Based upon the results from the scale analysis of the vertical equation of motion for the atmosphere and forecast experience, especially numerical weather forecast experience, the multilevel perturbation method has been proposed. It can be proved that in the vertical perturbation equation the magnitude of the perturbation decreases with n, n=1, 2, …. But, on the other hand, since the magnitude of truncation error is proportional to that of the discretized term, in the vertical equation of motion or the perturbation equation the truncation error of the discretized vertical pressure gradient force (VPGF) is the largest vertical truncation error, which can represent the total vertical truncation error in the equation. Hence the decrease in the magnitude of VPGF with n is favorable for reduction of the total vertical truncation error and use of the method can limit such an error within a permissible range, to make computation of VPGF more correct and describe convective activity better than ever. Therefore the impact of the largest truncation error on the tendency of w would be unable to distort or cover up the contribution of vertical Coriolis force term and that of curvature term to the tendency. Further, it can be proved that the solution of level n perturbation balance equation is a hydrostatic perturbation. The method is, substantially, a tool to be used to deduct the hydrostatic part from level n perturbation, in order to reduce the level n+1 perturbation in magnitude. In this way a high level perturbation VPGF may easily come closer to the other terms without perturbation in magnitude. Furthermore, because the sum of VPGF and gravity does not vary with n, the physical properties of the vertical equation of motion would not change. Therefore, the method may be named the multilevel perturbation hydrostatic deduction method. It can diminish the difference between the improved basic state vertical profile and the actual atmospheric vertical profile quite small, and under hydrostatic equilibrium in the vast majority of the model atmosphere there would be almost no vertical sound waves.