Application of Parallel Diffusive Wave Model to Inundation Analysis Using OpenMP

Abstract

The work that simulates 2-dimensional inundation event by numerical model needs much computational cost. It is further compounded by using increasingly available high-resolution digital topographic data and conducting flood studies over larger domains using GIS or Lidar etc. In recent years, MPI and OpenMP parallel technique to analyze simultaneously using multi-core cpu has considered as a way to decrease the processing time. In this study, 2-dimensional model is developed to simulate inundation and examined the applicability in the protected area by levee or downstream area of dam. For this purpose the source code of diffusive wave inundation program is rewritten using shared-memory parallel OpenMP technique and applied to hypothetical protected area and real dam downstream basin. The application results to hypothetical protected area are compared with original sequential model and distributed memory parallel MPI model. Through the review of results, this model shows that the flow parameters such as inundation patterns, magnitude and direction of velocity vector, are consistent with MPI and sequential model and it confirmed the correctness of this model. Also this model has better computing characteristics in aspect of elapsed time, processing speed-up, efficiency than MPI and sequential model. When the studied model using OpenMP is applied to real dam break case, in case of 12 cores, elapsed time is decreased to 12~14% compared with the original program.