Thermal infrared images observed by satellites are affected by atmospheric absorption and emission. Due to a lack of data, previous studies used at most one atmospheric vertical profile to retrieve surface temperature. In this study, atmospheric correction was performed using MODTRAN, with which the effects of the spatial distribution of atmosphere were considered. Five summer scenes and four winter scenes of Landsat-5/TM and -7/ETM+ (Kanto scene, Path 107/ Row 35), aerological data at Tateno, and regional objective analysis data with ca. 20 km spatial resolution were used. Effects of atmospheric correction were evaluated from the surface temperature estimated by upward longwave radiation at Tateno. Also, the effects of spatial variations of atmospheric transmissivity on surface temperature were evaluated.
The regional objective analysis data were adequate for input to MODTRAN. The differences among atmospheric corrections estimated from aerological data at Tateno and regional objective analysis data, were large in summer (ca.
-5.2-12.0°C) and small in winter (ca.
-1.9-0.9°C).The differences increased with both surface temperature and atmospheric water vapor content. Considering the accuracy of MODTRAN (ca.
1-2°C), it is necessary to take into account the spatial distribution of the atmosphere in summer. In winter, aerological data at Tateno represented the condition of the whole of the Kanto scene; therefore, it is not necessary to consider the spatial distribution of the atmosphere for atmospheric correction.
In summer, the spatial variation of atmospheric transmissivity of 0.01 was comparable to 1°Cof the surface temperature. While in winter, the spatial variation of atmospheric transmissivity of 0.04 was equivalent to 1°C of the surface temperature. These results were clarified by the present study for the first time.