Research And Verification Of Generating High Spatial And Temporal Resolution Surface Temperature Based On Fusing Multi-source Remote Sensing Data

Land surface temperature(LST) is an important parameter of water balance and energy balance. Thermal infrared remote sensing is an important means to obtain the surface temperature information of large area. Currently, the data can be divided into two categories: one with high spatial resolution/low temporal resolution, such as TM, ETM +, ASTER. This data can provide detail information but revisit periods are 16 days; another with low spatial resolution/high temporal resolution, such as AVHRR and MODIS. However, the spatial resolution of the data which has high temporal resolution and regional quality is below to 1 kilometer. Because of the irreplaceable value of the fusion of this two types of data in the application of the crop’s growth monitoring, many researches have focused on this question that how to fuse this two type of data together to get a suitable-spatial resolution and availabletemporal resolution data. However, to use this fusion data largely, are we need to validate the accuracy of them.Considering the spatial heterogeneity of land surface and the limitation of remote sensing products, in this paper,we propose a method which is called QSTARFM(quantitative spatial and temporal adaptive reflectance fusion model) to fuse this two types of LST data(one has high spatial resolution\lower temporal resolution, another has the coarse-spatial resolution but high temporal resolution) together to get this LST data which has high spatial and temporal resolution. After this, we evaluate the accuracy of the fusion LST data. In this paper, we choose the Middle Heihe River Basin as a test area and many multiple scales data in 2012 as the test data. By using this data, we propose the QSTARFM method to get high temporal and spatial LST data. Then validate the accuracy of the downscaling result systematically. Terminal, apply the method to the researches of evapotranspiration in cropland. The researchers are described briefly as bellows.(1) In the researches of get high temporal and spatial LST products, we used the multi-date ASTER data and MODIS data to describe the process of QSTARFM. By analyzing the proximity effect to confirm the suitable searching windows, searching the similar pixel one by one, calculating the weight of pixel, calculating the conversed coefficient of different scale to get high temporal and spatial scale LST data.(2) In the process of evaluate the high temporal and spatial LST data, we use the Analog Data, and ground measured LST data and standard remote sensing LST products to validate the accuracy of fusion LST products.(3) When using the fusion LST data in the research of evapotranspiration in cropland, we choose the LST data in two days(2012-08-18 and 2012-08-27) as targets. By using this downscaling LST data, other remote sensing image products and some meteorological parameters to build SEBAL model to get evaporation. Terminal, evaluate the accuracy of evaporation by using the eddy covariance system data.