Coupling strategy of atmospheric-oceanic general circulation model with ultra high resolution and its performance on the Earth simulator

Coupled global climate models (CGCM) provide the most powerful tool to reproduce main features of the observed climate. In stand-alone atmospheric and oceanic models, the computational efficiency has been progressed by tuning each of model codes respectively. Ordinarily supercomputers do not provide reasonable turnaround of CGCM run for century time scales at ultra high resolution. The great expense of running CGCM has been hesitated of development by limiting the number of calculations and by prohibiting the use of the reasonable resolution for satisfying physical requirement. The resource of the Earth Simulator might become to be able to carry out the huge scale simulation. Our objective here is to develop coupled global climate models for the Earth Simulator (CFES) with ultra high resolution to carry out century time integration within reasonable time without decrease of computational efficiency of the component models. It is composed of oceanic general circulation model for the Earth Simulator (OFES) and atmospheric general circulation model for the Earth Simulator (AFES). We provide a new coupling structure to transfer physical data from one component model to the other component through a coupler and back again. In the structure of coupling scheme, each component can run independently to avoid bias due to modeling the feedback timing. It allows us not to, worry about the sequential order of execution of component models. CFES was performed with fully paralleized implementation including I/0 interaction throughout coupling scheme. Due to the parallelization, CFES was able to control concurrent performance by changing the number of nodes which employed each component of atmospheric and oceanic models.