Plinius Conference Abstracts
Vol. 19, Plinius19-28, 2026, updated on 17 Jul 2026
https://doi.org/10.5194/egusphere-plinius19-28
19th Plinius Conference on Mediterranean Risks
© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.
Oral | Wednesday, 07 Oct, 16:45–17:00 (CEST)| Lecture room
An Artificial-Intelligence-based application for short-term river water-level prediction and flood early warning in Thessaly, Greece. 
Christos-Panagiotis Giannaklis1,3, Konstantinos Lagouvardos1, Vasiliki Kotroni1, Elias Dimitriou2, Anastasios Papadopoulos2, and Christos Giannaros3
Christos-Panagiotis Giannaklis et al.
  • 1Institute for Environmental Research and Sustainable Development, National Observatory of Athens, 15236 Athens, Greece
  • 2Hellenic Centre for Marine Research, Institute of Inland Waters, 19013, Anavissos Attikis, Greece
  • 3Laboratory of Meteorology and Climatology, Department of Physics, University of Ioannina, 45110 Ioannina, Greece

Mediterranean extreme weather events are often characterized by intense rainfall over short time periods and rapid hydrological response, making the short-term river water-level prediction a crucial component of flood early warning. In this work, we present the development and the pilot application of a data-driven operational prediction system for river water-level in multiple points in the region of Thessaly, Greece, based on artificial intelligence techniques.

The system exploits high temporal resolution observations from a regional network of meteorological and hydrometeorological stations in Thessaly, Greece, combining past rainfall data with river water-level data over a five-year period. The framework is designed to learn the nonlinear relationship between the antecedent rainfall and the subsequent evolution of the river stage (hydrological response). Several artificial-intelligence models, including gradient boosting machines and neural networks, are trained and evaluated to provide short-term water lever predictions at selected points of the network, with a prediction horizon up to 12 hours.

The preliminary results highlight the potential of artificial-intelligence techniques to real-time flood prediction in Mediterranean environments, providing valuable lead time for the local authorities and improved flood-risk preparedness.

Keywords: early warning systems, rainfall, river water level prediction, artificial intelligence

How to cite: Giannaklis, C.-P., Lagouvardos, K., Kotroni, V., Dimitriou, E., Papadopoulos, A., and Giannaros, C.: An Artificial-Intelligence-based application for short-term river water-level prediction and flood early warning in Thessaly, Greece. , 19th Plinius Conference on Mediterranean Risks, Murcia, Spain, 6–9 Oct 2026, Plinius19-28, https://doi.org/10.5194/egusphere-plinius19-28, 2026.