Author(s): Pasha Piroozmand; Davood Farshi

Linked Author(s): Davood Farshi

Keywords: Hydrometric Data Correction; Anomaly Detection; Time-Series Analysis; Hydrological Monitoring

Abstract: Ensuring high-quality, error-free hydrometric data is vital not only for precise hydraulic modeling but also for public transparency and informed decision-making in ecohydraulics and water management. Hydrometric data from Swiss rivers and lakes, collected by the Federal Office for the Environment (FOEN), undergoes multiple verification stages before becoming definitive. Currently, validation and correction are manual, making the process subjective and time-consuming. This study proposes an automated anomaly detection and correction method for hydrometric time-series data using a computationally efficient approach. The algorithm detects sensor anomalies, including point and step anomalies, by analyzing first -order differencing, and spectral characteristics of adjacent data segments. Anomalies are dynamically identified and corrected via interpolation, avoiding the need for computationally expensive machine learning models. Results from five Swiss hydrometric stations demonstrate significant errorreductions. The method achieved over 90% improvement in mean absolute error (MAE) for most stations, with Station Neuhausen showing 99.99% correction due to initially high discrepancies. Standard deviation reductions exceeded 99% at Station Bern and Neuhausen, confirming improved data stability. Maximum error reductions ranged from 8.94% to 99.99%, highlighting the method's effectiveness in mitigating extreme anomalies. This study provides a scalable and efficient solution for automating hydrometric data correction, reducing reliance on manual validation.

DOI: https://doi.org/10.64697/iahr.proc.ise2026.abs.324

Year: 2026