TM热波段图像的地表温度反演算法与实验分析

目前利用LandsatTM热波段数据反演地表温度有3种算法:辐射传导方程法、单窗算法和单通道算法。辐射传导方程法由于计算过程复杂且需要实时大气剖面数据,因而实际应用较为困难。单窗算法和单通道算法对Landsat热波段反演地表温度能获得较高精度。单窗算法所需的大气参数包括近地表气温和大气水分含量,单通道算法所需的大气参数仅为大气水分含量。地表辐射率为这两种算法共有的关键参数。本文以福建省福州市为研究区,使用1989年6月15日LandsatTM数据,利用单窗算法和单通道算法对研究区进行地表温度反演,并将这两种算法的反演结果与研究区反演的亮度温度进行了比较,结果表明:(1)两种算法反演的结果总体趋势比较接近,但单窗算法的结果相对于单通道算法较低,二者相差约2.45℃;(2)两种算法的结果与亮度温度相比,单窗算法要高出约2.84℃,而单通道算法则要高出约5.28℃。

Comparison of Two New Algorithms for Retrieving Land Surface Temperature from Landsat TM Thermal Band

Landsat TM has a thermal infrared band (band 6) which can be used to retrieve land surface temperature (LST). So far, three methods have been proposed to retrieve LST from TM6 data, they are Radiative Transfer Equation (RTE), Mono-window Algorithm and Single-channel Method. Due to the complicated process of computation and the general unavailability of in situ atmospheric profile at the satellite pass (especially for the images in the past), it is quite difficult to put the RTE method into real world application. The advent of the other two new algorithms has made it possible to retrieve the real LST from Landsat TM thermal band and with relatively high accuracy. The mono-window algorithm requires two atmospheric parameters (the total atmospheric transmissivity and the atmospheric water vapor content) while the single-channel algorithm only requires one (the atmospheric water vapor content). The land surface emissivity is the common key parameter of these two algorithms. In this paper, we make a comparison between the LSTs retrieved from the mono-window algorithm and the single-channel algorithm over a study area in Fuzhou city, Fujian, China. The Landsat TM image used in this paper is acquired on June 15, 1989. The research results show that: (1) when compared to each other, the retrieved LSTs from these two algorithms have similar overall temperature distributions, but, the LST retrieved from the mono-window algorithm is about 2.45℃ higher than that from the single-channel algorithm; and (2) when compared to the corresponding pixel brightness temperature, a higher difference value is obtained. For the mono-window algorithm, it is about 2.84℃, and for the single-channel algorithm, about 5.28℃.