Application of the Machine Learning LightGBM Model to the Prediction of the Water Levels of the Lower Columbia River

Due to the strong nonlinear interaction with river discharge, tides in estuaries are characterised as nonstationary and their mechanisms yet be fully understood. It remains highly challenging accurately predict estuarine water levels. Machine learning methods, which offer a unique ability simulate unknown relationships between variables, have been increasingly used large number of research areas. This study applies LightGBM model predicting levels along lower reach Columbia River. The inputs consist discharges from two upstream rivers (Columbia Willamette Rivers) tide characteristics, including range at estuary mouth (Astoria) constituents. is optimized selected parameters. results show that can achieve high prediction accuracy, root-mean-square-error values level being reduced 0.14 m correlation coefficient skill score ranges 0.975–0.987 0.941–0.972, respectively, statistically better than those obtained physics-based models such tidal harmonic analysis (NS_TIDE). importance subtide constituents interacting discharge clearly revealed results.