AbstractStudy RegionThe study region is the Magdalena River basin in Colombia. The basin was divided into three distinct regions (Andean, Caribbean, and Pacific) and analyzed across different elevations.Study FocusThe study proposes a Spatiotemporal Non-Linear Dynamics Assessment (SNLDA) framework to compare ERA5-Land reanalysis data with in-situ rain gauge observations. It specifically examines the constraints imposed by nonlinear dynamical processes and their associated space-time complexities on the representation of precipitation, particularly in a tropical region. The SNLDA framework incorporates three main components: (i) standard performance metrics (e.g., correlations, RMSE, and dry spell duration), (ii) rainfall spatiotemporal objects (characterizing precipitation events through attributes such as volumes and start-end centroids), and (iii) non-linear dynamics complexity (reconstructing dynamical behavior from time series and evaluating attractors properties, including the Hurst and Lyapunov exponents). These elements were analyzed both individually and in combination. Daily ERA5-Land information (0.1°x0.1°) and in-situ rain gauge data comprising 558 stations from 1980 to 2020 were used, enriched by an Inverse Distance Weighting (IDW) interpolation (0.1°x0.1°) to facilitate comparison across spatial scales.New Hydrological Insights for the RegionOverall, ERA5-Land overestimates precipitation, producing shorter, more frequent events while poorly representing extreme wet and dry spells.Andean region: ERA5-Land overestimates rainfall, with largest errors at low elevations, driven by unresolved spatiotemporal object volumes displacements and nonlinear processes.Caribbean region: ERA5-Land shows the highest errors in nonlinear dynamics and extremes, despite lower annual bias and RMSE.Pacific region: ERA5-Land strongly overestimates precipitation volumes and RMSE, while nonlinear errors remain low; these biases are mainly driven by spatiotemporal objects displacement.

Model predictive control (MPC) is an optimal control strategy suited for flood control of water resources infrastructure. Despite many studies on reservoir flood control and their theoretical contribution, optimisation methodologies have not been widely applied in real-time operation due to disparities between research assumptions and practical requirements. To address this gap, we include practical objectives, such as minimising the magnitude and frequency of changes in the existing outflow schedule. Incorporating these objectives transforms the problem into a multi-objective nonlinear optimisation problem that is difficult to solve in real time. Additionally, it is reasonable to assume that the weights and some parameters, considered the operators’ preferences, vary depending on the system state. To overcome these limitations, we propose a framework that converts the original intractable problem into parameterised linear MPC problems with dynamic optimisation of weights and parameters. This is done by introducing a model-based learning concept. We refer to this framework as Parameterised Dynamic MPC (PD-MPC). The effectiveness of this framework is demonstrated through a numerical experiment for the Daecheong multipurpose reservoir in South Korea. We find that PD-MPC outperforms standard MPC-based designs without a dynamic optimisation process for the objective weights and model parameter. Moreover, we demonstrate that the weights and parameters vary with changing hydrological conditions.

Study region: Madeira River basin, southwestern Amazonia Study focus: This study investigates spatial and temporal changes in precipitation, evaporation, and streamflow, and their relationship with deforestation in the Madeira River basin, the largest Amazonian sub-basin. We applied Mann-Kendall trend analysis, change-point detection, and correlation analysis across multiple spatial scales, using satellite, reanalysis, and observed data from 1981 to 2015. These methods enabled us to detect long-term trends, identify shifts, and quantify the relationships between forest loss and hydrological changes New hydrological insights for the region: The basin experienced an average deforestation rate of 2810 km² per year from 2001 to 2020, predominantly in the Brazilian portion. Between 1981 and 2016, we observed statistically significant negative trends in precipitation, evaporation, and streamflow, especially in the most deforested areas during the wet season. Correlation analysis (2001–2015) showed a statistically significant and positive relationship between forest area and evaporation in wet months (r = 0.73, p < 0.1) and a negative correlation between forest area and streamflow during the same season (r = –0.6, p < 0.1). These findings highlight the critical role of forests in modulating hydrological processes, supporting the hypothesis that deforestation may reduce evaporation, alter moisture recycling, and slow the water cycle. While our results are robust, we acknowledge that factors such as climate variability and land management practices may also influence hydrological changes and should be considered in future research.

Long-term water resource management involving multipurpose coordination requires robust decision-making in water infrastructure cases to cope with various types of uncertainties. Traditional robust optimization methods generally do not explicitly propagate input or parametric uncertainties into estimates of the robustness of solutions, which limits their ability to address uncertainty comprehensively across solution spaces. In this study, we introduce an explicit robust decision-making framework that blends multiobjective search, probabilistic analysis of robustness, and diagnostic verification tools to identify robust optimal solutions to external uncertainty. The proposed framework is illustrated on four diverse robustness formulations, which capture a wide variety of stakeholder attitudes from highly risk-averse to risk-neutral, for the primary operating objectives (hydropower production, water diversion, and hydrological alteration degree) in China's Hanjiang cascade reservoir system. By analyzing the Pareto front propagated from inflow uncertainty, it is found that optimal robust policies with a significantly higher degree of hydrological alteration are preferred in most formulations to achieve relatively lower joint uncertainty of hydropower and water diversion. These policies also yield sufficiently stable model performance in the case of an out-of-sample streamflow set during diagnostic verification. Furthermore, a comparative analysis of four different formulations suggests that a composite normalized robustness indicator (NRI) developed in this study to integrate various robustness metrics can achieve an effective balance for all considered objectives. These findings highlight the benefits of explicit robust optimization for managing hydrological uncertainties in multipurpose cascade reservoirs.

While drought impacts are widespread across the globe, climate change projections indicate more frequent and severe droughts. This underscores the pressing need to increase resistance and resilience to drought. The strategic application of Preventive Drought Management Measures (PDMMs) is a suitable avenue to reduce the likelihood of drought and ameliorate associated damages. In this study, we use an optimisation approach with a multicriteria decision-making method to allocate PDMMs for reducing the severity of agricultural and hydrological droughts. The results indicate that implementing PDMMs can reduce the severity of agricultural and hydrological droughts, and the obtained management scenarios (solutions) highlight the utility of multi-objective optimisation for PDMMs planning. However, examined management scenarios also illustrate the trade-off between managing agricultural and hydrological droughts. PDMMs can alleviate the severity of agricultural droughts while producing opposite effects for hydrological droughts (or vice versa). Furthermore, the impact of PDMMs displays temporal and spatial variabilities. For instance, PDMMs implementation within a specific subbasin may mitigate the severity of one type of drought in a given month yet exacerbate drought conditions in preceding or subsequent months. In the case of hydrological droughts, the PDMMs may intensify streamflow deficits in the intervened subbasins while alleviating the hydrological drought severity downstream (or vice versa). These complexities emphasise a customised implementation of PDMMs, considering the basin characteristics (e.g., rainfall distribution over the year, soil properties, land use, and topography) and the quantification of PDMMs' effect on the severity of each type of drought.

Kazakhstan is recently experiencing an increase in drought trends. However, low-capacity probabilistic drought forecasts and poor dissemination have led to a drought crisis in 2021 that resulted in the loss of thousands of livestock. To improve drought forecasting accuracy, this study applies Machine Learning and Deep Learning (ML and DL) algorithms to capture the sequences of drought events using a non-contiguous drought analysis (NCDA). Precipitation, 2-m temperature, runoff, solar radiation, relative humidity, and evaporation were collected from the ERA5 database as input variables. Combinations of inputs were used to build ML models, including seven classifiers (Logistic, K-NN, Kernel SVM, Decision Tree, Random Forest, XGBoost, and GRU). The output events were defined by standardized precipitation index (SPI) and SPEI indicators as binary classes. Weekly time series from 1991 to 2021 for each cell were used to forecast a lead time from 1 week to 6 months. GRU provided 97–99% accuracy in more volatile regions while Random Forest and XGBoost showed 94–99% accuracy at a lead time of 6 months. The accuracy evaluation was based on the confusion matrix and F1 score to analyze the stage change capture. This study demonstrates the effectiveness of using ML and DL algorithms for drought forecasting, with potential applications for other regions.

Extreme peak runoff forecasting is still a challenge in hydrology. In fact, the use of traditional physically-based models is limited by the lack of sufficient data and the complexity of the inner hydrological processes. Here, we employ a Machine Learning technique, the Random Forest (RF) together with a combination of Feature Engineering (FE) strategies for adding physical knowledge to RF models and improving their forecasting performances. The FE strategies include precipitation-event classification according to hydrometeorological criteria and separation of flows into baseflow and directflow. We used ∼ 3.5 years of hourly precipitation information retrieved from two near-real-time satellite precipitation databases (PERSIANN-CCS and IMERG-ER), and runoff data at the outlet of a 3391-km² basin located in the tropical Andes of Ecuador. The developed models obtained Nash-Sutcliffe efficiencies varying from 0.86 to 0.59 for lead times between 1 and 6 h. The best performances were obtained for peak runoffs triggered by short-extension precipitation events (<50 km²) where infiltration- or saturation-excess runoff responses are well learned by the RF models. Conversely, the forecasting difficulty is associated with extensive precipitation events. For such conditions, a deeper characterization of the biophysical characteristics of the basin is encouraged for capturing the dynamic of directflow across multiple runoff responses. All in all, the potential to employ near-real-time satellite precipitation and the use of FE strategies for improving RF forecasting provides hydrologists with new tools for real-time runoff forecasting in remote or complex regions.

Water utilities are urged to decrease their real water losses, not only to reduce costs but also to assure long-term sustainability. Hardware- and software-based techniques have been broadly used to locate leaks; within the latter, previous works that have used data-driven models mostly focused on single leaks. This paper presents a methodology to locate multiple leaks in water distribution networks employing pressure residuals. It consists of two phases: one is to produce training data for the data-driven model and cluster the nodes based on their leak-flow-rate-independent signatures using an adapted hierarchical agglomerative algorithm; the second is to locate the leaks using a top-down approach. To identify the leaking clusters and nodes, we employed a custom-built k-nearest neighbor (k-NN) algorithm that compares the test instances with the generated training data. This instance-to-instance comparison requires substantial computational resources for classification, which was overcome by the use of high-performance computing. The methodology was applied to a real network located in a European town, comprising 144 nodes and a total length of pipes of 24 km. Although its multiple inlets add redundancy to the network increasing the challenge of leak location, the method proved to obtain acceptable results to guide the field pinpointing activities. Nearly 70% of the areas determined by the clusters were identified with an accuracy of over 90% for leak flows above 3.0 L/s, and the leaking nodes were accurately detected over 50% of the time for leak flows above 4.0 L/s.

Conceptual hydrological models imply a simplification of the complexity of the hydrological system; however, they lack the flexibility in reproducing a wide range of the catchment responses. Usually, a trade-off is done to sacrifice the accuracy of a specific aspect of the system behavior in favor of the accuracy of other aspects. This study evaluates the benefit of using a modular approach, “The fuzzy committee model” of building specialized models to reproduce specific responses of the catchment. We also assess the applicability of using predicted runoff from specialized models to form a fuzzy committee model. In this paper, weighting schemes with power parameter values are investigated. A thorough study is conducted on the relation between the fuzzy committee variables (the membership functions and the weighting schemes), and their effect on the model performance. Furthermore, the Fuzzy committee concept is applied on a conceptual distributed model with two cases, the first with lumped catchment parameters and the latter with distributed parameters. A comparison between different combinations of the fuzzy committee variables showed the superiority of all Fuzzy Committee models over single models. Fuzzy committee of distributed models performed well, especially in capturing the highest peak in the calibration data set; however, it needs further study of the effect of model parameterization on the model performance and uncertainty.

Successful modelling of the groundwater level variations in hydrogeological systems in complex formations considerably depends on spatial and temporal data availability and knowledge of the boundary conditions. Geostatistics plays an important role in model-related data analysis and preparation, but has specific limitations when the aquifer system is inhomogeneous. This study combines geostatistics with machine learning approaches to solve problems in complex aquifer systems. Herein, the emphasis is given to cases where the available dataset is large and randomly distributed in the different aquifer types of the hydrogeological system. Self-Organizing Maps can be applied to identify locally similar input data, to substitute the usually uncertain correlation length of the variogram model that estimates the correlated neighborhood, and then by means of Transgaussian Kriging to estimate the bias corrected spatial distribution of groundwater level. The proposed methodology was tested on a large dataset of groundwater level data in a complex hydrogeological area. The obtained results have shown a significant improvement compared to the ones obtained by classical geostatistical approaches.