Research On Landsat Surface Temperture Data Reconstruction Strategy Based On Spatio-temporal Fusion Technology

Land surface temperature is the key parameter in the process of land surface system on the regional and global scales,and it is the direct driving factor of surface-atmosphere energy exchange.The surface temperature with high time resolution and high altitude resolution is of great significance to the study of climate change,surface heat balance,drought monitoring,urban heat island effect and so on.However,there is no thermal infrared satellite remote sensing platform with high spatial and high temporal resolution,which greatly limits the application of continuous monitoring of ground objects.How to use the existing satellite remote sensing data to obtain high-time-series surface temperature information based on current science and technology is a crucial issue.This paper takes the Gujiao Area of northwest Taiyuan City,Shanxi Province as the study area.Taking the MODIS and Landsat-8 satellite data of the medium resolution imaging spectrometer as the main data sources,two different fusion strategies are proposed to generate the surface temperature: first fusion-later calculation and first calculation-later fusion.Thus,the MODIS temperature data is reduced to the spatial resolution of Landsat-8.The experimental process adopts the STARFM(Spatial and Temporal Adaptive Reflection Fusion Model)proposed and two typical data reconstruction model algorithms of ESTRFM(Enhanced Spatial and Temporal Adaptive Refelectance Fusion Model)proposed.And the applicability of the two fusion methods to the temperature is analyzed and analyzed Research indicates:The conclusions of this paper are as follows: the two reconstruction algorithms are applicable to surface temperature fusion prediction from the overall analysis,and the fusion accuracy of STARFM is higher than that of ESTARFM fusion algorithm.From the point of view of fusion strategy,the accuracy of "first fusion-later calculation" fusion strategy is higher than that of "first calculation-later fusion".From the point of view of data source,the data fusion accuracy of the time when the seasonal phase is similar and the temperature is similar is higher than that of the seasonal phase,and the fusion accuracy is large when the temperature difference is large.From the point of view of time,when the vegetation type changes,the temperature sensitivity is high and the time sensitivity is low.