Performance Evaluation of MR-CSC-DMD in River Model Experiment with Groynes

This study applies a variant approach of dynamic mode decomposition (DMD), a multi-resolution convolutional-sparse-coded DMD (MR-CSC-DMD) method, to analyze the dynamic modes of riverbed changes, and newly evaluates the performance for the presence of groynes. In an authors' previous work, MR-CSC-DMD was used to predict riverbed undulation from water surface observations, where the dynamic mode was analyzed without any groynes. An important mission in the project, however, is to elucidate the impact of Groynes' placement on river channel. To address this gap, this work applies MR-CSC-DMD to data collected by a river experimental setup with groynes. By analyzing the time series data of the water surface and riverbed undulation, a mathematical time evolution model of riverbed changes is developed by taking the groynes into account. It is shown that the presence of groynes affects the prediction performance of MR-CSC-DMD.