ATMOSPHERIC CORRECTION OF LANDSAT TM THERMAL BAND USING 
SURFACE ENERGY BALANCE 
Alain VIDAL*, Claire DEVAUX-ROS*, M. Susan MORAN** 
* CEMAGREF-ENGREF Remote Sensing Lab., BP 5095, 34033 MONTPELLIER Cedex 1, France 
** USDA-ARS U.S. Water Conservation Lab., 4331 E. Broadway Rd., Phoenix Az. 85040, USA 
ABSTRACT 
Thermal infrared (IR) data of Landsat Thematic Mapper (TM) have been hardly used during 
the last decade, probably due to the difficulties met when trying to correct them for atmospheric effects. A new 
method for correcting these data has been designed, based on surface energy balance estimation of known wet 
and dry targets included in the TM image to be corrected. This method, only using the image itself and local 
RMS error on corrected temperatures is on order of 1°C. 
KEY-WORDS : thermal infrared, Landsat Thematic Mapper, atmospheric correction, energy balance, 
forestry and environment monitoring, 120-m resolution thermal IR data of Landsat Thematic Mapper (TM) 
have been hardly used during the last decade. This may be explained by their low frequency (16 days) and by 
the difficulties met when trying to correct them for atmospheric effects. These effects can lead to errors on 
estimated blackbody temperatures of more than 10°C above 40°C (Wukelic et al., 1989). Published results on 
such atmospheric correction were usually based on ground measurements of surface temperatures and/or 
atmospheric profiles using radiosondes, coupled with linear or more complex radiative transfer models (Moran 
et al., 1989 ; Wukelic et al., 1989). Such measurements are not necessarily available on the day of Landsat 
overpass, particularly when using archived data. 
1985) and especially those based on the concept of Crop Water Stress Index (see Jackson et al., 1981, and more 
recently, Moran et al., 1994a, on sparse canopies) have shown that, for a given surface under given climatic 
conditions, surface temperature ranges from a minimum (corresponding to maximal évapotranspiration : 
LE=LEp) to a maximum (corresponding to no-evapotranspiration : LE=0). The objective of this study is to use 
these surface energy balance considerations on well-known targets (with minimal and maximal surface 
temperature) of the Landsat TM image to remove atmospheric effects. 
2 - THEORY AND METHOD 
2.1. Estimation of surface temperature from Landsat TM thermal IR data 
Wukelic et al. (1989) have shown that uncorrected surface temperature Tu (K) can be 
obtained with : 
- xiuiuwij wvpivoovu in in TT.vui .ai .fim . rva tnt muijjc ui is ai i hum lincili lui a uMiai i auge 
meteorological data, has been tested and validated on various surfaces : agricultural, forest and rangeland. The 
crop, forest, rangeland 
1 - INTRODUCTION 
Despite the existing demand of high resolution thermal infrared (IR) data in agriculture, 
Major studies on surface energy balance and on its relation with surface temperature (Perrier, 
( 1 )