Abstract. Intermittent rivers and ephemeral streams (IRES) constitute a large fraction of global river networks, provide important ecosystem services, and are increasing in number with climate change. Yet, observing stage and calculating discharge in IRES can be technologically and methodologically challenging. To address this problem, we develop a method to classify relative stage categories from field camera imagery, creating a time series of categorical flow states without the need for direct stage measurements. Specifically, we employ a Logistic Regression model to classify conditions of no water, low water levels, or high water levels for an ephemeral stream located in the upper Russian River watershed of California (US). We trained our algorithm using hourly field camera images from 2017–2023, and validated the image classifications with 15 min continuous stage observations. We then used image classifications to perform quality control on the continuous stage time series, which allowed us to identify when the stream was dry and when the sensor malfunctioned. Next, we compared the image classifications to publicly accessible modeled discharge from the NOAA National Water Model CONUS Retrospective Dataset. We discuss how in-situ monitoring including field cameras and the classification of field camera imagery, combined with surface meteorology and soil moisture observations, provides detailed hydrologic information important for understanding how climate affects IRES. Because the image classification approach is transferable to other ephemeral stream sites equipped only with field cameras, this methodology provides a low-cost option for observing relative stage on sparsely-measured IRES that can augment existing hydrologic modeling used by water managers.

Abstract. The Lancang-Mekong River (LMR) Basin is highly vulnerable to extreme hydrological events, including Drought-Flood Abrupt Alternation (DFAA). The efficacy of potential mitigation measures, such as reservoir operations, on DFAA under climate change remains poorly understood. This study investigates these dynamics using five Global Climate Models (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6). It employs the Revised Short-cycle Drought-Flood Abrupt Alteration Index (R-SDFAI), along with the Tsinghua Representative Elementary Watershed (THREW) model integrated with the developed reservoir module. The findings reveal that DFAA in the LMR Basin is primarily dominated by DTF (drought to flood), with probabilities of DTF exceeding those of FTD (flood to drought) at mild, moderate, and severe intensity levels. The increase in DTF probability for future periods is also significantly higher than that of FTD. Mild DTF and mild FTD account for 58 % to 90 % and 75 % to 100 % of their total probability in the future, making the mild-intensity events the most frequent DFAA. Reservoirs play a significant role in reducing DTF risks during both dry and wet seasons, though their effectiveness in controlling FTD risks, particularly during the dry season, is relatively weaker. Furthermore, there is a positive correlation between the reservoir's capacity to mitigate total DFAA risk and its total storage. Reservoirs display a stronger ability to regulate high-intensity FTD and high-frequency DTF events, and significantly reduce the monthly duration of DFAA. These insights provide valuable guidance for the effective management of water resources cooperatives across the LMR Basin.

Abstract. The seasonal snow cover in the European Alps is increasingly threatened by rising temperatures due to climate change. Still, downscaled climate projections are lacking for many regions. To address this gap, we developed a literature-based approach for projecting future snow depths, that is applicable to all locations where historical snow depth data is available. We harmonized heterogeneous literature on future snow depth and snow water equivalent by translating emission scenarios to corresponding temperature scenarios and standardizing seasonal periods. Then, we parameterized localized reduction curves based on elevation, temperature scenarios and local climatologies, such as mean snow cover length and mean maximum snow depth. This method was applied to four measurement stations in Switzerland under a +2 °C temperature scenario, revealing significant declines in snow depth and season length, especially at lower elevations. Validation against published data shows that the approach captures key trends in snow loss, despite the simplification of climate dynamics. This resource-efficient method provides a practical tool for estimating climate change related snow depth declines in snow dominated regions, which are lacking highly resolved climate projections, and can support decision-makers in developing adaptation strategies for climate-related challenges.

Abstract. Merging physics-based with data-driven approaches in hybrid hydrological modeling offers new opportunities to enhance predictive accuracy while addressing challenges of model interpretability and fidelity. Traditional hydrological models, developed using physical principles, are easily interpretable but often limited by their rigidity and assumptions. In contrast, machine learning methods, such as Long Short-Term Memory (LSTM) networks, offer exceptional predictive performance but are often criticized for their black-box nature. Hybrid models aim to reconcile these approaches by imposing physics to constrain and understand what the ML part of the model does. This study introduces a quantitative metric based on Information Theory to evaluate the relative contributions of physics-based and data-driven components in hybrid models. Through synthetic examples and a large-sample case study, we examine the role of physics-based conceptual constraints: can we actually call the hybrid model “physics-constrained”, or does the data-driven component overwrite these constraints for the sake of performance? We test this on the arguably most constrained form of hybrid models, i.e., we prescribe structures of typical conceptual hydrological models and allow an LSTM to modify only its parameters over time, as learned during training against observed discharge data. Our findings indicate that performance predominantly relies on the data-driven component, with the physics-constraint often adding minimal value or even making the prediction problem harder. This observation challenges the assumption that integrating physics should enhance model performance by informing the LSTM. Even more alarming, the data-driven component is able to avoid (parts of) the conceptual constraint by driving certain parameters to insensitive constants or value sequences that effectively cancel out certain storage behavior. Our proposed approach helps to analyse such conditions in-depth, which provides valuable insights into model functioning, case study specifics, and the power or problems of prior knowledge prescribed in the form of conceptual constraints. Notably, our results also show that hybrid modeling may offer hints towards parsimonious model representations that capture dominant physical processes, but avoid illegitimate constraints. Overall, our framework can (1) uncover the true role of constraints in presumably “physics-constrained” machine learning, and (2) guide the development of more accurate representations of hydrological systems through careful evaluation of the utility of expert knowledge to tackle the prediction problem at hand.

Abstract. In May 2024, an extraordinary precipitation event triggered record floods in southern Brazil, particularly impacting complex river–estuary–lagoon systems, and resulting in unprecedented impacts on the local population and infrastructure. As climate change projections indicate an increase in such events for the region, understanding these flooding processes is essential for better preparing cities for future events like the May 2024 flood. In this context, hydrodynamic modelling is an important tool for reproducing and analysing this past extreme event. This paper presents the first detailed hydrodynamic assessment of this unprecedented flood, the worst registered natural disaster in Brazilian history. We also performed the first validation of a detailed hydrodynamic model using new observations from the SWOT satellite. The study investigates the main mechanisms that governed the disaster and assesses scenarios for hydraulic flood control interventions currently under public debate, with a focus on the most populated areas of the Metropolitan region of Porto Alegre (RMPA) capital city. The results demonstrated that the model accurately represented the event, with average NSE, RMSE and BIAS of 0.82, 0.71 and −0.47 m, respectively, across the basin's main rivers. Furthermore, the simulated flood extent showed an 83 % agreement with high-resolution satellite images. Our analysis of the governing mechanisms showed that the Taquari River was mainly responsible for the peak in the RMPA, while the Jacuí River contributed most to the flood's duration. Additionally, the synchronization of the flood peaks from both rivers could have increased water levels by 0.82 m. Evaluated hydraulic interventions demonstrated that the effectiveness of the proposed measures varied by location, with a generally limited influence on RMPA water levels (lower than 0.38 m). By accurately assessing the May 2024 flood, this study enhances the understanding of a complex river–estuary–lagoon system, quantifies the impacts of adverse scenarios, and reveals the limitations of potential hydraulic structure interventions. Finally, modelling this unprecedented event offers valuable insights for future research and global flood management policies.

Abstract. Estimating runoff components, including surface flow, baseflow and total runoff is essential for understanding precipitation partition and runoff generation and facilitating water resource management. However, a general framework to quantify and attribute runoff components is still lacking. Here, we propose a general formulation through observational data analysis and theoretical derivation based on the two-stage Ponce-Shetty model (named as the MPS model). The MPS model characterizes mean annual runoff components as a function of available water with one parameter. The model is applied over 662 catchments across China and the contiguous United States. Results demonstrate that the model well depicts the spatial variability of runoff components with R2 exceeding 0.81, 0.44 and 0.80 for fitting surface flow, baseflow and total runoff, respectively. The model effectively simulates multi-year runoff components with R2 exceeding 0.97, and the proportion of runoff components relative to precipitation with R2 exceeding 0.94. By using this conceptual model, we elucidate the responses of surface flow and baseflow to available water and environmental factors for the first time. The surface flow is jointly controlled by precipitation and environmental factors, while baseflow is mainly influenced by environmental factors in most catchments. The universal and concise MPS model offers a new perspective on the long-term catchment water balance, facilitating broader application in large-sample investigations without complex parameterizations and providing an efficient tool to explore future runoff variations and responses under changing climate.

Abstract. Subsurface flow in preferential pathways in soils may transport water more rapidly than the soil matrix, may be quickly activated during precipitation events and enhance infiltration or interflow. Vertical pathways are particularly important for runoff generation. However, identifying these pathways is challenging because traditional methods such as piezometers, soil moisture sensors, or hillslope trenches do not adequately capture the spatial scale and frequency of preferential flow features, while other experimental techniques like dye tracing are labor-intensive and invasive. In this study, we introduce a novel method to identify the locations of preferential flow by analysing vertical soil profiles of stable water isotopes. Across four catchments, we drilled 100 soil cores (1–3 m deep) per catchment and analyzed the stable isotope composition of the soil water in 10–20 cm depth intervals to construct depth profiles. We employed clustering techniques to group soil-water isotope profiles and selected those that matched to a seasonal sampling date to establish a reference profile for each catchment using LOESS regression, representing profiles influenced solely by matrix infiltration. Deviations from these reference profiles were then used as indicators of being influenced by vertical or lateral preferential flow. Our results revealed evidence of preferential flow in all studied catchments. Especially in the alpine catchment with highly heterogeneous soils, many profiles showed distinct preferential flow features, including multiple, vertically independent pathways occurring at variable depths, even among adjacent profiles. These findings demonstrate the feasibility of using soil water isotope profiles to assess preferential flow pathways and highlight the substantial spatial and vertical variability of preferential flowpaths at hillslope and catchment scales.

Abstract. Floods are a recurring global threat, causing lives lost, property damage, and agricultural impacts. Accurate and timely flood inundation forecasts are crucial for effective disaster preparedness and mitigation. However, traditional flood forecasting methods often face challenges in terms of computational demands and data requirements, particularly when applied to large geographic areas. This study presents a novel approach to scaling a data-driven flood forecasting framework, Forecasting Inundation Extents using REOF (Rotated Empirical Orthogonal Function) (FIER), to large geographic regions. FIER leverages historical satellite imagery and streamflow data to predict flood inundation extents offering a solution in regions typically considered data-scarce for traditional hydrodynamic modelling (i.e., lacking detailed bathymetry and friction coefficients information). We demonstrate the effectiveness of applying FIER over a large geographic extent using watershed boundaries to create individual FIER models and then mosaicking the results geographically to provide large flood inundation predictions. The Upper Mississippi Alluvial Plain in the United States was used as a test region. We evaluated multiple buffer sizes, ranging from 0–50 km, for watersheds for generating the data-driven FIER models to reduce edge effects along watershed boundaries when mosaicking the individual FIER implementations. The FIER method using watersheds, coupled with different forecast lead times from the National Water Model operational streamflow forecasts, was used to accurately predict the extent of surface water for select flood and low flow use cases. Our results show that the scaled FIER approach using watersheds yields higher accuracies for different error metrics, including the Structural Similarity Index Measure (SSIM), RMSE, and MAE. We found that scaling FIER using a watershed approach yielded statistically significant better performance compared to the baseline area using the Kolmogorov–Smirnov test: this is particularly true when using buffer sizes for the watersheds of 0–10 km and when applying a cumulative distribution function (CDF) matching post-processing correction to the FIER outputs. This approach offers a promising solution for large-scale flood forecasting, particularly in data-scarce regions where data required for traditional hydrodynamic modelling is lacking or ungauged basins. Future research will focus on refining the framework to incorporate additional hydrological variables and improve the accuracy of long-range flood inundation predictions.

Abstract. Arid endorheic basins exhibit limited water availability shaped by strong precipitation and evaporation variability. Understanding these processes is crucial for sustainable water resources management in such fragile environments. This study examines how rainfall and evaporation drive the spatial and temporal dynamics of groundwater recharge and water balance in an arid endorheic basin, using the Salar del Huasco in the Chilean Altiplano as a case study. For this, we implemented a modified semi-distributed rainfall-runoff model integrated with a 40 year record (1980–2019) of satellite-derived precipitation and evaporation estimates. Results show that, on average over the catchment, about 12 % of total rainfall (17 mm yr−1) recharges the aquifers, with a ∼35 d lag between rainfall and peak groundwater recharge. Spatial analysis reveals that most water infiltrates and recharges the groundwater system at high elevations (∼65 % of total recharge), while low-lying wetlands, shallow lagoons, and riparian zones lose most of the water via evaporation (up to 950 mm yr−1). Our findings highlight that when summer rainfall ceases, groundwater becomes the main water source supporting high evaporation rates, while recharge reaches a minimum by the end of autumn that persists until the end of the year. These results suggest a trade-off between groundwater recharge and evaporation for available water during the dry season. Moreover, while the basin receives around 145 mm yr−1 of annual precipitation, evaporation reaches 230 mm yr−1. These values imply a substantial water loss or an unaccounted groundwater inflow, challenging the endorheic assumption of the basin's hydrogeological boundaries. Future research should revisit this assumption and incorporate fully coupled groundwater-surface water simulations to explicitly include interactions with lateral groundwater flows and groundwater levels, as well as with snow dynamics and vegetation processes currently omitted. Also, validating satellite-derived inputs against additional local observations is essential to strengthen the reliability of the water balance assessment. Nonetheless, these results provide a valuable framework and a first-approximation for quantifying water balance components in an arid basin, offering insights for water resources management in a context of water scarcity and climate change.

Abstract. The Central American mid-summer drought (MSD) is a defining precipitation pattern within the regional hydrologic system linked to water and food security. Past changes and future projections in the MSD show a strong sensitivity to how the MSD is defined. The question then arises as to whether multiple definitions should be considered to capture the uncertainty in projected impacts as climate warming continues and a need to understand the impacts on regional hydrology persists. This study uses an ensemble of climate models downscaled over Nicaragua using two methods, global warming levels up to +3 °C, and different definitions of the MSD to characterize the contributions to total uncertainty of each component. Results indicate that the MSD definition contributes the least to total uncertainty, explaining 5 %–9 % of the total. At the same time, evidence suggests a shift of the MSD to later in the year. As warming progresses, total uncertainty is increasingly dominated by variability among climate models. While not a dominant source of uncertainty, downscaling method adds approximately 8 %–18 % to total uncertainty. Future studies of this phenomenon should include an ensemble of climate models, taking advantage of archives of downscaled data to adequately capture uncertainty in hydrologic impacts. These findings provide critical guidance for future research aiming to inform water planning and adaptation efforts in the region: by identifying the dominant sources of uncertainty across warming levels, this framework helps prioritize where to focus modeling and monitoring efforts. In particular, water resource managers can use this information to design adaptive strategies that are robust to model spread and shifts in seasonal precipitation timing, rather than to definitional ambiguity. The projection uncertainty partitioning approach could serve as a template to quantify the relative importance of uncertainty for projections of other precipitation-driven phenomena in different geographic contexts.