Based on the remote sensing image data of 2005, 2010 and 2015, the spatial distribution of urban land surface temperature was studied by using the IB algorithm in the study area of Haidian District, Beijing. The data assimilation algorithm EnKF-3DVar and CA/Markov model integration were used to simulate the urban surface temperature in Haidian District by assimilating the spatial distribution data of the annual mean ozone concentration. The results showed that the urban surface temperature in Haidian District showed a downward trend in the past 10 years, and then showed a rising trend. But its overall showed a downward trend. The average temperature in 2015 was 31.1393℃. The prediction model of EnKF-3DVar can significantly improve the simulation precision of the model, and the Kappa coefficient of the predicted data in 2015 was 0.8216. Under the model of urban park green space, the urban surface temperature showed a decreasing trend. In the absence of urban green space park, urban surface temperature had a clear trend of expansion. The maximum temperature reached 56.1423℃, and the urban ecological green space had a great influence on the spatial distribution of urban surface temperature. Rational layout of urban green space was of great significance. The urban green space had a very large effect on the land surface temperature. In the process of urban green space construction, the construction of the green space network should be strengthened, and in the area of high land surface temperature in Haidian District, a large green plate should be built. The research result can provide technical support for the current and future urban green space planning and regional surface temperature mitigation.