Climate change and socioeconomic development are projected to increase global fluvial flood risk. Most studies have focused on large rivers or aggregated regional assessments, leaving limited understanding of how flood risk varies across river sizes. By integrating high-resolution hydrodynamic simulations, demographic projections and current flood defence levels across the contiguous USA, our projections suggest that smaller rivers tend to exhibit greater sensitivity to climate change-induced increases in flood intensity by 2050. Meanwhile, flood defences along small rivers are currently the weakest, potentially placing nearby populations at disproportionately higher risk. Restoring future flood risk along small rivers to historical levels would require the highest flood defence investments among all river sizes-up to similar to US$4 million (2005) per kilometre. The elevated flood risk for residents along small rivers appears to be primarily driven by increasing climate-related hazard severity and low defence levels, whereas flood risk along larger rivers is dominated by rising exposure. Together, these results indicate a potential mismatch between future flood defence needs and existing defence levels across river sizes, highlighting the urgent need to prioritize investments for communities along small rivers.

Flooding in underground spaces has become more prevalent in recent decades due to increasing urbanization and the intensification of rainfall caused by climate change. While previous research on flood simulations has primarily focused on 1D and 2D models, limited studies have utilized 3D simulations. This study presents a comparison between 1D simulations using the SWMM model and 3D simulations using ANSYS Fluent to model flooding in an underground basement. The comparison was conducted under various flood scenarios. The results revealed a significant discrepancy in the predicted water volumes between the two models. Specifically, the 1D model overestimated the total water volume by approximately 50

Rapid hydropower expansion and irrigation growth heighten water competition, further intensified by climate variability in transboundary basins. Existing models overlook spatial heterogeneity in irrigation demand, limiting their ability to capture water reallocations across systems. Here, we develop a hydrological modeling framework that integrates reservoir operations, irrigation withdrawals, and future climate projections to quantify Water-Energy trade-offs. A key innovation is the inclusion of a hydraulic infrastructure topology module that uses intelligent remote sensing canal detection technique to detect irrigation canals and establish river-reservoir-field connectivity. Historical simulations reveal that prioritizing hydropower generation can reduce downstream irrigation water availability by up to 14%, with dry-season impacts up to five times greater than those in the wet season. Under future scenarios (2021-2040), irrigation demand is projected to increase by 63-68%, largely driven by the expansion of irrigated areas. However, projected increases in dry-season precipitation under future climate change could mitigate these trade-offs, reducing average irrigation shortfalls to 7%. Our findings highlight how interdependencies between irrigation water and hydropower are reshaped by climate and infrastructure development, offering a new framework for evaluating adaptive resource management in transboundary river systems.

Hydrological modeling is essential for water resources management but often requires specialized expertise, creating a significant barrier to its broad application. To address this challenge, we developed HydroCraft, a webbased platform designed to democratize the modeling process. HydroCraft provides a comprehensive and integrated workflow for basin extraction, driving data generation, parameter calibration, and result visualization, enabling rapid modeling for any watershed globally. Key features of the platform include: an adaptive watershed delineation method that optimizes the distribution of computational units based on user-defined sub-basin area thresholds; the integration of three hydrological models (THREW, Xin'anjiang, and Miyun Hydrological Model) suitable for diverse climatic and geological conditions; and the incorporation of a Large Language Model (LLM) agent, which serves as a robust assistant to provide guidance and automated task execution. Users only need a device with a web browser and an internet connection to perform advanced hydrologic simulations. Two case studies (user-driven vs. LLM-agent-assisted) demonstrated the platform's efficiency and intelligence. The platform advances hydrological modeling toward greater intelligence and accessibility, providing support for applications such as water resource management and flood forecasting.

Coevolution of coupled human-water systems (CHWS) is critical for long-term sustainable water management, linking to Panta Rhei. However, the study of CHWS suffers from complexity brought by diverse natural and social science disciplines. In this study, we investigated the general landscape of the theoretical frameworks, methods and data in CHWS case studies. Our meta-analysis, encompassing 205 cases, draws on eight proposed theoretical frameworks in four typologies, quantifying the prevalence and geographical distribution of methods and data. Results demonstrated the analytical strength of sociohydrology for CHWS, underscoring the need to integrate multidisciplinary theoretical frameworks. A combination of qualitative and quantitative methods and data would help overcome the limitations of each method when used in isolation, broadening the research scope of disciplines. This requires sociohydrology to enhance its ability to integrate diverse research approaches. The uneven global distribution of CHWS research teams highlights the necessity of increasing collaboration and resource sharing across borders.

Abstract. Runoff threshold behavior is widely reported in event-based hydrological studies, but its interpretation and cross-catchment comparability remain unresolved due to variations in threshold metrics and values across climates, landscape structures, and observational focuses. This study synthesizes reported storm-runoff thresholds from experimental catchments worldwide by compiling the indicators used to detect nonlinearity, the dominant runoff generation mechanisms, their observed transition pathways under increasing wetness, and recurrent soil–geology fingerprints. Across mechanisms and climates, thresholds are identified using diverse (and often non-standardized) rainfall-based, state-based, and composite indicators. However, antecedent and within-event state variables (e.g., soil moisture, catchment storage, groundwater level) consistently provide better explanations for nonlinear runoff responses than rainfall metrics alone, indicating that threshold behavior is primarily controlled by the state of the catchment but is triggered by rainfall. Subsurface- and saturation-related mechanisms dominate the reported cases, particularly in humid environments. When mechanism shifts are explicitly documented, responses show a strong directional organization with increasing wetness, typically evolving from infiltration-excess overland flow to saturation-excess overland flow, and then to subsurface or groundwater-dominated pathways. Soil–geology network analysis further reveals that each dominant mechanism is associated with recurring combinations of soil depth, texture, permeability contrasts, lithology, and geological structure, forming structural fingerprints that regulate connectivity development. Overall, runoff thresholds are best understood as markers of hydrologic connectivity transitions within structurally constrained landscapes, rather than fixed rainfall exceedances. We propose a connectivity-based conceptual framework linking rainfall forcing, evolving states, structural controls, and mechanism transitions to support cross-catchment comparison, guide future observations, and improve the representation of nonlinear runoff responses in hydrological models.

Flood monitoring is critical for disaster management. Synthetic Aperture Radar (SAR) offers unique advantages for flood detection due to its all-weather, day-and-night capabilities. Machine learning methods require extensive labeled data, often difficult to obtain, while conventional thresholding algorithms based on random patch selection and assumed background distributions are unstable. This study proposes three novel threshold-based SAR flood extraction algorithms that integrate hydrological and hydrodynamic principles with high-precision DEM data and permanent water body datasets. The methodology was validated across 12 flood events, including the July 2023 Haihe River flood as a representative case. Key findings demonstrate that incorporating hydrological/ hydrodynamic principles improves flood mapping precision and F1 scores in validation datasets compared to baseline automated thresholding methods. Analysis further indicates that algorithm performance is strongly influenced by DEM resolution and uncertainty, with high-resolution, low-uncertainty DEMs (e.g., 2 m) substantially improving flood delineation, particularly for hydrodynamic model-based approaches. The algorithms are generally robust within the tested Haihe River Basin events, but extreme parameters can reduce precision. Overall, HOA, HMOA, and HMDA are effective but limited by data quality, threshold assumptions, and validation scope, highlighting the need for higher-resolution data and broader testing.

Water management interventions are designed to mitigate undesirable aspects of coupled human-water systems (CHWS). However, due to the nonlinear feedback mechanisms inherent in CHWS, these interventions sometimes lead to unintended consequences that exacerbate the very issues they aim to resolve. To develop a generalized understanding of the underlying mechanisms behind such unintended outcomes, this study conducts a meta-analysis of 37 case studies from around the world. We identified six core subsystems and defined a critical pathway showing how hydrological perturbations propagate in a CHWS and lead to unintended consequences of interventions. By analysing case storylines, we identified the critical pathways and harmonize them into prevalent critical pathways, which most frequently lead to unintended consequences for specific phenomena, together with the key variables. The results of this study can support more sustainable and resilient water management, as it is the critical pathways that must be altered to avoid unintended consequences.

An accurate representation of the soil moisture (SM) state is fundamental to both refining the physical basis of parameters and to reducing uncertainty of runoff simulation in conceptual hydrological models. However, the specific impacts and underlying mechanisms of incorporating passive satellite SM products into semi-distributed conceptual models remain unclear. To address this gap, we employed the semi-distributed conceptual hydrological model, Tsinghua Representative Elementary Watershed (THREW), in 12 diverse watersheds across China. We assimilated two multi-resolution (1 km and 36 km) SM products derived from the AMSR-E/2 series using an exponential filter. We designed and compared three calibration strategies: a runoff-only benchmark, and two joint scenarios using different SM product resolutions. Results demonstrate that the SM product reduces the overestimation of small-to-medium flood peaks by correcting the antecedent top-layer moisture of the model. This corrective effect is most pronounced during seasonal transitions of spring-to-summer and summer-to-autumn characterized by distinct wet-dry shifts of SM within the whole year mainly in humid subtropical monsoon climate regions. However, the simulation accuracy for large flood events was compromised when calibrating with satellite SM, primarily due to the inconsistent sensitivity to rainfall inputs between the satellite data and the modeled top-layer moisture. This limitation is effectively mitigated by replacing the stringent KGE metric evaluating SM simulation with weaker constraints: the correlation coefficient (R) and the ratio of standard deviations (SDR).

Abstract. Accurate field-scale soil moisture is essential for hydrological processes such as infiltration, land–atmosphere exchange, and agricultural water management. UAV-borne L-band radiometry offers a promising intermediate scale between in situ measurements and satellite observations, but retrieval remains ill-posed due to uncertainties in vegetation attenuation, surface temperature, and sub-footprint heterogeneity. This study develops an uncertainty-aware Bayesian retrieval framework that integrates dual-polarized UAV L-band brightness temperature with RGB and thermal infrared information through footprint-consistent priors. Optical fraction cover, thermal state, and texture descriptors are used to constrain vegetation optical depth and its uncertainty at the scale of the radiometric footprint. The method was evaluated over heterogeneous cropland in Pengzhou, China, using independent calibration (4 scenes, ∼1.3 ha) and validation datasets (6 scenes, ∼3.3 ha). The proposed approach reduced RMSE from ∼0.07 to ∼0.04 m3 m-3 and largely eliminated the systematic dry bias of the conventional τ–ω inversion. Analysis further shows that sub-footprint heterogeneity primarily increases uncertainty in vegetation attenuation, leading to representation error in soil moisture retrieval. These findings highlight that retrieval performance is fundamentally constrained by observation scale and surface heterogeneity. Overall, the study demonstrates that physically informed multi-source priors can improve both accuracy and interpretability, providing a pathway toward more reliable field-scale soil moisture estimation for hydrological applications.

Sub-daily high-resolution precipitation products are crucial for improving mitigation and adaptation of natural hazards. However, such products are often lacking in complex terrains, such as the Tibetan Plateau (TP). While convection-permitting models can accurately capture kilometer-scale sub-daily convective precipitation processes, they are computationally expensive. Here we propose a high-resolution Hourly Precipitation Downscaling Network (HPDNet), a deep-learning surrogate designed to approximate the downscaling behaviors of the Weather Research and Forecasting (WRF) model. HPDNet was trained using 10-year WRF simulations (9 km to 1 km) over the southeastern TP (similar to 0.11 million km(2)). To evaluate its transferability, the trained model was applied to two independent products, ERA5-Land and IMERG, without retraining. Results indicate that downscaled IMERG exhibits higher accuracy than downscaled ERA5-Land, consistent with the superior initial quality of raw IMERG observations (bias: 0.86 mm/d) relative to ERA5-Land (bias: 2.66 mm/d). In the training domain, downscaled IMERG achieves an approximately 22.8% reduction in RMSE and a 20.0% improvement in the Critical Success Index (CSI). Furthermore, the model was successfully extended to the entire southern TP (similar to 1.38 million km(2)), where downscaled IMERG maintains a 16.2% RMSE reduction, stemming from the model's ability to capture fundamental physical drivers. Interpretability analysis reveals that HPDNet's spatial attention concentrates on steep elevation gradients-the primary drivers of orographic precipitation. This focus allows the model to recover fine-scale spatial heterogeneities and orographic enhancement signals that raw products typically smooth out. HPDNet facilitates the efficient generation of multi-decadal, 1-km hourly precipitation estimates across the complex terrain of the southern TP.

Accurate hydrological simulation and credible climate-change projections in cold mountainous basins are hindered by complex phase transitions and strong cryospheric controls that exacerbate model equifinality and uncertainty in runoff component partitioning. This study advances a tracer-aided hydrological modeling framework by using stream-water stable isotopes to diagnose model structural deficiencies associated with frozen soils and to quantify how such deficiencies propagate into projections of hydrological sensitivity to climate change. We implement the tracer-aided Tsinghua Representative Elementary Watershed model (THREW-T) in a Tibetan Plateau cold catchment and compare two configurations: a baseline model lacking explicit frozen-soil processes (THREW-NoFS) and an enhanced version incorporating a simplified catchment-scale frozen-soil module with dynamically varying soil hydraulic properties (THREW-FS). While both configurations reproduce observed streamflow, isotope constraints expose a key limitation of THREW-NoFS: it cannot simultaneously capture baseflow dynamics and stream-water isotopic signatures, indicating missing freeze–thaw controls that effectively induce unrepresented seasonal variability in soil hydraulic behavior. Incorporating the frozen-soil module substantially improves the joint simulation of streamflow and isotopes and yields a more physically consistent runoff partitioning, characterized by reduced baseflow during dry seasons and increased subsurface runoff contributions during wet seasons. Frozen soils exert limited influence on annual discharge totals but markedly reshape runoff seasonality through altered surface–subsurface connectivity. Importantly, the isotope-informed structural correction changes projected climate sensitivity: both models suggest runoff decreases with warming and increases with precipitation intensification, yet THREW-NoFS produces systematically stronger sensitivities and tends to overestimate runoff responses because it provides more available water for evaporation and misrepresents surface–subsurface partitioning. These results demonstrate that tracer-aided hydrological models, when constrained by stable isotopes, offer a powerful pathway to diagnose frozen-soil impacts on model structure, reduce uncertainty in runoff component contributions, and generate more reliable projections of hydrological sensitivity to climate change in cryospheric basins.

Cold-region headwaters on the Tibetan Plateau are highly sensitive components of regional water and carbon cycles, yet the catchment-scale controls on dissolved organic carbon (DOC) export remain poorly resolved because runoff generation and flow pathways cannot be reliably inferred from streamflow alone. Here we developed THREW-IC, a distributed multi-tracer-aided hydrological model that couples cryospheric processes with stable isotope and DOC dynamics, and applied it to the Source Region of the Yangtze River, a permafrost-affected alpine basin. The model was jointly calibrated against discharge, stream-water δ¹⁸O, and DOC concentrations in stream water and groundwater. The results show that: (1) THREW-IC reasonably reproduces the main temporal dynamics of discharge, stream-water δ¹⁸O, and DOC concentrations in stream water and groundwater. (2) The combined tracer constraints identify internal runoff partitioning more effectively than discharge alone, with isotopes providing stronger constraints on surface–subsurface runoff partitioning and DOC observations more effectively constraining the groundwater-outflow parameter KKA. (3) Spatial correlation analysis showed that SCL was significantly associated with total DOC export and surface DOC export, but not with subsurface DOC export, highlighting the pathway-dependent coupling between terrestrial carbon availability and hydrological transport in controlling basin-scale DOC export. (4) In the 2015 sensitivity experiment, representing frozen-soil effects increased simulated annual total DOC export by approximately 18.6%, primarily through a 35.6% increase in subsurface DOC export while exerting negligible influence on surface DOC export. This effect was strongly seasonal, suppressing DOC export during the frozen low-flow period but enhancing it during the thaw and wet period through the seasonal redistribution of subsurface runoff. These results show that integrating conservative and reactive tracers within a distributed framework can reduce process uncertainty and improve mechanistic understanding of water–carbon coupling in cold alpine headwaters.

To enhance hydrologic modeling, the hydrology community has developed benchmark datasets (e.g. Model Parameter Estimation Experiment, MOPEX), providing standardized data for model evaluation and parameter estimation. However, these datasets primarily focus on modeling natural hydrologic processes, leaving a critical gap in understanding the role of human influences. Here, we introduce the Coupled Hydrology-Human Activity Information (CHHAI) dataset, a benchmark dataset that integrates coupled human-water data from regions across all continents, excluding Antarctica. CHHAI incorporates data from 25 regions that cover various human impacts such as reservoir management, flood protection, river management policies, land use changes, and water use. Each basin reflects distinct challenges, providing a diverse and globally representative resource for researchers studying these processes. By offering standardized datasets for modeling and analysis, CHHAI aims to enhance our understanding of interactions between people and water and to support the development of improved strategies for managing coupled human-water systems.

Accurate delineation of flood and surface-water extent is essential for hydrologic monitoring, exposure assessment, and event-based impact analysis. In practice, flood mapping must operate under heterogeneous and often incomplete observations, as optical imagery is frequently obscured by cloud cover during extreme precipitation, whereas synthetic aperture radar (SAR) imagery is affected by speckle noise and geometric distortions. To enhance robustness across sensing conditions, we develop an end-to-end, inputadaptive segmentation framework that unifies optical-only, SAR-only, and joint optical–SAR inference within a single architecture and allows incorporation of terrain information when available. The framework introduces an alignment-guided cross-modal fusion strategy to promote consistency between optical and SAR representations. We establish a unified multi-dataset benchmark across five public datasets and conduct 39 controlled experiments, systematically comparing lightweight convolutional networks and large-scale foundation models under consistent training protocols. Across multimodal settings, the proposed fusion improves intersection-over-union and reduces omission and commission errors, particularly in narrow inundated channels and complex urban environments. These results indicate that alignmentguided multimodal learning enhances cross-sensor consistency and supports more hydrologically reliable flood mapping across diverse regions and acquisition conditions.The source code of the proposed framework is publicly available at https://github.com/xiaojiayin/SegFlood.

Study region Mainland China. Study focus Investigating long-term persistence (LTP) in river flows is crucial for understanding hydrological variability and improving water resource management, yet systematic nationwide assessments remain limited, especially in China. In this study, 45 years of monthly runoff from 60 stations across China were analyzed to quantify basin-scale LTP and its seasonal variations using the Hurst coefficient (H) estimated by Least Squares Variance, Whittle Estimator, and Rescaled Range Analysis. Drivers of LTP were further examined using Spearman’s correlation, random forests (RF), and SHapley Additive exPlanations (SHAP). New hydrological insights for the region The national annual mean H was 0.710, with marked spatial and seasonal variability. Northern catchments exhibited stronger persistence than southern catchments, and winter showed the highest LTP (0.758), whereas spring and summer displayed weaker persistence (0.650). Catchment area, grassland, water bodies, unused land, and coarse- to medium-textured soils were generally positively correlated with H, indicating stronger hydrological persistence, whereas slope, climatic factors, forest land, and fine-textured soils tended to weaken it. RF and SHAP analyses revealed seasonal shifts in dominant controls, with forest land, soil texture, and catchment area exerting the greatest influence annually and in summer, while climatic factors gained prominence in other seasons. This study presents the first spatially and seasonally resolved assessment of LTP in Chinese rivers, offering new insights into streamflow predictability and catchment behavior.

Abstract. Hybrid hydrological models integrating embedded neural networks (ENNs) have demonstrated strong potential for improving streamflow prediction. However, how uncertainties induced by ENNs propagate through hydrological processes and internal system states remains poorly understood. This limits a mechanistic understanding of the stability and reliability of hybrid hydrological systems. This study therefore moves beyond performance evaluation and focuses on the internal uncertainty propagation structure of fully coupled hybrid hydrological systems. We investigate the internal uncertainty propagation mechanisms of hybrid hydrological models with different levels of ENN embedding across 31 cold-region basins. An ensemble-based framework driven by stochastic optimization was used to quantify uncertainty across hydrological fluxes and internal state variables. Multiple complementary metrics were employed to characterize uncertainty magnitude, ensemble consistency, coverage and spread. Results show that ENNs improve streamflow simulation performance but alter the internal uncertainty structure of hydrological systems. Uncertainty amplification is primarily localized within ENN-replaced hydrological processes, while propagation to physically based unreplaced processes remains limited. Instead, internal state variables with memory effects act as key “uncertainty reservoirs”, where uncertainty accumulates over time. This reveals a structured rather than uniform pattern of uncertainty propagation in hybrid hydrological systems. Furthermore, increased ensemble coverage is accompanied by wider uncertainty bands, indicating a trade-off between predictive reliability and sharpness induced by stochastic optimization. Overall, this study provides new insights into how ENNs reshape uncertainty propagation pathways in hydrological models. The results highlight the importance of jointly considering predictive performance, uncertainty structure and internal system stability in hydrological modeling.

As a result of climate change and global warming, vegetation greening can influence the hydrological processes of river basins. This study evaluated the possible impact of vegetation greening under the CMIP6 SSP585 scenario on the runoff volume till 2100 at key sections of the Yarlung Tsangpo-Brahmaputra River, which is renowned for its abundant hydropower resources and the complex relations among riparian countries and regions. From 2020 to 2100, with the greening vegetation, annual runoff at Bahadurabad can decrease by 3 to 31 billion m3 on average. Water yield decline brought by vegetation greening frequently happens in upper and middle reaches of the basin. The decrease in runoff also shows significant seasonal characteristics. The main peak of runoff reduction usually occurs during the monsoon season from June to October. The sensitivity analysis shows that the greening of areas with less current vegetation distribution is the more significant factor causing runoff reduction than the increasing lushness of existing vegetation. Dominance of vegetation coverage (FVC) and lushness (LAI) can also vary in different areas of the basin. The findings of this study underscore the importance of accurately predicting the spatial distribution of vegetation species and their temporal dynamics in hydrological modeling. The enhanced understanding of runoff tendency under climate change and vegetation evolution would help supporting more precise management of water resources in large transboundary river basins, promoting basin-wide cooperation, and increasing the efficiency of climate adaptation.

A compound wetness index combining antecedent soil moisture and event rainfall (S0 = ASI0 + P) is commonly used to analyze threshold-based runoff responses. However, it remains unclear how the depth of soil moisture integration affects threshold detection and predictive performance. Here, we examined the relationship between runoff thresholds and maximum integration depth (MID) across four experimental catchments spanning diverse climates and runoff generation mechanisms. Using segmented regression across multiple MIDs and monitoring sites, we identified runoff thresholds and evaluated the explanatory power of S0 for stormflow using post-threshold coefficients of determination (Rh2). Thresholds were consistently detected across integration depths, but explanatory power varies strongly among catchments. In storage-controlled systems, Rh2 increased with depth, indicating that deeper soil moisture integration better captured subsurface storage controls on stormflow initiation. In contrast, in connectivity-limited or shallow flow-dominated systems, shallow integration depths performed better because near-surface wetting controlled preferential flow or infiltration-excess runoff, whereas deeper integration included hydrologically inactive storage. These results show that the optimal integration depth is catchment dependent and closely linked to dominant runoff generation processes. Accounting for this depth dependence can improve threshold-based runoff analysis and support more process-consistent soil moisture monitoring and hydrological modeling.