Design and deployment of a dynamic-coupling tool for EFDC

Abstract

A dynamic-coupling tool designed to link several hydrodynamic models is presented. The tool is able to dynamically transfer time-series data among models that are geographically adjacent. Dynamic data transfer is implemented at the models' common boundaries. The Message Passing Interface (MPI) and a coupling code were used for implementing the dynamic link. Several issues that had to be overcome during the development of the tool (such as porting of the code to a Linux environment, MPI implementation, and compiler flags used for optimum performance) are discussed. The tool is applied to a test case in which three hydrodynamic models built with the Environmental and Fluid Dynamics Code (EFDC) are run with the dynamic-coupling tool in a Linux cluster. Run times were compared to a sequential run of the three models in a Windows environment. A speed up of 8.53 was achieved by exploring and finding an optimal combination of Intel Fortran compiler flags. © 2014 Springer International Publishing.