We used explainable machine learning that incorporates memory effects to study how plants respond to weather and drought. Using data from 90 sites worldwide, we show that memory plays a key role in regulating plant water stress. Forests and savannas rely on longer past conditions than grasslands, reflecting differences in rooting depth and water use. These insights improve our ability to anticipate ecosystem vulnerability as droughts intensify.

The Atlantic Meridional Overturning Circulation (AMOC) moderates the European climate. The AMOC is a tipping element and may collapse to a substantially weaker state under climate change. Such an event induces global and regional climate shifts. The European hydroclimate becomes drier under an AMOC collapse, this response is not considered in the standard hydroclimate projections. Our results indicate a considerable influence of the AMOC on the European hydroclimate.

Our study evaluated 24 precipitation datasets using a hydrological model at global scale to assess their suitability and accuracy. We found that MSWEP (Multi-Source Weighted-Ensemble Precipitation) V2.8 excels due to its ability to integrate data from multiple sources, while others, such as IMERG (Integrated Multi-satellitE Retrievals for Global Precipitation Mission) and GDAS (Global Data Assimilation System), demonstrated strong regional performances. This research assists in selecting the appropriate dataset for applications in water resource management, hazard assessment, agriculture, and environmental monitoring.

Green water stress can negatively affect agricultural production and is often mitigated through irrigation. In this global modelling study, we investigate where and to what extent the implementation of irrigation helps to compensate for green water stress but at the same time leads to an increase in blue water scarcity. Our findings highlight the need to consider both water stresses together, along with their dynamic interactions, for sustainable water management.

Trees in rocky landscapes often depend on hidden water deep underground, yet the role of bedrock cracks remains unclear. In a karst forest in southwest China, we traced water movement and seasonal use by trees using stable isotopes. We found that large soil-filled fractures can store water for months, providing a crucial supply in the early growing season and helping forests cope with wet–dry changes in rainfall patterns.

The influence of climate change on drought in the UK has gained attention recently. However, a probabilistic assessment of temperature’s nonstationary influences on hydrological drought characteristics, which could provide key insights into future risks and uncertainties, has not been conducted. This study evaluates changes across seasons and warming scenarios, finding that rare droughts may become more severe, while frequent summer droughts are shorter but more intense.

We present a framework using observed rainfall and temperature to generate realistic storms and simulate street-scale flooding for present and future climates. It integrates temperature-based rainfall scaling, storm-frequency estimation, and urban flood modeling, demonstrated in Beijing to assess changes in regional storm and flood depth, timing, and flow velocity. The workflow is data-light, physically grounded, and transferable worldwide.

Compound hot-dry and the dry-wet seesaw transition are extreme meteorological events driven by the interaction between rainfall, temperature and soil moisture. Using data covering New Zealand for the 1950-2021 period we model these interactions using both copulas and single variables. The copula approach detects more frequent and intense events, particularly across eastern regions of the country, while also better representing the event drivers and improving our understanding of their risk.

Arctic rivers move enormous amounts of water and carbon into the ocean, influencing global climate, but their flow is hard to predict because the region is remote and the frozen ground behaves in unusual ways. This research combines artificial intelligence with the physics of snow and permafrost to forecast river flow more accurately. Demonstrated on the Kolyma River, the new model outperforms existing approaches and provides a robust framework for understanding Arctic hydrological systems.

The GRACE (Gravity Recovery and Climate Experiment) and GRACE Follow-On satellites monitor changes in Earth's water storage by observing gravity variations. By integrating these observations into hydrological models through data assimilation, estimates of groundwater, soil moisture, and hydrological trends are improved, helping to monitor droughts, floods, and human water use. This review highlights recent advances in GRACE data assimilation, identifies key challenges, and discusses future directions with upcoming satellite missions.

Water transport in forest soils occurs both vertically and horizontally. We added a stable isotope label to a small forest plot and monitored its passage vertically into the soil and horizontally into stems of surrounding trees. The labelled water was detected in the upper layers of the soil and in trees up to 6.7 m away, but was mostly taken up by one tree adjacent to the plot. These results affect how we think about summing over individual trees to describe the water economy of a whole forest.

As glaciers retreat in mountain regions, heavy rainstorms increasingly control how much soil and sediment rivers carry downstream. We analysed rainfall and sediment data over 21 years in the Austrian Alps and found heavy precipitation events becoming more important for sediment movement, although total annual suspended sediment transport is declining as glaciers shrink. This shift may increase flood hazards, affecting ecosystems and water quality downstream.

Improving how rainfall-runoff models estimate evapotranspiration is key to better reproducing water partitioning under current conditions, and will increase model realism under future changing conditions. We tested how well different conceptual rainfall-runoff model equations simulate evapotranspiration using Australian catchment and flux tower data. We found one equation consistently worked better than the others. However, even this equation had flaws, pointing to missing vegetation processes.

Observations and reanalyses show widespread declines in annual precipitation but increases in one-day maxima in Southern Europe. Very high-resolution simulations over a representative Mediterranean sub-region isolate sea surface temperature effects: with all else unchanged, projected warmer seas mainly raise the frequency of heavy rainfall over land by intensifying previously moderate events. For the most extreme cases, mean overland totals change little because peak rainfall remains offshore.

This study introduces a new interpretable deep learning method that accurately predicts multi-depth soil moisture simultaneously without physical assumptions. The model provides insights into soil properties, while delivering precise predictions across diverse scenarios. Tested under various conditions, it outperforms traditional approaches, particularly when enhanced with basic physics. This tool can help improve water management by offering reliable and efficient soil moisture forecasts.

Evapotranspiration is an integral element of the water availability estimation in tropical regions like West Africa. Climate change is projected to impact the water cycle. In this study, we evaluate the importance of AET (actual evapotranspiration) simulation for future climate impact assessments. We highlight differences if AET is or is not integrated in the model calibration. Our work contributes to the reduction of uncertainties in hydrological climate impact studies.

Using global and regional datasets, we compared attention-based models and Long Short-Term Memory (LSTM) models to predict hydrologic variables. Our results show LSTM models perform better in simpler tasks, whereas attention-based models perform better in complex scenarios, offering insights for improved water resource management.

The analysis of nutrient dynamics during precipitation events using normalised cumulative loads (NCL) is a classic and frequently used method of water quality analysis. Our approach combines the classical pollutograph with hydrograph separation using stable water isotopes. The pollutograph is set up for each of the separated runoff components (pre-event water and event water) and the export dynamics of the runoff components of an event are compared.

We developed a new approach to improve the climate data used to study water and glaciers in high-mountain regions. By correcting unrealistic air temperatures and moisture while keeping the data physically consistent, we achieved much better agreement with measurements and more accurate simulations of snow, glacier change, and river flow. The method also preserved long-term warming trends, making it useful for studying future climate impacts in mountain areas.

Iceland relies heavily on hydropower, making it important to understand changes in streamflow. Rivers reflect a warming, wetter climate, with rainfall increasingly replacing snowfall. Annual streamflow has generally increased, most notably during the cool season, while summer flows have declined in some non-glacial rivers. Glacial rivers show increasing melt-season flows over 1973–2023 but declining trends over 1993–2023, reflecting glacier retreat and recent North Atlantic sea surface cooling.

Plants typically transpire less with rising atmospheric carbon dioxide, leaving more water in the ground for human use, but many future water scarcity assessments ignore this effect. We use a land surface model to examine how plant responses to carbon dioxide and climate change affect future water scarcity. Our results suggest that including these plant responses increases overall water availability for most people, highlighting the importance of their inclusion in future water scarcity studies.

Explore2 provides hydrological projections for 1,735 French catchments. Using QUALYPSO (Quasi-Ergodic Analysis of Climate Projections Using Data Augmentation), this study assesses uncertainties, including internal variability. By the end of the century, low flows are projected to decline in southern France under high emissions, while other indicators remain uncertain. Emission scenarios and regional climate models are key uncertainty sources. Internal variability is often as large as climate-driven changes.

Throughfall in forests varies in space and time creating distinct patterns. We developed a novel throughfall monitoring approach for continuous, automated measurement that features 60 self-built and cost effective throughfall samplers. Collected data show the potential of the approach to capture throughfall variability at small distances, among and within rainfall events and between different trees species.

We present the design of an actual evapotranspiration product which joined the Copernicus Land Monitoring Service portfolio in December 2025. The product relies on free and open data and advanced modelling methods and has global coverage with a 300 m pixel resolution. A prototype dataset was compared against measurements from 206 geographically distributed stations, achieving good results. Such product will find multiple uses, including in water resources management and food security fields.

This study assesses how spatial patterns, temporal trends, and record length in hourly precipitation data affect annual maximum intensities estimated with stationary and non-stationary models across a climatically and topographically diverse region. Comparing five gridded datasets, we find consistent spatial patterns but notable differences in intensities, with non-stationary estimates generally slightly lower.

The link between rainfall extremes and temperature has been widely studied and a scale dependence of the rate of increase with temperature is commonly reported. Here we investigate, with high resolution data, how rainfall variability across scales, and the associated extremes, change with temperature. We confirm scaling behaviour and find that the scale invariant maximum observable singularity increases on average with greater temperature. Connection with previous results are discussed.

When floods happen during or shortly after droughts, the impacts of each of the events can be magnified. In hydrological research, defining these events represents a challenging and important task in the process of understanding where and why they occur. We have used real-word examples to address some of these challenges and show different approaches influence outcomes. We make suggestions on when to use which approach and outline some pitfalls of which researchers should be aware.

This study presents a cost-effective method to reconstruct lake bathymetry using only digital elevation model (DEM) data, without relying on field measurements or optical imagery. By simulating lake-level decline and using shoreline topography, it accurately infers underwater terrain. Validation across 12 lakes on the Tibetan Plateau and Lake Mead demonstrates good accuracy and generalizability, enabling scalable lake storage estimation for hydrological and ecological applications.

This study focuses on hyperdroughts (HDs) in central Chile, defined as years with a regional rainfall deficit exceeding 75 %. Only five HDs occurred in the last century (1924, 1968, 1998, 2019, 2021), but they caused disproportionate environmental and social impacts. In some systems, the effects were larger than expected from those considering moderate droughts and dependent on the antecedent conditions. HDs have analogs from the remote past, and they are expected to increase in the near future.

Water resources face more challenges from climate change and human activities. We improved global water modeling by developing a high-resolution system using Soil and Water Assessment Tool (SWAT+), using automated reproducible workflow. This approach simplifies tracking the progress of global impact assessment modelling efforts. The global model will further help assess water stress hot-spots and inform sustainable water management as further improvements come.

This technical note evaluates the efficacy of the Self-Organizing Map (SOM) algorithm to represent concentration-discharge hysteresis patterns in watersheds. Through a proof-of-concept with sediment hysteresis, we show that the SOM algorithm accurately characterizes loop shape and can also be used to reveal the controls of hysteresis patterns in watersheds. We developed a Python package along with this technical note to support broader adoption of the SOM algorithm for hysteresis workflows.

An ecosystem services model was applied across Germany to estimate water supply and carbon sequestration. The results showed that total annual water discharge and carbon sequestration for Germany is 85 billion m³ and 106 Tg C, respectively. Furthermore, we found that croplands provide the largest amount of water, deciduous broadleaf forests sequester most of the carbon, and wetlands are very effective in absorbing carbon. During extreme events, we noticed a real impact on both services.

This study integrates Xinanjiang (XAJ) and Temporal Convolutional Network - Gated Recurrent Unit (TCN-GRU) via Random Forest (RF) for streamflow simulation. It combines XAJ’s physical modeling with TCN-GRU’s temporal analysis. Validated in four hydrologically diverse basins, the model achieves Nash-Sutcliffe Efficiency (NSE) 0.971–0.991, outperforming traditional models. Robust in flood/interval simulations, analysis identifies dew point temperature and evaporation as key factors through three interpretable methods.

To improve hydrological uncertainty estimation, recent studies have explored machine learning (ML)-based post-processing approaches. Among these, quantile random forests (QRF) are increasingly used for their balance between interpretability and performance. We develop a hydrologically informed QRF trained in a multi-site setting. Our results show that the regional QRF approach is beneficial, particularly in catchments where local information is insufficient.

This study evaluates the extrapolation performance of long short-term memory (LSTM) networks in rainfall–runoff modeling, specifically under extreme precipitation conditions. The findings reveal that the LSTM cannot predict discharge values beyond a theoretical limit and that this limit is well below the extremity of its training data. This behavior results from the LSTM's gating structures rather than saturation of the cell states alone.

Belgium has faced intense droughts in recent years, causing major losses across sectors. To assess their rarity, we used a hydrological model to reconstruct fifty years of soil moisture in the country. We show that 2011–2020 experienced the most severe droughts since 1971, with nearly 30% of the decade under drought. We also show that rainfall-based indicators underestimate soil moisture droughts, so including soil-moisture monitoring can give decision-makers a clearer picture of drought risks.

This study presents a method to make the spatial resolution of global Water Storage Compartments (WSCs) compatible with terrestrial water storage (TWS) data from Gravity Recovery And Climate Experiment (GRACE) missions. The method compares the spatial structure of the WSCs and TWS by considering the correlation between neighboring grid cells. An isotropic Gaussian filter with an optimal filter width of 250 km is found to be the most suitable, ensuring compatibility for consistent comparison with GRACE data in hydrological applications.

The development of carbon processes in the water quality model WSFS (Watershed Simulation and Forecasting System)-Vemala presents a significant advancement in simulating both total organic and inorganic carbon dynamics, burial and emissions through a river/lake network. The addition of organic acids to the total alkalinity definition improved pH simulations and thus the simulation of CO₂ emissions in the acidic and organic rich waters of Finland. The new Vemala model provides a robust foundation to support water management in the future.

We studied an arid, endorheic basin in the Chilean Altiplano to understand how rainfall and evaporation affect groundwater and water availability. Using a rainfall-runoff model and 40 years of satellite data, we found that much of the water evaporates and less reaches the aquifers than expected. Our results challenge the idea that the basin is fully closed and suggest that current water budget estimates may need revision, an urgent task under a changing climate.

We investigated how glaciers helped sustain summer river flow during the severe snow shortages of 2022 and 2023 in the Italian Alps. Using high-resolution modeling and long-term data, we found that glacier melt contributed up to three times more than in average years. This enhanced melt partly compensated for low snow and precipitation, showing the critical role of glaciers in supporting water resources during droughts.

This study compares three hydrological modeling approaches for assessing climate change impacts on water systems. It evaluates the conventional method alongside a fully- and semi-asynchronous methods, which excels in capturing extreme events but faces challenges with event timing. The results highlight the potential of the semi-asynchronous method as an innovative and robust tool for hydrological modeling under climate change.

We improved the widely used SWAT (Soil and Water Assessment Tool) model to better predict how pesticides move through the environment. We added a new process that considers how plants take-up chemicals from the soil. Testing this updated model in two catchments showed very good prediction capabilities and a reduction of chemicals in river water by up to 17 % due to the plant uptake. The enhanced model offers a valuable tool for assessing the environmental impacts of agricultural management.

We developed a literature-based approach for projecting future snow depths, which was applied to four measurement stations in Switzerland under a +2 °C temperature scenario, revealing significant declines in snow depths. Validation against published data shows that the approach captures key trends in snow loss. This resource-efficient method provides a practical tool for estimating climate change related snow depth declines, which are lacking highly resolved climate projections.

Automatic water samplers help to collect water for isotope analysis, but if the samples are stored in open bottles for long periods of time, their composition can change due to evaporation. We have tested various barriers to prevent this and found that the tube-dip-in method as an evaporation barrier is the most practical solution. Our results highlight the importance of bottle design and handling, especially for small sample volumes, to ensure accurate water analysis.

A five-year study of hot springs at a fault intersection on the southeastern Tibetan Plateau developed an anomaly detection model that links synchronous changes in water chemistry to earthquakes with magnitude ≥4. The model combines multiple components to improve accuracy of earthquake timing forecasting and identify reliable predictors. Stronger or closer earthquakes show more components with synchronous anomalies, providing a valuable reference for real-time forecasting in high-risk areas.

Forests can change the kinetic energy of rain below them. We built a new model that breaks down the canopy into layers, and tracks two types of raindrop: direct splashes and water dripping from leaves. The model was validated through nine rainfall events. The canopy doesn't always reduce the rain's force, and sometimes it increases it, depending on the specific structure of the leaves and branches.

LSTM (Long Short-Term Memory) networks are excellent deep-learning tools to reproduce stream temperature observations, but their performances over the range of extreme (summer) stream temperature values have been overlooked. We tested strategies to improve the LSTM performances over the highest 10 % values of stream temperature observations. We found that the best strategy is to train the LSTM models at several locations with input variables that include static, catchment and reach attributes.

We present a neural network model that estimates river discharge based on gridded elevation, precipitation, and solar radiation. Some instances of our model produce more accurate forecasts than the European Flood Awareness System (EFAS) when simulating discharge with lead times of 50 days on the Neckar river network in Germany. It consists of multiple components that are designed to model distinct sub-processes. We show that this makes the model behave in a more physically realistic way.

Many existing datasets for hydrologic analysis tend to treat catchments as single spatially homogeneous units focusing on daily data and typically do not support more complex models. This paper introduces a dataset that goes beyond this set-up by (1) providing data at a higher spatial and temporal resolution, (2) specifically considering the data requirements of all common hydrologic model types, and (3) using statistical summaries of the data aimed at quantifying spatial and temporal heterogeneity.

Climate change impacts on a hydrologic phenomenon termed the mid-summer drought (MSD), affecting water and food security of smallholder farmers in Central America, depends on how the MSD was defined. We used many climate model projections and MSD definitions to quantify the sources of uncertainty in future MSD characteristics in Nicaragua for varying levels of global warming. We found that the definition can change where the MSD occurs but it does not add substantially to the total uncertainty.