The traditional EVAD technology with linear hypothesis results in relatively poorrepresentativeness mean wind field information in the calculating for upper levels with high nonlinearity. The magnitude of rootmeansquare errors of the radial velocities at equidistance rings from the 2order Fourier series fitting can indicate the level of nonlinearity. The larger the RMS, the larger the nonlinearity is. Setting the proper threshold of RMS can eliminate the relatively large nonlinear rings, which makes the rings in calculation meet the linear assumption. When the nonlinearity of winds is large, the 3 m/s threshold can avoid the bad mean wind instead of “ND” value. The mean RMS of all rings can reflect the representativeness of the mean wind field for the real wind field. The rationality of the threshold 3 m/s needs to be verified by more cases.