Land surface emissivity (LSE) has already been recognized as a crucial parameter for the determination of land surface temperature (LST). There is an ill-posed problem for the retrieval of LST and LSE. And laboratory-based emissivity is measured in natural constant conditions, which is limited in the application in thermal remote sensing. To solve the above problems, the coupling of LST and LSE is explored to eliminate temperature effects and improve the accuracy of LES. And then, the estimation accuracy of LST from passive remote sensing images will be improved. For different land surface materials, the coupling of land surface emissivity and land surface temperature is various. This paper focuses on studying concrete surface that is one of the typical man-made materials in urban. First the experiments of measuring concrete surface emissivity and concrete surface temperature in natural conditions are arranged reasonably and the suitable data are selected under ideal atmosphere conductions. Then to improve the determination accuracy of concrete surface emissivity, the algorithm worked on the computer of Fourier Transform Infrared Spectroradiometer (FTIR) has been improved by the most adapted temperature and emissivity separation algorithm. Finally the coupling of concrete surface temperature and concrete surface emissivity is analyzed and the coupling model of concrete surface temperature and concrete surface emissivity is established. The results show that there is a highest correlation coefficient between the second derivative of emissivity spectra and concrete surface temperature, and the correlation coefficient is -0.925 1. The best coupling model is the stepwise regression model, whose determination coefficient (R2) is 0.886. The determination coefficient (R2) is 0.905 and the root mean squares error (RMSE) is 0.292 1 in the validation of the model. The coupling model of concrete surface temperature and concrete surface emissivity under natural conditions provides a new approach for improving the estimation accuracy of LST from passive remote sensing images.