Abstract: By analyzing statistics and dynamics in this text, we can come to such conclusions: (1) The ECMWF medium-term numerical forecast can forecast medium-term activity of subtropical high, and the accuracy rate of forecast cannot have large improvement by transnational corrections. (2) The important cause that ECMWF medium-term numerical forecast made errors in 1998 is that astronomical tide is not included in the model. (3) Two indexes are found from which it can be judged that ECMWF medium-term numerical forecast will make errors for disregarding the astronomical tide in the model : ① when the 54.7° line under the moon of nodal month astronomical singularities coincides with the trough-line of subtropical jet flow from 50°E to 150°E on 500 hPa at 2000 L. T.The same day, and is approximately vertical (α>60 °) with the isotherm, then the day in 0～2 days after nodal month astronomical singularities is defined as initial day, then in three days after the initial day, ECMWF medium-term numerical forecast of north latitude of 588 line on 120°E will make continuance errors by two latitudes (7/9). Otherwise, it won't make continuance error (13/13), ② otherwise, if the 54.7° line is in the range of low pressure between two high pressures, then there is dispersed error on the day of nodal month astronomical singularities (5/7). On the contrary, there is not any error (6/6).