Applying Geo-data Mining to Analysis Spatial Variance Characters of Urban Land Surface Temperature
运用遥感数据挖掘解析城市地表温度的空间变异规律

In this paper,the main city of Shanghai is chosen as the study target,based on Landsat ETM+ images.Using Mono-window Algorithm to calculate land surface temperature,spatial variance of land surface temperature is analyzed by applying the Geostatistical method.The results show that the spatial difference of land surface temperature is significant and asymmetric distribution of east-west direction.The land surface temperature of Pudong area is obvious lower than Puxi area.The lowest temperature zone is located at the junction of the southern part of Pudong District and Nanhui District,but the highest temperature zone is distributed in Puxi area,including city centre along the Huangpu River,the southern of Baoshan District and northern of Putuo District.The land surface temperature is between highest and lowest temperature in other districts.From the view of the spatial variance of land surface temperature,at the lesser spatial scales,spatial heterogeneity caused by random factors is larger proportion of the total,and spatial autocorrelation is unconspicuous.With the scale augmenting,the spatial heterogeneity caused by random factors share decrease and spatial autocorrelation increase.When the grain size is 180m and 540m,the land surface temperature has the moderate degree of the spatial autocorrelation,and when the grain size increases to 1080m,the spatial autocorrelation of land surface temperature is obvious.