Land surface temperature retrieval from Landsat 8 OLI/TIRS images based on back-propagation neural network

Abstract

Land surface temperature (LST) is an important parameter related to the environmental assessment concerning the surface energy, water balance, greenhouse effect, etc. at local and global scales. With the rapid development of the remote sensing technology, various methodologies have been developed to retrieve LST from space-based remote sensing images. Due to the ill-posed problem, the LST retrieval is still a challenge. In this research, a so-called multiple band reflectance (MBR)-LST model has been proposed based on the back-propagation neural (BPN) network, which can be employed to retrieve the LSTs from Landsat 8 Operational Land Imager (OLI)/TIRS images as well as produce continuous spatial LST distributions with a spatial resolution of 30 m. Experiments conducted in two randomly selected areas in mainland China proved that the proposed MBR-LST model has yielded a much better performance than the traditional radiative transfer equation (RTE) method with respect to both the accuracy and stability for the LST retrievals. Moreover, another significant advantage of the proposed MBR-LST is the generic nature – once trained by the sample data in the whole region of mainland China, the proposed MBR-LST model can be utilized for the accurate LST-retrievals in any area of mainland China.