To resolve the problem caused by incompatibility between the dynamic model and observational data in the assimilation, the effect of model compatible part of observational information on the assimilation is explored by utilizing the NCCo model and a method of separating the model compatible part of information from the observational data. The researches show that the predictions are rather poor when the observational SSTA data are assimilated directly by the simple ocean-atmospheric coupled model, as the NCCo model, because the assimilated initial fields contain too much small scale disturbances which are incompatible with the dynamic features of model and grow up rapidly in the integrating process. However, the effect of assimilation could be greatly improved by just assimilating the model compatible part of observational data whether the assimilation circumstance is uncoupled or not, because the initial fields primarily contain ENSO scale information, which matches the dynamical spatial and temporal scales of the NCCo model.