The Ucayali River is the major Andean conveyor of sediment to the Amazon. By combining field measurements, satellite data and modelling over decades, we show that floodplains trap 36% of this sediment while recycling 22% back into the river during flood recession. During high waters, flooding reduces the river's capacity to carry sand, capturing 14% at peak discharge. These findings show that floodplains act as dynamic regulators of sediment transport in large tropical rivers.

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.

Understanding how baseflow contributes to river flow is essential for managing water resources. We studied a widely used method for separating baseflow and found that a key parameter was often estimated too simply. Using symbolic regression and data from 855 catchments, we uncovered new formulas that greatly improve accuracy and reveal how soil, snow, and catchment size jointly influence baseflow estimation.

The quality of large-domain hydrologic model simulations is often quantified with so-called accuracy metrics. Here we use simple benchmarks to provide relevant context for these accuracy metrics. Results show that areas where the model cannot beat the benchmarks do not always align with areas where the accuracy metrics are low. This suggests that model improvements are possible in regions that under more typical model evaluation approaches (i.e., without benchmarks) might not be obvious.

Identifying drivers is crucial for process understanding and predictions. In Hydrometeorological systems, many variables are closely related, and common methods often rely on correlation. We describe theoretically distinct methods of discovering cause-effect relations from data. We evaluate them in a large simulated environment. Results show that finding cause-effect relations provides a parsimonious picture and to obtain robust predictions, especially under changing environmental conditions.

This paper highlights the importance of discharge assimilation (DA) for artificial intelligence (AI)-based operational discharge forecasting. Using two public datasets from France and the USA, simulated discharge from two rainfall-runoff models, and a multilayer perceptron for implementation, we evaluate three DA strategies under both deterministic and probabilistic forecasting approaches. Results show that DA is crucial and that model performance may decrease between the two forecasting cases.

Climate change and human activities jointly shape river droughts, yet their combined impacts remain uncertain. We pair a data-driven river model with scenario-based climate projections to assess future water shortages in China’s Upper Hanjiang River. We also evaluate improved reservoir operating rules. Results show rising risk of prolonged drought, while refined reservoir operations ease short events but cannot offset long-term deficits, informing resilient water-energy planning.

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.

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.

Convection-permitting regional climate models, such as ICON-CLM at 3 km resolution, have great potential for improved hydroclimatic simulations. The studied model run shows lower bias in air temperature and summer global radiation, as well as in the frequency of wind speed over the Weiße Elster catchment in East Central Germany. Due to a pronounced overestimation of the intensity and frequency of heavy rainfall, however the discharge estimates are skewed, with no apparent added value.

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 mapped dominant hydrologic processes across the United States by analyzing observed streamflow dynamics. Using random forest models and interpretable machine learning techniques, we predicted processes in data-scarce regions and identified key drivers such as climate, soil and geology, land cover, topography, and human influence. The resulting maps of dominant processes and their drivers reveal strong regional patterns that guide hydrologic model selection and water resource management.

We developed a satellite-based irrigation advisory system that operates weekly, helping water providers make informed, science-based decisions. It estimates crop water needs using satellite data combined with rainfall and past irrigation and can also be used to simulate future cropping patterns under policy changes or reduced water supply. Co-developed with stakeholders, it is scalable to other regions with similar water management challenges.