Yongchao Zhu

This study proposes a method for retrieving Antarctic sea ice concentration (SIC) with spatiotemporal consistency constraints, utilizing multi-mode GNSS-R data acquired by the Chinese Tianmu-1 (TM-1) satellite constellation. A deep learning architecture integrating Neural Networks (ResCNet) is developed. The model employs a dual-stream input mechanism to effectively combine high-dimensional texture features from Delay-Doppler Maps (DDMs) with auxiliary GNSS-R features. A weighted mean square error loss function is introduced to enhance sensitivity in marginal ice zones and extreme SIC regions. Furthermore, two data fusion strategies, including cross-system consistency constrained fusion strategy (CCCFS) and multi-system spatiotemporal complementary fusion strategy (MSCFS), are designed to mitigate observational limitations of the single system. These are coupled with a spatial autocorrelation correction (SAC), which significantly suppresses random noise and improves spatiotemporal consistency of SIC distribution in the retrieval results. Compared against the Ocean and Sea Ice Satellite Application Facility (OSISAF) SIC product, the experimental results indicate that: (1) In single-system retrieval, the ResCNet demonstrates superior generalization performance, with GPS performing best, achieving a Root Mean Square Error (RMSE) of approximately 4.866%. (2) In multi-mode data fusion, the MSCFS enhances both coverage and accuracy. Fusing data from BDS, GPS, and Galileo achieves a RMSE of 4.196% and yields a correlation coefficient (R) of 0.990. (3) Applying SAC further significantly lowers the RMSE to 3.81% for GPS and 3.57% for the multi-system fused result. This study systematically validates for the first time the effectiveness of TM-1 multi-mode data in Antarctic SIC retrieval, demonstrating its immense potential as a new low-cost, high spatiotemporal resolution means of sea ice monitoring.

On 28 March 2025, a devastating Mw 7.7 earthquake struck the central and southern segments of the Sagaing fault in Myanmar and led to the largest surface rupture length of over 500 km ever documented for a strike-slip event. Here, the Synthetic Aperture Radar images from Sentinel-1A/B satellites were processed using the pixel offset tracking (POT) method to retrieve high-resolution coseismic displacements of the 2025 Mw 7.7 Mandalay earthquake. By integrating Differential Interferometric Synthetic Aperture Radar, aperture interferometry, and POT observations, we quantified the near-field on-fault and off-fault deformation patterns and constrained fault geometry and slip distribution of the Mandalay earthquake. The results show that the rupture of the Mandalay earthquake is dominated by right-lateral strike-slip motions with the maximum slip of ∼6 m located in the shallowest part. The shallow slip deficit of the Mandalay earthquake is small (<15%), consistent with the generally small diffuse deformation on the off-fault damage zones. We also inverted the interseismic coupling distribution on the Sagaing fault constrained by updated Global Navigation Satellite Systems velocities. The result indicates that the coseismic slip of the Mandalay earthquake did not penetrate deeper than the regional seismogenic layer, exhibiting high spatial correlation with interseismic strongly coupled patches on the Sagaing fault. The rupture of the Mandalay earthquake nucleated in a transitional area with heterogeneous stress distribution caused by a rapid change of fault coupling depth on the fault plane. Coulomb stress analysis indicates increased loading on the southern Nay Pyi Taw and northern Sagaing fault segments, implying elevated seismic hazard in these areas. Considering the strong coupling of adjacent segments and the dense population in central Myanmar, heightened attention to future earthquake potential is warranted.

Against the backdrop of the increasing frequency of extreme hydrological events and persistent over-extraction of groundwater, the North China Plain (NCP) is facing significant land subsidence. This study systematically analyzed the surface subsidence response patterns and mechanisms of the NCP during extreme rainfall events by integrating Gravity Recovery and Climate Experiment (GRACE) data, Global Navigation Satellite System (GNSS) observations, environmental load models, well data, and precipitation records. The main findings are as follows: (1) From 2002 to 2020, the groundwater storage change (GWSC) in most of the study area declined at an average rate of trend about 5 cm/yr, while from 2021 to 2024, influenced by heavy rainfall recharge, GWSC recovered with a mean rate of trend about 7 cm/yr; (2) During the extreme rainfall event from 1 July to 31 August 2023, the environmental loading model effectively captured the vertical deformation caused by hydrological loading, showing general consistency with GNSS monitoring results in spatial distribution. Most GNSS stations experienced rapid subsidence during the event (GNSS: 5 mm, model: 2 mm), followed by a gradual rebound after the extreme rainfall, consistent with elastic theory; (3) The deformation at the TJBH station exhibited anomalies attributable to porous elastic effects; (4) Integrated well data confirmed that rainfall recharge primarily influences shallow groundwater. This study reveals the multiple mechanisms underlying extreme hydrological induced land subsidence in the NCP.

Water body extraction, a critical task in environmental monitoring and resource management, has gained extensive research and application in recent years through the use of remote sensing image semantic segmentation. However, due to the complexity of spectral information of ground objects in remote sensing images, water extraction faces challenges such as inaccurate edge extraction and low extraction accuracy. This article proposed an improved model, namely TA-TransUNet, to address these issues, particularly poor edge recognition and difficulty in identifying water bodies such as lakes and ponds. First, the model uses an encoder-decoder architecture, with a Transformer-based encoder that enhances the ability to capture global features. Second, the Triplet Attention module is introduced into the model to improve the processing and understanding of input data by integrating local, global, and cross-channel attention mechanisms, thereby enhancing the performance of water extraction. At last, the designed module is applied to the up-sampling stage of the TransUNet decoder, enhancing the expressive power of different features by fusing spatial, channel, and dimensional features. By merging up-sampled features with those from the encoder, the model integrates local details and global semantic information, improving edge detection accuracy and distinguishing complex land cover types. To validate the effectiveness of the proposed model, two publicly available datasets, GID and LoveDA, were used to test the performance of the model. Experimental results showed that TA-TransUNet outperforms the state-of-the-art algorithm models in edge detection and spatial detail capture.

With the growing demand for high-precision water body extraction in complex environments, traditional segmentation models continue to face challenges such as insufficient boundary detection and limited multi-scale feature representation. To address the limitations of fixed dilation rates in conventional convolutional networks, which constrain the adaptability of receptive fields across varying spatial scales, this paper proposes a novel segmentation module: the Frequency-Selective Deformable Convolution. The Frequency-Selective Deformable Convolution consists of two key components: a Frequency-Selection module and a deformable convolution unit. The frequency-selection module performs frequency-domain decomposition via Fourier transform and applies learnable masks to selectively enhance or suppress different frequency bands, enabling the network to focus on the frequency-domain features of key spatial structures. The deformable convolution component introduces content-aware sampling through learned offsets, dynamically adjusting the receptive field in the spatial domain based on local geometric variations. This collaborative mechanism of the frequency-spatial dual domain not only strengthens the feature representation of multi-scale spatial structures through frequency-domain filtering but also improves the capture of geometric details via adaptive sampling in the spatial domain, ultimately enhancing the model’s capability for multi-scale and structurally adaptive feature modeling. We integrate Frequency-Selective Deformable Convolution into three mainstream architectures, namely UNet, Attention-UNet, and TransUNet, to construct enhanced models. Experimental results on high-resolution, multi-source remote sensing imagery demonstrate that the FSDC-introduced models significantly improve water body extraction performance, especially in terms of boundary accuracy, omission suppression, and false positive reduction. These findings underscore the potential of frequency-aware, adaptive convolutional mechanisms for advancing remote sensing applications such as water resource monitoring, ecological management, and land cover analysis.