Accurate soil moisture (SM) estimates with high spatial resolution are highly desirable for agricultural, hydrological, and environmental applications. This study developed a two-step reconstruction approach to obtain a high-quality and high-spatial-resolution (0.05°) SM dataset from microwave and model-based SM products, combining Bayesian three-cornered hat (BTCH) merging and machine/deep learning downscaling algorithms. Firstly, a three-cornered hat (TCH) method was used to analyze the uncertainty of seven SM products on four main land cover types in the Pearl River Basin (PRB). On this basis, the SM products with low uncertainty were merged using the BTCH method. Secondly, two machine/deep learning algorithms (random forest, RF, and long short-term memory, LSTM) were applied to downscale the merged SM data from 0.25° to 0.05° based on the relationship between SM and auxiliary variables. The overall performance of RF and LSTM downscaling models with/without antecedent precipitation were compared. The merged and downscaled SM results were validated against in situ observations and the China Meteorological Administration (CMA) Land Data Assimilation System (CLDAS) SM data. The results indicated the following: (1) The BTCH-based SM estimate outperformed the parent products and the AVE-based SM estimate (the arithmetic average), indicating that BTCH is a fusion approach that can effectively reduce data uncertainties and optimize weights. (2) The optimal time scale for the cumulative effect of precipitation on SM was 35 days during 2015–2020 in the PRB. SM estimations using RF and LSTM downscaling algorithms both had substantial improvement by considering the antecedent precipitation variable, both at the 0.25° and 0.05° spatial scales. Feature importance assessment also revealed the most important role of antecedent precipitation (30.01%). Moreover, the LSTM model with antecedent precipitation performed slightly better than the RF model with antecedent precipitation. (3) The downscaled SM results all mitigated the overestimation inherent in the original SM data, though they were inevitably limited by the performance of the original SM data and difficult to surpass. The developed two-step reconstruction approach was effective in generating an accurate SM dataset at a finer spatial scale for wide regional applications.

Refined urban flood risk assessment serves as a fundamental safeguard for urban sustainability. However, most studies based on scenario analysis method tend to rely on a single risk evaluation criterion, with limited consideration of applicability differences arising from underlying computational principles. Furthermore, as flood events are inherently dynamic spatial–temporal processes, most studies often overlook the three-dimensional characteristics of flood risk, particularly the connectivity of risk in physically adjacent spaces. To address these issues, this paper proposes a comprehensive flood risk assessment framework that integrates the spatial–temporal characteristics of disaster-causing factors. An improved analysis method for grid-scale flood assessment is proposed based on the comprehensive mechanical analysis method and the drowning factor. In addition, a quantitative approach for characterizing the spatial aggregation of urban flood risk is established using risk thresholds and aggregation area thresholds. These methods are then integrated through a combination weighting–cluster analysis framework for comprehensive flood risk assessment. The results show that the improved analysis method can better reflect the change in risk of flow velocity and water depth combined. Spatiotemporally, the Yinshan Road and western section of the Dongzhong Road, exhibiting high localized risk, moderate overall risk, high risk on the time scale and high spatial agglomeration status, are comprehensively assessed as extremely high-risk flooded zones. The proposed framework effectively characterizes the spatial–temporal distribution of disaster-causing factors, providing a scientific basis for disaster prevention and contributing to urban sustainability.

Accurately characterizing the spatiotemporal distribution of tropical cyclone rainfall is fundamental to enhancing coastal flood defense capabilities. Taking western Guangdong Province as the study case and utilizing hourly rainfall data from 2005 to 2025, this paper employs the Locally Linear Embedding algorithm, K-means clustering, and composite analysis to classify the spatiotemporal patterns of tropical cyclone rainfall and to explore the linkage mechanisms between each pattern and the synoptic fields. The results indicate that, during 2005–2025, a total of 49 tropical cyclones made landfall in western Guangdong, predominantly during the period from June to October. The tropical cyclone rainfall can be categorized into three spatiotemporal patterns: Pattern 1 is a bimodal migrating type, characterized by a southwest-to-northeast migration of the rainfall center and the strongest spatial variability; Pattern 2 is a unimodal balanced type, featuring the most spatially uniform distribution and a broad areal influence; Pattern 3 is a bimodal locked type, with the rainstorm center persistently anchored over the Yangjiang, the highest mean rainfall intensity, and the strongest spatial concentration. For Pattern 1, tropical cyclones with northwest and northward tracks allow the right-front quadrant of heavy precipitation to dominate the migration of the rainstorm center. For Pattern 2, the dispersed low-level wind field jointly suppress the migration and concentration of the rainstorm center. For Pattern 3, the superposition of high-value moisture transport and funnel-shaped topographic forced lifting appears to be a primary contributing factor to the stable maintenance of the rainstorm center. The findings of this paper can provide scientific evidence for tropical cyclone rainfall forecasting and flood defense in western Guangdong.