In recent decades, extreme floods and droughts have occurred frequently around the world, which seriously threatens the social and economic development and the safety of people’s lives and properties. Therefore, it is of great scientific significance to discuss the causes and characteristic quantization of extreme floods and droughts. Here, the terrestrial water storage change (TWSC) derived from the Gravity Recovery and Climate Experiment (GRACE) and its Follow-On (GRACE-FO) data was used to characterize the floods and droughts in the Yangtze River basin (YRB) during 2003 and 2020. To reduce the uncertainty of TWSC results, the generalized three-cornered hat and least square methods were used to fuse TWSC results from six GRACE solutions. Then combining precipitation (PPT), evapotranspiration, soil moisture (SM), runoff, and extreme climate index data, the influence of climate change on floods and droughts in the YRB was discussed and analyzed. The results show that the fused method can effectively improve the uncertainty of TWSC results. And seven droughts and seven floods occurred in the upper of YRB (UY) and nine droughts and six floods appeared in the middle and lower of YRB (MLY) during the study period. The correlation between TWSC and PPT (0.33) is the strongest in the UY, and the response time between the two is 1 month, while TWSC and SM (0.67) are strongly correlated with no delay in the MLY. The reason for this difference is mainly due to the large-scale hydropower development in the UY. Floods and droughts in the UY and MLY are more influenced by the El Niño-Southern Oscillation (ENSO) (correlation coefficients are 0.39 and 0.50, respectively) than the Indian Ocean Dipole (IOD) (correlation coefficients are 0.19 and 0.09, respectively). The IOD event is usually accompanied by the ENSO event (the probability is 80%), and the hydrological hazards caused by independent ENSO events are less severe than those caused by these two extreme climate events in the YRB. Our results provide a reference for the study on the formation, development, and recovery mechanism of regional floods and droughts on a global scale.

In recent years, the surge in marine activities has increased the frequency of submarine pipeline failures. Detecting and identifying the buried conditions of submarine pipelines has become critical. Sub-bottom profilers (SBPs) are widely employed for pipeline detection, yet manual data interpretation hampers efficiency. The present study proposes an automated detection method for submarine pipelines using deep learning models. The approach enhances the YOLOv5s model by integrating Squeeze and Excitation Networks (SE-Net) and S2-MLPv2 attention modules into the backbone network structure. The Slicing Aided Hyper Inference (SAHI) module is subsequently introduced to recognize original large-image data. Experimental results conducted in the Yellow Sea region demonstrate that the refined model achieves a precision of 82.5%, recall of 99.2%, and harmonic mean (F1 score) of 90.0% on actual submarine pipeline data detected using an SBP. These results demonstrate the efficiency of the proposed method and applicability in real-world scenarios.

GNSS buoys are essential for real-time elevation monitoring in coastal waterways, yet the vertical coordinate time series are frequently contaminated by complex non-stationary noise, and existing denoising methods often rely on empirical parameter settings that compromise reliability. This paper proposes GVMD-NLM, a hybrid denoising framework optimized by an improved Grey Wolf Optimizer (GWO). The method introduces an adaptive convergence factor decay function derived from the Sigmoid function to automatically determine the optimal parameters (K and $\alpha$) for Variational Mode Decomposition (VMD). Sample Entropy (SE) is then employed to identify low-frequency effective signals, while the remaining high-frequency noise components are processed via Non-Local Means (NLM) filtering to recover residual information while suppressing stochastic disturbances. Experimental results from two datasets at the Dongguan Waterway Wharf demonstrate that GVMD-NLM consistently outperforms SSA, CEEMDAN, VMD, and GWO-VMD. In Dataset One, GVMD-NLM reduced the RMSE by 26.04% (vs. SSA), 17.87% (vs. CEEMDAN), 24.28% (vs. VMD), and 13.47% (vs. GWO-VMD), with corresponding SNR improvements of 11.13%, 7.00%, 10.18%, and 5.05%. In Dataset Two, the method achieved RMSE reductions of 28.87% (vs. SSA), 17.12% (vs. CEEMDAN), 18.45% (vs. VMD), and 10.26% (vs. GWO-VMD), with SNR improvements of 10.48%, 5.52%, 6.02%, and 3.11%, respectively. The denoised signal maintains high fidelity, with correlation coefficients (R) reaching 0.9798. This approach provides an objective and automated solution for GNSS data denoising, offering a more accurate data foundation for waterway hydrodynamics research and water level monitoring.