Abstract: In order to solve the problem of large uncertainty in the estimation of vertical ground shaking parameters, a prediction model of vertical ground shaking parameters for the 8 January 2022 Menyuan M_w6.7 earthquake in Qinghai is established based on the NGA-West strong ground shaking database using stochastic adaptive neuro-fuzzy inference technique (ANFIS). Based on the confidence test using the actual observation data from the National Seismic Intensity Rapid Reporting and Early Warning Project , the distribution of the peak vertical ground shaking acceleration (PGA) of the Menyuan area is given to explore the directional effect, site amplification effect, and its causative mechanism of the strong ground motion. The research results show that: (1)The ANFIS-based vertical ground motion intensity prediction model demonstrated satisfactory performance in forecasting the vertical PGA during the Menyuan M_w6.7 earthquake, yielding a correlation coefficient (R) of approximately 0.809 with observed values and an RMSE of 0.046, indicating its high reliability and applicability in predicting moderate-to-strong destructive earthquakes in China. (2)The predicted vertical PGA contours of the Menyuan M_w6.7 earthquake exhibited an elliptical pattern, with the major axis well-aligned with the seismogenic fault’s strike, and the maximum vertical PGA at the epicentral zone reached approximately 376.3 Gal. The vertical PGA decayed with increasing fault distance while displaying notable directivity effects and near-field saturation behavior under strong-motion conditions. (3)The site amplification effect on vertical PGA was relatively weaker than that on horizontal motion. With decreasing , the peak amplification factor of vertical PGA relative to bedrock conditions(V_s30=1000m/s) ranged between 1.14 and 1.27, while under strong-motion conditions(M_w=7.0), soft soil sites (V_s30=100m/s) exhibited an amplification coefficient of ~0.79, indicating pronounced seismic motion reduction due to soft soil effects.