Rainfall-runoff modeling in ungauged basins continues to be a great hydrological research challenge. A novel approach is the Long-Short-Term-Memory neural network (LSTM) from the Deep Learning toolbox, which few works have addressed its use for rainfall-runoff regionalization. This work aims to discuss the application of LSTM as a regional method against traditional neural network (FFNN) and conceptual models in a practical framework with adverse conditions: reduced data availability, shallow soil catchments with semiarid climate, and monthly time step. For this, the watersheds chosen were located on State of Ceará, Northeast Brazil. For streamflow regionalization, both LSTM and FFNN were better than the hydrological model used as benchmark, however, the FFNN were quite superior. The neural network methods also showed the ability to aggregate process understanding from different watersheds as the performance of the neural networks trained with the regionalization data were better with the neural networks trained for single catchments.

Inter-basin water transfers are the root of many conflicts, and water scarcity accentuates them. Those conflicts involve the priority of water use between regions. The Jaguaribe Metropolitan system, located in the Brazilian semiarid region, presents conflicts amongst different water users: irrigated perimeters, industry, and households. This paper analyzed the Jaguaribe Metropolitan water transfer during the 2012–2018 drought by considering environmental and societal aspects. Changes in consumption and users’ drought perception were assessed. The results showed that the drought was longer and more severe in the region that provided water (i.e., Jaguaribe) than in the region that received it (i.e., FMR). Jaguaribe irrigators were aware of the beginning of the drought, but it did not result in immediate consumption control. On the other hand, drought perception was delayed in the FMR. The results of this study suggested that the water allocation decision-making process should include not only the water demands but also the characteristics of the drought and how people perceive it. The main strategy for improving water governance seems to be promoting integrated regional planning and the empowerment of participatory management.

Predicting in ungauged basins (PUB) depends on the modelling uncertainties in the donor catchments (DCs). However, PUB is normally limited to a unique outcome, which may be quite uncertain, mainly in semi-arid basins where streamflow variability is high. Our goal is to assess the uncertainty in the parameter regionalization for streamflow prediction in semi-arid ungauged basins (UB). We used Differential Evolution Adaptive Metropolis (DREAM) for a parameter calibration that considers its intrinsic uncertainty. A basin similarity regionalization was performed to transfer the parameters to UB, considering one, three and five DCs. Leave-one-out cross-validation was applied for 28 gauged catchments. Regionalization performance had average Nash-Sutcliffe efficiency above 0.50. The approach considering one DC performed better than the others. The model of a poorly monitored catchment performed better using the transferred parameters from long-recorded, similar catchments than using those calibrated in the catchment itself. The developed regionalization may be a relevant tool for water management in drylands.

Streamflow forecasting plays an important role in ensuring the reliable supply of electricity in countries heavily reliant on hydropower. This paper proposes a novel framework that integrates various hydrological models, climate models, and observational data to develop a comprehensive forecasting system. Three families of models were employed: seasonal forecasting climate models integrated with hydrological rainfall-runoff models; stochastic or machine learning models utilizing endogenous variables, and stochastic or machine learning models that consider exogenous variables. The hyper-multimodel framework could successfully increase the overall performance of the scenarios generated through the use of the individual models. The quality of the final scenarios generated was directly connected to the performance of the individual models. Therefore, the proposed framework has potential to improve hydrological forecast for the Brazilian electricity sector with the use of more refined and calibrated individual models.

Synthetic streamflow scenario generation is particularly important in countries like Brazil, where hydroelectric power generation plays a key role and properly handling the uncertainty of future streamflow is crucial. This paper showcases a collaborative effort within the Brazilian electrical sector to enhance streamflow scenario models, focusing on horizons up to one year. Five institutions proposed diverse methodologies, and their effectiveness was evaluated using a comparative framework. The results reveal the strengths and areas for improvement in each model. GHCen emerged as the top performer, excelling in both short-term and moving average analyses, while the PARX model demonstrated superior performance in specific regions. The PAR(p)-A, which is the official methodology in Brazil, was the second-best model in the moving average analysis. This research offers valuable insights for countries facing similar hydrothermal scheduling and scenario generation challenges.