Time Spectrum Authenticity Validation Of Land Surface Temperature Inversion Products From The Domestic Hj Satellite Thermal Infrared Channel

High-precision Land Surface Temperature products have a wide range of application scenarios.LST is of great significance to forestry breeding,agricultural farming,weather forecasting,urban heat island research and other fields.The time continuity and spatial coverage of the surface temperature obtained by traditional ground observations are limited,which cannot reflect the true temporal and spatial distribution of the surface temperature.For the acquisition of large-scale surface temperature information,thermal infrared remote sensing is the only possible means to meet this demand.The HJ-1B thermal infrared channel data has high time resolution and is easy to obtain,providing us with a new data source for obtaining the surface temperature.At present,although the HJ-1B satellite is used to invert the surface temperature and the quality inspection of its results has been carried out,the MODIS surface temperature products are usually used to compare and analyze the inversion results.There is no quantitative analysis,which limits the application of surface temperature inversion products.In response to this problem,this article takes part of multiple underlying surfaces in the Heihe River Basin as the research area,uses HJ-1B data from 2012 to 2018 as the research data,and uses universal single-channel algorithm to invert the surface temperature of the different underlying surface research areas.The inversion results were verified using the ground observation site data of the Heihe River Basin and MODIS surface temperature data.Due to the long data time series used in this study,the dynamic time warping method in the similarity measurement is introduced when evaluating the accuracy of the time-spectrum dimension of time-series data.To verify whether DTW(dynamic time warping)and other evaluation indicators are consistent in the evaluation of timing accuracy,provide a reliable basis for the application of HJ-1B satellite surface temperature inversion products.The main content and relevant conclusions of this study are summarized as follows:(1)Use a universal single-channel algorithm to invert the surface temperature of the HJ-1B satellite data covering the study area,and use the measured data from the ground station to verify and analyze the accuracy of the inversion results.The results show that the inversion product has the best accuracy in the desert and Gobi areas,with a relative deviation of less than1 K and a root mean square error of about 0.5K.For vegetation and crop areas,the deviation is within 2K,and the root mean square error is 1?2K.For wetland areas with complex land cover types,the relative deviation is within 3K,and the root mean square error is within 3?4K.(2)Verification of the inversion results using MOD11A1 surface temperature shows that the inversion accuracy is the highest for sites where the underlying surface is desert or desert,with a deviation of 1?2K.For sites where the underlying surface is vegetation or crops,the difference between the inversion result and the verification data is about 2K.For sites where the underlying surface is wetland,the inversion accuracy is low.The verification accuracy of MODIS LST products is slightly lower than that of ground-based verification,but the difference between the verification results of the two methods is between 1 and 2K,and the results are consistent.(3)The underlying surface in the study area is rich in types and contains a lot of vegetation.Analyze the inversion results of different underlying surfaces with seasonal changes.Through the analysis of the time spectrum of the inversion results,it is found that the sites with vegetation and crops on the underlying surface have lower inversion accuracy during the vegetation growing season,and higher inversion accuracy in other seasons.It may be because the surface emissivity changes greatly during the vegetation growing season,which has an impact on the calculation of the inversion parameters.Therefore,when performing the inversion prediction of the surface temperature of the vegetation coverage area,the images of the non-growing season of vegetation should be selected.(4)The research data has a long time sequence,and there is a time mismatch between the HJ-1B satellite data and the verification data.In response to this situation,dynamic time warping is introduced to evaluate the time sequence similarity.The results show that dynamic time warping is consistent with other existing evaluation indicators and can evaluate data that does not match the timing.This method can make full use of abundant verification data.Therefore,we can try to use DTW to evaluate the timing accuracy of remote sensing images.In summary,the surface temperature is retrieved from the HJ-1B satellite data,and the results are verified and analyzed.The results show that the accuracy of the universal singlechannel algorithm can meet the requirements as a whole.The accuracy is not ideal for areas where the type of land cover is more complex and changes a lot.The scale mismatch of the verification data and the inversion data will affect the verification results.These issues will be further studied in the next work.