In order to improve the accuracy of yield separation, the article attempted to apply the HP filter decomposition method to separating climatic output from long time-sequence grain yield. The paper first considered three methods, HP filter, 5-year moving average and Logistic fitting to construct the trend equations for the grain output respectively, based on the climatic output from 1961 to 2011. The three methods were compared by the trend changes and the fluctuation range. The results showed that the equations established by the three methods were all significant for the significance level 0. 01. The output trend simulated by HP filter was coincided with the fact and appeared an obvious fluctuation in the research period, which reflected the characteristic of the actual yield. Meanwhile, the output trend simulated by 5-year moving average method was similar to that of HP filter method, but the changing range of the trend change was relatively slow. The output simulated by Logistic fitting was obviously higher than the actual yield and retained a growing status during the research period, which was different to the actual yield. It showed that, HP filter has more advantages to model the separating climatic output from long time-sequence grain yield with, on one hand, the operating process is more easy and the climatic output simulated by this method agrees with the fact, on the other hand, the data had strong inclusiveness.