A machine learning technique is implemented for retrieving space-borne Global Navigation Satellite System Reflectometry (GNSS-R) wind speed. Conventional approaches commonly fit a function in a predefined form to matchup data in a least-squares (LS) sense, mapping GNSS-R observations to wind speed. In this study, a feedforward neural network is trained for TechDemoSat-1 (TDS-1) wind speed inversion. The input variables, along with the derived bistatic radar cross-section $σ⁰, are selected after investigating the wind speed dependence and the model performance. When compared to an LS-based approach, the derived model shows a significant improvement of 20% in the root mean square error (RMSE). The proposed neural network demonstrates an ability to model a variety of effects degrading the retrieval accuracy such as the different levels of the effective isotropic radiated power (EIRP) of GPS satellites. For example, the derived Mean Absolute Error (MAE) of the satellite with SVN 34 is decreased by 32% using the machine-learning-based approach.

For the first time, a rain signature in Global Navigation Satellite System Reflectometry (GNSS‐R) observations is demonstrated. Based on the argument that the forward quasi‐specular scattering relies upon surface gravity waves with lengths larger than several wavelengths of the reflected signal, a commonly made conclusion is that the scatterometric GNSS‐R measurements are not sensitive to the surface small‐scale roughness generated by raindrops impinging on the ocean surface. On the contrary, this study presents an evidence that the bistatic radar cross section σ⁰ derived from TechDemoSat‐1 data is reduced due to rain at weak winds, lower than ≈ 6 m/s. The decrease is as large as ≈ 0.7 dB at the wind speed of 3 m/s due to a precipitation of 0–2 mm/hr. The simulations based on the recently published scattering theory provide a plausible explanation for this phenomenon which potentially enables the GNSS‐R technique to detect precipitation over oceans at low winds.

Reliable, accurate, and timely information about oceans is important for many applications, including water resource management, hydrological cycle monitoring, environmental studies, agricultural and ecosystem health applications, economy, and the overall health of the environment. In this regard, remote sensing (RS) systems offer exceptional advantages for mapping and monitoring various oceanographic parameters with acceptable temporal and spatial resolutions over the oceans and coastal areas. So far, different methods have been developed to study oceans using various RS systems. This urges the necessity of having review studies that comprehensively discuss various RS systems, including passive and active sensors, and their advantages and limitations for ocean applications. In this article, the goal is to review most RS systems and approaches that have been worked on marine applications. This review paper is divided into two parts. Part 1 is dedicated to the passive RS systems for ocean studies. As such, four primary passive systems, including optical, thermal infrared radiometers, microwave radiometers, and Global Navigation Satellite Systems, are comprehensively discussed. Additionally, this article summarizes the main passive RS sensors and satellites, which have been utilized for different oceanographic applications. Finally, various oceanographic parameters, which can be retrieved from the data acquired by passive RS systems, along with the corresponding methods, are discussed.

This study presents the development and a systematic evaluation study of GNSS reflectometry wind speeds. After establishing a wind speed retrieval algorithm, UK TechDemoSat-1 (TDS-1) derived winds, from May 2015 to July 2017, are compared to the Advanced Scatterometer (ASCAT). ERA-Interim wind fields of the European Centre for Medium-range Weather Forecasts (ECMWF) and in situ observation from Tropical Atmosphere Ocean buoy array in the Pacific are taken as reference. One-year averaged TDS-1 global winds demonstrate small differences with ECMWF in a majority of areas as well as discuss under- and overestimations. The pioneering TDS-1 winds demonstrate a root-mean-squared error (RMSE) and bias of 2.77 and $-$ 0.33 m/s, which are comparable to the RMSE and bias derived by ASCAT winds, as large as 2.31 and 0.25 m/s, respectively. Using buoys measurements as reference, RMSE and bias of 2.23 and $-$ 0.03 m/s for TDS-1 as well as 1.40 and $-$ 0.68 m/s for ASCAT are obtained. Utilizing rain microwave–infrared estimates of the Tropical Rainfall Measuring Mission, rain-affected observation of both ASCAT and TDS-1 are collected and evaluated. Although ASCAT winds show a significant performance degradation resulting in an RMSE and bias of 3.16 and 1.03 m/s, respectively, during rain condition, TDS-1 shows a more reliable performance with an RMSE and bias of 2.94 and $-$ 0.21 m/s, respectively, which indicates the promising capability of GNSS forward scattering for wind retrievals during rain. A decrease in TDS-1-derived bistatic radar cross sections during rain events, at weak winds, is also demonstrated.

The NASA Cyclone GNSS (CYGNSS) mission provides one Delay Doppler Map (DDM) per second along observational tracks. To account for spatiotemporal correlations within adjacent DDMs in a track, a deep hybrid CNN-LSTM model is proposed for wind speed prediction. The model combines convolutional and pooling layers to extract features from DDMs in one track, which are then processed by LSTM as a sequence of data. This method leads to a test RMSE of 1.84 m/s. The track-wise processing approach outperforms the architectures that process the DMMs individually, namely based on Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN), and a network based solely on fully connected (FC) layers, as well as the official retrieval algorithm of the CYGNSS mission with RMSEs of 1.92 m/s, 1.92 m/s, 1.93 m/s, and 1.90 m/s respectively. Expanding on the CNN-LSTM model, the CNN-LSTM+ model is proposed with additional FC layers parallel with convolutional and pooling layers to process ancillary data. It achieves a notable reduction in test RMSE to 1.49 m/s, demonstrating successful implementation. This highlights the significant potential of track-wise processing of GNSS-R data, capturing spatiotemporal correlations between DDMs along a track. The hybrid deep learning model processing the data sequentially in one track learns these dependencies effectively, improving the accuracy of wind speed predictions.