Study On Near Surface Urban Heat Island Effect Of Beijing City Based On Remote Sensing

Monitoring of urban heat island effect by remote sensing data is one of important means to monitor urban heat island effect. The traditional remote sensing monitoring method is mainly focused on the analysis of urban land surface temperature or brightness temperature, which reflects the surface urban heat island effect. There are few studies focusing on urban heat island effect represented by near surface temperature. This paper takes Beijing as the study area, and uses the Landsat5/TM remote sensing image of Beijing in September 8, 2004 and July 26,2011 to extract the land surface temperature (LST) from remote sensing images by single channel algorithm. Normalized difference vegetation index (NDVI),modified normalized difference water index (MNDWI), surface albedo (Albedo) and normalized building index (NDBI) were calculated, with the addition of Altitude and land surface temperature (LST), which were taken into random forest model and multivariate regression model to obtain the near surface high temperature in Beijing.Comparing with multivariate regression model, the effect of near surface air temperature retrieved by random forest model is better, which is selected to analyze urban heat island effect in Beijing. Based on the analysis of the spatial distribution and temporal variation characteristics of near surface heat island effect in Beijing City, the influence of the underlying surface and vegetation on the urban heat island effect was discussed. The main conclusions are as follows:(1) The forest model is innovatively introduced to estimate the near surface air temperature, and the surface temperature data with high accuracy are obtained. The mean absolute error (MAE) and root mean square error (RMSE) of the near surface temperature in 2004 is 1.11 ℃ and 2.02 ℃ respectively; The mean absolute error(MAE) and root mean square error (RMSE) of the near surface temperature in 2011 is 0.82 ℃ and 1.09 ℃ respectively. Compared with the random forest model, the mean absolute error (MAE) of the multiple regression statistical method is 1.18 ℃and the root mean square error (RMSE) is about 1.82 ℃ .In the process of calculating of near surface air temperature with the random forest model in September 8, 2004 and July 26, 2011, land surface temperature (LST) is of greatest impact to the near surface estimation model error. The altitude and normalized building index (NDBI) is of minimum impact to random forest model error.(2) Analysis of urban heat island effect by near surface temperature that is calculated by random forest model Beijing urban heat island effect is obvious, The center of the city is a high temperature area, the surrounding suburbs are lower temperatures,Beijing city in 2011, urban heat island effect is more serious than urban heat island effect in 2004.1n 2011, Haidian District, Chaoyang District, Fengtai District has become a high temperature zone and sub high temperature zone,Fangshan District, Daxing District and Pinggu also have high temperature and sub high temperature phenomenon .Low temperature areas are located on the edge of the city such as Yanqing and Mentougou.(3) The relationship between the urban heat island intensity and the underlying surface can be summarized as follows: the heat island intensity of urban area is the highest, the bare land is the second, and the distribution of forest land is the lowest.To further explore the vegetation and impervious face of urban heat island effect’s influence,there is significant negative correlation between the normalized difference vegetation index and the intensity of heat island. With the increase of normalized difference vegetation index, the intensity of heat island effect gradually weakens.There is a positive correlation between the heat island intensity and the urban impervious surface coverage. With the increasing of impervious surface coverage,the heat island intensity gradually increases.