Land Surface Temperature (LST) Retrieval And Land Surface Emissivity (LSE) Estimation From Landsat TM Data

The paper mainly focused on the technology of land surface temperature (LST) retrieval and land surface emissivity (LSE) estimation from only single thermal infrared band of Landsat TM data in Hangzhou City of Zhejiang Province using the Jimenez-Munoz et al.'s generalized single-channel method, the Qin et al.'s mono-window algorithm and the Malaret et al.'s absolutely land surface temperature method, and LSE estimated by NDVI method, with emphasis on the determination of seneral important parameter and equation expression in LSE estimation method and the comparation of LST retrieval. The LST was retrieved from TM6 data in case study area. The results showed as following:(1) The LSE of Landsat/TM 6 channel was estimated by NDVI method from TM3 and TM4. It is necessary to carry out the atmospheric correction in order to obtain at-surface emissivity. In this paper, we used the Chavez method-COST model because external data to the satellite is not needed.The MODIS LSE products (error:0.01) in the same time as TM were used to validating estimated TM LSE .The results found the accuracy of TM LSE estimated by NDVI method is 0.01, and showed the NDVI method was feasible.(2) In order to compare the accuracy between the different retrieved LST algorithms, MODIS LST products which the accuracy was 1K (under 0.5K for uniformity area) was used to verify the retrieved LST results. Then the extracted three uniformity regions was compared with MODIS LST,including forest, cropland and urban areas , and the extracted LST values were carried out a linear regression with the TM LSTs retriveded by Jimenez-Munoz et al.'s , Qin et al.'s and Malaret et al.'s ,with zero intercept function. The findings from this investigation showed that the TM LST results obstained by Jimenez-Munoz et al.' generalized single-channel method was better than other algorithms.(3) Moreover, the study also tried to apply the MODIS LST improving the retrieved TM LST results using Jimenez-Munoz et al.' generalized single-channel method, and obstained a total linear regression with zero intercept function. The results corrected by the function wascompared with MODIS LST. The error was 0.6K between TM LST and MODIS LST for uniformity forest area. The results showed the corrected function improved the retrieved land surface temperature from Landsat TM.