水力发电学报 ›› 2021, Vol. 40 ›› Issue (2): 121-130.doi: 10.11660/slfdxb.20210212
基于Landsat卫星遥感资料的河流水温反演研究
Derivation of river surface temperature from Landsat thermal infrared data
摘要： River surface temperature (RST) is a crucial factor of river environment and ecology. Currently, methods to retrieve the surface temperature of large waterbodies in oceans and shallow lakes are relatively mature, but RST estimation lacks widely accepted approaches since its accuracy depends on the boundary effect of river banks and the stochastic factors of river channels. This study converts Landsat thermal infrared images to RST data using the radiation transfer model for two study areas, Cuntan and Miaohe in the Three Gorges reservoir region, and eliminates the boundary effect of the data through an enhancing water body retrieval method. We adopt three data windows of different types for RST data extraction and compare them with in-situ measurements. Results show that the removal of the boundary effect reduces the errors in RST by 0.2 ℃; and that compared with the single raster selection, the square data window selection method reduces the errors by 0.1 ℃. The RMSEs for the two study areas are comparable, and Landsat-5 that is more accurate than Landsat-7 has errors in the range of 0.87 - 0.91 ℃. This study helps RST analysis in river engineering and promotes application of Landsat thermal infrared images.