Synthetic Aperture Radar (SAR), with its ability to function under any weather conditions and at any time of day, along with multi-polarization and frequent revisit capabilities, plays a crucial role in flood monitoring. However, SAR images face challenges such as coherent speckle noise, feature mixing, terrain undulation, and adverse weather, making flood monitoring difficult. To address these challenges, this paper proposes a high-precision flood mapping method from Sentinel-1 dual-polarization SAR data. We begin by generating false-color images through polarization combination and apply them to a multiscale segmentation approach, overcoming the limitations of single-polarization scattering and effectively reducing speckle noise. Digital elevation model and reference water datasets are integrated into the segmentation process to mask terrain shadowing and permanent water. To reduce feature mixing effects, the optimal SAR image with minimal feature mixing is selected for flood mapping using the Gaussian Mixture Model. In the subsequent two-step classification process, fuzzy sets of texture features are incorporated to assist in categorizing uncertain regions, further reducing interference from feature mixing and enhancing flood recognition accuracy. Additionally, integrating pixel-level and object-level analyses minimizes errors caused by improper segmentation. The proposed method is compared with several well-established algorithms, and the results demonstrate that our method outperforms the others in flood mapping accuracy. Analysis of years of flooding on the Leizhou Peninsula shows that Sentinel-1 SAR has the potential to effectively monitor the occurrence and development of floods.

Severe convective weather is hugely destructive, causing significant loss of life and social and economic infrastructure. Based on the U-Net network with the attention mechanism, the recurrent convolution, and the residual module, a new model is proposed named ARRU-Net (Attention Recurrent Residual U-Net) for the recognition of severe convective clouds using the cloud image prediction sequence from FY-4A data. The characteristic parameters used to recognize severe convective clouds in this study were brightness temperature values TBB₉, brightness temperature difference values TBB₉−TBB₁₂ and TBB₁₂−TBB₁₃, and texture features based on spectral characteristics. This method first input five satellite cloud images with a time interval of 30 min into the ARRU-Net model and predicted five satellite cloud images for the next 2.5 h. Then, severe convective clouds were segmented based on the predicted image sequence. The root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and correlation coefficient (R²) of the predicted results were 5.48 K, 35.52 dB, and 0.92, respectively. The results of the experiments showed that the average recognition accuracy and recall of the ARRU-Net model in the next five moments on the test set were 97.62% and 83.34%, respectively.

The interferometric microwave radiometer (IMR) utilizes interferometric synthetic aperture technique to achieve high spatial resolution in low-frequency microwave remote sensing, addressing the challenges of deploying large-scale passive sensors in space. IMR measures spatial harmonics of scene brightness temperature, known as visibility, which are then used in inversion algorithms to reconstruct the target brightness temperature. In 1-D IMR, interferometric synthetic aperture technique is applied only in the cross-track direction, resulting in higher resolution compared with the coarser along-track direction determined by the real antenna aperture. Current research focuses on cross-track inversion, which has yielded promising results; however, the low along-track resolution remains a significant limitation for its overall application. This paper introduces the BG-inspired 1-D IMR resolution enhancement method, inspired by real aperture microwave radiometer techniques, to address along-track resolution limitation. The study utilizes the L-band 1-D IMR of the Active Passive Microwave Imager (MICAP) aboard the Chinese Ocean Salinity Satellite as an example. Results from synthetic test images and hardware-in-the-loop simulation demonstrate that the proposed method enhances along-track resolution and provides the flexibility to optimize for either higher image quality or radiometric resolution comparable to the traditional 1-D IMR inversion method. Additionally, it improves the accuracy of salinity measurements in coastal areas.

Deep Neural Network (DNN), equipped with powerful distinctive generalization ability, is one of the most popular retrieval tools for processing ocean remote sensing data from spaceborne microwave radiometers. Considering the signal received by microwave radiometers at different frequencies changes with atmospheric effects, such as rainfall and water vapor in both horizontal (H) and vertical (V) polarization (pol.), and differential signals between H and V pol, these physical factors cannot be ignored in high wind speed retrieval. This study focuses on the integration of physically-driven statistical functions with a residual neural network (ResNet) to derive a specialized type of deep neural network (DNN) framework for high wind speed retrieval. Specifically, a Dynamic Contribution-Matrix ResNet (DCM-ResNet) model is proposed, which dynamically adjusts the contribution ( C ) matrix coefficients within the model. These coefficients, reflecting the physical information of radiation transfer, are used to automatically regulate the model’s estimation of high wind speeds. Experiments are conducted using brightness temperatures from the Advanced Microwave Scanning Radiometer-2 and wind speeds from the Stepped-Frequency Microwave Radiometer. The experimental results show that the maximum wind speed can reach up to 60 m/s with a root mean square error (RMSE) of 2.92 m/s. Additionally, the wind speed RMSE remains stable within 6 m/s as the rainfall rate increases. The contributions of different frequency bands to the C -matrix results are closely related to radiative transfer.

Synthetic Aperture Radar (SAR) has unique advantages in sea surface wind field retrieval due to its all-weather observation capability and high spatial resolution. However, the effectiveness of wind measurement is vulnerable to limitations such as saturation of backscattering signals from the sea surface and attenuation of heavy rainfall under extreme weather conditions. Considering that the response of physical factors (e.g., radar backscattering coefficient at different polarizations) to wind varies in different wind speed ranges, in this study, based on 36 scenes of dual-polarized Sentinel-1 SAR images of hurricanes from 2016 to 2023, and using wind speed observations from the airborne Stepped-Frequency Microwave Radiometer (SFMR) as ground truth values, we proposed a partition and fusion residual transfer network (PFRTNet) model. In the training process, two separate residual neural networks were pretrained under low-to-medium wind speeds (< 30 m/s) and high winds (≥ 30 m/s), and combined with the idea of transfer learning. An attention mechanism was then introduced to achieve adaptive feature fusion. The optimal model inputs were determined by feature importance analysis and sensitivity experiments, which consists of 14 features including SAR measured physical parameters, image texture features, hurricane morphological information, as well as environmental and geographic factors. The PFRTNet model can effectively alleviate the underestimation of high wind speed caused by sample imbalance, and significantly improve the accuracy of hurricane wind retrieval. Independent evaluation results based on an additional 9 SAR images demonstrate that the algorithm achieves a root mean square error (RMSE) of 2.81 m/s under high wind conditions. Compared with other machine learning methods and traditional empirical algorithms, PFRTNet has significant performance advantages, with a RMSE reduction of approximately 2.0 m/s or more, confirming its robustness in retrieving hurricane wind speeds and finer-scale structural features.

Correction for the impact of direct Sun contamination is a crucial task in the data processing of interferometric microwave radiometer (IMR). The evident presence of solar radiation is observed in the brightness temperature images derived from the Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) payload onboard the Soil Moisture and Ocean Salinity (SMOS) satellite, significantly impacting the data quality to retrieve sea surface salinity. This paper introduces a novel sinusoid correction method for correcting the direct Sun contamination. By leveraging the characteristics of the solar disk, the proposed method simulates and compensates for the contribution of direct solar impact in the spatial frequency domain based on the response pattern of small point sources within the solar disk. The proposed method exhibits a reduced dependency on the precise solar position information and demonstrates resistance to radio frequency interference, and validations through a simulated IMR and data from in-orbit SMOS confirm the reduction of the direct solar impact in brightness temperature images. The proposed sinusoid correction method outperforms the single and multiple source methods, used in the SMOS operational data processing, especially for the central regions around the location of direct Sun.