Product Fusion Of Land Surface Temperature Based On Generalized Three-cornered Hat Method

Land surface temperature(LST)is a very important characteristic physical quantity that characterizes the change of land surface process.It is the direct driving factor of the energy exchange between surface and atmosphere.It is widely used in the research fields of land surface energy balance,climate change and resource and environment monitoring.Accurate estimation of LST is essential for a better understanding of changes in regional climate characteristics and plays an important role in agriculture,forestry,ecology,global change and other fields.In particular,the daily average and monthly average surface temperature data have been the focus of research by scholars at home and abroad,so many daily average and monthly average surface temperature products have been produced.These LST products have different advantages and limitations due to the observation error of data sources,the imperfection of forecasting methods,the existence of random factors,and the difference of algorithms between products,so that LST products are uncertain.However,it is important to make a comprehensive assessment of the uncertainty of land surface temperature products in the use of land surface temperature products.Therefore,this study takes the global land as the research area and adopts a precision evaluation method independent of measured data,the Three-Cornered Hat Method(TCH),to analyze the uncertainty of four land surface temperature products.The seasonal changes of their uncertainties were discussed in different time Windows,and the LST products were fused by Bayes-based triangular hat fusion method(BTCH)to improve the estimation of LST.In order to further illustrate the reliability of TCH method and the advantages of BTCH.196 sites around the world were used to evaluate the results of BTCH fusion and individual LST products to demonstrate the reliability of the TCH method,and to further demonstrate the superiority of the BTCH fusion method by comparing the accuracy of BTCH and simple average results.The main conclusions are as follows:(1)Zhan and Liu have similar spatial distribution of uncertainty,and have greater uncertainty in the northern agro-pastoral ecotone.The uncertainty of ERA5 is higher than that of the other three products on the whole,especially in the polar region.Chen has a low uncertainty on the whole.(2)For the uncertainties obtained in different time Windows,the overall uncertainty in autumn and winter is higher than that in spring and autumn.The BTCH results based on five different time Windows and site verification show that the fusion results in the 6-month window are the best.It is compatible with numerical problems caused by samples and can also make use of seasonal variations of product uncertainty of land surface temperature,so as to achieve the effect of reducing random errors.(3)Assessment of BTCH,LST products,and simple arithmetic averaging integration using global land measurement sites.The results show that BTCH is better than a single LST product at 196 sites,and both simple averaging and BTCH can improve LST product.Overall,the BTCH results are slightly better than the simple average results,but with the completeness of the sample size of the sites,the BTCH results are significantly better than the simple average results.