Recent  

We generate operational forecasts of daily maximum stream water temperature for the next month at 54 stations in Switzerland with our best performing data-driven model. The average forecast error is 0.38 °C for 1 day ahead and increases to 0.90 °C for 1 month ahead given the uncertainty in the meteorological variables influencing water temperature. Here we compare the skill of several models, how well they can forecast at new and ungauged stations, and the importance of different model inputs.

Stream water temperature (SWT) is a key indicator of water quality that benefits public use and aquatic life health. Advances in computer modeling have helped improve our understanding of SWT dynamics, but challenges remain. Recently, scientists have begun to use machine learning (ML) in SWT modeling, but it is unclear how ML is increasing our understanding of SWT causes and effects. This work reviews the application of ML in SWT modeling and discusses where there is still room for improvement.

The subsurface of sandy beaches under high energy conditions where tides, waves and storms constantly reshape the beach surface are globally common and relevant for the alteration of solute fluxes across the land-sea continuum. Our generic modelling study shows that a dynamic beach morphology paired with aquifer properties results in spatio-temporal variable flow and solute transport regimes in the subsurface and enhances mixing reactions.

This paper introduces a statistical method to estimate probable maximum precipitation (PMP), addressing flaws in current approaches. The method accounts for uncertainty and incorporates climate change impacts, using the Pearson Type I distribution. Tested at two meteorological stations in Québec, it offers an objective solution for more reliable PMP estimates, crucial for infrastructure like dams exposed to extreme weather.

We present a new method for extracting water from soil for stable isotope analysis. The reason for initiating the development was the need to easily and accurately extract sufficient amounts of water from soil samples. The developed apparatus for this method has a medium throughput, high accuracy and high precision. The method is suitable for experiments and studies where high precision is required to distinguish between different pools of water and where mere trend detection is not sufficient.

A warming climate alters the freshwater availability over land, and due to related tree cover change and potential forestation this availability can be further enhanced or negated. We find that large-scale change in tree cover counteracts climate-driven changes on a global scale, whereas regionally the climate and tree cover impacts can differ extensively. Current ecosystem restoration projects should account for the effects of (re)forestation on (non-)local water availability.

This study explores six methods to improve the ability of Long Short-Term Memory (LSTM) neural networks to predict peak streamflows, crucial for flood analysis. By enhancing data inputs and model techniques, the research shows LSTM models can match or surpass traditional hydrological models in simulating peak flows. Tested on 88 catchments in Quebec, Canada, these methods offer promising strategies for better flood prediction.

The fate of liquid water from melting snow in winter and spring is difficult to understand in the mountains. This work uses uncommon methods to accurately track the dynamics of snowmelt and infiltration at different depths in the ground and at different altitudes. The results show that melting snow quickly infiltrates into the upper layers of the soil but is also quickly transferred into the surface layer of the soil along the slopes towards the river.

Hydrology models rely on simplistic static approaches to precipitation phase partitioning. We evaluate model skill changes for a suite of snow metrics by transitioning to a more accurate dynamic partitioning. We found that the transition resulted in a better match between modeled and observed metrics, with a 50 % reduction in model bias, emphasizing the need for the hydrological modeling community to adopt dynamic partitioning.

The release of carbon from landscapes into streams is one important component within the global carbon cycle. We measured the concentrations of dissolved organic carbon (DOC), one of the forms in which carbon can be present, in the streams of three nested forested subcatchments over 12 months. The export of DOC is closely linked to water flow processes within the subcatchments, but the interplay of soils, vegetation, topography and microclimate results in distinct seasonal DOC release patterns.

We produced a new interpolated map of stable isotopes in groundwater in southern Germany and compared it to local precipitation. Interestingly, discrepancies were found between those two compartments of the hydrological cycle, highlighting different recharge patterns and evaporation processes in the northern and southern part of the study area. This research provides insights to understand the different groundwater recharge patterns on a large scale and eventually for groundwater management.

Through a survey, involving six-water dependent sectors, the relevance and impact of drought vulnerability factors for sectors and societies in forested cold climates was studied. The results show that the relevance and impact of vulnerability factors differ across sectors and how governance processes and policies are important for drought risk. The results offer unique insights into the dynamics of drought vulnerability, valuable for risk assessments, drought plans and increasing resilience.

In hydrology, domain dependence (DD) of spatial Principal Component patterns is a rather unknown feature of the widely applied Principal Component Analysis. It easily leads to wrong hydrological interpretations. DD reference patterns enable to differentiate from the effect. Here, we (1) explore the DD effect, (2) present two methods to calculate DD reference patterns and (3) discuss considering DD. Scripts with an introduction to the DD effect and an implementation of both methods are provided.

This study introduces a method for estimating High-Resolution Snow Water Equivalent (HR-SWE) using Low-Resolution Climate Data (LR-CD). By applying a data-driven approach, we utilize historical weather patterns from LR-CD to estimate HR-SWE maps. Our approach uses statistical relationships between LR-CD and HR-SWE data to provide HR-SWE estimates for dates when HR-SWE data is unavailable. This method improves water resource management and climate impact assessments in regions with limited data.

Coastal-urban areas are highly prone to flooding from rainfall, storm surge, and their combination. We improve a coastal model and use it to quantify flooding from Hurricane Ida in the Jamaica Bay watershed of NYC, creating a flood map and flooded area estimation. Experiments with shifted storm tracks and rainfall timing at high tide show that Ida, already the worst rainfall in NYC, could have been worse. This highlights the area's vulnerability and the need for thorough flood risk analysis.

This study introduces RDV-Copula, a new method to simplify complex vine copula structures by reducing dimensionality while retaining essential data. Applied to Shifeng Creek in China, RDV-Copula captured critical spatial-temporal relationships, demonstrating high synchronization probabilities and significant flood risks. Notably, it was found that increasing structure complexity does not always improve accuracy. This method offers an efficient tool for analyzing and simulating multisite flows.

Earth scientists assume that evaporation depends on the gradient in water vapor concentration (Fick’s law) but this is only true in an isothermal system. Temperature gradients can impact evaporation via the Soret effect. Here we evaluate the relative magnitude of the Soret effect and find that it is at least two orders of magnitude smaller than classical concentration-dependent mass (‘Fickian’) diffusion. This result justifies the standard practice of assuming evaporation follows Fick’s law.

Models help to understand natural systems and are used to predict changes based on scenarios e.g. climate change. To simulate water temperature and deduce impact on water quality in lakes, 1D hydrodynamic models are often used. There are several such models which differ in their assumptions and mathematical process description. This study examines the performance of four such models on a global dataset of 73 lakes and relates the performance to the models structure and the lake characteristics.

Future drought changes in Australia –the driest inhabited continent on Earth– have remained stubbornly uncertain. We assess future drought changes in Australia using projections from climate and hydrological models. We show an increasing probability of drought over highly-populated and agricultural regions of Australia in coming decades, suggesting potential impacts on agricultural activities, ecosystems and urban water supply.

Water budget non-closure is a widespread phenomenon among multisource datasets, which undermines the robustness of hydrological inferences. This study proposes a Multisource Datasets Correction Framework grounded in Physical Hydrological Processes Modelling to enhance water budget closure, called PHPM-MDCF. We examined the efficiency and robustness of the framework using the CAMELS dataset, and achieved an average reduction of 49 % in total water budget residuals across 475 CONUS basins.

Estimating extreme streamflows is essential for identifying flood-prone areas and designing safe river infrastructure. However, most river sections lack direct measurements of streamflows. Hydrologists use spatial interpolation and hydrological modeling to estimate streamflows in these unmeasured areas, but these estimates come with uncertainties. Our study introduces a new model to better account for these uncertainties, improving the accuracy of predicting extreme streamflows.

This study compares Long Short-Term Memory (LSTM) neural networks with traditional hydrological models to predict future streamflow under climate change. Using data from 148 catchments, it finds that LSTM models, which learn from extensive data sequences, perform differently and often better than traditional hydrolgical models. The continental LSTM model, which includes data from diverse climate zones, is particularly effective for understanding climate impacts on water resources.

We proposed an innovative model to improve flood prediction and management, mitigating the impact of extreme inundations. Using advanced deep learning techniques, the proposed model accurately forecasts long-lasting river flood events in negligible computational time. The results show significant improvements in prediction accuracy, providing communities and policymakers with a valuable tool for flood management and early warning systems.

This study estimated the water balance of Lake Titicaca using an integrated modeling framework that considers natural hydrological processes and net irrigation consumption. The proposed approach was implemented at a daily scale for a period of 35 years. This framework is able to simulate lake water levels with good accuracy over a wide range of hydroclimatic conditions. The findings demonstrate that a simple representation of hydrological processes is suitable for use in poorly-gauged regions.

To understand how eco-hydrological variables evolve jointly and why, this study develops a framework using correlation and causality to construct complex relationships between variables at the system level. Causality provides more detailed information that the compound causes of evolutions regarding any variable can be traced. Joint evolution is controlled by the combination of external drivers and direct causality. Overall, the study facilitates the comprehension of eco-hydrological processes.

This study compared global satellite and one reanalysis precipitation dataset to assess diurnal variability. We found that all datasets capture key diurnal precipitation patterns, with maximum precipitation in the afternoon over land and early morning over the ocean. However, there are differences in the exact timing and amount of precipitation. This suggests that it is better to use a combination of datasets for potential applications rather than relying on a single dataset.

Data-driven techniques have shown the potential to outperform process-based models in rainfall-runoff simulations. Hybrid models, combining both approaches, aim to enhance accuracy and maintain interpretability. Expanding the set of test cases to evaluate hybrid models under different conditions we test their generalization capabilities for extreme hydrological events.

Scientists and engineers need better indices to frame the hydrologic effects of land development. Existing approaches are not able to reflect the interactions due to the spatial arrangement of distinct land patches, which affect how much runoff is generated and how fast it can travel downstream, impacting flood response. Our novel, GIS-based modeling framework explicitly considers these aspects and is applicable to a wide range of problems, including peak-flow predictions in ungauged basins.

Flood peak distributions indicate how likely the occurrence of an extreme flood is at a certain river. If the distribution has a so-called heavy tail, extreme floods are more likely than might be anticipated. We find heavier tails in small compared to large catchments, and that spatially variable rainfall leads to a lower occurrence probability of extreme floods. Spatially variable runoff does not show an effect. The results can improve estimations of occurrence probabilities of extreme floods.

Rainfall and evaporation are two key variables influencing when droughts occur and their severity. We use the latest climate simulations for Australia to see how changes to rainfall and evaporation influence future droughts for Australia and show increases are likely over most of Australia, especially in the south. Increases to evaporation are shown to be larger than changes to rainfall over most of the continent. We show that keeping emissions to lower levels can work to mitigate these impacts.

Hydrometeorological forecasting is crucial for managing water resources and mitigating extreme weather impacts, yet current long-term forecast products are often embedded with uncertainties. We develop a deep learning based modelling framework to improve 30-day rainfall and streamflow forecasts by combining advanced neural networks and outputs from physical models. With the forecast error reduced by up to 32%, the framework has the potential to enhance water management and disaster preparedness.