Performance Evaluation and Applicability Assessment of Non-Contact Discharge Estimation Methods Using Multi-Line Surface Velocity Time Series

In this study, the performance and applicability of non-contact discharge estimation methods were evaluated using surface velocity time series obtained from a Doppler radar-based multi-line surface velocimetry system (PAVOs). The focus was on improving cross-sectional representativeness beyond conventional single-line observations, and complementing contact-based discharge measurements during high-stage and peak-flow periods in the flood season. The quality of the surface velocity data was improved by optimizing operational parameters and removing outliers using a Hampel filter. Discharge was estimated using two methods: (i) the mid-section method, which reflects multi-line surface velocity distributions, converts surface velocities to mean velocities, and integrates discharge over subsections, and (ii) the index-velocity method, which combines stage-area and index velocity-mean velocity relationships. The estimated discharges were compared with the reference discharge (Qref) and their performance was assessed using the Nash–Sutcliffe efficiency (NSE), correlation coefficient (r), mean absolute error (MAE), percent bias (PBIAS), and peak discharge error (peak_err). The results showed that both methods generally reproduced the reference discharge variations under medium- to high-stage conditions, although their applicability differed depending on the stage conditions. These findings suggest that multi-line non-contact surface velocity measurements can support peak flood discharge estimation and continuous discharge monitoring by complementing the limitations of conventional single-line observations and more effectively reflecting cross-sectional velocity variability.