文中使用四维变分同化技术将海温观测资料同化到 Zebiak- Cane模式中 ,通过优化模式的初始场提高了模式的预报技巧。通过用理想场进行检验 ,说明所建立的同化伴随模式是正确的。用文中建立的四维变分同化模式以 1 997年 1月为初始场所做的预报结果与实况相比 ,结果较好。这对今后 ENSO预报打下了良好的基础 The four-dimensional variational assimilation technique is used to assimilate the SST data into the Zebiak-Cane model, and the prediction technique of the model is improved through the initial field of the optimized model. By testing with ideal field, it shows that the established assimilation pattern is correct. The prediction results of the four-dimensional variational assimilation model established in this paper with the initial location of January 1997 are better than those of the actual situation. This laid a good foundation for future ENSO forecast