A Hybrid Approach to Combine Physically Based and Data-Driven Models in Simulating Sediment Transportation

Abstract

The objective of this study is to develop a methodology for hybrid modelling of sedimentation in a coastal basin or large shallow lake where physically based and data driven approaches are combined. This research was broken down into three blocks. The first block explores the possibility of approximating a physically based model using a data driven model to predict suspended particulate matter (SPM) concentrations. It identifies the necessary input variables and data manipulation operations to expose maximum information to a data driven modelling tool. The second block investigates the effect of specifying time varying open boundary condition instead of fixed boundary conditions on the simulation of (SPM) along the Dutch coast. First, a methodology for generating time varying open boundary conditions using a data driven model is developed. Local hydrodynamic and meteorological conditions are used as input variables. Secondly the improvement of the time varying open boundary conditions (OBC) on the simulation results is studied The third block examines how knowledge gained in modelling sediment transport along the Dutch coast can be transfer to understand sediment transport in Lake Victoria. This part lays ground for more accurate and reliable modelling of sediment transport in Lake Victoria in future through transfer of methods developed in the preceding sections.