Abstract: To improve the accuracy of secondary radiation noise prediction, this study investigates the environmental protection acceptance monitoring data of three typical geological cities(Cities A, B, and C) using a combination of field measurements and empirical prediction methods. The results indicate that the average difference between the maximum Z-weighted vibration level(VLzmax) and the secondary radiation noise level(Laeq) during train passage is generally consistent with the correction parameters in the Guidelines. However, this difference becomes more discrete in softer soil. A vibration amplification phenomenon is observed in the area 20～30 m horizontally from the tunnel. Regarding ground attenuation, the attenuation of the Z-weighted vibration level is more significant in test sections with softer soil. This study confirms the accuracy of the empirical prediction method in the Guidelines to a certain extent, providing a theoretical basis for future environmental impact assessments of urban rail transit. It is recommended that cities combine local geological conditions to develop prediction models with regional characteristics when conducting relevant environmental impact predictions and environmental protection acceptance monitoring.