Improved estimation of aerosol optical depth from Landsat TM/ETM+ imagery over land

The radiation from the sun to satellites in the sky is always modulated twice by atmosphere. Aerosol is one of the most active components in atmosphere and it usually contaminates the remotely sensed imagery severely so that most remotely sensed imagery cannot be used without atmospheric effect correction. However, the remotely sensed imagery is always the coupling of atmosphere and land surface information, which makes it very difficult to decouple the remotely sensed information to retrieve accurate atmospheric information and land surface information respectively from remotely sensed imagery alone. Based on the physical mechanism of radiative transfer model, many researchers assumed the specific surface condition, such as dark objects and invariant objects, so the atmospheric information like aerosol optical depth (AOD) can be decoupled from remotely sensed information. Since these assumptions are just for some specific surface conditions, the atmospheric information of many other surface conditions, such as sparsely vegetated areas and snow covered areas, are not available. In this study, we develop a new aerosol estimation algorithm that can effectively estimate the spatial distribution of atmospheric aerosols from TM/ETM+ imagery under general atmosphere and surface conditions. The estimated aerosol optical depth from this algorithm is validated by Aerosol Robotic Network (AERONET) measurements. The case study in Beijing, China indicates that this algorithm can retrieve aerosol optical depths from TM/ETM+ imagery reasonably well. Moreover, this algorithm has the potential to be applied to some new satellite images with moderate to high spatial resolution, such as Huan Jing (HJ) and China-Brazil Earth Resources Satellite (CBERS) series of China.