An algorithm was developed to retrieve aerosol optical depth (AOD) from AVHRR over land for climate change studies. The multiple regression method was employed by using spatially and temporally collocated MODIS AOD and AVHRR Level lb measurements to generate regression coefficients from 2008 to 2011. Validation with AERONET AOD measurements and comparison with MODIS AOD product have been conducted to evaluate the uncertainty of AVHRR AOD retrievals over land from 2003 to 2007. It is found that the distribution pattern of the seasonal mean AOD is reasonable. The AOD from AVHRR has a tendency to be consistent with that from MYD04. Compared AVHRR with AERONET in Beijing and Xianghe, the correlation coefficients are larger than 0.6. These results show that the multiple regression method has the potential to retrieve a long-term AOD climatology data from AVHRR over land. Error sources about this retrieving algorithm were also discussed.