Water quality monitoring plays a crucial role in protecting aquatic ecosystems and ensuring the safety of water resources, and high-frequency (minutes to hours) monitoring provides data for modeling hydrology and biogeochemical dynamics of aquatic environments. However, these datasets are often affected by noise, which obscures the underlying concentration–discharge (C–Q) relationship and limits their interpretability. In this study, we investigated the application of two hybrid reconstruction models: Weighted Median Filter combined with Ensemble Empirical Mode Decomposition (WMF-EEMD) and Variational Mode Decomposition (WMF-VMD), which are data-driven signal processing models for improving the quality of high-frequency turbidity time series. Storm-event turbidity signals, paired with radar-derived water level data from a small watershed, were decomposed, selectively filtered using entropy-based criteria, and iteratively reconstructed to enhance signal quality. The study found that both hybrid models can effectively reduce high-frequency noise while preserving critical hydrological patterns. EEMD reconstruction captured a wide range of fluctuations, from high-frequency noise to low-frequency trends, with high variability across metric ranges, while VMD offered greater consistency and efficiency, as evidenced by faster computational time, higher average R 2 , and lower JSD, with slight shifts in the signal's central tendencies. These models provide scalable, automated frameworks to improve the reliability of high-frequency sensor data used in water quality monitoring and management systems. • Introduces two hybrid frameworks for denoising high-frequency turbidity data. • Weighted Median Filter is effective in suppressing outliers and zero values. • The framework improves water-level and turbidity relationship. • EEMD captures multi-scale variability, while VMD ensures stability. • Provides scalable and automated method to enhance water quality monitoring system.
Badrudeen et al. (Fri,) studied this question.