The aim of this study is to improve the water distribution uniformity of a vertical impact sprinkler and explore the design method of the drive spoon blades. The width of straight blades (h1), the width of curved blades (h2) and number of blades (s) were chosen as the experiential variables. The suitable ranges of three variables for response surface method were determined initially by one-factor experimental design method, and 17 different drive spoons were designed according to response surface methodology. The results showed that in the one-factor experimental condition, the CU (Christiansen’s uniformity coefficient) values first increased and decreased slightly when h1 exceeded 3 mm with the increase of h1 within the variation range of the experimental factor. The CU values firstly increased and then decreased with the increase of h2. The CU values decreased rapidly when s was less than 3 or greater than 6. The relationship between CU values and h1, h2 and s was established using response surface methodology. The p-values for h1, h2 and s were 0.0359, 0.0092, 0.0212, and all of the selected factors were significant on CU. The order of parameters affecting CU were h2, h1 and s. The ideal parameters for the drive spoon blades were h1 = 6 mm, h2 = 4 mm, and s = 3. CU was greatly improved after the optimization of structure for the drive spoon blades, which increased to 87.96% from 73.12%. After optimization, the application rates within 1 to 5 m were improved and increased from 10% to 15% with an average of 10.7% under different operating pressures. The maximum application rates decreased from 9.3, 9.3, 9.4 and 8.4 mm·h−1 to 8.5, 8.4, 8.5 and 7.9 mm·h−1 with operating pressures of 300, 400, 500 and 600 kPa, respectively. The maximum application rates in the overlap area were decreased from 18, 16, 16 and 15 mm·h−1 to 16, 14, 14 and 12 mm·h−1 with operating pressures of 300, 400, 500 and 600 kPa, respectively.