| The development of remote sensing technology provides a means for the acquisition of regional surface temperature data.Currently,the surface temperature images frequently referred to are mainly divided into two categories according to their spatial resolution:high spatial resolution surface temperature data and low spatial resolution surface temperature data.High temporal and high spatial resolution can not be used to solve the problem,can be high spatial and temporal resolution of the sub-pixel surface temperature data,so downscaling research has become a hot spot in the field of remote sensing data applications.In this paper,based on the research area which includes the Yingke irrigation district of Heihe River Basin in Zhangye City of Gansu Province and the urban area of Beijing,and using the Yingke irrigation district for the surface temperature downscaling,we primarily apply a total of 5 phases of ASTER data from July to September 2012 downloaded from Heihe watershed allied telemetry experimental resaerch and the MODIS data of the same period downloaded from the NASA official website to our research.Beijing urban area is worked on the study of the application of the surface temperature downscaling method to the urban heat island effect.It mainly uses the sunny,cloudy or cloudless images from June to September in every year of 1985 to 2015 downloaded by the USGS,as well as the missing year's images generated by the MODIS data from the NASA official website from June to September of the missing year.Specific research results are as follows:(1)Comparison of land surface temperature downscaling method and assimilation study:Based on a total of 5 phases of ASTER data and MODIS data from July to September 2012,we use three means,namely the statistical experience method?the enhanced spatial and temporal adaptive reflectance fusion model(ESTARFM)and the flexible spatiotemporal data fusion method(FSDAF),to generate 90 meters?180 meters?270 meters?360 meters?450 meters of five scales and different dates of the sub-pixels surface temperature,and finally put forward the assimilation strategy for generating sub-pixel surface temperature contrary to the three methods used in the paper.From the visual point of view,the three methods are effective to improve the resolution of the original MODIS LST products,effectively improve the texture of MODIS temperature products,the three methods of statistical experience to reduce the scale of the results of the lowest precision,serious block phenomenon,and ESTARFM method and FSDAF method is high accuracy,and with the original ASTER LST products difficult to distinguish from the naked eye.(2)Contrast analysis and verification of the downscaling methods:The land surface temperature of the subpixel predicted by the three downscaling methods used in the paper is verified and analyzed by the ground point temperature observation data and the ASTER temperature product data respectively.The accuracy of the verification is expressed by the three factors:R2(correlation coefficient)?RMSE(root mean square error)and MAE(mean absolute error).The results of the ground point temperature observation data show that the ESTARFM method predicts that the generated subpixel LST and the ground observation value R2 are higher than 0.6128,while the FSADF method predicts that the generated subpixel LST and the ground observation value R2 are higher than 0.5055.Verification with the ground point found that most of the site value is higher than the actual value of ground observation.ASTER temperature product verification results show that the statistical accuracy of the statistical method is the lowest,and the predicted sub-pixel LST value and the ASTER LST product value are higher than 0.25,MAE is lower than 5.55K and RMSE are lower than 7.14K.The accuracy of ESTARFM method is the highest,and the predicted sub-pixel LST value and ASTER LST product value are higher than 0.88(2012-08-18),MAE is lower than 1.73K(2012-08-02),RMSE are lower than 2.37K(2012-08-02);FSDAF method of precision center,R2 were higher than 0.85(2012-08-18-90),MAE were lower than 1.77 K(2012-08-02-90),RMSE are lower than 2.42K(2012-08-02-90).The accuracy of the statistical method is significantly lower than that of the FSARF method.The accuracy of the ESTARFM method is slightly higher than that of the FSDAF method.(3)Analysis of temporal and spatial variation of urban heat island effect in Beijing:By using the ESTARFM method for spatial and temporal fusion in the year of lack of Landsat surface temperature data,the summer surface temperature data of the continuous phase of the 31st annual period in 1985-2015 was obtained,and the supervised classification of the five Landsat data is combined with the relative percentage temperature grading strategy(RPGS)to obtain the surface temperature classification criterion.According to the criterion,the data of 31 periods are classified,and then the influence of the development of urbanization on the temporal and spatial changes of urban heat island effect in Beijing is studied from both qualitative and quantitative aspects.The results show that the whole study area has no large-scale accumulation of high-temperature area,replaced by sporadic distribution of small heat island area;plant area on the heat distribution is much greater than the residential area on the thermal distribution of the impact;most areas are building The temperature of the low-density and low-vegetation-covered areas is much higher than the temperature of the buildings with high sparsities and high vegetation coverage areas.. |
PDF Full Text Request