Application of RNN on GNSS Reflectometry Sea level monitoring

ABSTRACT

Global Navigation Satellite System reflectometry (GNSS-R) can be used to monitor the sea level with the signal-to-noise ratio (SNR) using the spectral analysis approach where the additional frequencies in SNR lead to small errors. This study is the first attempt to apply the Recurrent Neural Networks (RNN) approach to reduce the effect of additional frequencies in SNR. The output from the RNN approach is processed by the spectral analysis to estimate the frequency. To analyse the number of effective results, all results are processed by the outlier removal method which removes the outliers only depending on the results under low elevation angles. The data set used in the experiment is sampled from the SC02 GPS station. The experiment is set up by using abnormal elevation angle ranges to observe the performance. The results from the RNN-based method, detrend SNR method and EMD-based method are compared to the reference sea level data. RNN models are trained differently in three experiments based on 7-day historical sea level data, 7-day tide prediction data and 1-month tide prediction data. The results show that the RNN increases the temporal resolutions for all experiments and provides sufficient accuracy as compared to the detrend SNR method and EMD-based method. Using tide prediction data instead of sea level data as the training data improves temporal resolutions and correlations. RNN can be applied as the data preparation and denoising method for SNR spectral analysis sea level monitoring. Compared to the EMD-based method, the RNN-based denoising method is more suitable for short SNR data records.

KEYWORDS:

• Global Navigation Satellite System reflectometry (GNSS-R)
• signal to noise ratio (SNR)
• sea level monitoring
• Recurrent Neural Network (RNN)

Acknowledgement(s)

The GNSS data were obtained from the Plate Boundary Observatory operated by UNAVCO for EarthScope. The Friday Harbor tide gauge data were provided by the National Oceanic and Atmospheric Administration (https://tidesandcurrents.noaa.gov/). Thanks are due to Ms. Zhaoyi Zhang for assistance with the writing and editing.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the corresponding author [BF. Guo] upon reasonable request by email: [email protected].

Additional information

Funding

This work was financially supported by the Open Fund Project of Key Laboratory of Ocean Observation Technology, MNR (Grant No. 2021klootB05) and National Natural Science Foundation of China (Grant No.41704010).