To investigate the temporal and spatial evolution of ground temperature, using a variety of temperature sensors comprehensively, continuous high-precision temperature data of 0 to 100 m in three boreholes in Nanjing City were obtained, and the change in ground temperature for 3 months was simulated based on the one-dimensional heat conduction equation. By analyzing the temporal and spatial variation laws of measured and simulated ground temperature, the simulation error of shallow ground temperature field under different spatial and temporal accuracy modeling conditions is explored. Compared with the average thermal diffusion coefficient, when the layered thermal diffusion coefficient is used, the shallow simulated temperature of 0-4 m reduces the error by 2%-4%. Compared with the results of high-precision boundary simulation, the results don’t show much difference when the boundary is the daily average temperature, while the accuracy of monthly average temperature is reduced by 2%-3% between 0-5 m, and the influenced depth on the homogeneous layer is smaller. For the simulation of the ground temperature at the shallow temperature layer, the use of segmented thermal diffusion efficient at the depth of 0-5 m and the daily average temperature as the boundary condition can reduce the monitoring cost while ensuring the accuracy of the results. It also provides a reference for the long-term prediction and evaluation of shallow ground temperature in Nanjing.