Development and application of distributed rainfall-runoff model using MPI technique

Abstract

A lumped model, in general, is expressed by ordinary differential equations, which does not take into account the spatial variability of processes, input, boundary conditions and watershed geometric characteristics. For this reason, recently distributed run-off models have been developed to represent the variability in physical watershed characteristics such as topography, land use, and rainfall properties. However, the distributed rainfall-runoff model requires a lot of time and effort to generate input data. Also, it takes a lot of time to calculate the discharge using a numerical analysis based on kinematic wave theory in runoff process. Therefore, most river basins that use the distributed model are of limited scales, such as small river basins. Especially, the necessity for an integrated watershed management system has been increasing due to changes in watershed management concepts and discharge calculations for whole river basins, including the upstream and downstream of dams. In this study, a distributed rainfall-runoff model using Message Passing Interface (MPI) technique was developed to decrease the calculation time needed for the application of large scale watersheds. This model, which uses a parallelized based K-water hydrologic & hydraulic Distributed RUnoff Model (K-DRUM), can simulate temporal changes and the spatial distribution of flood discharge by taking into consideration radar grid rainfall and grid based hydrological parameters. The model was applied to the Geum River Basin, which include the Yongdam and Daecheong Dam Watersheds located on the Korean Peninsula. The results were calculated between a single domain and a divided small domain. They were compared to analyze the application effects of the parallelization technique. As a result, a maximum of 10 times the amount of calculation time was saved by using parallelization code rather than a single processor. Also, problems related to the running time and inaccurate settings that typically occurred using the existing trial and error method were solved by applying an auto calibration method in setting initial soil moisture conditions. As the result, the calculation results showed a good agreement with the observed data.