Abstract: Surface fitting of Urban Heat Island (UHI) intensity provides a strong promise for deeply analyzing the spatial pattern, morphology and the evolution features of UHI. The surface fitting algorithm based on the support vector machine (SVM) technique was introduced. With the LST (Land Surface Temperature) products from MODIS and after elimination of the data covered by clouds, 757 imageries (274 and 483 ones for day and night, respectively) in Beijing metropolitan area from the year 2006 to 2008 were fitted one by one using SVM. The sensitivity analysis and accuracy assessment both indicate that this algorithm is of high accuracy and capable to be used in depicting the spatial pattern of UHI. Furthermore, by virtue of it, mapping of urban surface temperature becomes possible. The application results also show that, during daytime, the UHI intensity in Beijing is weakly and positively correlated with its rural background temperatures linearly; while during nighttime, the circumstance is on the opposite side. In addition, from an annually perspective, UHI capacities during daytime and nighttime in time series are both generally subjected to periodic sinusoidal variation; however, the amplitude of annual variation of UHI capacity during nighttime is far less than that during daytime. This is because different driving factors dominate different patterns of UHI temporally and spatially in distinct seasons and illumination conditions. The support vector machine (SVM) fitting of UHI intensity model presented in this paper pays more attention on the overall spatial features of UHI approximately, while ignoring the noise caused by random factors and the details of a weak surface temperature change; hence it is a powerful tool for investigating the spatial pattern of temperature distribution in the analysis of urban thermal environment.