Methodology Development For Land Surface Temperature Retrieval From Landsat 8 Thermal Infrared Data Over Mountainous And Cloud-covered Areas

Land surface temperature(LST)plays the crucial role for energy dynamic balance,which is widely utilized in hydrological research,meteorological prediction,evapotranspiration research,crop disaster assessment and urban heat island effect.LST with high spatial resolution is of great significance for ecological environment monitoring and regional or global scale climate change research.As the satellite sensors are improved and developed,the acquisition of high-resolution data provides an effective data support for the study of LST retrieval.Aiming at the current situation of topographic effect correction and the lack of LST contaminated by cloud and taking Landsat 8 thermal infrared data,a single channel algorithm based on the thermal radiative transfer equation is developed to retrieve the mountain surface temperature corrected by topographic effect,and the high spatial resolution LST under cloud is reconstructed from the annual temperature cycle model and the principle of spatial similar pixels.The contents and conclusions of this study are as follows:(1)Based on the traditional thermal infrared radiative transfer equation,the terrain parameters such as slope,aspect,and sky-view factor(SVF)are introduced to construct the mountain thermal radiative transfer model,which not only follows the thermal radiative energy transfer mechanism,but also aims at the geometry structure of mountainous area.By parameterizing the terrain parameters such as SVF in the mountain radiation transfer model,the influence of reduced atmospheric downward radiation and increased adjacent terrain radiation on mountain surface temperature retrieval is quantitatively analyzed.The results show that the key parameters affecting the retrieval of land surface temperature include SVF,atmospheric water vapor content,target pixel emissivity and average LST of adjacent pixels.(2)A single channel algorithm based on thermal infrared radiative transfer equation is developed to retrieve the mountain LST corrected by topographic effect.The results show that the terrain effect is greater in the valley where the SVF is smaller,and the LST retrieved by the traditional model can be overestimated by 1 K compared with the mountain radiation transfer equation.When the SVF is less than0.7,topographic effect correction must be considered in mountain LST retrieval.In addition,in the three sub regions with different degrees of topographic relief,the mountain radiative transfer model constructed has good spatial consistency with the brightness temperature obtained by DART forward simulation,and their RMSE are all less than 0.25 K.(3)The annual average surface temperature,annual amplitude and annual phase are obtained by inputting the LST retrieved from Landsat 8 into the annual temperature cycle model through the single channel algorithm,and the model reference LST is obtained according to the day of year.Based on the principle of spatial similar pixels,the difference between the LST of Landsat 8 and the model reference LST under clear-sky condition is used to deduce the difference between the LST contaminated by cloud condition and the model reference LST to get the LST under cloud.According to the simulation results under clear-sky conditions,when the search radius ranges from 0.5 km to 5 km,the RMSE is less than0.95 K;and when the search radius is fixed on 5 km,the RMSE is less than 1.3 K with different missing regions.