Retrieving land surface temperature from Landsat TM using different atmospheric products as ancillary data

Abstract

Land surface temperature (LST) is one of the most important variables associated with the main land surface processes. Previous researches demonstrated the atmospheric heterogeneity at regional scale and advocated that it is vital to get atmospheric parameters properly to retrieve LST precisely by using satellite remote sensing, especially for sensors provided with one thermal infrared channel (e.g. Landsat TM/ETM+). In this paper, MODIS standard atmospheric products (MOD 05 and MOD 07) were selected primarily due to its proximate passing over time to Landsat TM and its relatively high spatial resolution, compared with other ones. Different types of the total atmospheric water vapour (AWV) content were extracted from MOD 05 and MOD 07. At the same time, an interpolating procedure was proposed to estimate the land surface air temperature from MOD 07 air temperature profile. Then these parameters' performances were test in improving the accuracy of LST retrieved from the TM thermal band by means of the single-channel algorithm and the mono-window algorithm respectively, followed by discussions on LST validation processes. The validation results suggest that different atmospheric products and the interpolated parameters are useful for obtaining LST accurately from the single thermal band when local soundings are not available. In particular, the AWV content estimated from the near infrared (NIR) data, stored in MOD 05 provided relatively more accurate LST results when it was input into the single-channel algorithm with the root mean square error (RMSE) is about 1.3K. However, it is not within the scope of this study to find a superior retrieval method. At last, several issues related to the MODIS standard atmospheric products were suggested to be resolved, to enable more scientific and practical uses of this kind of products for atmospheric correction.