Cloud detection is absolutely necessary in the processing of satellite remote sensing data. Through

analyzing meteoric characteristics of cloud in different spectra and integrating the spectral characteristics of

MODIS (Moderate Resolution Imaging Spectroradiometer), the authors put forward a new cloud detection method based on

multi-spectrum synthesis. Taking into account such factors as visible reflectance, infrared brightness temperature

and window brightness temperature difference, the algorithm can gradually build a cloud detection mask and finally

obtain an entire cloud processing result which can discriminate cloud from clear sky. The algorithm was applied in

different periods and different scenes to make validation and analysis. The results show that the cloud detection is

ideal, especially for thin cirrus which is invisible in the visible band. This technique can promote the use of

MODIS data and improve the accuracy of retrieving.