Özlem Baydaroğlu , Marian Muste, Atiye Beyza Cikmaz , Kyeongdong Kim, Ehab Meselhe, Ibrahim Demir

Flood waves propagating in natural streams are highly complex flow situations with intricate dependencies among flow variables that are still not well understood. The most relevant information for deciphering these dependencies is extracted from continuous in-situ measurements acquired through high sampling frequency with special attention to the magnitude and timing of the hydrograph peaks. The new generation of acoustic instruments can produce experimental evidence if the useful nonlinear, nonstationary signals are extracted from the inherently noisy dataset acquired on site. This study presents a new screening protocol to smoothen streamflow data from the unwanted influences and noise generated by flow perturbations and observational fluctuations. The protocol is flexible and robust combining quantitative (statistical fitness parameters) and qualitative (domain expert judgments) evaluations. The screening protocol is tested with 18 smoothing methods applied to 118 datasets to identify the optimal data conditioning candidates. Sensitivity analyses are conducted to assess the validity, generality, and scalability of the smoothing procedures. The main goal of this analysis is to set a mathematical foundation for the empirical results that can lead to unified, general conclusions on principles or protocols for unsteady flows propagating in open channels, formulating practical guidance for future data acquisition and processing, and using the in-situ data to better support numerical and data-driven modeling efforts.