Development and evaluation of a rip current forecasting system for channel-type rip currents on a low-energy headland beach

Abstract

A 34-day field observation was conducted on a headland beach in South China with a typical channel rip current system. The Delft3D model was applied to reproduce hydrodynamic conditions and extract rip currents, while a Generalized Additive Model (GAM) was used to capture and quantify the nonlinear interactions among environmental factors. By optimizing interaction terms, a comprehensive model was constructed, and a hybrid rip current warning system combining physical modelling and statistical analysis (NGRWS) was proposed. Results showed that rip currents in the study area were generally of low energy and persistent, being jointly controlled by multiple factors, with wave direction (WaveDir), significant wave height (SWH), and water level (WL) as the maindrivers. Using limited offshore environmental variables, the system achieved 6-h forecasts with errors <0.3 cm/s, correctly predicting 87% of rip events with false and missed alarm rates of 5% and 8%. A storm causing coastal erosion further provided evidence that the system has certain robustness, although recalibration remains necessary over time. In addition, while quantifying the contributions of environmental factors, the GAM also revealed potential cognitive blind spots under traditional empirical judgments. This study supports improved risk avoidance for beach users.