Urban green has diverse forms and functions. It provides many kinds ofecosystem services for cities to maintain urban sustainability. However, urban land resources are always so limited and precious that too much land for urban green is not applicable. It is thus very important to maximize ecosystem services of limited urban greenby rational allocation of the existing urban green spaces. In this study, a case study was carried on part of Beijing. QuickBird (QB) data of Beijing on July 5, 2002 was selected for urban green extraction, and ETM + data of Landsat 7 on July 9, was chosen to retrieve land surface temperature (LST). Correlation analysis was investigated between urban green LST, old island intensity and different urban green pattern. 6 different urban green types-Woodland,Shrubland,Grassland,Wetland Ⅰ, Wetland Ⅱ, and Cropland were interpreted, and three kinds of pattern parameters were analyzed, including size, shape, and adjacency relationships. Size was represented by area of urban green patches. Shape was indicated by shape index SI=P/(2 √πA ), where, P and A is the parameter and area of green patches respectively. Adjacent relations were presented both by relative green patch numbers in Neighborhood- Neighborhood Mean(N_MN) as well as Green Percentage(GP) in the Neighborhood buffers. Neighborhood Mean was calculated from urban green vector data on Neighborhood mean tool of Patch Analysis. 15 m to 90 m buffers were created based on selected green vector data for counting Green Percentage(GP_15,GP_30, …,GP_90) and the mean LST in the Neighborhood buffers. Cool island intensity CII (CII_15,CII_30,…,CII_90)was derived by subtracting urban green LST from mean LST in the Neighborhood buffers. GIS Mapping and statistical analysis was carried out after all urban green pattern parameters, LST and CII data prepared.Results show that park with Waterland presented cooler island than park simply with vegetation, and that different pattern parametersof different urban green types affect green LST differently. The LST of Main green types in the downtown including Woodland,Waterland (Ⅰ and Ⅱ) and Grassland all have asignificantly negative correlation to its area,and Waterland (Ⅰ and Ⅱ),has the strongest correlated coefficient; while for Shape index, only LST of Woodland, Waterland (Ⅰ and Ⅱ) showed significant negative correlation to SI and only LST of Cropland in the suburban area and all Woodlandshowed positive correlation to N_MN. The temperature of the Shrubland showed no correlation with its Area or SI or N_MN. Woodland buffer analysis showed that the temperature of Woodland was not only affected by its size and shape, but also by adjacent Green Percentage, and the strongest correlation negative was with the adjacent Green Percentage.The CoolIslandIntensity (CII)of neighborhood buffers was mainly affected by the Green Percentage of the adjacent 15-30mbuffer of both sides.So CII_75 was most negatively correlated to GP_60, GP_75 and GP90,and CII_90 was strongly correlated GP_75 and GP_90. These results might contribute to urban green planning and management and green pattern analysis.