidently applied to SSH inversion based on TDS-1 data.

Global Sea Surface Height Measurement From CYGNSS Based on Machine Learning

“…The energy is affected by the delay effect and the Doppler effect, which is caused by the relative high-speed movement between the receiver and the transmitter, the dynamically changing direction, and the considerable satellite altitude and spacing. The integral Doppler reflects the magnitude and direction of the change in geometric distance [21]. CYGNSS L1 data directly provides the location of the BRCS DDM specular reflection point Doppler column and the peak Doppler column, as well as the vector velocities of the transmitter and receiver satellites.…”

“…One problem of ML methods is the relatively poor generalization ability. By using appropriate and valid data as well as adding DDM-related features, CYGNSS data has a better generalization performance compared to TDS-1 data in SSH inversion [21].…”

“…Compared with the DTU15 model, the RMSE was about 0.23 m and the correlation coefficient was about 0.75. In the same year, Zhang et al [21] analyzed TDS-1 data distributed over a vast sea area. They proposed two spaceborne SSH inversion models, PCA-SVR and CNN, and the inversion results of the models had an average MAE below 1 m compared to DTU15.…”

“…In contrast to airborne or ground-borne SSH inversion analysis, which lacks space-time diversity, this paper utilizes the high-resolution spaceborne GNSS-R data provided by CYGNSS, to analyze a wide range of sea areas. Compared with TDS-1 data [21], CYGNSS has 8 GNSS-R receiving satellites and a larger number of valid samples, so CYGNSS is expected to have better accuracy, especially better generalization performance, which will be discussed below.…”

Flying Target Detection Technology Based on GNSS Multipath Signals

“…Furthermore, in recent years, national navigation satellite systems have been developed rapidly, the number of global navigation satellites has become more abundant, and remote sensing technology using GNSS signals has become increasingly advanced. At present, this technology has realized engineering applications in the fields of sea surface altitude measurement [ 8 , 9 ], effective wave height measurement at sea level [ 10 , 11 ], the remote sensing of wind fields at sea level [ 12 , 13 , 14 , 15 ], the remote sensing of seawater salinity [ 16 , 17 , 18 ], and tidal detection [ 19 , 20 , 21 ]. In land surface remote sensing, numerous breakthroughs have also been made for measuring quantities such as soil moisture [ 22 , 23 , 24 ], snow thickness [ 25 ], and vegetation cover [ 26 ].…”