The global navigation satellite system (GNSS) radio occultation (RO) technique is an atmospheric sounding technique that originated in the 1990s. The data provided by this approach are playing a consistently significant role in atmospheric research and related applications. This paper mainly summarizes the applications of RO to numerical weather prediction (NWP) generally and specifically for tropical cyclone (TC) forecast and outlines the prospects of the RO technique. With advantages such as high precision and accuracy, high vertical resolution, full-time and all-weather, and global coverage, RO data have made a remarkable contribution to NWP and TC forecasts. While accounting for only 7% of the total observations in European Centre for Medium-Range Weather Forecasts’ (ECMWF’s) assimilation system, RO has the fourth-largest impact on NWP. The greater the amount of RO data, the better the forecast of NWP. In cases of TC forecasts, assimilating RO data from heights below 6 km and from the upper troposphere and lower stratosphere (UTLS) region contributes to the forecasting accuracy of the track and intensity of TCs in different stages. A statistical analysis showed that assimilating RO data can help restore the critical characteristics of TCs, such as the location and intensity of the eye, eyewall, and rain bands. Moreover, a non-local excess phase assimilation operator can be employed to optimize the assimilation results. With denser RO profiles expected in the future, the accuracy of TC forecast can be further improved. Finally, future trends in RO are discussed, including advanced features, such as polarimetric RO, and RO strategies to increase the number of soundings, such as the use of a cube satellite constellation.

Based on deep learning, this paper proposes a new hybrid neural network model, a recurrent deep neural network using a feature attention mechanism (FA-RDN) for GNSS-R global sea surface wind speed retrieval. FA-RDN can process data from the Cyclone Global Navigation Satellite System (CYGNSS) satellite mission, including characteristics of the signal, spatio-temporal, geometry, and instrument. FA-RDN can receive data extended in temporal dimension and mine the temporal correlation information of features through the long-short term memory (LSTM) neural network layer. A feature attention mechanism is also added to improve the model’s computational efficiency. To evaluate the model performance, we designed comparison and validation experiments for the retrieval accuracy, enhancement effect, and stability of FA-RDN by comparing the evaluation criteria results. The results show that the wind speed retrieval root mean square error (RMSE) of the FA-RDN model can reach 1.45 m/s, 10.38%, 6.58%, 13.28%, 17.89%, 20.26%, and 23.14% higher than that of Backpropagation Neural Network (BPNN), Recurrent Neural Network (RNN), Artificial Neural Network (ANN), Random Forests (RF), eXtreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR), respectively, confirming the feasibility and effectiveness of the designed method. At the same time, the designed model has better stability and applicability, serving as a new research idea of data mining and feature selection, as well as a reference model for GNSS-R-based sea surface wind speed retrieval.

Global Navigation Satellite System Reflectometry (GNSS-R) is a burgeoning remote sensing observation technology that can retrieve global sea surface wind speeds using satellite signals reflected from the sea surface. The improvement of data quality and the accumulation of data volume of this technology provides data support for constructing interdisciplinary-based retrieval models. This article constructs a hybrid deep neural network model based on deep learning for wind speed retrieval, which can receive and perform feature mining on the entire delay waveform while simultaneously supporting multiple auxiliary features input and achieving joint fitting. Then a bias correction method based on cumulative distribution function (CDF) matching is introduced to mitigate bias, especially at high wind speeds. We verify the contribution of different attribute features in wind speed retrieval by designing a feature ablation analysis. The fluctuation variation of the retrieval accuracy in the time dimension and the retrieval results distribution in space are compared and analyzed. The root mean square error (RMSE) of retrieval results is 1.486m/s and can reach 1.399m/s under the 94.25% wind speed condition. After bias correction based on CDF matching, the retrieval accuracy at high wind speed is improved by 7.19%. Besides, this model also has good temporal stability and can reproduce large-scale wind fields while effectively mitigating retrieval bias on a global scale, showing great potential for operational applications.

In radio occultation (RO) data processing and data assimilation, the forward model (FM) is used to calculate bending angle (BA) from refractivity (N). The accuracy and precision of forward modeled BA are affected by refractivity profiles and FM methods, including Abel integral algorithms (direct, exp, exp_T, linear) and methods of interpolating refractivity during integral (log-cubic spline and log-linear). Experiment 1 compares these forward model methods by comparing the difference and relative difference (RD) of the experimental value (forward modeled ECMWF analysis) and the true value (BA of FY3D RO data). Results suggested that the exp with log-cubic spline (log-cubic) interpolation is the most accurate FM because it has better integral accuracy (less than 2%) to inputs, especially when the input is lower than an order of magnitude of 1 × 10⁻² (that is, above 60 km). By contrast, the direct induced a 10% error, and the improvement of exp T to exp is limited. Experiment 2 simulated the exact errors of an FM (exp) based on inputs on different vertical resolutions. The inputs are refractivity profiles on model levels of three widely used analyses, including ECMWF 4Dvar analysis, final operational global analysis data (FNL), and ERA5. Results demonstrated that based on exp and log-cubic interpolation, BA on model level of ECMWF 4Dvar has the highest accuracy, whose RD is 0.5% between 0–35 km, 4% between 35–58 km, and 1.8% between 58–80 km. By contrast, the other two analyses have low accuracy. This paper paves the way to better understanding the FM, and simulation errors on model levels of three analyses can be a helpful FM error reference.