Abstract: The magnitude of dual polarization observations is much smaller than the reflectivity factor (Z_H), and the accuracy of the observations is the basis for the application of dual polarization radar data. Anomalies in dual polarization observations caused by ground have largely been overlooked in the past. For the first time, utilizing the observation data from 93 CINRAD radars upgraded with dual polarization, the characteristics and scope of ground influence are determined by the anomalous areas after long–term data accumulation. Statistical analysis is then conducted to reveal the overall proportion of the influence in CINRAD radars, the main sources of the influence, the relationship between the influence and radar types and observation modes, the temporal variations of the influence, the nature of the influence and the distribution characteristics of dual polarization observations. Results are as follows: (1) The ground can cause significant deviations of the accumulated values of dual polarization observations from the precipitation interval, the differential reflectivity factor (Z_DR) is decreased by more than 1 dB, the correlation coefficient (ρ_HV) falls below 0.92, and the standard deviation of differential phase (Φ_DP) is greater than 10°. This characteristic served as the main basis for identifying ground influence. (2) The ground influence is not limited to the near–ground region. As the elevation angle increases, the sidelobe reception of ground signals can also produce similar effects. The proportion of influence in the near–ground region can reach 14% to 32%, while at high elevation angles it reaches 4% to 7%, far more severe than the beam blockage. (3) The ground influence is mainly caused by mountains and buildings with strong scattering capabilities. When the site is below the terrain height, surrounding mountains are the main contributors; when the site is close to the terrain height, buildings within urban areas are the main contributors. (4) The ground influence can be quantitatively described by the superposition of random errors and systematic errors, leading to increased uncertainty in the dual polarization observations within the affected area at a specific time. Moreover, the influence is more severe in the mainlobe region than in the sidelobe region.