Development of Integrated DHSVM-Flood2D-GPU modeling framework – Case of ACT River Basin

Abstract

The objective of this study is to develop an integrated and computationally-efficient hydrologic-hydraulic modelling framework that couples a spatially distributed hydrologic model (DHSVM) with a high-resolution GPU-accelerated flood model (Flood2D-GPU). Taking Alabama-Coosa-Tallapoosa (ACT) river basin as a test bed, hydrologic simulation is conducted using a 90m resolution DHSVM hydrologic model that can be driven by observed precipitation or simulated meteorology from Weather Research Forecasting (WRF) model. The DHSVM hydrograph along with 10m resolution digital elevation and estimated bed roughness are then used to conduct high-resolution flood simulation using an accelerated Flood2D-GPU model. A model to model comparison is performed for the Carrollton City, GA, located inside the ACT river basin by examining the difference between the maximum flood inundation extents obtained from Flood2D-GPU and a commonly-used single-CPU Flo-2D model. The comparison shows that Flood2D-GPU can largely reduce the required simulation time while being able to provide similar results to Flo-2D. Our next step is to scale up the simulation framework for a larger domain and for more parallel ensemble members utilizing high-performance computing. We expect that this integrated DHSVM-Flood2D-GPU modelling framework can support hydroclimate researchers evaluating the flood risks on our sensitive energy-water infrastructures in a more efficient manner.