UPV develops AI system to predict water contamination in the Júcar river basin
The new tool combines artificial intelligence and hydrological models to assess risk in surface water bodies.
By Empar Soler i Martí
••3 min read
AI
Generic image of a water quality assessment system with technology elements and landscape.
Researchers from the IIAMA-UPV have developed a system that combines artificial intelligence with hydrological and water quality models to assess the risk of contamination in the surface water bodies of the Júcar river basin.
The methodology allows for the estimation of biochemical oxygen demand (BOD), phosphorus, and ammonium concentrations, and calculates the probability of exceeding regulatory limits in each water body. This information helps identify periods and areas of higher risk and guide preventive measures.
The research, published in Journal of Hydrology, is led by Alberto Mena, Rafael J. Bergillos, Javier Paredes-Arquiola, Abel Solera, and Joaquín Andreu from IIAMA-UPV, along with Gerald Corzo from IHE Delft in the Netherlands.
“
"The results highlight the relationship between hydrological variability and water quality response, and demonstrate the importance of incorporating uncertainty into predictive modeling."
To develop and evaluate the system, the team used meteorological information from the State Meteorological Agency and hydrological and water quality data from the SIA-Júcar system, managed by the Júcar Hydrographic Confederation. Comparison with records from 129 monitoring stations shows that the system reproduces the main spatial pollution patterns.
“
"The accuracy rate reaches 90% for BOD, 84.4% for phosphorus, and 85.3% for ammonium."
The methodology employed neural networks to predict monthly water contributions and generate 1,000 possible scenarios per sub-basin to represent uncertainty. In addition to estimating pollutant concentrations, the tool calculates the probability of exceeding regulatory limits and the magnitude of that increase.
This information allows for the identification of periods and water bodies with higher risk and guides the adoption of preventive measures. "By quantifying both the probability and the magnitude of exceedances, the system allows for a more robust and complete assessment of water quality dynamics under uncertain conditions," explain the authors.
The study shows that uncertainty increases during episodes of higher contributions and can accumulate downstream. Incorporating this effect helps interpret predictions, especially when concentrations approach regulatory limits.
“
"The evaluation registers an average false alarm rate of only 3.1%, a very favorable result for its potential application in early warning systems."
Therefore, the tool should complement measurements from control networks and support monitoring and decision-making. The work has received support from the Ministry of Science, Innovation and Universities, the State Research Agency, the European Association for Biodiversity, and the European Union through the NBSPLUS project. Alberto Mena has also received funding from the Ministry and the State Research Agency, and the European Social Fund Plus.