Calibrating a 2D high-performance hydrodynamic model for fluvial process modelling along the Mekong River

The Vietnamese Mekong Delta (VMD), representing a highly complex hydrodynamic system, plays a major role in food security and socio-economic development in Vietnam. With ongoing climate change and rapid urbanization, the VMD is increasingly vulnerable to flood risk from multiple sources, e.g. driven simultaneously by fluvial, pluvial and coastal processes. It is essential to develop reliable modelling tools to simulate such compound flooding processes to support hazard risk assessment and management to inform the development of policies and effective strategies to sustain the delta development.

To support reliable compound flood modelling and risk assessment in VMD, it is important to accurately predict the fluvial processes along the Mekong River. For large river and river network modelling, one-dimensional (1D) and quasi two-dimensional (2D) hydrodynamic models are commonly used. However, modelling overbank flow and flooding process over floodplains is out of the capability of these 1D or even quasi 2D models. These 1D or quasi 2D models are then integrated with a 2D inundation model through one-way coupling to predict the flooding processes in floodplains. The resulting one-way coupled models neglect the dynamic interactions between the flows in the river and floodplain as well as upstream and downstream domains, inevitable introducing model uncertainties that are difficult to quantified and controlled. Ideally, we can use a full 2D hydrodynamic model to simulate the entirely fluvial flooding process spreading from the river channels over to the floodplains. However, this approach has not been widely reported for large-scale application due to the prohibited computational cost of a 2D hydrodynamic model.

In this work, we explore the possibility of calibrating a fully 2D hydrodynamic model, the High-Performance Integrated hydrodynamic Modelling System (HiPIMS), to reproduce high, medium and low flood conditions along the middle and lower reaches of Mekong River of 55 km, starting from the Kratie gauge in Cambodia to avoid tidal influence. The model is driven by inflow at Kratie and calibrated using the measurements of both water level and discharge available at 4 gauge stations (Can Tho, My Thuan, Chau Doc, Tan Chau). The Nash-Sutcliffe efficiency (NSE) is used to quantify prediction errors to support the model calibration process.