An experiment was conducted successfully to retrieve the leaf temperature of grass canopy by applying LSF conceptual model on data obtained on August 6

1999 from AVHRR and synchronized ground observations with the known soil surface temperatures beforehand.The difference between the retrieved leaf temperatures from these two different data sources is less than 0.2k

which means that the grass canopy leaf temperature can be retrieved satisfactorily through LSF conceptual model combined with satellite data.Further analysis on the retrieved leaf temperatures under different leaf and soil emissivity conditions reveals that the estimated accuracy of component emissivity shows little impact on the retrieved results

whereas the estimated accuracy of the precipitable vapor in the whole layer of the air column could affect the retrieved leaf temperatures significantly.The precipitable water vapour estimated based on the sounding data and dew-point temperature can lead to an error up to 1.1k in the retrieved leaf temperatures.When there exists sufficient vapour in the air

the variance of leaf temperature is only 1.22k in spite of the quite intensive solar radiation during 11.00—17.00 on fine days.In contrast

the variance of soil surface temperature can be 9.5k and the variance of air temperature at the height of 1.5m is 6.0k

therefore the variance of leaf temperature is much less than that of the soil surface temperature and air temperature.The experimental results also show that in the southeastern Tibetan plateau

the land surface temperature can be retrieved with quite high accuracy by Becker and Li’s split window algorithm.