Mena et al. (2026) A deep learning-based probabilistic framework for streamflow prediction and water quality assessment

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-06-26
• Authors: Alberto Mena, Rafael J. Bergillos, Javier Paredes-Arquiola, Gerald Corzo, Abel Solera, Joaquín Andreu-Álvarez
• DOI: 10.1016/j.jhydrol.2026.135934

Research Groups

• Instituto de Ingeniería del Agua y Medio Ambiente (IIAMA), Universitat Politècnica de València, Spain
• Coastal and Urban Risk Resilience, IHE Delft, the Netherlands

Short Summary

This study presents an integrated modelling framework for water quality assessment, combining deep-learning based inflow prediction, stochastic uncertainty propagation, hydrological spatial disaggregation, and process-based pollutant modelling. The framework is applied to the Júcar River Basin (eastern Spain), a highly regulated Mediterranean system characterized by strong hydrological variability and significant water quality pressures.

Objective

• To develop an integrated modelling framework for water quality assessment that combines deep-learning based inflow prediction, stochastic uncertainty propagation, hydrological spatial disaggregation, and process-based pollutant modelling.
• To apply the proposed approach to the Júcar River Basin (eastern Spain) and evaluate its performance in reproducing pollutant concentrations and water chemical status.

Study Configuration

• Spatial Scale: The study focuses on the Júcar River Basin, a Mediterranean river basin with complex hydrological and water quality dynamics.
• Temporal Scale: The study uses daily meteorological inputs to drive the HBV hydrological model, and monthly observed streamflow data were used to train the LSTM models.

Methodology and Data

• Models used:
  • Long Short-Term Memory (LSTM) neural networks for subbasin inflow prediction.
  • Hydrological module EvalHid of the Aquatool Decision Support System (HBV model) for hydrological modelling using the HBV rainfall-runoff model.
  • Rapid Response to Environmental Status (RREA) water quality model for pollutant concentration simulations.
• Data sources:
  • Meteorological data, including precipitation and temperature series, were provided by the Spanish State Meteorological Agency (AEMET).
  • Hydrological and water quality data were retrieved from the JRB Information System (SIA-Júcar), an official platform managed by the JRB Authority.

Main Results

• The proposed framework successfully reproduces the main spatial and temporal patterns of pollutant concentrations while explicitly quantifying the uncertainty associated with hydrological predictions.
• Water chemical status classifications show high agreements with observations, achieving accuracies of 90% for BOD, 84.4% for phosphorus, and 85.3% for ammonium.

Contributions

• The proposed approach offers a robust and transferable methodology for integrating probabilistic hydrological forecasting with water quality assessment.
• The framework enables identification of periods with elevated risk of regulatory threshold exceedance and provides a more robust basis for water chemical status assessment and supports risk-informed classification and management.

Funding

• This research was funded by the Spanish Ministry of Science, Innovation and Universities (PID2020-120333GB-I00).
• Additional funding was provided by the European Union's Horizon 2020 research and innovation programme under grant agreement No. [grant number].

Citation

@article{Mena2026deep,
  author = {Mena, Alberto and Bergillos, Rafael J. and Paredes-Arquiola, Javier and Corzo, Gerald and Solera, Abel and Andreu-Álvarez, Joaquín},
  title = {A deep learning-based probabilistic framework for streamflow prediction and water quality assessment},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135934},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135934}
}