Heat supply, automobile exhaust, industrial production and decrease of thermal inertia in winter caused by the decrease of vegetation coverage leads to an obvious difference in the distribution of the land thermal field in the winter compared with other seasons. The Urban thermal field distribution in the winter directly affects the spread of air pollutants, which has important implications for analyzing the contribution of the thermal field to particulate air pollution. Atmospheric transmissivity and atmospheric upwelling/downwelling radiance in simulations are first calculated using the moderate spectral resolution atmospheric transmittance algorithm and computer model (MODTRAN). Then, we solve the radiative transfer model of the thermal infrared band by constructing a look-up table. In addition, the accuracy estimation is performed using the simulated data, showing that when the error range of emissivity and water vapor content are confined to ±0.005 and ±0.6, respectively, the temperature retrieval error are less than 0.348 and 2.117 K, respectively indicating the high retrieval accuracy of the method. In addition, the long-term sequenced Landsat TM and ETM+ data were selected to retrieve land surface temperature (LST) during 1985-2015. The analysis of the temporal and spatial distribution of thermal fields in Beijing show that the spatial and temporal variations are observable. The spatial variation covers four levels: high temperature is distributed within the second ring, low temperature loops are distributed between the second and the fifth ring, high temperature is distributed in the outer suburb areas and the lowest temperature is distributed in the western mountainous areas. Meanwhile, the temporal variation of thermal field distribution changed a great deal during the rapid development in the past 3 decades: the low temperature loop expanded from the third to the sixth ring; the intensity and scope of the heat island effect within the second ring increased gradually.