Cloudy-sky Land Surface Temperature Determination By Integrating Microwave And Optical Remote Sensing

Land surface temperature?LST?is a key parameter in the research of hydrological,climatology,and agriculture.As an important indicator of energy balance at Earth's surface and greenhouse effect,the quantitative retrieval of LST plays an important role in the study of hydrology,meteorology and ecology.With the rapid development of remote sensing,the widely used satellite data makes it possible to obtain the LST accurately.Current methods of retrieving LST mainly based on the thermal infrared and microwave remote sensing.Thermal infrared?TIR?measurements have been widely used with the advantage of high retrieval accuracy,mature algorithms,and relatively high spatial resolution.However,thermal infrared remote sensing is unable to penetrate clouds,and is heavily influenced by atmosphere.Thus,under cloudy condition,few LST measurements are available,which severely limits its practical applications.Passive microwave?PMW?remote sensing is attractive for retrieving land surface temperature,especially under cloudy condition.With the superiority of penetrating clouds,not relaying on the sun as a source of radiation,passive microwave remote sensing provides an alternative way for estimating LST with multi-polarizations and all-weather conditions.This paper is mainly about utilizing the advantages of optical and microwave remote sensing to inverse the land surface temperature.On the basis of briefly introducing the present situation of surface temperature retrieval and related concepts and theories.The typical inversion algorithms are summarized.Aiming at the shortcomings in previous studies,a series of researches on LST retrieval were carried out.?1?Validation of algorithm proposed by Holmes.Holmes utilized vertical polarization channel of 36.5GHz as the main parameter,setting₍,36.5?1?=259.8as the threshold to calculate the surface temperature by linear regression equation for the brightness temperature of₍,36.5?1?>259.8.In this paper,we evaluate the algorithm over the continental of the United States by inputing AMSR-E data,MODIS LST products,and ground measurements acquiring from SURFRAD sites to the algorithm.The validation results laid good foundation for further research.?2?Proposed a new LST retrieval algorithm.Data is the basis of the algorithm,we collected AMSR-E data,MODIS land surface temperature\emissivity products,NDVI products and ground measurements as main data of the training dataset.According to the pass-time of AMSR-E?1:30 AM/PM?,we divided the training dataset into two parts,one for day-time and the other for night-time.For each temporal part,we further extracted data of different weather conditions,including clear and cloudy conditions.In the previous studies,some researches only use the data under clear-sky condition,which is not representative for cloudy condition.In order to propose a representative method,we introduced ground measurements from 50 sites of Ameri-Flux to represent cloudy condition and MODIS LST products for clear condition.To suppress the error caused by seasonal change,we classified the land surface types by introducing NDVI products.Overall,the accuracy of the algorithm has been refined as compared to previous studies.?3?LST fusion method.The thermal infrared measurements are highly affected by clouds and atmosphere.This severely limits the application of TIR measurements,on average,60%of the land is covered by clouds.Clouds would not only induce spatial incontinuity but also cause temporal gaps,which severely limits their practical uses.As a complementation,passive microwave measurements are able to penetrate the clouds and,to some extent the rain.Aiming at making full use of the advantages of these two measurements,we proposed a fusion method by utilizing the cloud fraction to fuse TIR and PMW data together.This method provided potential to obtain continuous LST products.?4?Validation of the retrieval algorithm proposed in this paper.Under cloudy condition,we directly compared the retrieval LST with the in situ LST measurements at the satellite overpass time.Because of the large difference in spatial scale,the sites with homogeneous environment and flat terrain are selected as far as possible.Under clear conditions,cross validation is used to evaluate the algorithm based on the surface temperature products of MODIS.Due to the cloud contamination,the land surface temperature products are discontinuous.We compared temperature products before and after fusion to analyze the feasibility of the fusion method.Better agreements were observed over the dense vegetated area where NDVIs are greater than 0.6,with root mean standard error?RMSE?of 2.13-2.59K,followed by sparsely vegetated area.Higher discrepancies were from the area where NDVIs are less than 0.2.Overall,the algorithm accuracy has been refined as compared to previous studies which provides potential to be applied at global scale in all-weather condition.The temperature fusion algorithm combines the advantages of microwave and optical remote sensing and provides a reference for the collaborative applications of multi-sensor data.