Study And Implementation Of Parallel LASG/IAP Coupled Climate Model Based On Software Framework

The simulation and prediction of the climate is an important research area in thepresent world. With the progress of the climate science and the computer technique,the coupledclimatemodelhasbecomethetrendofthe climateprediction.The software framework technique supplies a reusable design and a partimplementation through distilling common features of the specific domain anddistinguishing the variable requirements. The use of the framework can enhance theefficiency of the application development and make the application more reusable,more reliable, and more portable. The technique is an effective approach to dealingwiththecomplexityofthecoupledclimatemodel.The construction of the coupled climate model is realized by coupling andintegrating existing independent climate models in most cases. Based on the analyseof LICOM1.0 ocean model and IAP4.0 atmosphere model, this thesis proposes adesign of a sea-air coupled climate model which is formed by integrating the twomodels withESMF.Data fields in each component models are usually distributed on different gridsevenindifferentphysical forms.Andseparatingcomponentmodelsfromthecouplinglogic is necessary in the modularization design. Consequently, this thesis designs amodel-independent coupler to centralize the coupling logic. Its main function is theprocessing ofdatafieldstransmissionandtransformationbetweencomponentmodels.The long-time integral in climate prediction applications requires flux conservation indata exachanges to avoiderror accumulation. So the conservation remapping of datafields between different grids is most important in the coupler. This thesis proposes asphere-outspreading andinterpolation remapping (SOIR) algorithm. Compared to theexisting surface integral redistribute remapping algorithm, the new algorithmenhances the precision while predigesting the realization and not increasingcomputationalcomplexity.Resultsofexperimentsshowthatthecostofcouplingtakesupasmallproportionof the total cost of model computation. The sea-air coupled climate model has arelativelyhighparallelperformance.